U.S. flag

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

  • Publications
  • Account settings
  • Advanced Search
  • Journal List
  • Br J Cancer
  • v.124(4); 2021 Feb 16

A case-control study to evaluate the impact of the breast screening programme on mortality in England

Roberta maroni.

1 Centre for Cancer Prevention, Wolfson Institute of Preventive Medicine, Queen Mary University of London, Charterhouse Square, London, EC1M 6BQ UK

Nathalie J. Massat

Dharmishta parmar, amanda dibden, jack cuzick, peter d. sasieni.

2 Faculty of Life Sciences and Medicine, Cancer Prevention Group, School of Cancer and Pharmaceutical Sciences, King’s College London, Guy’s Campus, Great Maze Pond, London, SE1 9RT UK

Stephen W. Duffy

Associated data.

Data were saved on the servers of the Barts Cancer Institute, Queen Mary University of London, in a folder with restricted access to D.P. A clean, anonymised version of the data was produced and made available to R.M., A.D. and S.W.D. with restricted access to the staff of the Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis at Queen Mary University of London. The data were obtained via the Office for Data Release at Public Health England. We do not have authority to share the data with others, but requests for access to data will be forwarded to the Office for Data Release.

Over the past 30 years since the implementation of the National Health Service Breast Screening Programme, improvements in diagnostic techniques and treatments have led to the need for an up-to-date evaluation of its benefit on risk of death from breast cancer. An initial pilot case-control study in London indicated that attending mammography screening led to a mortality reduction of 39%.

Based on the same study protocol, an England-wide study was set up. Women aged 47–89 years who died of primary breast cancer in 2010 or 2011 were selected as cases (8288 cases). When possible, two controls were selected per case (15,202 controls) and were matched by date of birth and screening area.

Conditional logistic regressions showed a 38% reduction in breast cancer mortality after correcting for self-selection bias (OR 0.62, 95% CI 0.56–0.69) for women being screened at least once. Secondary analyses by age group, and time between last screen and breast cancer diagnosis were also performed.

Conclusions

According to this England-wide case-control study, mammography screening still plays an important role in lowering the risk of dying from breast cancer. Women aged 65 or over see a stronger and longer lasting benefit of screening compared to younger women.

Following an evaluation of several randomised controlled trials (RCT) 1 that showed an overall reduction in mortality from breast cancer in women undergoing mammography screening, the National Health Service Breast Screening Programme (NHS BSP) was launched in the United Kingdom (UK) in 1988. At the time, it aimed to offer free routine screening to every woman aged 50–64 once every three years. It now invites women aged 50–70, with an age extension to younger and older women (47–73 years) being trialled. 2

Over the last thirty years, major advances have been made in the fields of cancer screening, treatment, and management (including effective adjuvant systemic therapies 3 and two-view mammography 3 , 4 ), with resulting lengthening of survival times after a breast cancer diagnosis. 5 Despite recent reductions in breast cancer mortality, breast cancer is still the cancer with the highest incidence 6 and the second most common cause of cancer death 7 in females in the UK.

Case-control studies are a useful tool to evaluate screening programmes in settings where lack of equipoise would mean that RCTs would be unethical, or as in this case, where the RCTs have already been done, but there remains a need to ensure that the service is delivering the expected clinical benefit. Case-control studies also overcome some limitations associated with other observational designs by taking into account changes in cancer incidence and use of treatments over time and adjusting for any imbalances in other factors that could affect breast cancer mortality.

Taking as an example a case-control study 8 that resulted in policy change within the NHS cervical screening programme by altering age at first screen and the screening interval, we designed a similar study focussing on the NHS BSP with the aim of:

  • Evaluating the effect of mammography screening in the NHSBSP on breast cancer mortality
  • Evaluating the effect of mammography screening on breast cancer incidence, and incidence of late stage disease
  • Estimating overdiagnosis
  • Analysing the interplay of early detection, pathology, and treatment on fatality of breast cancer.

The study protocol and results from two pilot studies have been published previously. 9 – 11 This paper reports on the first objective above (breast cancer mortality), making use of England-wide data. Effects on incidence etc. will be reported in future papers.

Definition of cases and controls

As the main objective was to evaluate the effect of mammography screening on breast cancer mortality, cases were defined as women whose primary cause of death was breast cancer, who were diagnosed at age 47 years or older and died at age 89 years or younger in 2010–2011. We chose the lower limit of 47 as there is a major trial of screening in ages 47–49 ongoing, 2 so substantial numbers of women have been screened in this age group. We chose the upper limit of 89 because above this age we would not expect a major effect of screening taking place mainly at ages 50–70, because we were less confident of the cause of death in the very old, and because screening is essentially aimed at preventing premature mortality, which one might reasonably interpret as death below age 90 years. Only diagnoses occurring after 1990 were included in the analysis. Their matched controls were women sampled from the general population of those invited for screening (99.9% of women eligible for screening in England 12 ) and alive at the time of their corresponding case’s death. Controls may have been diagnosed with breast cancer, but not before their case’s date of diagnosis. Where possible, two controls were selected per case and matched on date of birth (within one month of the case’s) and screening area at date of diagnosis.

For the purposes of the statistical analysis, controls were assigned a date of pseudodiagnosis, equal to the diagnosis date of their corresponding matched case. To be eligible as a case or a control, a woman had to have had at least one invitation to screening prior to the date of diagnosis/pseudodiagnosis.

The primary endpoint was to estimate, among those invited to breast screening, the effect of ever attending breast screening on mortality from breast cancer. Changes in this effect over time were also investigated. Secondary endpoints included the effect of measures of screening intensity, such as time between last screen and diagnosis/pseudodiagnosis, and their estimations in different age subgroups.

Data selection and linkage

Cases were identified from the National Cancer Registration and Analysis Service (NCRAS) database accessed through the Office for Data Release of Public Health England (PHE). This database contains Office for National Statistics date and cause of death data. NHS Digital used the National Health Application and Infrastructure Services (NHAIS) system to identify matched controls and provided breast and cervical screening histories within.

We excluded any breast screens occurring outside the usual call/recall system of the national screening programme. All the screening histories of the study subjects were considered up to and including their date of diagnosis/pseudodiagnosis.

The data were processed according to the NHS Information Governance guidelines. 13

Sample size

Sample size calculations for the pilot study showed that, assuming an OR for breast cancer mortality of 0.7 and a number of discordant pairs of 33%, two controls per case with 800 breast cancer deaths and 1600 controls would confer more than 90% power to detect such an effect size at the 5% significance level using a two-sided test. 10 As the data for this main phase encompassed the whole of England, we had ample power, not only for the primary outcome (8288 cases and 15,202 controls after exclusions), but also for subgroup analyses.

Statistical analysis

Data were analysed using Stata version 13 14 by matched (conditional) logistic regression with death from primary breast cancer as the outcome. Date of birth and screening area were accounted for by the matching process.

Ineligible subjects were excluded (see Fig.  1 ). For some of these, this resulted in a matched set containing only a case, or only controls, which could then no longer be used in the matched logistic regression. Sensitivity analyses using unmatched logistic regression and controlling for age at diagnosis/pseudodiagnosis and screening area were performed on the same dataset with fewer exclusions; in this case, the inclusion criteria considered were the same, but the fact that a case or a control was excluded did not imply discarding that matched set.

An external file that holds a picture, illustration, etc.
Object name is 41416_2020_1163_Fig1_HTML.jpg

Asterisk indicates that these records were excluded for being in a 1:1 matched set where the case or the control was excluded or for being in a 1:2 matched set where the case or both controls were excluded. Hash indicates that these become 1:1 matched sets in the final dataset. Note: some records may be excluded for more than one reason.

Case-control studies used to evaluate population screening programmes are subject to a type of bias known as non-compliance or self-selection bias, which is based on the assumption that people who are already ill may be less likely to attend screening and those who do attend may be more health conscious, and therefore healthier, than those who do not take up the invitation. This may confer an artificially greater protective effect for screening, which was corrected in our analyses using a variant of the method by Duffy et al. 15

The effect of self-selection bias was estimated using data available on cervical screening attendance for the women in the study, on the basis that any observed protective effect of cervical screening on breast cancer death cannot be due to cervical screening (which does not include breast examination) and is therefore likely to be caused by self-selection bias. In particular, the odds ratio (OR) uncorrected for self-selection is an estimate of the relative risk:

An unbiased estimate of the effect of screening on risk of dying from breast cancer would be (refer to Duffy et al. 15 ):

The OR for death from breast cancer associated with attendance at cervical screening, i.e. the self-selection correction factor, can be considered an approximate estimate of the relative risk:

Therefore, we obtain an estimate of θ by dividing γ by φ . The fundamental assumption here is that the populations choosing to attend or not to attend cervical cancer screening have the same risk of dying of breast cancer a priori as those choosing or not choosing to attend breast cancer screening. We do not assume that the effects of self-selection are the same in the two programmes. This is referred to as our first method of correction in the Results section.

As there is considerable uncertainty in the extent of self-selection, and of course decisions to attend at two separate screening programmes are likely to be confounded with each other, we also corrected for this using the method of Duffy et al. 15 . This method estimates the effect of participation in screening in those who would participate if invited as:

where p is the proportion of the invited population who participate in screening and D r is the a priori relative risk of dying of breast cancer for someone who chooses not to attend compared to an uninvited general population member. We estimated D r as 1.19 (95% CI 1.11–1.27), from the cohort study of Johns et al. 16 Thus, this correction was based on a prospective estimate of the extent of self-selection bias in a cohort of 988,090 women in the NHS Breast Screening Programme. We estimated p as 73.4% from the annual report of the National Programme. 12 This method, referred to as our second method of correction in the Results section, also yields an estimate of the effect of invitation to screening as follows: 15

More details on the methods are available in the published study protocol 9 and pilot study analysis. 10

The study dataset had a total of 9550 cases and 17,993 controls. There were 1107 sets with matching ratio 1:1 (1 case to 1 control) and 8443 sets with matching ratio 1:2 (1 case to 2 controls). Records of 1262 cases and 2791 controls (15% of the total) were excluded for various reasons before the statistical analysis (see study flow diagram in Fig.  1 ). This left a final dataset of 8288 cases and 15,202 controls, divided into 1,374 matched sets of size 1:1 and 6914 of size 1:2.

Sensitivity analyses using unconditional logistic regression were performed including subjects without a matched case or control, leaving us with 8479 cases and 16,794 controls.

Table  1 shows patient demographics and screening histories. Median age at first diagnosis was 64 years for both cases and controls and median age at death for cases was 71 years. Whilst the distributions of the number of screening invitations in the two study groups were comparable, differences can be noted in screening attendance, with 72% of the cases versus 82% of the controls attending their first screening invitation; 64% of the cases versus 76% of the controls attending their last screening invitation before diagnosis/pseudodiagnosis; and 21% of the cases versus 12% of the controls never being screened. Median time between last screen and date of diagnosis/pseudodiagnosis for compliers was also slightly longer for cases. From the data available on cervical screening history up to the date of diagnosis/pseudodiagnosis, it can be noted that 22% of the cases compared with 19% of the controls never had a cervical screen.

Patient demographics and screening history by case-control status.

Table  2 summarises the main results without and with correction for self-selection bias. Using data from cervical screening attendance, the self-selection correction factor was estimated to be 0.78 (95% CI 0.73–0.84). The primary endpoint, the association between attending one or more screens and death from breast cancer, had a resulting OR = 0.49 (95% CI 0.45–0.53) and, when corrected for self-selection, had OR = 0.62 (95% CI 0.56–0.69) by our first method and OR = 0.63 (95% CI 0.55–0.71) by our second. Using the second method, the estimate of the effect of invitation to screening was a 26% reduction in breast cancer mortality (OR = 0.74, 95% CI 0.68-0.81). The unmatched logistic regression on the larger dataset for sensitivity analyses showed a similar effect of screening on breast cancer mortality both before and after controlling for age at diagnosis/pseudodiagnosis and screening area (in both cases, uncorrected OR = 0.55, 95% CI 0.51–0.59).

Results of the matched logistic regression evaluating the association between screening attendance and breast cancer mortality.

a Self-selection correction performed using our first method (variant of Duffy et al. 15 ), with the OR of 0.78 associated with participation in cervical screening.

b Self-selection correction performed using our second method (Duffy et al. 15 ).

In order to analyse changes of the effect of screening over time, we excluded women diagnosed before year 2000 (13% of the total records), which led to a corrected OR of 0.56 (95% CI 0.51–0.63) for the effect of ever attending mammographic screening on breast cancer mortality. Women diagnosed from year 2003 onwards had an even larger benefit from being screened (OR corrected by first method = 0.53, 95% CI 0.47–0.59). The estimated effect continued to increase as we restricted the year of diagnosis/pseudodiagnosis further in time (Supplementary Fig.  1 ).

Table  3 shows how the effect of screening varies depending on how much time has passed between a woman’s last screen and her diagnosis/pseudodiagnosis. Screen-detected cancers (assumed to be cancers diagnosed within three months of screening) showed a positive association with breast cancer fatality, after self-selection bias correction by our first method (OR = 1.93, 95% CI 1.68–2.22), while women screened in any other time interval were at reduced risk of dying from breast cancer. This was lowest for women screened in the last year (OR corrected by our first method = 0.19, 95% CI 0.17–0.23) and gradually increased, while still conferring a beneficial effect to screening, for women screened further back in time with respect to their date of diagnosis/pseudodiagnosis. Results using our alternative correction for self-selection were very similar (Table  3 ). Note that the time is from screening to diagnosis, not to death. The Table shows risk of subsequently dying of breast cancer increasing by the time between the screen and diagnosis/pseudodiagnosis.

Results of the matched logistic regressions evaluating the association between time since last screening attendance and breast cancer mortality.

A similar analysis is shown in Table  4 and Fig.  2 for different time intervals after stratifying for three different age categories at diagnosis/pseudodiagnosis (younger than 60 years, between 60 and 64 years, and 65 years or older). The results show that the protective effect of a screen is greater and lasts longer in the oldest group. The benefit of attending screening in the three years prior to diagnosis/pseudodiagnosis, the recommended interval for screening in the NHS BSP, is shown in the final row of Table  4 , and shows close to a halving of risk with screening within the recommended interval, following self-selection correction by our first method (OR = 0.51, 95% CI 0.46–0.57). Results using our second method of correction were very similar to those using the first (Supplementary Table  1 ). The estimated effect of invitation to screening within the last 36 months using our second method was a 33% reduction in breast cancer mortality (OR = 0.67, 95% CI 0.61–0.73).

Results of the matched logistic regressions evaluating the association between time since last screening attendance and breast cancer mortality, stratified by age at diagnosis/pseudodiagnosis.

a Self-selection correction performed using our first method (variant of Duffy et al 15 ), with the OR of 0.78 associated with participation in cervical screening.

An external file that holds a picture, illustration, etc.
Object name is 41416_2020_1163_Fig2_HTML.jpg

Note: the coordinates on the x -axis are the midpoints of the time intervals: 0–3, 3–6, 6–18, 18–36, 36–54 and 54–72 months.

Despite the many improvements in treatments, diagnostic procedures and technologies over the last thirty years, and changes in baseline rate of breast cancer mortality, our data showed an overall reduction in the risk of dying from breast cancer of ~38% for women attending at least one mammography screen, after adjusting for self-selection bias. This is in line with the results obtained from the pilot phase of the study, 10 in which a mortality reduction of 39% was seen for women attending screening in London (deaths occurring in 2008–2009). Using the same calculation method as in the review by the Independent UK Panel on Breast Cancer Screening UK Independent Review, 17 this would correspond to approximately nine breast cancer deaths prevented for every 1,000 women attending screening at ages 50–69 years, larger than but in the same general scale as the six deaths estimated from the UK Independent review.

It should be noted that there is a wide range of estimates of the absolute mortality benefit of mammography screening 18 – 21 some finding considerably smaller benefits than above. The size of the estimated effect depends on sources used and assumptions made. However, it has been shown to depend more crucially on whether the effect pertains to screening per se or to invitation to screening only, and on the timescale envisaged. 22 Screening prevents deaths not this year or next, but 5, 10, 15 or 20 years from now. Considering the effect of screening on 10-year mortality will considerably underestimate the absolute benefit. Nevertheless, it should be acknowledged that while the body of evidence, randomised and observational, points to a substantial reduction in breast cancer mortality with screening, there is sufficient variation that different views are still possible.

Our first method of correction for self-selection caused a decrease of about 25% in the estimated protective effect of screening for women having at least one mammogram. The second method yielded similar results. This is a greater correction than the one estimated in the pilot phase, 10 where self-selection only played a minor role, despite the fact that the final risk reduction is very similar. London has a lower coverage than the rest of England for both breast and cervical screening, which is largely explained by factors like deprivation and ethnicity. 23 Such variations in coverage might be one of the causes for the different impact of self-selection between the two phases of the study. For example, a larger population of non-participants, such as in London, may be less different in health status than a smaller population. In the Swedish two-County trial, 24 where only 15% of the population were non-participants, the rate of death from breast cancer in this population was very high. It is also worth noting that, during the early 21st century, breast screening attendance was rapidly increasing in London, and the socioeconomic gradient in attendance was reducing with time nationally. 25 , 26

Case-control studies tend to give higher estimates of benefit than other evaluations, largely because they assess the effect of actually being screened rather than simply being invited to screening. 19 , 27 It should be noted that with our second correction for self-selection bias, we were able to estimate the effect of invitation, giving a 26% breast cancer mortality reduction, similar to the effect observed in the randomised trials in this age group and to the prospectively estimated effect of a 25% reduction in the Copenhagen screening programme. 28 As a comparison, in the review by the Independent UK Panel on Breast Cancer Screening, 17 a meta-analysis of 11 RCTs found that the relative risk reduction of breast cancer mortality for women invited to screening was 20%. Furthermore, in the same report, the panel stated that the case-control studies that they had analysed seemed to inflate the benefit of screening compared to the trials and postulated that this may have been caused by some residual bias unaccounted for by the authors. We believe that our adjustments for self-selection bias has largely accounted for this and that the greater effect of screening in this study is due to technical improvements in mammography since the RCTs were carried out, accompanied by improved treatment and strong quality assurance measures in the NHS BSP. 11

The greater benefit of screening observed for women diagnosed after year 2000 was similar to the pilot study, 10 but here we were able to restrict the analysis to later years of diagnosis and see the benefit getting larger (data not shown). We could conjecture that this improvement was due to the introduction of better procedures in the NHS BSP, such as two-view mammography at every attendance in year 2000 4 ; however, there may be a bias in comparing different times since diagnosis as we only have data on deaths in years 2010–2011. In the first place, cases diagnosed before 2000 have a long survival by definition, and there might therefore be an over-representation of screen-detected cancers. In other words, it is more likely that a case diagnosed before year 2000, for example, who had a breast cancer for more than 10–11 years before dying from it, had a screen-detected cancer rather than a symptomatic one. This confers a bias against screening in the analysis of cancers diagnosed prior to the year 2000. In the second place, there will be a bias in favour of screening if the analysis is restricted to cancers diagnosed within a short time before death, i.e. if we only consider women (pseudo)diagnosed a few years before 2010–2011. We are therefore unable to make any definitive conclusions on the impact of any improvements in the NHS BSP over time.

As shown in RCTs of breast screening, 24 measures of the benefit of screening are largely influenced by the consequent reduction in mortality from symptomatic cancers. This is due to the fact that screen-detected cancers (defined as the ones diagnosed within three months of a screen), despite being less fatal overall, represent a larger proportion of the cancer-related deaths in the immediate period after a screen as it can be seen from the spike in excess mortality in Fig.  2 .

The duration of the benefit of attending screening appears to be greater in older women (Table  4 and Fig.  2 ). Women aged 65 or more see the greatest and longer lasting benefit, which might suggest that they could be screened less often than younger women. This result is in agreement with the impact of ageing on breast cancer biology 29 and is also potentially important in light of the recent incident in the NHS BSP, where a number of women aged 69 and 70 years did not receive the scheduled invitation to their last screening appointment. 30 The exact number affected has been debated but an Independent Review concluded that 5000 women were not invited as scheduled, and that a further 62,000 could be interpreted as having missed their final invitation as defined in the service specification. 30 Our findings suggest that the effect of a delayed screen in older women has a lesser consequence for increased risk of breast cancer mortality than it would have had in younger women. While three years is a longer interval than other programmes in Europe and North America, and further slippage of the interval should be avoided if at all possible, these results could also be used as guidelines for screening units at times of capacity constraints, with the provision that all women receive an opportunity for a final screen around or shortly after age 70. There is interest in stratified screening and these results may inform further thinking on this subject.

A limitation of the study is the retrospective design and the potential for self-selection bias. We have corrected for this in two different ways and for one of these, an effect of invitation to screening was derived which was consistent with trials results and prospective studies for this age group. However, it must be acknowledged that there remains some uncertainty about the extent of self-selection bias. Furthermore, case-control studies for cancer screening programmes are subject to an inherent type of anti-screening bias known as screening opportunity bias. 27 As most of the controls do not have a breast cancer diagnosis, the only way they can be exposed to screening is if they attended a mammography appointment in the past. Cases, on the other hand, may have had a screen in the past, but some of them will also have an additional screen for when their cancer was diagnosed. This induces an artificially higher retrospective probability of screening exposure among cases. Screening opportunity bias was corrected for in the pilot study, 10 where a 10–15% increase in mortality reduction was seen following this, but here we preferred to keep a conservative approach and not adjust for it. To minimise biases with respect to age and opportunity to be screened, we matched very closely for age. This meant that in 1107 cases out of 9550, we could only find one control.

Although the effect of the NHS BSP in preventing breast cancer mortality has been assessed several times, 31 – 34 we are aware of only one other case-control study conducted using national data. 34 The latter relies on data up to year 2005 (diagnoses and deaths took place between 1991 and 2005), while ours uses more recent data up to year 2012, arguably more in the epoch of effective adjuvant systemic therapies. It is of interest that our more recent case base shows similar results in terms of the reduction in risk of breast cancer death with screening. In any case, we suggest that it would be of interest to repeat this type of analysis for years thereafter, to ensure that the programme continues to deliver its aims even with the introduction of new diagnostic technologies (e.g. digital mammography). Before the establishment of the NHS BSP in 1987, it was suggested that a routine case-control assessment could and should be part of an ongoing evaluation of a mass screening programme. 35 For this reason, we believe that this exercise should be held on a two-yearly basis.

The results of further national case-control studies (1) evaluating the effect of the NHS BSP on breast cancer incidence and incidence of late stage disease, (2) estimating overdiagnosis, and (3) analysing the interplay of early detection, pathology and treatment on fatality of breast cancer will be published shortly.

To conclude, this study showed that the breast screening programme in England continues to play an important role in the control of breast cancer. The effect of screening within the NHS BSP in England is stronger and longer lasting in women aged 65 or over, but it remains highly relevant for younger women.

Supplementary information

Acknowledgements.

Data for this study is based on information collected and quality assured by the PHE National Cancer Registration and Analysis Service. Access to the data was facilitated by the PHE Office for Data Release. We would like to thank Rachael Brannan from the PHE Office for Data Release and David Graham from NHS Digital for their help with the data selection and matching of cases and controls. This work uses data provided by patients and collected by the NHS as part of their care and support.

Author contributions

R.M. oversaw the first draft of the manuscript draft and submission. N.J.M., J.C., P.D.S. and S.W.D. designed the study. D.P. contributed to data collection and cleaning. A.D. assisted with data interpretation. R.M. and S.W.D. analysed the data and produced the figures. All the authors critically reviewed the paper.

Ethics approval and consent to participate

The study protocol was reviewed and approved by the Department of Health. Ethical approval was obtained from the London Research Ethics Committee of the National Research Ethics Service (reference: 12/LO/1041), and by the National Information Governance Board Ethics and Confidentiality Committee (reference: ECC 6–05 (e)/2012). The ethics committee agreed that informed consent to participate for the study subjects was not necessary. The study was performed in accordance with the Declaration of Helsinki.

Data availability

Competing interests.

P.D.S. reports personal fees from GRAIL Bio outside the submitted work. J.C. and S.W.D. are members of the editorial board of the British Journal of Cancer. The remaining authors declare no competing interests.

Funding information

his research is funded by the National Institute for Health Research (NIHR) Policy Research Programme, conducted through the Policy Research Unit (PRU) in Cancer Awareness, Screening and Early Diagnosis, PR-PRU-1217-21601. The PRU is a collaboration between researchers from seven institutions (Queen Mary University of London, University College London, King’s College London, London School of Hygiene and Tropical Medicine, Hull York Medical School, Durham University, and Peninsula Medical School). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. The funding body was not involved in design, data collection, analysis or interpretation. The funding body had sight of the paper prior to publication but has not had input to its content.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Roberta Maroni, Nathalie J Massat

These authors jointly supervised this work: Peter D Sasieni, Stephen W Duffy

Supplementary information is available for this paper at 10.1038/s41416-020-01163-2.

  • Introduction
  • Conclusions
  • Article Information

WUSM indicates Washington University School of Medicine.

eAppendix. Analytic Framework

eFigure 1. Top Row Shows Histogram of Mammographic Density of the Two Breasts on the Original Volumetric Scale and the Corresponding QQ Plot for the Normality of Residuals; Bottom Row Shows Mammographic Density on the Box-Cox–Transformed Scale and the Corresponding QQ Plot for the Normality of Residuals With Improved Fit

eFigure 2. Three Types of Correlations Using Box-Cox–Transformed Breast Densities in the Control Women: R 1  = Correlation Within the Same Breast Over Time; R 2  = Inter-breast Correlation Within the Same Woman; R 3  = Cross-Correlation Between Breasts at Different Time Points

eFigure 3. Scatter Plot of Box-Cox Transformation vs Original Scale Mammographic Density

eFigure 4. Volumetric Cut Points for the Original Scale MD to BI-RADS Levels

eFigure 5. Illustration for Change in MD Stratified by a) BI-RADS A (MD <3.5%) and b) BI-RADS D (MD >15.5%; Bottom Row) for the Case Breast in the Case Women and Control Women Over Time

eReferences.

Data Sharing Statement

  • Association of Breast Density With Risk of Breast Cancer JAMA Oncology Comment & Response December 1, 2023 Junyu Long, MD, PhD; Xinting Sang, MD, PhD; Haitao Zhao, MD

See More About

Select your interests.

Customize your JAMA Network experience by selecting one or more topics from the list below.

  • Academic Medicine
  • Acid Base, Electrolytes, Fluids
  • Allergy and Clinical Immunology
  • American Indian or Alaska Natives
  • Anesthesiology
  • Anticoagulation
  • Art and Images in Psychiatry
  • Artificial Intelligence
  • Assisted Reproduction
  • Bleeding and Transfusion
  • Caring for the Critically Ill Patient
  • Challenges in Clinical Electrocardiography
  • Climate and Health
  • Climate Change
  • Clinical Challenge
  • Clinical Decision Support
  • Clinical Implications of Basic Neuroscience
  • Clinical Pharmacy and Pharmacology
  • Complementary and Alternative Medicine
  • Consensus Statements
  • Coronavirus (COVID-19)
  • Critical Care Medicine
  • Cultural Competency
  • Dental Medicine
  • Dermatology
  • Diabetes and Endocrinology
  • Diagnostic Test Interpretation
  • Drug Development
  • Electronic Health Records
  • Emergency Medicine
  • End of Life, Hospice, Palliative Care
  • Environmental Health
  • Equity, Diversity, and Inclusion
  • Facial Plastic Surgery
  • Gastroenterology and Hepatology
  • Genetics and Genomics
  • Genomics and Precision Health
  • Global Health
  • Guide to Statistics and Methods
  • Hair Disorders
  • Health Care Delivery Models
  • Health Care Economics, Insurance, Payment
  • Health Care Quality
  • Health Care Reform
  • Health Care Safety
  • Health Care Workforce
  • Health Disparities
  • Health Inequities
  • Health Policy
  • Health Systems Science
  • History of Medicine
  • Hypertension
  • Images in Neurology
  • Implementation Science
  • Infectious Diseases
  • Innovations in Health Care Delivery
  • JAMA Infographic
  • Law and Medicine
  • Leading Change
  • Less is More
  • LGBTQIA Medicine
  • Lifestyle Behaviors
  • Medical Coding
  • Medical Devices and Equipment
  • Medical Education
  • Medical Education and Training
  • Medical Journals and Publishing
  • Mobile Health and Telemedicine
  • Narrative Medicine
  • Neuroscience and Psychiatry
  • Notable Notes
  • Nutrition, Obesity, Exercise
  • Obstetrics and Gynecology
  • Occupational Health
  • Ophthalmology
  • Orthopedics
  • Otolaryngology
  • Pain Medicine
  • Palliative Care
  • Pathology and Laboratory Medicine
  • Patient Care
  • Patient Information
  • Performance Improvement
  • Performance Measures
  • Perioperative Care and Consultation
  • Pharmacoeconomics
  • Pharmacoepidemiology
  • Pharmacogenetics
  • Pharmacy and Clinical Pharmacology
  • Physical Medicine and Rehabilitation
  • Physical Therapy
  • Physician Leadership
  • Population Health
  • Primary Care
  • Professional Well-being
  • Professionalism
  • Psychiatry and Behavioral Health
  • Public Health
  • Pulmonary Medicine
  • Regulatory Agencies
  • Reproductive Health
  • Research, Methods, Statistics
  • Resuscitation
  • Rheumatology
  • Risk Management
  • Scientific Discovery and the Future of Medicine
  • Shared Decision Making and Communication
  • Sleep Medicine
  • Sports Medicine
  • Stem Cell Transplantation
  • Substance Use and Addiction Medicine
  • Surgical Innovation
  • Surgical Pearls
  • Teachable Moment
  • Technology and Finance
  • The Art of JAMA
  • The Arts and Medicine
  • The Rational Clinical Examination
  • Tobacco and e-Cigarettes
  • Translational Medicine
  • Trauma and Injury
  • Treatment Adherence
  • Ultrasonography
  • Users' Guide to the Medical Literature
  • Vaccination
  • Venous Thromboembolism
  • Veterans Health
  • Women's Health
  • Workflow and Process
  • Wound Care, Infection, Healing

Others Also Liked

  • Download PDF
  • X Facebook More LinkedIn

Jiang S , Bennett DL , Rosner BA , Colditz GA. Longitudinal Analysis of Change in Mammographic Density in Each Breast and Its Association With Breast Cancer Risk. JAMA Oncol. 2023;9(6):808–814. doi:10.1001/jamaoncol.2023.0434

Manage citations:

© 2024

  • Permissions

Longitudinal Analysis of Change in Mammographic Density in Each Breast and Its Association With Breast Cancer Risk

  • 1 Division of Public Health Sciences, Department of Surgery, Washington University School of Medicine in St Louis, St Louis, Missouri
  • 2 Department of Radiology, Washington University School of Medicine in St Louis, St Louis, Missouri
  • 3 Channing Division of Network Medicine, Brigham and Women’s Hospital, Boston, Massachusetts
  • Comment & Response Association of Breast Density With Risk of Breast Cancer Junyu Long, MD, PhD; Xinting Sang, MD, PhD; Haitao Zhao, MD JAMA Oncology

Question   Is change in mammographic breast density associated with the development of breast cancer, and does this change diverge from the expected decrease in density with age?

Findings   In this nested case-control cohort study of 947 women attending breast screening during up to 10 years, a decrease in breast density was observed in all women regardless of subsequent breast cancer development. The rate of density change was significantly slower in the breast in which cancer was later diagnosed.

Meaning   This study found that evaluating longitudinal changes in breast density from digital mammograms may offer an additional tool for assessing risk of breast cancer and subsequent risk reduction strategies.

Importance   Although breast density is an established risk factor for breast cancer, longitudinal changes in breast density have not been extensively studied to determine whether this factor is associated with breast cancer risk.

Objective   To prospectively evaluate the association between change in mammographic density in each breast over time and risk of subsequent breast cancer.

Design, Setting, and Participants   This nested case-control cohort study was sampled from the Joanne Knight Breast Health Cohort of 10 481 women free from cancer at entry and observed from November 3, 2008, to October 31, 2020, with routine screening mammograms every 1 to 2 years, providing a measure of breast density. Breast cancer screening was provided for a diverse population of women in the St Louis region. A total of 289 case patients with pathology-confirmed breast cancer were identified, and approximately 2 control participants were sampled for each case according to age at entry and year of enrollment, yielding 658 controls with a total number of 8710 craniocaudal-view mammograms for analysis.

Exposures   Exposures included screening mammograms with volumetric percentage of density, change in volumetric breast density over time, and breast biopsy pathology-confirmed cancer. Breast cancer risk factors were collected via questionnaire at enrollment.

Main Outcomes and Measures   Longitudinal changes over time in each woman’s volumetric breast density by case and control status.

Results   The mean (SD) age of the 947 participants was 56.67 (8.71) years at entry; 141 were Black (14.9%), 763 were White (80.6%), 20 were of other race or ethnicity (2.1%), and 23 did not report this information (2.4%). The mean (SD) interval was 2.0 (1.5) years from last mammogram to date of subsequent breast cancer diagnosis (10th percentile, 1.0 year; 90th percentile, 3.9 years). Breast density decreased over time in both cases and controls. However, there was a significantly slower decrease in rate of decline in density in the breast that developed breast cancer compared with the decline in controls (estimate = 0.027; 95% CI, 0.001-0.053; P  = .04).

Conclusions and Relevance   This study found that the rate of change in breast density was associated with the risk of subsequent breast cancer. Incorporation of longitudinal changes into existing models could optimize risk stratification and guide more personalized risk management.

Mammographic breast density is a well-established and strong risk factor for breast cancer. 1 , 2 Consistent cross-sectional international data suggest that a decrease in breast density with age is a universal intrinsic biological process in women that is associated with decreasing circulating hormone levels (estrogen and progestin). 3 Although the assessment of breast density in the radiographic-screening era was qualitative and based on the radiologist’s subjective evaluation of the presence of dense glandular tissue, widespread use of digital mammography created the potential for the quantitative assessment of breast density. The association of breast density with breast cancer risk is consistent across a range of methods used to estimate density as the percentage of density (amount of dense glandular tissue compared with total breast tissue). 4 , 5 Volumetric density is an established risk factor for breast cancer and is incorporated into existing risk assessment models, including Tyrer-Cuzick model, version 8. 6 However, to our knowledge, there are limited data on the association between change in density over time and risk of subsequent breast cancer diagnosis. Retrospective data from Korea show that change in the category of Breast Imaging Reporting and Data System (BI-RADS) breast density is associated with change in risk across 3 mammograms, 7 consistent with an earlier meta-analysis. 8 However, this study did not use volumetric analysis of density.

We hypothesized that there is a difference in the rate of change in breast density in women who subsequently develop breast cancer compared with those who do not. Using longitudinal, quantitative data captured for up to 10 years, we evaluated the association between rate of change in mammographic density and risk of breast cancer development.

Ethical approval for this prospective nested case-control cohort study was obtained from the Washington University in St Louis institutional review board. Participants provided informed written consent. We followed the Strengthening the Reporting of Observational Studies in Epidemiology ( STROBE ) reporting guideline.

Women were recruited and provided consent for follow-up from November 3, 2008, to April 30, 2012, through the mammography service at the Joanne Knight Breast Health Center at Washington University in St Louis, Missouri. As previously described, more than 50% of eligible women presenting for screening mammography accepted the invitation to enroll, completed a risk factor questionnaire, and consented to follow-up. 9 Screening mammograms were obtained from 12 153 women. Of these women, 1672 were excluded for a history of any cancer (except nonmelanoma skin cancer) or a diagnosis of breast cancer within 6 months of registration for the study, for a total of 10 481 women ( Figure 1 ).

Follow-up was conducted by linking to electronic health records, a tumor registry, and a death register. Routine screening mammograms were obtained every 1 to 2 years. A total of 8908 of the 10 481 women in our cohort (85%) had either a mammogram or clinic visit within the last 3 years of follow-up (through October 31, 2020). 9 The median number of mammograms for each woman in the cohort was 4 (range, 1-10), with an SD of 2.4.

We sampled 2 control participants for each case patient according to age at cohort entry and year of enrollment. Incident breast cancers were identified through record linkage to pathology and tumor registries. We identified 347 cases and 694 controls. After linkage to screening mammogram files, we excluded women with breast implants and those without screening mammograms retrieved, and we retained 289 cases and 658 controls with mammograms taken through October 31, 2020, resulting in a total number of 8710 craniocaudal-view mammograms for analysis. All analyses performed in this study used the nested case-control cohort for efficiency of image data processing.

Women self-reported breast cancer risk factors on entry to the cohort. These risk factors were drawn from established and validated measures. 10 The questionnaire at entry assessed height, current weight, parity, age at first birth, cessation of menses (yes or no), age at menopause (natural or with surgical removal of the uterus, with or without removal of the ovaries), age at hysterectomy, family history of breast cancer (mother, sister, or both), history of biopsy-confirmed benign breast disease, current alcohol intake, and race and ethnicity.

The volumetric density was estimated from each digital craniocaudal-view mammogram with an automated pixel-thresholding algorithm implemented at Washington University on processed images. The skin around the breast is automatically removed with a boundary detection algorithm before estimation of the volume of dense glandular tissue. The volumetric percentage of density is then estimated with the volume of dense glandular tissue divided by the total breast volume; use of the percentage of density normalizes the difference in breast size across women and is consistent with other density estimation methods in the literature. 11 , 12 The correlation between volumetric density generated from our automated algorithm with Volpara, version 1.5.0 (Volpara Health) was 0.81 according to an out-of-sample study with 375 women recruited from the mammography service at Washington University with a mean (SD) age of 47 (4.8) years (see eFigure 4 in Supplement 1 for distribution of volumetric density in this cohort). 13 , 14

To evaluate hypotheses, we fitted a linear mixed-effects model with separate records for each breast to accommodate longitudinal correlated continuous breast density data. 15 We fitted a term for breast density in each breast to evaluate differences between cases and controls at entry to the cohort, and then an interaction with time for each breast was used to evaluate the second hypothesis of change in density varying over time (or repeated mammograms with aging) between cases and the controls. With this model, we accounted for correlations within a woman between both her breasts and separately for density measured on the same breast over time. We used Box-Cox transformation to normalize the distribution of breast density and performed model checking and evaluation of residuals (eFigure 1, eFigure 3, and the eAppendix in Supplement 1 for more details). In addition, to replicate standard practice, we also reported a conventional analysis using the mean of density in the 2 breasts.

Of the 947 women, 141 were Black (14.9%), 763 were White (80.6%), and 20 were of other race or ethnicity (2.1%); 23 patients did not report this information (2.4%). Race and ethnicity were self-reported. The mean (SD) age of the patients was 56.67 (8.71) years at entry. The mean (SD) interval between screening mammograms was 1.3 (0.7) years (10th percentile, 1.0 year; 90th percentile, 2.0 years). For the cases, the mean (SD) time from last mammogram to subsequent breast cancer diagnosis was 2.0 (1.5) years (10th percentile, 1.0 year; 90th percentile, 3.9 years).

The breast cancer risk factors assessed at entry for the women in this study, stratified by case and control status, are presented in Table 1 . Most women in both the case and control groups were postmenopausal (209 and 482 women, respectively) and parous (228 and 508 women, respectively). There was no important difference between cases and controls in qualitative breast density assigned by the radiologist at enrollment mammogram (BI-RADS A and B categories [not dense] vs BI-RADS C and D categories [dense]). The correlation of the 2 breasts in the controls at cohort entry was 0.86, and the 2-year correlation within each breast was 0.78 (eFigure 2 in Supplement 1 ).

When the mean volumetric density of the 2 breasts was used, breast density at entry was significantly higher for cases compared with controls (estimate = 0.140; 95% CI, 0.033-0.246; P  = .01) ( Table 2 and Figure 2 ). Breast density decreased over time in both groups of postmenopausal women. Women with higher body mass index had lower breast density (estimate = −0.041; 95% CI, −0.049 to −0.033; P  < .001), and women with a history of biopsy-confirmed benign breast disease had higher breast density (estimate = 0.224; 95% CI, 0.129-0.318; P  < .001).

When the mean density of both breasts was used, change in density over time did not differ between cases and controls (represented by follow-up time × case status, estimate = 0.018; 95% CI, −0.004 to 0.039; P  = .11). However, when density change in each breast was analyzed separately, there was a significant difference in the rate of density change over time in the breast that developed cancer compared with that in controls (estimate = 0.027; 95% CI, 0.001-0.053; P  = .04) ( Table 3 ). Mammograms for breasts that subsequently developed cancer demonstrated a significantly slower rate of decrease in density than mammograms for breasts that did not later develop cancer in the control women.

To illustrate this significant change in the rate of decrease in density during the follow-up interval, in Figure 2 , we show the volumetric density for breasts that subsequently developed cancer (ie, case breast within the cases) compared with the volumetric density for breast mammograms from controls. The decrease in breast density over time was significantly slower for the case breast during the follow-up interval compared with the breasts within the controls. Plots stratified by BI-RADS show similar patterns on the original volumetric density scale and the Box-Cox–transformed scale (eFigure 5 in Supplement 1 ).

In this prospective cohort study designed to evaluate breast cancer risk over time, we used case-control sampling for efficient processing of 8710 prospectively ascertained digital mammographic images. We observed that breast density was higher at entry to the study for women who would later develop breast cancer compared with controls who remained cancer free. Volumetric breast density decreased significantly over time in both groups. To reflect the underlying biology of breast cancer development, we modeled each breast independently to allow for different rates of change for the breast that would develop breast cancer and the contralateral breast in a case patient that remained free from breast cancer. The rate of decrease in density for the breast that developed breast cancer was significantly slower than for controls who did not develop breast cancer. This observation, using longitudinal repeated measures, reflects the dynamic changes in breast tissue that differ significantly between women who develop breast cancer and those who do not. These data suggest that longitudinal changes in breast density may be used to refine the assessment of risk of breast cancer and to inform precision prevention strategies.

Breast density remains one of the most readily available summary measures from screening mammograms that may be used to assess future breast cancer risk. 16 Although density is an accepted intermediate marker of breast cancer risk, 2 studies of change in density are limited. Previous studies have used reader-dependent, subjective measures of density and categorical density data (fatty, scattered, heterogeneously dense, and extremely dense in accordance with BI-RADS). 7 , 17 - 21 Previous studies have also been based on analysis of digitized radiographic images, 7 , 22 , 23 which are of lower image quality than standard digital mammograms. 24 , 25 Furthermore, changes in breast density in these studies were often assessed from comparison of only 2 time points. Despite these limitations, a recent meta-analysis of 4 studies reported that a change in BI-RADS breast density category is associated with an increased risk of breast cancer. 8

Here, we moved beyond these limitations to draw on the full data of digital mammograms to capture changes within each breast over time. Analysis of mammographic images of each breast independently allows for the study of the specific breast in which cancer later develops. An additional strength of this study is the use of a median of 4 (range, 1-10) or more digital mammograms per participant during a follow-up period of 10 years. Follow-up mammograms from 2008 to 2020 reflect current clinical practice. The cohort was generated from a population undergoing routine screening and reflects the broader community. 9 The mean (SD) time from most recent mammogram to breast cancer diagnosis was 2.0 (1.5) years, and we excluded mammograms within 6 months of diagnosis to avoid bias that can be induced if density is spuriously increased owing to undetected cancer. 4 , 26 We used craniocaudal views because epidemiologic studies show that associations of density with breast cancer are stronger for craniocaudal views than for mediolateral oblique views. 27

Once mammography screening is begun from the age of 45 years (American Cancer Society) or 50 years (US Preventive Services Task Force), repeated either annually or biannually, women accumulate a series of mammograms. Therefore, the available data are naturally composed of longitudinal images for each breast. Traditional models focus on fixed prediction for some future interval conditional on information gathered at baseline. However, given the available longitudinal data, dynamic prediction that can be updated immediately after mammography conditional with the woman’s most up-to-date individualized density trajectory in each breast could improve overall breast cancer risk classification over time. These repeated measures can be further refined for routine practice and incorporated into a dynamic risk classification and hence guide personalized prevention services according to the woman’s level of risk. Further research is needed to define the clinical association between incorporation of density change over time and any subsequent risk reduction strategies that may be recommended as a result. 28 , 29 Finally, refining the trade-off of number of images and interval between images to assess the change over time will also inform translation to clinical practice. 30

In this prospective cohort study using repeated images and established statistical approaches to repeated-measures analysis, 15 we advanced one-time assessment of breast density, which has already been established as a key component for risk stratification. 31 A systematic review shows that, in models predicting women’s risk of breast cancer that were published from 2007 to 2019, the addition of a onetime measure of breast density significantly increased discriminatory accuracy in 7 of 11 studies. The increase in the area under the curve ranged from 0.03 to 0.14. 31 The current findings open the potential for dynamic prediction modeling to account for breast-specific density changes over time.

Much ongoing research focuses on mammography analysis to allow for an earlier diagnosis of breast cancer, when it is the most treatable. 32 The potential to use the data embedded in mammographic images and make full use of longitudinal measures will facilitate optimizing risk stratification to guide more personalized risk reduction. 33

The limitations of the study include a population that was predominantly White (763 individuals [80.6%]) and analysis of digital images from only 1 manufacturer (Hologic). However, other studies have shown breast density is a robust risk factor across race and ethnicity and mammography vendors. 3 , 4 , 26

Using longitudinal digital mammograms of each breast during up to 10 years, this prospective cohort case-control study found that women with a slower decrease in breast density had a higher risk of developing breast cancer. These dynamic changes in density over time may be used to refine risk stratification and guide more individualized screening and prevention approaches.

Accepted for Publication: January 27, 2023.

Published Online: April 27, 2023. doi:10.1001/jamaoncol.2023.0434

Open Access: This is an open access article distributed under the terms of the CC-BY License . © 2023 Jiang S et al. JAMA Oncology .

Corresponding Author: Graham A. Colditz, MD, DrPH, Division of Public Health Sciences, Department of Surgery, Washington University School of Medicine, 660 S Euclid Ave, MSC 8100-0094-02, St Louis, MO 63110 ( [email protected] ).

Author Contributions: Drs Jiang and Colditz had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Jiang, Rosner, Colditz.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: Bennett, Rosner, Colditz.

Statistical analysis: Jiang, Rosner, Colditz.

Obtained funding: Colditz.

Administrative, technical, or material support: Colditz.

Supervision: Bennett, Rosner, Colditz.

Conflict of Interest Disclosures: Dr Jiang reported receiving grants from the National Cancer Institute (NCI) and the Breast Cancer Research Foundation (BCRF) during the conduct of the study and reported having a patent pending for automated volumetric density assessment by pixel thresholding for individual digital mammograms. Dr Colditz reported receiving grants from NCI and BCRF during the conduct of the study and reported having a patent pending for automated volumetric density assessment by pixel thresholding for individual digital mammograms. No other disclosures were reported.

Funding/Support: This work was supported by BCRF grant 21-028 and in part by the NCI (R37 CA256810).

Role of the Funder/Sponsor: The BCRF and NCI had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 2 .

  • Register for email alerts with links to free full-text articles
  • Access PDFs of free articles
  • Manage your interests
  • Save searches and receive search alerts

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • View all journals
  • My Account Login
  • Explore content
  • About the journal
  • Publish with us
  • Sign up for alerts
  • Open access
  • Published: 05 October 2017

Biomarkers of inflammation and breast cancer risk: a case-control study nested in the EPIC-Varese cohort

  • Claudia Agnoli 1 ,
  • Sara Grioni 1 ,
  • Valeria Pala 1 ,
  • Alessandra Allione 2 , 3 ,
  • Giuseppe Matullo 2 , 3 ,
  • Cornelia Di Gaetano 2 , 3 ,
  • Giovanna Tagliabue 4 ,
  • Sabina Sieri 1   na1 &
  • Vittorio Krogh 1   na1  

Scientific Reports volume  7 , Article number:  12708 ( 2017 ) Cite this article

3158 Accesses

52 Citations

30 Altmetric

Metrics details

  • Cancer epidemiology
  • Epidemiology

Breast cancer (BC) is the leading cause of cancer death in women. Adipokines, and other inflammation molecules linked to adiposity, are suspected to be involved in breast carcinogenesis, however prospective findings are inconclusive. In a prospective nested case-control study within the EPIC-Varese cohort, we used conditional logistic regression to estimate rate ratios (RRs) for BC, with 95% confidence intervals (CI), in relation to plasma levels of C-reactive protein (CRP), tumor necrosis factor-alpha (TNF-α), interleukin-6, leptin, and adiponectin, controlling for BC risk factors. After a median 14.9 years, 351 BC cases were identified and matched to 351 controls. No marker was significantly associated with BC risk overall. Significant interactions between menopausal status and CRP, leptin, and adiponectin were found. Among postmenopausal women, high CRP was significantly associated with increased BC risk, and high adiponectin with significantly reduced risk. Among premenopausal women, high TNF-α was associated with significantly increased risk, and high leptin with reduced risk; interleukin-6 was associated with increased risk only in a continuous model. These findings constitute further evidence that inflammation plays a role in breast cancer. Interventions to lower CRP, TNF-α, and interleukin-6 and increase adiponectin levels may contribute to preventing BC.

Introduction

Breast cancer is the commonest cancer and leading cause of cancer death in women worldwide, with an estimated 1.7 million cases and over 520,000 deaths in 2012, accounting for 25% of all female cancers and 15% of all female cancer deaths 1 .

As long ago as 1863, Rudolf Virchow proposed that cancers originate at sites of chronic inflammation 2 . It is now clear that chronic inflammation is associated with several human cancers and that pro-inflammatory cytokines and other immunomodulatory molecules can be produced by cells in cancerous tissue to favor tumor growth, infiltration and metastasis 3 . C-reactive protein (CRP), an acute-phase protein of hepatic origin that is a sensitive yet nonspecific marker of the inflammatory response, has been associated with breast cancer risk in some 4 , 5 , 6 , 7 but not all studies 6 . Several adipokines (immunomodulatory proteins produced by adipose and other tissues) are also suspected to play a role in breast carcinogenesis 8 . In particular, altered levels of tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), adiponectin, and leptin have important roles in promoting inflammation in obesity and chronic inflammatory diseases 9 , and may play a role in carcinogenesis 8 , 10 . However prospective studies on markers of inflammation and breast cancer risk have produced conflicting results 11 , 12 , 13 .

We carried out a case-control study to prospectively assess whether pre-diagnostic levels of CRP, TNF-α, IL-6, leptin, and adiponectin in plasma are associated with risk of developing breast cancer.

Baseline characteristics of cases and controls, by tertiles of plasma level of inflammatory biomarkers, are shown in Table  1 . Women in the highest tertiles of all biomarkers tended to be older and have higher BMI (lower BMI for high adiponectin). Alcohol intake decreased with increasing adiponectin levels. Women in the highest tertiles were also less educated (not for increasing levels of adiponectin), less likely to be smokers (not for increasing levels of IL-6) or sometime oral contraceptive users, and more likely to be postmenopausal. Age at menarche was higher for increasing levels of CRP and IL-6. Lastly, while women with highest CRP levels were less likely to be nulliparous, the opposite was the case for women with highest adiponectin levels.

Table  2 shows RRs of developing breast cancer by tertiles of plasma markers. None of the markers was significantly associated with risk. Table  3 shows results for postmenopausal and premenopausal women separately. Significant interactions between menopausal status and plasma levels were found for CRP (tertiles model), leptin (tertiles and continuous models), and adiponectin (tertiles model).

Among postmenopausal women, high levels (third tertile) of CRP were associated with significantly increased risk (RR 2.42; 95% CI: 1.17–5.00) compared to the first tertile, fully-adjusted model; while high levels (third tertile) of adiponectin were associated with significantly reduced risk (RR 0.37; 95% CI: 0.19–0.72) compared to the first tertile, fully-adjusted model. None of the other biomarkers was significantly associated with breast cancer risk in postmenopausal women.

Among premenopausal women, high TNF-α was associated with significantly increased breast cancer risk in the tertile model only (P trend = 0.017); and high IL-6 was associated with increased risk in the continuous model only (RR 1.58; 95% CI: 1.02–2.46). By contrast, high plasma leptin was associated with significantly reduced risk, both in the tertile-based model (RR 0.43; 95% CI: 0.20–0.89, third vs. first tertile) and for a 1 standard deviation increase in leptin (RR 0.71; 95% CI: 0.50–0.99). None of the other biomarkers was significantly associated with breast cancer risk premenopausal women.

In this nested case-control study, none of the inflammatory biomarkers analyzed was associated with breast cancer risk in the overall population. However, there were significant interactions between menopausal status and levels of CRP, leptin, and adiponectin. Among postmenopausal women, high CRP was associated with increased breast cancer risk, and high adiponectin was associated with decreased risk. And among premenopausal women, high TNF-α and IL-6 were associated with increased risk, and high leptin was associated with decreased risk.

Our finding of a direct association between plasma CRP and risk of postmenopausal breast cancer is in line with the findings of two recently published meta-analyses 4 , 14 . The first 4 , which examined 12 prospective studies, found that risk increased significantly by 7% overall and by 6% in postmenopausal women, for each doubling of CRP concentration. The other study 14 analyzed 15 cohort and case-control studies, and included premenopausal women, finding that risk increased by 16% for each natural log unit increase in CRP; however when postmenopausal and premenopausal women were analyzed separately, the risk increase was significant only in postmenopausal women.

CRP is an established systemic marker of inflammation, being produced by the liver in response to cytokines (including IL-6 and TNF-α) produced by cells in inflamed tissue 15 , 16 . The lack of association between CRP and breast cancer risk in premenopausal women suggests that inflammation plays little or no role in premenopausal breast cancer.

However this conclusion is opposed by our finding that high TNF-α was associated with increased breast cancer risk in premenopausal women. Our finding, in fact, contrasts with the results of the only four prospective studies we are aware of to have investigated TNF-α and breast cancer risk 12 , 17 , 18 : three found no significant association between TNF-α and breast cancer; while a fourth case-control study on postmenopausal women nested in the Malmö Diet and Cancer Cohort 19 found an association between high TNF-α and reduced breast cancer risk.

TNF-α is a major mediator of inflammation: its induction, for example by tissue damage, induces a cascade of other inflammatory cytokines, chemokines, growth factors and endothelial adhesins which recruit and activate a range of cells at the site of tissue damage to promote healing 20 , 21 . However, when produced chronically, TNF-α seems to act as a tumor promoter, contributing to the tissue remodeling and stromal development necessary for tumor growth and spread 20 , 21 . Recent data suggest that TNF-α is involved in carcinogenesis at least in part because it activates nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) (reviewed in 22 ), which is responsible for inducing the expression of genes associated with cell proliferation, apoptosis, inflammation, metastasis, and angiogenesis 23 . TNF-α can also stimulate the activity of the inducible nitric oxide synthase (iNOS), which is implicated in cellular changes leading to malignancy (transformation of normal cells, growth of transformed cells, angiogenesis and metastasis of malignant cells 24 ).

Although our finding that TNF-α is associated with increased breast cancer risk in premenopausal women is not supported by previous studies, the association could be real since TNF-α has been found to stimulate the enzymes of estrogen synthesis 25 . High TNF-α could therefore promote breast cancer by this mechanism, which is likely to be more important in premenopause − before the fall in estrogen synthesis heralded by the menopause.

We found that plasma IL-6 was associated with increased breast cancer risk among premenopausal women, but only in the continuous model. IL-6 has been reported to activate Janus kinase (JAK) and signal transducer and activator of transcription 3 (STAT3) pathways 22 to promote a cellular microenvironment that may promote cancer growth. However, no association between increased IL-6 levels and breast cancer risk has been reported in previous studies, specifically the case-cohort study nested within the Women’s Health Initiative Observational Study 12 , the British Women’s Heart and Health Study cohort study and Caerphilly Cohort 26 , and the Health Aging and Body Composition prospective cohort study 17 .

The inverse association we found between plasma leptin and breast cancer risk in premenopausal women is supported by the findings of a case-control study by Harris et al . 27 on premenopausal women. By contrast the case-control study of Touvier et al . 13 found no association between leptin and breast cancer risk. Other prospective studies that investigated plasma leptin and breast cancer mainly involved postmenopausal women, and found no association 12 , 28 or a direct association 11 , 29 . Leptin is thought to be involved in promoting breast cancer in obese women by stimulating the conversion of aromatizable androgens (androstenedione and dehydroepiandrosterone) to estradiol. This occurs not only in adipose but also in breast epithelial cells, particularly when levels of circulating estrogens decline, as they do in postmenopausal women 30 , 31 . This mechanism might explain why high leptin was associated with non-significantly increased breast cancer risk in our postmenopausal women, even after adjusting for BMI. However, in premenopausal women, high leptin may lower breast cancer risk, since leptin is involved in the regulation of ovarian folliculogenesis 32 and at high levels may reduce follicular estradiol secretion 33 . Very high leptin levels have been reported in women with chronic anovulation 34 , a condition that may be associated with reduced breast cancer risk 35 .

As regards adiponectin, we found decreased breast cancer risk with increasing levels in postmenopausal women, in agreement with the findings of a 2013 meta-analysis that examined 17 observational studies (4 nested case-control studies and 9 case-control studies) 36 , and found no association of adiponectin with breast cancer risk overall, but decreased risk in postmenopausal women. A 2014 meta-analysis 37 which examined 15 observational studies (6 prospective and 9 case-control studies), found a 5% reduction in overall breast cancer risk for 3 μg/ml increments in adiponectin, but no significant associations in post- or premenopausal women examined separately. A 2015 case-cohort study on postmenopausal women found no association between plasma adiponectin and breast cancer risk 12 . A 2016 meta-analysis of 107 epidemiological studies 38 found that circulating adiponectin levels were lower in patients with various cancers than controls. However, other studies have found that increased adiponectin levels correlate with cancer progression (reviewed in 39 ); and in patients with viral infections or chronic inflammation, increased levels of adiponectin predict cancer development 39 . Obesity is protective against breast cancer in premenopause but increases breast cancer risk in postmenopause. Adiponectin levels are low in obesity, so a presumed cancer-promoting effect of low adiponectin in premenopause may be masked by concomitant and protective obesity. In postmenopause, obesity is not protective so low levels of adiponectin may be “freed” to exert a cancer promoting effect, possibly explaining our finding that high adiponectin was associated with lowered breast cancer risk in postmenopausal women only 40 .

Strengths of our study are its prospective design, relatively large sample size, and availability of detailed information on lifestyle that made it possible to control for confounding effects. A limitation is that we assessed variables at baseline only and do not know to what extent they may have changed subsequently. Limited data indicate that adipokine levels in a single blood sample are useful biomarkers of inflammation in population-based studies 41 .

Another limitation is that the relation between the circulating levels of the biomarkers we examined and their activity in breast or adipose is unknown, and it is possible that plasma levels may be a poor surrogate for local activity. For example leptin and adiponectin seem to function primarily in a paracrine manner, so circulating levels are unlikely to reflect biological activity in the breast 42 . Another possible limitation is that samples were collected, stored at −196 °C, and analyzed up to 20 years later. There may have been differential decay of the analytes over that period 43 , 44 . However, unless analyte decay varied with initial concentration (which seems unlikely), this will not bias analyte-risk associations.

To conclude, the findings of this case-control study nested in the EPIC-Varese cohort suggest that high levels of CRP and low levels of adiponectin may increase the risk of postmenopausal breast cancer, while high levels of TNF-α and IL-6, and low levels of leptin may increase breast cancer risk among premenopausal women. Further research is required to elucidate the mechanisms by which leptin can influence the etiology of premenopausal breast cancer; interventions to lower CRP and increase adiponectin levels might help reduce the risk of developing postmenopausal breast cancer, while interventions to lower TNF-α and IL-6 levels might help reduce the risk of developing premenopausal breast cancer.

Study population and data collection

The case-control study was nested within the 9378 women, resident in Varese Province, northern Italy, who were recruited in 1992-1997 (age 35–69 years) to the European Prospective Investigation into Cancer and Nutrition (EPIC)-Varese study, and gave blood samples on recruitment.

At baseline, detailed information was collected on reproductive and medical history, physical activity, alcohol consumption, smoking, education, and socioeconomic variables, using a standardized lifestyle questionnaire. Diet over the previous year was investigated using a food frequency questionnaire specifically developed to capture local dietary habits. Weight, height, and blood pressure were measured, and a 30 mL fasting blood sample was collected. The blood samples were divided into 0.5 mL aliquots of plasma, serum, red blood cells, and buffy coat, on the day of collection, and stored in liquid nitrogen at −196 °C 45 .

Ethics Statement

The study protocol was approved by the ethics committee of the Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. At baseline, participants signed a written informed consent to use clinical data for research. Consent forms were stored with barcode ID for subject identification. The ethics committee approved this consent procedure. The study protocol and informed consent procedure met the requirements of Italian legislation and the Declaration of Helsinki of 1975, as revised in 2008.

Breast cancer cases and selection of control women

The women were followed-up to December 31, 2009 (median 14.9 years), through the Varese section of the Lombardy Cancer Registry, characterized by high data completeness and quality. A total of 362 new breast cancers was identified.

For each case, one matched control was chosen, using an incidence density sampling protocol, from appropriate risk sets consisting of cohort members alive and free of cancer at the time of diagnosis of the index case. Matching criteria were age at recruitment (±5 years), date of recruitment (±180 days), menopausal status (postmenopausal, premenopausal, and perimenopausal at baseline), and analysis of inflammatory markers in the same batch.

Analysis of plasma samples

Plasma samples were analyzed using Luminex multiplex technology, which determines multiple analytes in a single microwell plate, using antibody kits purchased from Bio-Rad (Bio-Plex, TNF-α, IL-6, leptin, and adiponectin) or Merck (CRP) 46 . All the analyses were performed in duplicated and results with a intra-assay %CV > 20% were discarded. The contents of each well were read by Bio-Plex 100 System array reader (Bio-Rad Laboratories, California, USA), which identifies and quantifies each analyte based on bead color and fluorescent signal intensity. Instrument calibration procedure was performed daily by Bio-Plex Calibration Kit (Bio-Rad) for optimal performance and reproducibility of results. The data were processed using Bio-Plex Manager software (version 6.1) using five-parametric curve fitting and converted to pg/ml.

All kits supplied lyophilized standards that were reconstituted and diluted at 7 serial concentrations following manufacturer’s instructions (standard curves). Standards included all recombinant proteins tested and were considered as positive controls for the procedure. Standard diluent buffers alone were used as negative controls.

Statistical methods

Plasma levels of inflammatory molecules were grouped into tertiles based on the distribution in controls. Baseline characteristics of cases and controls, according to tertiles of plasma inflammatory biomarkers, were summarized as means and standard deviations (continuous variables) or frequencies (categorical variables). Conditional logistic regression models were used to estimate rate ratios (RRs) for breast cancer with 95% confidence intervals (CIs), with lowest tertile as reference. The significance of linear trends was assessed by treating each tertile as a continuous variable in the model and performing the Wald test. RRs were also calculated for 1 standard deviation increments in plasma concentration as a continuous variable. We ran a minimally adjusted model, adjusted for age (continuous) and BMI (<25 kg/m 2 , 25- < 30 kg/m 2 , ≥30 kg/m 2 ), and a fully-adjusted model, with the following additional covariates: age at menarche (<15 years, ≥15 years), parity (nulliparous, 1–2 children, >2 children), oral contraceptive use (never, sometime), education (≤8 years, >8 years), smoking status (never, former, current), and alcohol consumption (continuous).

We analyzed all women, and postmenopausal and premenopausal women separately. P values for interaction of inflammatory markers with menopausal status were estimated by adding the product of tertile of plasma concentrations and menopausal status to the model and applying the Wald test.

We excluded seven cases and their matched controls because a plasma sample was not available for the case or the control. We excluded four additional cases and controls because confounder variables were missing for the case or control. The analyses were therefore performed on 351 cases and 351 matched controls, total 702 women – 334 postmenopausal, 360 premenopausal, and 8 perimenopausal. All statistical tests were two-sided, differences were considered significant for P  < 0.05. The analyses were performed with Stata version 14.0 (College Station, TX, USA).

Data availability

The data that support the findings of this study are held by the corresponding author, however their availability is restricted: for ethical reasons, the Ethical Committee does not allow open/public sharing of data pertaining to individuals. However aggregated data are available to other researchers, upon request.

Torre, L. A. et al . Global cancer statistics, 2012. CA Cancer J. Clin. 65 , 87–108 (2015).

Article   PubMed   Google Scholar  

Coussens, L. M. & Werb, Z. Inflammation and cancer. Nature 420 , 860–867 (2002).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Crusz, S. M. & Balkwill, F. R. Inflammation and cancer: advances and new agents. Nat. Rev. Clin. Oncol. 12 , 584–596 (2015).

Article   CAS   PubMed   Google Scholar  

Chan, D. S., Bandera, E. V., Greenwood, D. C. & Norat, T. Circulating C-Reactive Protein and Breast Cancer Risk-Systematic Literature Review and Meta-analysis of Prospective Cohort Studies. Cancer Epidemiol. Biomarkers Prev. 24 , 1439–1449 (2015).

Frydenberg, H. et al . Pre-diagnostic high-sensitive C-reactive protein and breast cancer risk, recurrence, and survival. Breast Cancer Res. Treat. 155 , 345–354 (2016).

Wang, G. et al . A prospective follow-up study of the relationship between C-reactive protein and human cancer risk in the Chinese Kailuan Female Cohort. Cancer Epidemiol. Biomarkers Prev. 24 , 459–465 (2015).

Wulaningsih, W. et al . Prediagnostic serum inflammatory markers in relation to breast cancer risk, severity at diagnosis and survival in breast cancer patients. Carcinogenesis 36 , 1121–1128 (2015).

van Kruijsdijk, R. C., van der, W. E. & Visseren, F. L. Obesity and cancer: the role of dysfunctional adipose tissue. Cancer Epidemiol. Biomarkers Prev. 18 , 2569–2578 (2009).

Fantuzzi, G. Adipose tissue, adipokines, and inflammation. J. Allergy Clin. Immunol. 115 , 911–919 (2005).

Rose, D. P. & Vona-Davis, L. Biochemical and molecular mechanisms for the association between obesity, chronic inflammation, and breast cancer. Biofactors 40 , 1–12 (2014).

Gross, A. L., Newschaffer, C. J., Hoffman-Bolton, J., Rifai, N. & Visvanathan, K. Adipocytokines, inflammation, and breast cancer risk in postmenopausal women: a prospective study. Cancer Epidemiol. Biomarkers Prev. 22 , 1319–1324 (2013).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Gunter, M. J. et al . Circulating Adipokines and Inflammatory Markers and Postmenopausal Breast Cancer Risk. J. Natl. Cancer Inst . 107 , (2015).

Touvier, M. et al . Association between prediagnostic biomarkers of inflammation and endothelial function and cancer risk: a nested case-control study. Am. J. Epidemiol. 177 , 3–13 (2013).

Guo, L. et al . C-reactive protein and risk of breast cancer: A systematic review and meta-analysis. Sci. Rep. 5 , 10508 (2015).

Article   ADS   PubMed   PubMed Central   Google Scholar  

Volanakis, J. E. Human C-reactive protein: expression, structure, and function. Mol. Immunol. 38 , 189–197 (2001).

Gauldie, J. et al . Interferon beta 2/B-cell stimulatory factor type 2 shares identity with monocyte-derived hepatocyte-stimulating factor and regulates the major acute phase protein response in liver cells. Proc. Natl. Acad. Sci. USA 84 , 7251–7255 (1987).

Il'yasova, D. et al . Circulating levels of inflammatory markers and cancer risk in the health aging and body composition cohort. Cancer Epidemiol. Biomarkers Prev. 14 , 2413–2418 (2005).

Krajcik, R. A., Massardo, S. & Orentreich, N. No association between serum levels of tumor necrosis factor-alpha (TNF-alpha) or the soluble receptors sTNFR1 and sTNFR2 and breast cancer risk. Cancer Epidemiol. Biomarkers Prev. 12 , 945–946 (2003).

CAS   PubMed   Google Scholar  

Dias, J. A. et al . Low-Grade Inflammation, Oxidative Stress and Risk of Invasive Post-Menopausal Breast Cancer - A Nested Case-Control Study from the Malmo Diet and Cancer Cohort. PLoS. One. 11 , e0158959 (2016).

Article   PubMed   PubMed Central   Google Scholar  

Balkwill, F. & Mantovani, A. Inflammation and cancer: back to Virchow? Lancet 357 , 539–545 (2001).

Balkwill, F. Tumor necrosis factor or tumor promoting factor? Cytokine Growth Factor Rev. 13 , 135–141 (2002).

Prieto-Hontoria, P. L. et al . Role of obesity-associated dysfunctional adipose tissue in cancer: a molecular nutrition approach. Biochim. Biophys. Acta 1807 , 664–678 (2011).

Karin, M. Nuclear factor-kappaB in cancer development and progression. Nature 441 , 431–436 (2006).

Article   ADS   CAS   PubMed   Google Scholar  

Geller, D. A. & Billiar, T. R. Molecular biology of nitric oxide synthases. Cancer Metastasis Rev. 17 , 7–23 (1998).

Purohit, A., Newman, S. P. & Reed, M. J. The role of cytokines in regulating estrogen synthesis: implications for the etiology of breast cancer. Breast Cancer Res. 4 , 65–69 (2002).

Heikkila, K. et al . Associations of circulating C-reactive protein and interleukin-6 with cancer risk: findings from two prospective cohorts and a meta-analysis. Cancer Causes Control 20 , 15–26 (2009).

Harris, H. R., Tworoger, S. S., Hankinson, S. E., Rosner, B. A. & Michels, K. B. Plasma leptin levels and risk of breast cancer in premenopausal women. Cancer Prev. Res. (Phila) 4 , 1449–1456 (2011).

Article   CAS   Google Scholar  

Stattin, P. et al . Plasma leptin and breast cancer risk: a prospective study in northern Sweden. Breast Cancer Res. Treat. 86 , 191–196 (2004).

Ollberding, N. J. et al . Prediagnostic leptin, adiponectin, C-reactive protein, and the risk of postmenopausal breast cancer. Cancer Prev. Res. (Phila) 6 , 188–195 (2013).

Catalano, S. et al . Leptin enhances, via AP-1, expression of aromatase in the MCF-7 cell line. J. Biol. Chem. 278 , 28668–28676 (2003).

Pasqualini, J. R. et al . Concentrations of estrone, estradiol, and estrone sulfate and evaluation of sulfatase and aromatase activities in pre- and postmenopausal breast cancer patients. J. Clin. Endocrinol. Metab 81 , 1460–1464 (1996).

Brannian, J. D. & Hansen, K. A. Leptin and ovarian folliculogenesis: implications for ovulation induction and ART outcomes. Semin. Reprod. Med. 20 , 103–112 (2002).

Falk, R. T. et al . Interrelationships between serum leptin, IGF-1, IGFBP3, C-peptide and prolactin and breast cancer risk in young women. Breast Cancer Res. Treat. 98 , 157–165 (2006).

Brewer, C. J. & Balen, A. H. The adverse effects of obesity on conception and implantation. Reproduction. 140 , 347–364 (2010).

Terry, K. L., Willett, W. C., Rich-Edwards, J. W. & Michels, K. B. A prospective study of infertility due to ovulatory disorders, ovulation induction, and incidence of breast cancer. Arch. Intern. Med. 166 , 2484–2489 (2006).

Liu, L. Y. et al . The role of adiponectin in breast cancer: a meta-analysis. PLoS. One. 8 , e73183 (2013).

Macis, D., Guerrieri-Gonzaga, A. & Gandini, S. Circulating adiponectin and breast cancer risk: a systematic review and meta-analysis. Int. J. Epidemiol. 43 , 1226–1236 (2014).

Wei, T., Ye, P., Peng, X., Wu, L. L. & Yu, G. Y. Circulating adiponectin levels in various malignancies: an updated meta-analysis of 107 studies. Oncotarget (2016).

Hebbard, L. & Ranscht, B. Multifaceted roles of adiponectin in cancer. Best. Pract. Res. Clin. Endocrinol. Metab 28 , 59–69 (2014).

Barb, D., Williams, C. J., Neuwirth, A. K. & Mantzoros, C. S. Adiponectin in relation to malignancies: a review of existing basic research and clinical evidence. Am. J. Clin. Nutr. 86 , s858–s866 (2007).

PubMed   Google Scholar  

Lee, S. A. et al . Intra-individual variation of plasma adipokine levels and utility of single measurement of these biomarkers in population-based studies. Cancer Epidemiol. Biomarkers Prev. 16 , 2464–2470 (2007).

Vona-Davis, L. & Rose, D. P. Adipokines as endocrine, paracrine, and autocrine factors in breast cancer risk and progression. Endocr. Relat Cancer 14 , 189–206 (2007).

de Jager, W., Bourcier, K., Rijkers, G. T., Prakken, B. J. & Seyfert-Margolis, V. Prerequisites for cytokine measurements in clinical trials with multiplex immunoassays. BMC. Immunol. 10 , 52 (2009).

Flower, L., Ahuja, R. H., Humphries, S. E. & Mohamed-Ali, V. Effects of sample handling on the stability of interleukin 6, tumour necrosis factor-alpha and leptin. Cytokine 12 , 1712–1716 (2000).

Palli, D. et al . A molecular epidemiology project on diet and cancer: the EPIC-Italy Prospective Study. Design and baseline characteristics of participants. Tumori 89 , 586–593 (2003).

Allione, A. et al . Anticoagulants used in plasma collection affect adipokine multiplexed measurements. Cytokine 80 , 43–47 (2016).

Download references

Acknowledgements

We thank A. Evangelista and D. Del Sette for technical support, and Don Ward for help with the English. The Italian Association for Cancer Research (AIRC) provided financial support for this study.

Author information

Sabina Sieri and Vittorio Krogh jointly supervised this work.

Authors and Affiliations

Epidemiology and Prevention Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy

Claudia Agnoli, Sara Grioni, Valeria Pala, Sabina Sieri & Vittorio Krogh

Medical Sciences Department, University of Torino, Torino, Italy

Alessandra Allione, Giuseppe Matullo & Cornelia Di Gaetano

Italian Institute for Genomic Medicine (IIGM), Torino, Italy

Lombardy Cancer Registry Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy

Giovanna Tagliabue

You can also search for this author in PubMed   Google Scholar

Contributions

C.A., S.S., and V.K. conceived and designed the experiments. C.A., S.S., V.K., S.G., A.A., and G.M. performed the experiments. C.D.G. and G.T. contributed analysis tools. C.A., S.S., V.K., and V.P. wrote the paper. All authors reviewed the manuscript.

Corresponding author

Correspondence to Vittorio Krogh .

Ethics declarations

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Agnoli, C., Grioni, S., Pala, V. et al. Biomarkers of inflammation and breast cancer risk: a case-control study nested in the EPIC-Varese cohort. Sci Rep 7 , 12708 (2017). https://doi.org/10.1038/s41598-017-12703-x

Download citation

Received : 15 June 2017

Accepted : 18 September 2017

Published : 05 October 2017

DOI : https://doi.org/10.1038/s41598-017-12703-x

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Dietary inflammatory index and risk of breast cancer: evidence from a prospective cohort of 67,879 women followed for 20 years in france.

  • Mariem Hajji-Louati
  • Amandine Gelot
  • Francesca Romana Mancini

European Journal of Nutrition (2023)

Associations of aspirin and other anti-inflammatory medications with breast cancer risk by the status of COX-2 expression

  • Lusine Yaghjyan
  • A. Heather Eliassen
  • Rulla M. Tamimi

Breast Cancer Research (2022)

Prognostic significance of cachexia in advanced non-small cell lung cancer patients treated with pembrolizumab

  • Tatsuya Yoshida
  • Yuichiro Ohe

Cancer Immunology, Immunotherapy (2022)

Circulating inflammatory biomarkers, adipokines and breast cancer risk—a case-control study nested within the EPIC cohort

  • Manon Cairat
  • Sabina Rinaldi
  • Laure Dossus

BMC Medicine (2022)

Inflammatory biomarkers and risk of breast cancer among young women in Latin America: a case-control study

  • Emma Fontvieille
  • Mathilde His
  • Hailey Loucks

BMC Cancer (2022)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

  • Explore articles by subject
  • Guide to authors
  • Editorial policies

Sign up for the Nature Briefing: Cancer newsletter — what matters in cancer research, free to your inbox weekly.

breast cancer case control study

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

A case-control study of breast cancer risk factors in 7,663 women in Malaysia

Roles Formal analysis, Visualization, Writing – original draft

Affiliations Department of Applied Mathematics, Faculty of Engineering, University of Nottingham Malaysia Campus, Semenyih, Selangor, Malaysia, Cancer Research Malaysia, Subang Jaya, Selangor, Malaysia

ORCID logo

Roles Writing – review & editing

Roles Methodology, Project administration, Supervision

Affiliation Cancer Research Malaysia, Subang Jaya, Selangor, Malaysia

Roles Project administration

Roles Data curation

Roles Investigation

Roles Resources

Affiliation Department of Surgery, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia

Affiliations Department of Surgery, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia, Biomedical Imaging Department, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia

Affiliation Biomedical Imaging Department, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia

Roles Methodology

Affiliations Sime Darby Medical Centre, Subang Jaya, Selangor, Malaysia, Jeffrey Cheah School of Medicine and Health Sciences, Monash University Malaysia Campus, Subang Jaya, Selangor, Malaysia

Roles Investigation, Methodology, Resources

Affiliation Department of Paediatrics, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia

Roles Conceptualization, Funding acquisition, Methodology, Project administration, Supervision

Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Supervision

Affiliation Sime Darby Medical Centre, Subang Jaya, Selangor, Malaysia

  •  [ ... ],

Roles Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing – review & editing

* E-mail: [email protected]

  • [ view all ]
  • [ view less ]
  • Min-Min Tan, 
  • Weang-Kee Ho, 
  • Sook-Yee Yoon, 
  • Shivaani Mariapun, 
  • Siti Norhidayu Hasan, 
  • Daphne Shin-Chi Lee, 
  • Tiara Hassan, 
  • Sheau-Yee Lee, 
  • Sze-Yee Phuah, 

PLOS

  • Published: September 14, 2018
  • https://doi.org/10.1371/journal.pone.0203469
  • Reader Comments

Table 1

Breast cancer risk factors have been examined extensively in Western setting and more developed Asian cities/countries. However, there are limited data on developing Asian countries. The purpose of this study was to examine breast cancer risk factors and the change of selected risk factors across birth cohorts in Malaysian women.

An unmatched hospital based case-control study was conducted from October 2002 to December 2016 in Selangor, Malaysia. A total of 3,683 cases and 3,980 controls were included in this study. Unconditional logistic regressions, adjusted for potential confounding factors, were conducted. The breast cancer risk factors were compared across four birth cohorts by ethnicity.

Ever breastfed, longer breastfeeding duration, a higher soymilk and soy product intake, and a higher level of physical activity were associated with lower risk of breast cancer. Chinese had the lowest breastfeeding rate, shortest breastfeeding duration, lowest parity and highest age of first full term pregnancy.

Conclusions

Our study shows that breastfeeding, soy intake and physical activity are modifiable risk factors for breast cancer. With the increasing incidence of breast cancer there is an urgent need to educate the women about lifestyle intervention they can take to reduce their breast cancer risk.

Citation: Tan M-M, Ho W-K, Yoon S-Y, Mariapun S, Hasan SN, Lee DS-C, et al. (2018) A case-control study of breast cancer risk factors in 7,663 women in Malaysia. PLoS ONE 13(9): e0203469. https://doi.org/10.1371/journal.pone.0203469

Editor: Natarajan Aravindan, University of Oklahoma Health Sciences Center, UNITED STATES

Received: March 6, 2018; Accepted: August 21, 2018; Published: September 14, 2018

Copyright: © 2018 Tan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data collected in this study are compliant with the Data Protection Act in Malaysia and can only be shared with research groups that contact Cancer Research Malaysia directly. All requests for data should be sent to Joanna Lim at the Data Access Committee of Cancer Research Malaysia using the following email address: [email protected] .

Funding: This study was supported by grants from Newton-Ungku Omar Fund [grant no: MR/P012930/1] and Wellcome Trust [grant no: v203477/Z/16/Z]. The Malaysian Breast Cancer Genetic Study was established using funds from the Malaysian Ministry of Science, and the Malaysian Ministry of Higher Education High Impact Research Grant [grant no: UM.C/HIR/MOHE/06]. The Malaysian Mammographic Density Study was established using funds raised through the Sime Darby LPGA tournament and the High Impact Research Grant. Additional funding was received from Yayasan Sime Darby, PETRONAS and other donors of Cancer Research Malaysia. The Newton-Ungku Omar Fund (grant no: MR/P012930/a), https://www.britishcouncil.my/programmes/newton-ungku-omar-fund was used to establish the cohort; Wellcome Trust (grant no: v203477/Z/16/Z), https://wellcome.ac.uk/funding , was used to establish the cohort; and Malaysian Ministry of Higher Education High Impact Research Grant (grant no: UM.C/HIR/MOHE/06, https://www.mohe.gov.my/en/initiatives-2/187-program-utama/penyelidikan/548-research-grants-information , was used to establish the cohort.

Competing interests: The authors have declared that no competing interests exist.

Breast cancer risk factors have been examined extensively and the common ones include early age of menarche, late age of menopause, short breastfeeding duration, late age of first full term pregnancy, nulliparity and low parity [ 1 – 5 ]. However, most of these studies were conducted predominantly in developed countries in a Western setting. Although a limited number of studies examining women living in Asian countries also supported the association of these common risk factors with breast cancer [ 6 – 10 ], they were conducted in the more developed Asian cities/countries, or have been limited to sample sizes of several hundred women and mostly limited to one ethnicity. Therefore, there is a need to conduct a more extensive study with a larger sample size to determine whether these risk factors also play a similar role among Asian populations in developing countries, as this evidence should contribute importantly to the development of appropriate strategies for breast cancer prevention and control in Asia.

Malaysia offers a unique opportunity to examine breast cancer risk factors in Asian populations because of its multi-cultural and multi-religious setting, both of which might influence lifestyle and reproductive characteristics, and hence, breast cancer risk. Notably, the three main ethnicities in Malaysia, namely, Malay, Chinese and Indian, represent the three largest ethnic groups in Asia. Breast cancer is the most common cancer among Malaysian women and accounted for 31% of total female cancers [ 11 ]. The age-adjusted breast cancer incidence in Malaysia is 47.4/100,000, about half of that in North America [ 12 ]. Chinese have the highest incidence (59.9/100,000) followed by Indians (54.2/100,000) and Malays (34.9/100,000) [ 11 ]. Like many developing Asian countries, Malaysia is undergoing a transition toward a Westernized diet that is high in fat and sugar, an increasingly sedentary lifestyle [ 13 ] and also experiencing changes in reproductive characteristics [ 14 ]. Thus, there is an urgent need to examine the impact of these changes on breast cancer risk.

In this paper, we report the association between clinical, exogenous hormonal, menstrual, reproductive, anthropometric and lifestyle factors with breast cancer from a hospital-based case-control study of 7,663 women in Malaysia. We also present the change of selected breast cancer-related factors across birth cohorts and their implication for breast cancer in Malaysia and potentially other developing Southeast Asian countries.

Materials and methods

The study was approved by the Independent Ethics Committee, Ramsay Sime Darby Health Care (reference nos: 201109.4 and 201208.1), and the Medical Ethics Committee, University Malaya Medical Centre (reference no: 842.9). All participants provided written informed consent. The study was performed in accordance with the Declaration of Helsinki.

The Malaysian Breast Cancer Genetic Study (MyBrCa), initiated in 2002, is a hospital-based case-control study of breast cancer risk factors. The study participants are recruited from two participating hospitals in Selangor, Malaysia: University Malaya Medical Centre (UMMC), a public hospital, and Subang Jaya Medical Centre (SJMC), a private hospital. All patients diagnosed clinically with breast carcinoma are eligible for inclusion as cases. Cases from UMMC were recruited since October 2002, and from SJMC, since September 2012. Controls are healthy women between ages 40 and 74 with no personal history of breast cancer and recruited in the Malaysian Mammography Study (MyMammo) at UMMC and SJMC. At SJMC, MyMammo is a subsidized opportunistic mammogram screening programme that was initiated in 2011; while at UMMC, MyMammo started recruitment in 2014 from patients attending routine opportunistic screening in UMMC.

All participants were interviewed by trained interviewers at the hospitals. The participants completed questionnaire that included items related to demographics, personal and family history of cancers, history of breast surgery, menstrual and reproductive history, use of oral contraceptive and hormone replacement therapy (HRT), breast cancer diagnosis (cases only) and history of and motivation of attending mammography screening (controls) only. The participants provided a blood sample that was processed and stored.

Statistical analysis

To date, a total of 4,056 cases and 4,145 controls were recruited and interviewed. Only participants recruited before 1 January 2017 were included in this study. After removing duplicates, males and non-breast cancer cases, the remaining cohort consists of 3,683 cases and 3,980 controls.

Ever had breast surgery was defined as whether the participant had surgery for a benign lump or cyst in the breast. Women who had sisters/mothers/daughters with breast cancer were categorized as having a first-degree family history of breast cancer. Ever used oral contraceptives and HRT was defined as at least one month of usage. Post-menopausal status was defined as no menses for the past one year. The participants were categorized as parous if they had at least one full term pregnancy (live or still birth). BMI was calculated as dividing weight (kg) by the square of height (m). Soy products intake included the consumption of tofu, fermented soybeans, tofu pudding, and soy products other than soymilk. The participants reported their average duration of strenuous, moderate and gentle physical activity of three periods: childhood (before 18 years old), young adulthood (18–30 years), and the recent years. Weekly metabolic equivalent (MET)-hours were obtained by multiplying the corresponding MET value of each intensity of physical activity (7, 4, 3 for strenuous, moderate and gentle activities, respectively) with the average time spent on physical activity [ 15 ].

Cases and controls were compared using chi-square tests for categorical variables and t-tests for continuous variables. Unconditional logistic regressions were conducted to assess the association between risk factors and breast cancer, adjusting for potential confounders and other risk factors. The first models were adjusted for age, ethnicity, and hospital; for history of breast surgery, and anthropometric and lifestyle variables, the models were adjusted for age, ethnicity and education, and only participants from private hospital were included. In the second models, other breast cancer risk factors such as age of menarche, age of menopause, ever had full term pregnancy, first degree family history of breast cancer, and age of first full term pregnancy were added when appropriate. Conditional logistic regression using hospital-, ethnicity- and age- (±5 years) matched samples and unconditional logistic regressions stratified by pre- and post-menopausal status were also conducted. However, the results were similar to the unconditional and unstratified analysis thus they are not reported here.

The participants were categorized based on their year of birth into four birth cohorts: those born before 1949, between 1950–59, between 1960–69, and after 1969, and their breast cancer risk factors were compared across the birth cohorts. To compare across ethnicity, analysis of variances (ANOVAs) were conducted for continuous variables while chi-square tests were conducted for categorical variables. To determine whether there was a change of trend in the selected variables across birth cohorts, trend analyses were conducted by entering the birth cohort variable as continuous parameter in the regression models.

All analyses were conducted using R [ 16 ].

Table 1 is the demographic comparisons of cases and controls. Controls were significantly older than cases, with mean ages of 54.0 years and 50.8 years, respectively (p<0.001) and significantly more controls had received secondary education. There were significantly more Chinese among the cases.

thumbnail

  • PPT PowerPoint slide
  • PNG larger image
  • TIFF original image

https://doi.org/10.1371/journal.pone.0203469.t001

We conducted unconditional logistic regression to examine the association of clinical, exogenous hormonal, menstrual and reproductive factors with breast cancer ( Table 2 ). Compared with those who had never had breast surgery, participants who had breast surgery to remove cysts and lumps were 2.3 times (95% CI = 1.82–2.83) more likely to develop breast cancer after adjusting for demographics and other risk factors. First-degree family history of breast cancer was associated with 19% increased risk of breast cancer after adjusting for demographics and other risk factors. Post-menopausal women had a 52% increased risk of breast cancer after adjusting for demographics and other risk factors. The use of oral contraceptives and HRT were not significantly associated with breast cancer risk after adjustment of other breast cancer risk factors.

thumbnail

https://doi.org/10.1371/journal.pone.0203469.t002

Of the menstrual and reproductive factors examined, breastfeeding had the strongest protective effect against breast cancer ( Table 2 ). Among parous women, those who ever breastfed had 35% lower risk in the fully adjusted models; compared with those who did not breastfeed, the reduction of risk for those who breastfed between 1–12 months and those who breastfed more than 12 months was 30% and 70% respectively.

We also examined the association between anthropometric and lifestyle factors and breast cancer ( Table 3 ). A higher BMI was associated with a lower risk of breast cancer; those who are overweight (BMI = 23.0–27.4kg/m 2 ) had 33% reduced risk and those who are obese (BMI ≥ 27.5kg/m 2 ) had 53% reduced risk, after controlling for other risk factors ( Table 3 ). Those who consumed one cup or more soymilk per week and soy products once or more per week had 75% and 60% reduction in breast cancer risk, respectively. We did not find any significant association between smoking status and breast cancer. Women who drink less than 1 glass of alcohol per week and 1 glass per week or more had 55% and 48% reduced risk of breast cancer. It is noteworthy that the prevalence of those who reported alcohol intake in our cohort is low at 14%. A higher level of physical activity during childhood, young adulthood and recent period were also significantly associated with reduced risk of breast cancer before and after adjusting for other risk factors.

thumbnail

https://doi.org/10.1371/journal.pone.0203469.t003

Change of risk factors by birth cohorts

We examined the change of all risk factors across birth cohorts of both controls and cases in the three major ethnic groups in Malaysia and here we report the variables that had significantly changed across birth cohorts. Fig 1 showed the change of parity, age of first full term pregnancy, breastfeeding rate, breastfeeding duration and total soy intake. Compared with Indians and Malays, Chinese have the lowest parity, oldest age of first full term pregnancy, lowest breast feeding rate and shortest breastfeeding duration (p<0.001). All ethnic groups were experiencing significant reduction in parity (p<0.001 for all races) and significant increase of age of first full term pregnancy (p<0.001 for Chinese and p<0.001 for Malays and Indians) across birth cohorts.

thumbnail

https://doi.org/10.1371/journal.pone.0203469.g001

All ethnic groups had significant increase of breastfeeding rate across birth cohorts (p<0.001). The increase was more noticeable among Chinese; there was an increase from 50% among the oldest cohort to 79% among the youngest cohort. Only Chinese had a significant increase of breastfeeding duration across birth cohorts (p<0.001); however, breastfeeding duration among Chinese remained low compared with Malays and Indians.

There was a significant decrease of total soy intake among Chinese (p<0.001) and Malay (p<0.05) across birth cohorts. Compared with Malays and Indians, Chinese consumed significantly less soy products (p<0.001). However, the number of Malays and Indians who reported their intake of soy products were small compared with that of Chinese.

In this hospital-based case-control study of 7,663 Malaysian women, we showed that a higher breastfeeding rate and duration, soy intake and level of physical activity were associated with a reduced risk of breast cancer among Southeast Asian women. Although Southeast Asian countries are experiencing a substantial increase in the burden of breast cancer, there have been limited studies in risk factors for breast cancer in these populations. Before the Malaysian Breast Cancer Genetic study, the previous largest study on breast cancer risk factors in Southeast Asia was from Indonesia and included 526 cases and 1,052 controls [ 17 , 18 ]. A large scale prospective cohort study that followed 35,303 women in which 629 developed breast cancer has been conducted in Singapore, however, it focused mainly on soy intake and breast cancer risk and was limited to Chinese only [ 19 ]. Our current study included a large sample size and examined a wide range of breast cancer risk factors.

The strongest predictor of breast cancer in our study was breastfeeding, and the inverse association between breastfeeding and breast cancer risk is well documented [ 6 , 7 , 20 – 22 ]. Our study also showed an increasing trend of breastfeeding across birth cohorts in all ethnicity; however, among Chinese the breastfeeding rate and duration were still relatively low. The low breastfeeding rate and short breastfeeding duration may contribute to the highest breast cancer incidence (59.9/100,000) among Chinese in Malaysia compared with Indians (54.2/100,000) and Malays (34.9/100,000) [ 11 ]. Thus, the results of our study could be helpful in public health strategies to reduce risk of breast cancer through modifiable lifestyle choices including breastfeeding.

Our study also found that higher intake of soymilk and soy products is associated with lower risk of breast cancer. Soy is a major food in many parts of Asia and retrospective cross-sectional cohort studies in China and Japan show that increased soy protein intake is associated with reduced breast cancer risk in pre- and post-menopausal women [ 23 , 24 ]. A study in Singapore showed that increased soy intake was significantly associated with reduced breast cancer risk among pre-menopausal women but not post-menopausal women [ 9 ] while another study in China found no significant association between soy protein intake and breast cancer risk. While there is some heterogeneity across these Asian studies, meta-analyses of observational studies in both Caucasian and Asian countries have consistently shown that high soy intake is associated to a lower risk of breast cancer, particularly among Asian women [ 25 – 30 ]. Given that our results shows declining soy intake across birth cohorts, future studies are required to confirm the benefit of soy in reducing population risk of breast cancer, as well as to also identify effective strategies to increase soy intake among Asian women, for whom a soy intervention may be an affordable and acceptable strategy for breast cancer prevention.

Another lifestyle factor that is shown to be associated with decreased breast cancer risk in our study is physical activity. This is consistent with the latest World Cancer Research Fund report which showed strong probable evidence that regular physical activity of various intensity decreases the risk of breast cancer among post-menopausal women while among pre-menopausal women, regular vigorous physical activity is associated with reduced risk [ 31 ]. A recent systematic review evaluated 80 studies and found that moderate-vigorous physical activity is associated with lower breast cancer risk among pre-menopausal (RR = 0.80, 95% CI = 0.74–0.87) and post-menopausal cohort studies (RR = 0.79, 95% CI = 0.76–0.84) [ 32 ]. Another systematic review that examined the dose-response between physical activity and major non-communicable diseases, which included breast cancer, found that compared with insufficiently active women, the reduction of risk of breast cancer among the low active, moderately active and highly active was 3%, 6% and 14% respectively [ 33 ]. Compared with other populations, Malaysian women have a higher prevalence of physical inactivity [ 34 ] and in our study there was no significant change of physical activity across birth cohorts. Thus, there is a need to construct innovative strategy to increase the level of physical activity in order to reduce future breast cancer risk.

In our study, two risk factors were associated with breast cancer risk in the contradictory direction. First, the consumption of alcohol was associated with a decreased risk of breast cancer in our study. The association of alcohol consumption with increased breast cancer risk has long been established [ 35 ]. However, in our study, only 6% reported an intake of more than 1 glass of alcohol per week, which is low compared with other populations. The second risk factor that had a contradictory association with breast cancer in our study was a higher BMI, which was associated with a lower risk of breast cancer after adjustment for major breast cancer risk factors. Past studies have shown that a higher BMI is associated with increased risk among post-menopausal women and reduced risk among pre-menopausal women [ 36 ]. However, when stratified by menopausal status, our analysis showed that a higher BMI was still significantly associated with lower breast cancer risk in both pre- and post-menopausal women. More studies need to be conducted among the Malaysian women to further explore the link between BMI and breast cancer risk.

In addition, our study did not find a significant association between parity, age of first full term pregnancy, age of menarche and menopause and breast cancer, which is inconsistent with other studies [ 2 – 5 , 8 , 10 , 37 – 41 ]. Our study also found only a slight association between first-degree family history of breast cancer and increased risk of breast cancer risk, while other studies show that family history is strongly associated with increased breast cancer risk [ 20 , 21 , 39 , 41 – 44 ].

Since this is a hospital-based case-control study rather than population-based, it might be subject to selection bias. The two hospitals where our participants were recruited were located in urban areas and rural Malaysian women were not included. However, it is noteworthy that these hospitals treat more than 10% of the breast cancer cases in Malaysia. The controls of our study were enriched for women who had a family history of breast cancer because they were participants of an opportunistic mammography screening programme.

In conclusion, our study shows that breastfeeding, soy intake and physical activity are modifiable risk factors for breast cancer; and with the increasing incidence of breast cancer there is an urgent need to educate the women about lifestyle intervention they can take to reduce their risk of breast cancer.

Supporting information

S1 file. survey questions..

This file contains the questionnaire items used in this study.

https://doi.org/10.1371/journal.pone.0203469.s001

Acknowledgments

We want to thank Pui-Yoke Kwan, Norhashimah Hassan, Peter Choon-Eng Kang, In-Nee Kang, Kah-Nyin Lai, Hanis Hasmad, Jin-Tong Ng, Dr. Gaik-Theng Toh, Nancy Geen-See Tan, Dr. Suhaida Selamat, Dr. Rohaya Mohd Kasim, Dr. Malkit Kaur Dhillon, Dr. Thin-Chai Liu, Ernie Azwa, Hanani Che Halim, Leelavathy Krishnan, Don-Na Tan, Sweet-Lin Goh, Nur Naquiah Kamaruddin, Faridah Bakri, the participants of this study, and all staff at Cancer Research Malaysia, University Malaya, and Sime Darby Medical Centre who assisted in recruitment and interviews.

  • View Article
  • Google Scholar
  • PubMed/NCBI
  • 11. Lim GCC, Rampal S, Yahaya H. Cancer incidence in Peninsular Malaysia, 2003–2005: The Third Report of the National Cancer Registry, Malaysia: National Cancer Registry; 2008.
  • 16. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing: Vienna, Austria; 2016.
  • 31. World Cancer Research Fund International. Diet, Nutrition, Physical Activity and Breast Cancer. London: American Institute for Cancer Research; 2017.
  • Open access
  • Published: 07 February 2024

Are better AI algorithms for breast cancer detection also better at predicting risk? A paired case–control study

  • Ruggiero Santeramo 1 , 2 ,
  • Celeste Damiani 1 , 3 ,
  • Jiefei Wei 4 ,
  • Giovanni Montana 2 , 4 &
  • Adam R. Brentnall 1  

Breast Cancer Research volume  26 , Article number:  25 ( 2024 ) Cite this article

656 Accesses

1 Altmetric

Metrics details

There is increasing evidence that artificial intelligence (AI) breast cancer risk evaluation tools using digital mammograms are highly informative for 1–6 years following a negative screening examination. We hypothesized that algorithms that have previously been shown to work well for cancer detection will also work well for risk assessment and that performance of algorithms for detection and risk assessment is correlated.

To evaluate our hypothesis, we designed a case-control study using paired mammograms at diagnosis and at the previous screening visit. The study included n = 3386 women from the OPTIMAM registry, that includes mammograms from women diagnosed with breast cancer in the English breast screening program 2010–2019. Cases were diagnosed with invasive breast cancer or ductal carcinoma in situ at screening and were selected if they had a mammogram available at the screening examination that led to detection, and a paired mammogram at their previous screening visit 3y prior to detection when no cancer was detected. Controls without cancer were matched 1:1 to cases based on age (year), screening site, and mammography machine type. Risk assessment was conducted using a deep-learning model designed for breast cancer risk assessment (Mirai), and three open-source deep-learning algorithms designed for breast cancer detection. Discrimination was assessed using a matched area under the curve (AUC) statistic.

Overall performance using the paired mammograms followed the same order by algorithm for risk assessment (AUC range 0.59–0.67) and detection (AUC 0.81–0.89), with Mirai performing best for both. There was also a correlation in performance for risk and detection within algorithms by cancer size, with much greater accuracy for large cancers (30 mm+, detection AUC: 0.88–0.92; risk AUC: 0.64–0.74) than smaller cancers (0 to < 10 mm, detection AUC: 0.73–0.86, risk AUC: 0.54–0.64). Mirai was relatively strong for risk assessment of smaller cancers (0 to < 10 mm, risk, Mirai AUC: 0.64 (95% CI 0.57 to 0.70); other algorithms AUC 0.54–0.56).

Conclusions

Improvements in risk assessment could stem from enhancing cancer detection capabilities of smaller cancers. Other state-of-the-art AI detection algorithms with high performance for smaller cancers might achieve relatively high performance for risk assessment.

Introduction

Breast cancer is one of the most prevalent diseases and common cause of death in women worldwide, despite improvements in treatment and mammography screening coverage [ 1 ]. Early detection of breast cancer remains an important objective for population health. In recent years, improvements in AI and deep learning technologies have helped to improve the technical accuracy of cancer detection methods [ 2 , 3 , 4 ], including for breast cancer detection [ 5 , 6 ]. Real-world evaluations of how these tools might be used effectively in practice are ongoing [ 7 ]. Coupled with these developments for cancer detection are ongoing studies to evaluate risk-based breast cancer screening. This paradigm aims to personalise early detection by directing more intensive screening to those at greatest risk of death from cancer [ 8 ]. Retrospective evidence suggests that using AI based on mammograms for risk assessment might be more informative over a 1–6-year period than classical models, and therefore a potentially important component for new risk-based strategies [ 9 , 10 , 11 ].

One algorithm that has been evaluated in multiple settings for risk assessment is called Mirai [ 12 ]. This deep-learning algorithm was trained for risk assessment using a large US cohort. A limitation of this method compared with classical breast cancer risk models using epidemiological risk factors [ 13 ] is that the AI model is largely a black box, and it is not clear why it appears to perform well. In an earlier analysis that we conducted of this algorithm using data from the OPTIMAM repository, we noted that it performed quite well as a cancer detection algorithm, that it was a good predictor of advanced (Stage 2+) cancers, and that the performance and probability of cancer increased as mammograms were taken closer to diagnosis [ 10 ]. Other algorithms trained for cancer detection have also shown potential utility for risk assessment, including one called GMIC (globally aware multiple instance classifier) [ 12 , 14 ]. We therefore hypothesized that algorithms that have previously been shown to work well for cancer detection will also work well for risk assessment and further that performance of algorithms for detection and risk assessment is correlated. In this paper, we report a study to test this hypothesis directly. This is done by evaluating the performance of different algorithms for both detection and risk assessment using paired mammograms on the same women attending screening in the English program, taken at the time cancer was detected at screening, and at a previous screening appointment three years prior to diagnosis when cancer was not detected or detectable by radiologists.

Patients attended the National Health Service Breast Screening Program (NHSBSP) between February 2010 and September 2019 at sites that are part of the OPTIMAM mammography image database (OMI-DB, see supplementary material and Halling-Brown et al. [ 15 ]). The source data are accessible for other research groups. Our manuscript reports new work that has not been undertaken or reported previously using data from this database [ 16 ]. Patients were eligible for inclusion if they had standard four-view mammography of ‘for presentation’ type and had normal or malignant episode outcomes. Screening episodes were excluded if the mammograms were not from Hologic machines, or if the woman was not 46–74 year at the time of their screening mammogram, or the woman had breast implants. All mammograms were taken using Hologic Lorad Selenia or Hologic Selenia Dimensions Mammography Systems, following requirements for the MIRAI algorithm [ 17 ].

Four open-source algorithms were applied. They were designed for breast cancer risk assessment, or breast cancer detection. The first algorithm (Mirai version 0.3.1) is a risk assessment algorithm. It is composed of four underlying modules that process each mammogram separately, then combine information before estimating risk annually for 5 years [ 18 ]. In our study, we exclusively provided image data as input to the examined algorithm. However, it is also equipped to process additional risk factor data if available. The other selected algorithms were initially designed for the purpose of cancer detection. These were: GMIC [ 14 ], and two others from a New York (NY) group that we call: NY [ 19 ], and NY-H, where H denotes heatmap and the algorithm is an extension of NY model using heatmaps [ 19 ]. Both the NY and NY-H models, along with GMIC, employ deep convolutional neural networks. Among them, GMIC operates with the most intricate architecture. It produces pixel-level saliency maps highlighting potential malignant areas. This innovation finds application in screening mammography interpretation, specifically in predicting the presence of benign or malignant lesions. In contrast, NY-base and NY-H introduce a deep convolutional neural network for breast cancer screening examination classification, employing a two-stage architecture and training approach that strategically combine multiple input views. All three algorithms were trained and evaluated on the same dataset with image-level labels comprising over 225,000 examinations and more than 1 million images. Due to their development process, we a priori expected GMIC to perform better than NY-H for cancer detection and NY-H to perform better than NY. The reason for including all three algorithms is that the expected variation in performance is helpful to test our hypothesis on the correlation between algorithm performance for detection and risk assessment.

Study design

The target population of our study was women who attended the NHS Breast Screening Program 2010–2019. The primary endpoint was diagnosis of invasive or insitu carcinoma, with biopsy-confirmed cancer diagnoses as recorded on the National Breast Screening System (NBSS). The main predictor variable evaluated was probability of cancer according to the algorithm. For Mirai we used 3 year risk, corresponding to the triennial population screening program. Age range was restricted because most women have screening age 50 to 70 years in England; with some starting aged 47 years or ending aged 73 years during the study epoch due to the age-extension trial (ISRCTN33292440); opportunistic screening is available for those older than 73 years. The case–control study reported is a sub-study of the case–control study reported by [ 10 ]. Only screen-detected cancers were included from the earlier work, restricting analysis to matched pairs (case and control) with screening mammograms taken both at diagnosis and 3 years prior to diagnosis (or pseudo diagnosis). The primary focus was on the relationship between performance of predictions by AI algorithms at the time of detection and for risk assessment 3 years prior to detection.

We determined that sample size was sufficient for this analysis before running the algorithms. This was because previous analysis showed a very strong statistical relationship between the Mirai algorithm for both detection and risk [ 10 ].

Statistical analysis

All analysis was adjusted for the site where mammography was done, the model of the mammography device, and where appropriate, age. Predictive performance was measured using the area under the curve (AUC) associated with the algorithm prediction, after adjustment for matching factors with 95% confidence intervals from Wilson’s method [ 20 ]. Heterogeneity was assessed using likelihood-ratio tests for interaction based on conditional logistic regression models. The strength of association between risk of breast cancer at diagnosis with risk of breast cancer 3 year prior to diagnosis was evaluated graphically, sub-categorized by tumor size and type (invasive, size unknown; DCIS; invasive: 0 to < 10 mm; 10 to < 20 mm; 20 to < 30 mm; 30 mm+), and by grade (1–2 vs. 3) and estrogen-receptor (ER) status (postive/negative). These cutpoints have been used previously [ 10 ]. This was done by estimating model performance using women with mammograms at diagnosis and 3 years prior to diagnosis. A further analysis was conducted by sub-categorizing the data based on the age at which patients were diagnosed with cancer. This stratification involved the utilization of age subgroups (< 55, 55 to 59, 60 to 64, 65 to 69, and 70+), which were chosen to keep the age range in each group relatively constant.

Women were included in this analysis from the previous case–control study if they had mammograms at detection (or pseudo detection if controls) and their previous screening visit 3 years earlier. This led to complete data on n = 3386 cases and controls, matched 1:1. Basic demographic characteristics of those included in the study, including matching variables, are shown in Table 1 .

figure 1

Spearman correlation between algorithms when used on mammograms 3 year prior to diagnosis (risk) or at diagnosis (det)

figure 2

Receiver operating characteristics for detection and risk by algorithm

Figure  1 reports Spearman correlation between the algorithms in controls and cases. In controls, there was only a weak correlation between the algorithm prediction and age. Correlation was moderate to good between algorithms. Mirai was more correlated with itself for detection and risk than the other three algorithms; the NY and NY-H for risk were more correlated with each other than with Mirai or GMIC, as might be expected due to their similar architecture and process of development. It is also of interest that there was a higher correlation between Mirai and age, than the other algorithms. Because age is a strong risk factor for breast cancer, and Mirai was trained for risk assessment, it appears that it uses age from mammography scans for risk assessment, whereas the other algorithms developed for cancer detection do not. A different correlation structure was observed in the cases. Here the algorithms were more correlated between themselves for risk or detection, than with themselves for the other set of mammograms. Overall, there was again moderate to good correlation between the algorithms, but also some variation indicating lack of agreement in the rank ordering of patients.

The association between algorithm performance for risk and detection is examined further in Table 2 and Fig. 3 . Model performance for detection was consistently better for larger cancers across all algorithms. In addition, both within cancer types and size, and across algorithms, there was a pattern whereby better performance for cancer detection was associated with better performance for risk assessment. A similar but less strong pattern was seen by age subgroup (Additional file 1 : Fig. S1). On average, the algorithm performed best for risk in the age groups where it performed best for detection. Similar findings were observed by ER subgroups (Additional file 1 : Fig. S2) and for cancer grade 1 and 2 (Additional file 1 : Fig. S3). However, there was very little association seen for grade 3 cancers, suggesting that different radiological features are observed at detection than at the previous screening mammogram in faster-growing grade 3 cancers.

Figure  2 plots receiver operating characteristics (ROCs) for risk and detection by algorithm. The same ordering of performance by algorithm for both is seen, with Mirai having strongest performance for both risk assessment and detection, and the NY algorithm the weakest.

figure 3

Association between AUC for diagnosis and AUC for risk by algorithm (Mirai black, GMIC red, NY green, NY-H blue) overall ( \(+\) , with width of the bars corresponding to the 95% CI) and by type and size of cancer when detected

Our analysis suggests that algorithms that perform better for cancer detection are also likely to perform better for risk assessment. We found evidence that “easier to detect” larger cancers at diagnosis are also likely to be given a higher probability of malignancy three years prior to diagnosis, i.e., for risk assessment.

There are several implications of our findings. Firstly, improvements for risk assessment algorithms for breast cancer might be gained by improving their performance for cancer detection. For example, analysis by subgroup of cancer size showed that GMIC was broadly comparable with Mirai at detection of larger cancers (> 10 mm) and DCIS, but worse for smaller cancers both at diagnosis and at risk assessment. To achieve similar performance for risk assessment therefore, one might suggest additional training or developments of the GMIC algorithm to focus on smaller cancers—improvements to risk assessment are likely to follow. Secondly, our results suggest that state-of-the-art algorithms for breast cancer detection might be considered to be repurposed for risk assessment. In time, AI for mammography is likely to become implemented in national screening programs such as the UK. Such developments could then enable routine risk assessment to help drive new risk-based screening regimens. Thirdly, our findings help to explain why Mirai works well for risk assessment: it is finding early signs of cancer. These are likely most visible in the larger cancers at screen detection because they are more likely to have been there at the previous screen than smaller cancers which might have only developed in the interval between screens. Fourthly, our results suggest that, more generally, deep learning computer vision algorithms are able to discern intricate patterns in breast scans, which are not currently acted upon by radiologists. Their ability to extract latent insights from visual data only without the use of any classical risk factors suggests that the development of a more sophisticated diagnostic models should yield better results for risk assessment.

Strengths of our study include the paired design, whereby the mammograms at detection and earlier screening rounds were on the same women. Our design has not been used before to assess correlation between performance for detection and risk assessment. It is also little applied in other work on AI algorithms for breast cancer risk assessment, where most publications have focused on cancer following a single screening visit for risk assessment. Using paired data lets us test our hypothesis more reliably than indirect comparisons of performance for risk and detection using samples of different women. Another strength is that this study was an external validation assessment of all the algorithms, with no training or fine tuning done. This helps to ensure a reliable evaluation.

There are several limitations to our study. Firstly, although some algorithms produce heatmaps that can provide a more in-depth view of the inner mechanisms, the algorithms were applied as a “black box”, and we do not know if the higher risk was due to a suspicious area in the region where the cancer was found, or something else. For example, an alternative explanation for the findings might be that the algorithms identify a field effect in the breast, not a specific pattern associated with breast cancer. Secondly, our study is a retrospective and observational case–control study. The area is largely lacking evaluation through more prospective designs, and the retrospective nature of this work makes it at risk of bias including related to the decision to seek publication of results. Thirdly, the analysis is limited by when and where screening mammograms were recorded (e.g., it was based on women attending the English screening program, but we do not know the race or ethnicity of those included). Fourthly, we were unable to compare directly with other domains or risk models, including family history and polygenic risk scores; or the other risk factors that may be added to Mirai. Fifthly, we were limited by availability of code to run pre-defined algorithms for risk or detection. Other algorithms may perform differently, and this is worth further investigating. Lastly, it is important to note that this study specifically focused on mammograms acquired via Hologic machines, which may constrain the applicability of the findings to other types of mammogram machines.

In conclusion, this study evaluated whether the performance of an AI model for detection is associated with its performance for risk assessment. We did this using four open-source algorithms. The analysis suggests that algorithms that excel at cancer detection also perform well for risk assessment. The correlation between the ability to detect cancer in mammograms and the ability to assess the risk of developing cancer suggests that improvements in risk assessment algorithms could be obtained by focusing on improving their capabilities for cancer detection. For instance, algorithms may need additional training on detecting smaller cancers to achieve better performance in risk assessment. More generally, current state-of-the-art detection algorithms might be repurposed for risk assessment. This could enable the AI technologies currently being trialled to aid cancer detection using mammograms to play a vital role in future risk-based screening programs. For example, it might be advisable to recommend more frequent screening for higher-risk patients [ 21 ]. Finally, the paired mammograms in our study were about 3 years apart as per the standard breast cancer screening interval in the UK. Therefore, the evidence reported is most relevant to short-term breast cancer risk, perhaps due to the detection of indolent breast cancers not detected by the human eye. To better inform long-term mammography screening patterns, developing breast cancer risk prediction models over a longer time horizon would be useful. This extension could provide a comprehensive understanding of breast cancer risk dynamics and contribute to refining strategies for effective and personalized long-term screening.

Availability of data and materials

The images and data used in this publication are derived from the OPTIMAM imaging database [ 15 ], we would like to acknowledge the OPTIMAM project team and staff at the Royal Surrey NHS Foundation Trust who developed the OPTIMAM database, and Cancer Research UK who funded the creation and maintenance of the database.

Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71:209–49.

Article   PubMed   Google Scholar  

Huang S, Yang J, Fong S, Zhao Q. Artificial intelligence in cancer diagnosis and prognosis: opportunities and challenges. Cancer Lett. 2020;471:61–71.

Article   CAS   PubMed   Google Scholar  

Santeramo R, Withey S, Montana G. Longitudinal detection of radiological abnormalities with time-modulated LSTM. In: Deep learning in medical image analysis and multimodal learning for clinical decision support: 4th international workshop, DLMIA 2018, and 8th international workshop, ML-CDS 2018, held in conjunction with MICCAI 2018, Granada, Spain, September 20, 2018, proceedings, vol 4. Springer; 2018. p. 326–33.

Chattopadhyay A, Maitra M. MRI-based brain tumour image detection using CNN based deep learning method. Neurosci inform. 2022;2:100060.

Article   Google Scholar  

Shen L, Margolies LR, Rothstein JH, Fluder E, McBride R, Sieh W. Deep learning to improve breast cancer detection on screening mammography. Sci Rep. 2019;9:12495.

Article   PubMed   PubMed Central   ADS   Google Scholar  

Nassif AB, Talib MA, Nasir Q, Afadar Y, Elgendy O. Breast cancer detection using artificial intelligence techniques: a systematic literature review. Artif Intell Med. 2022;127: 102276.

Nijnatten TJ, Payne NR, Hickman SE, Ashrafian H, Gilbert FJ. Overview of trials on artificial intelligence algorithms in breast cancer screening—a roadmap for international evaluation and implementation. Eur J Radiol. 2023;167: 111087.

Harkness EF, Astley SM, Evans DG. Risk-based breast cancer screening strategies in women. Best practice and research. Clin obstet Gynaecol. 2020;65:3–17.

Google Scholar  

Arasu VA, Habel LA, Achacoso NS, et al. Comparison of mammography AI algorithms with a clinical risk model for 5-year breast cancer risk prediction: an observational study. Radiology. 2023; 307: e222733.

Damiani C, Kalliatakis G, Sreenivas M, et al. Evaluation of an AI model to assess future breast cancer risk. Radiology. 2023;307: e222679.

Lauritzen AD, Euler-Chelpin MC, Lynge E, et al. Assessing breast cancer risk by combining AI for lesion detection and mammographic texture. Radiology. 2023;308: e230227.

Yala A, Mikhael PG, Strand F, et al. Toward robust mammography-based models for breast cancer risk. Sci Transl Med. 2021;13:eaba4373.

Brentnall AR, Cuzick J. Risk models for breast cancer and their validation. Stat Sci. 2020;35:14–30.

Article   MathSciNet   PubMed   PubMed Central   Google Scholar  

Shen Y, Wu N, Phang J, et al. An interpretable classifier for high-resolution breast cancer screening images utilizing weakly supervised localization. Med Image Anal. 2021;68: 101908.

Halling-Brown MD, Warren LM, Ward D, et al. Optimam mammography image database: a large-scale resource of mammography images and clinical data. Radiol Artif Intell. 2021;3: e200103.

OPTIMAM. Publications. https://medphys.royalsurrey.nhs.uk/omidb/publications/

Yala A. Mirai. 2021. https://github.com/yala/Mirai

Yala A, Mikhael PG, Strand F, et al. Toward robust mammography-based models for breast cancer risk. Sci Transl Med. 2021;13.

Wu N, Phang J, Park J, et al. Deep neural networks improve radiologists’ performance in breast cancer screening. IEEE Trans Med Imaging. 2019;39:1184–94.

Article   PubMed   PubMed Central   Google Scholar  

Brentnall AR, Cuzick J, Field J, Duffy SW. A concordance index for matched case-control studies with applications in cancer risk. Stat Med. 2015;34:396–405.

Article   MathSciNet   PubMed   Google Scholar  

Brentnall AR, Atakpa E, Hill H, et al. An optimization framework to guide choice of thresholds for risk-based cancer screening. NPJ Digit Med. 2023;6:223.

Download references

The work of RS, JW and CD was supported by Cancer Research UK (Grant reference: C49757/A28689, awarded to GM and ARB).

Author information

Authors and affiliations.

Wolfson Institute of Population Health, Queen Mary University of London, Charterhouse square, London, EC1M 6BQ, England, UK

Ruggiero Santeramo, Celeste Damiani & Adam R. Brentnall

Warwick Manufacturing Group, University of Warwick, Coventry, CV4 7AL, England, UK

Ruggiero Santeramo & Giovanni Montana

Fondazione Istituto Italiano di Tecnologia (IIT), 16163, Genoa, Italy

Celeste Damiani

Department of Statistics, University of Warwick, Coventry, CV4 7AL, England, UK

Jiefei Wei & Giovanni Montana

You can also search for this author in PubMed   Google Scholar

Contributions

RS contributed to conceptualization, methodology, validation, investigation, formal analysis, visualization, writing—original draft; CD contributed to investigation and methodology; GM contributed to methodology, supervision, and funding acquisition; AB contributed to conceptualization, methodology, formal analysis, writing—original draft, visualization, supervision, project administration, funding acquisition; All contributed to writing—review & editing.

Corresponding authors

Correspondence to Ruggiero Santeramo , Giovanni Montana or Adam R. Brentnall .

Ethics declarations

Ethics approval and consent to participate.

The data used in this manuscript were fully anonymized and made available for research following ethical approval from the NHS Health Research Agency (REC reference: 19/SC/0284, IRAS reference: 265403).

Competing interests

AB declares income from Cancer Research UK arising from commercial use of the Tyrer-Cuzick (IBIS) breast cancer risk evaluator.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1.

. Supplementary Figures, Methods and Tables.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Santeramo, R., Damiani, C., Wei, J. et al. Are better AI algorithms for breast cancer detection also better at predicting risk? A paired case–control study. Breast Cancer Res 26 , 25 (2024). https://doi.org/10.1186/s13058-024-01775-z

Download citation

Received : 02 October 2023

Accepted : 20 January 2024

Published : 07 February 2024

DOI : https://doi.org/10.1186/s13058-024-01775-z

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Breast cancer
  • Risk assessment
  • Deep learning
  • Medical imaging

Breast Cancer Research

ISSN: 1465-542X

breast cancer case control study

  • - Google Chrome

Intended for healthcare professionals

  • Access provided by Google Indexer
  • My email alerts
  • BMA member login
  • Username * Password * Forgot your log in details? Need to activate BMA Member Log In Log in via OpenAthens Log in via your institution

Home

Search form

  • Advanced search
  • Search responses
  • Search blogs
  • Use of hormone...

Use of hormone replacement therapy and risk of breast cancer: nested case-control studies using the QResearch and CPRD databases

Linked practice.

Risk of breast cancer with HRT depends on therapy type and duration

  • Related content
  • Peer review
  • Yana Vinogradova , senior research fellow in medical statistics 1 ,
  • Carol Coupland , professor of medical statistics in primary care 1 ,
  • Julia Hippisley-Cox , professor of clinical epidemiology and general practice 2
  • 1 Division of Primary Care, University Park, University of Nottingham, Nottingham NG2 7RD, UK
  • 2 Nuffield Department of Primary Care Health Sciences, University of Oxford, Oxford, UK
  • Correspondence to: Y Vinogradova Yana.Vinogradova{at}nottingham.ac.uk
  • Accepted 17 September 2020

Objective To assess the risks of breast cancer associated with different types and durations of hormone replacement therapy (HRT).

Design Two nested case-control studies.

Setting UK general practices contributing to QResearch or Clinical Practice Research Datalink (CPRD), linked to hospital, mortality, social deprivation, and cancer registry (QResearch only) data.

Participants 98 611 women aged 50-79 with a primary diagnosis of breast cancer between 1998 and 2018, matched by age, general practice, and index date to 457 498 female controls.

Main outcome measures Breast cancer diagnosis from general practice, mortality, hospital, or cancer registry records. Odds ratios for HRT types, adjusted for personal characteristics, smoking status, alcohol consumption, comorbidities, family history, and other prescribed drugs. Separate results from QResearch or CPRD were combined.

Results Overall, 33 703 (34%) women with a diagnosis of breast cancer and 134 391 (31%) controls had used HRT prior to one year before the index date. Compared with never use, in recent users (<5 years) with long term use (≥5 years), oestrogen only therapy and combined oestrogen and progestogen therapy were both associated with increased risks of breast cancer (adjusted odds ratio 1.15 (95% confidence interval 1.09 to 1.21) and 1.79 (1.73 to 1.85), respectively). For combined progestogens, the increased risk was highest for norethisterone (1.88, 1.79 to 1.99) and lowest for dydrogesterone (1.24, 1.03 to 1.48). Past long term use of oestrogen only therapy and past short term (<5 years) use of oestrogen-progestogen were not associated with increased risk. The risk associated with past long term oestrogen-progestogen use, however, remained increased (1.16, 1.11 to 1.21). In recent oestrogen only users, between three (in younger women) and eight (in older women) extra cases per 10 000 women years would be expected, and in oestrogen-progestogen users between nine and 36 extra cases per 10 000 women years. For past oestrogen-progestogen users, the results would suggest between two and eight extra cases per 10 000 women years.

Conclusion This study has produced new generalisable estimates of the increased risks of breast cancer associated with use of different hormone replacement preparations in the UK. The levels of risks varied between types of HRT, with higher risks for combined treatments and for longer duration of use.

Introduction

Hormone replacement therapy (HRT) (also known as hormone therapy (HT) or menopausal hormonal therapy (MHT)) is prescribed to relieve the symptoms of menopause, which can be life changing. HRT is used by millions of women, sometimes over extended periods. A range of hormone combinations are available, each with different efficacy and side effects. HRT can bring several improvements to quality of life, and it can prevent osteoporosis. Concerns about adverse effects, particularly the increased risk of breast cancer associated with HRT, 1 has, however, resulted in a substantial decrease in HRT use over the past 17 years. 2 Breast cancer is the most common cancer in women, with more than 55 000 women in the UK affected each year, 3 so different drug use scenarios might result in substantial differences in the number of women who develop breast cancer, even though the risk differences between hormones might seem relatively small. Current clinical guidelines recommend use of HRT for no longer than five years and have signalled that more information is needed about the risks of breast cancer associated with different types of HRT. 4 5

Randomised trials using enrolled participants are now impractical to investigate the risks of breast cancer associated with HRT because of the numbers required and the length of follow-up. Trials would also be difficult to justify ethically, given the known harms that are associated with some types of HRT. Earlier trials have been limited, focusing on specific age groups possibly unrepresentative of women likely to request HRT or selecting specific types of HRT 6 (the largest being the Women’s Health Initiative trial), 7 or, because the design failed to distinguish between treatment types, looking only at overall effect or association. 8 9 10 Observational studies are more feasible but require access to large datasets covering lengthy time periods; so far only the Million Women Study has approached the requisite power. 10 A recent meta-analysis published after our study commenced, pooled information from 24 prospective observational studies to provide more comprehensive data on the details of exposure and breast cancer risks for the most commonly prescribed oestrogens and progestogens. 11 This meta-analysis reported that the risk of breast cancer is increased for both oestrogen only and oestrogen-progestogen current users, with, respectively, a 17% and 60% increase for 1-4 years of use and a 33% and 108% increase for 5-14 years of use. The results also showed a remaining increased risk even after discontinuation of HRT. As with many meta-analyses, however, the included studies were conducted in different settings, had different selection criteria, and had different definitions of exposure, so the data and original study designs were heterogeneous. The study provided information for the most commonly used HRT preparations, albeit with notably smaller statistical power for dydrogesterone—a progestogen previously found to be associated with a low increased risk of breast cancer. 12 At publication, the focus of publicity was the higher than expected associations with breast cancer risks than had been suggested by earlier trials. The Medicine and Healthcare products Regulatory Agency subsequently raised an HRT drug safety alert specific to breast cancer, but this has since been questioned as having caused “considerable anxiety,” particularly for women who might need HRT for reasons other than menopausal symptoms. 13

Our study focused on exposure to all the commonly prescribed types of HRT in the UK over the past 20 years in a representative primary care population. We assessed the differences in risks associated with the individual component hormones used in HRT, including dydrogesterone. Our findings are based on prospectively collected electronic health records from the two largest UK primary care databases linked to secondary care data sources. We analysed these separately and then combined the results. In contrast with data and analytical designs used in studies included in the recent meta-analysis, 11 our data were homogeneous, and the analytical approach was common. This has allowed us to gain a realistic picture of exposure in the UK to component hormones used in HRT, and the associations with increased breast cancer risk of specific treatments, providing consistently derived information for patients and doctors.

Study design

Full details for this study are available in the published protocol. 14 To summarise, we undertook a nested case-control study using the two largest UK primary care databases, QResearch and Clinical Practice Research Datalink (CPRD) GOLD, and utilised linked data from Hospital Episode Statistics (HES), Office for National Statistics (ONS) mortality data, and (QResearch only) cancer registry data. We included all general practices that had contributed data for at least three years and from these we identified two open cohorts of women aged between 50 and 79 and registered with the general practice between 1 January 1998 and 31 December 2018. We excluded women with already diagnosed breast cancer or records of mastectomy at the cohort entry date, and, to ensure completeness, any with fewer than three years of medical records.

Selection of cases and controls

Across both databases, we identified all cases between 1 January 1998 and 31 December 2018. From the QResearch database, we identified all cases of incident breast cancer using general practice, hospital admission, mortality, and cancer registry records. From CPRD, when practices were linked, we used general practice, hospital admission (up to 31 December 2017), and mortality data records (up to 13 February 2018) to identify cases, and, when not linked, general practice records only. Each case was matched to a maximum of five controls by year of birth and general practice using incidence density sampling. 15 For each case in any data source, the date of the first breast cancer record became the index date for their matched controls. QResearch and CPRD GOLD use different computer systems to collect records from practices, and as patients can be registered with only one practice, there was no overlap of cases and controls.

Exposure to HRT

We extracted prescription information for all oestrogens, progestogens, and tibolone from practice records. Symptoms indicative of developing breast cancer before diagnosis could have resulted in cessation of HRT. To minimise this source of possible protopathic bias, we excluded prescriptions issued in the year before the index date. 16

Exposure to HRT was taken as the date from when a woman received her first prescription containing systemic oestrogen (oral, subcutaneous, or transdermal) indicated to treat menopausal symptoms. If a woman received no prescription that contained a progestogen after this date, she was classified as an oestrogen only therapy user. If a woman received any prescription that contained a progestogen, she was classified as a combined therapy user. We also included topical oestrogen preparations (vaginal pessaries or cream) and tibolone, because both are commonly prescribed to menopausal women.

A large proportion of women switched between different combinations of oestrogens and progestogens, so we analysed each hormonal preparation as a separate exposure. For oestrogen only users, we distinguished between types, doses, and application method, whereas for combined therapy users we analysed combinations of any oestrogen, concentrating on progestogen type and application method. If combined therapy users had also used oestrogen only therapy, we analysed the women as oestrogen-progestogen users but adjusted the combined exposure results to account for periods of oestrogen only treatment. For all treatments, the reference category was no exposure (never users) to HRT.

At the time of our study, two types of oestrogen (conjugated equine oestrogen and estradiol) and four types of progestogen (norethisterone acetate, levonorgestrel, medroxyprogesterone, and dydrogesterone) were commonly prescribed in the UK and were included in our analyses. Of these, sufficient data were available for estradiol, estradiol-norethisterone, and oestrogen-levonorgestrel to facilitate separate analysis of application methods—oral, transdermal, or injection, and (for levonorgestrel) intrauterine. We investigated two daily dosage levels of oestrogen: low (0.625 mg/day or less for oral conjugated equine oestrogen, 1 mg/day or less for oral estradiol, and 50 mg or less for transdermal estradiol) and high (all other dosage levels). Median dosages for each oestrogen and for each woman were also calculated and analysed.

We have not specified the type of oestrogen for combinations with progestogens, but in our data conjugated equine oestrogen was by far the most commonly prescribed drug in combination with medroxyprogesterone (only 16% of prescriptions included estradiol) and levonorgestrel (only 5% included estradiol). Estradiol was the only oestrogen prescribed in combination with norethisterone and dydrogesterone.

Our data showed that HRT prescriptions were frequently issued for three months, so we assessed the durations of use by summing the lengths of prescriptions in days, including gaps of fewer than 90 days between prescriptions. Most (79%) repeated prescriptions were, however, issued within 30 days. We then categorised durations of use as never (0), less than 1 year, 1-2 years (≥1 and <3), 3-4 years (≥3 and <5), 5-9 years (≥5 and <10), and 10 years or more. Excluding prescriptions in the past year, the gap between the end of the last prescription and the index date was categorised as 1-2 years (>1 and <2), 2-4 years (≥2 and <4), 5-9 years (≥5 and <10), and 10 years or more.

Because some women discontinued HRT more than a year before the index date, and associated breast cancer risks might have diminished noticeably, we investigated two recency related exposures: recent, if the women had a prescription more than one year and less than five years before the index date (this includes current users of HRT at one year before the index date), and past, if their last prescription ended before that period (≥5 years before). Using these, we analysed different durations of exposures in relation to the recency of the last prescription.

Confounders

Analyses were all adjusted by the same factors—those that might have affected a doctor’s prescribing decision for HRT or might have affected a woman’s decision to take HRT or are associated with an increased breast cancer risk. 3 11 14 The data for confounders were derived from practice or hospital records, and data for drugs were from practice records only. To minimise protopathic bias, records of confounders had to be from at least a year before the index date. Confounders included lifestyle factors (smoking status, alcohol consumption, body mass index (BMI), and Townsend fifth as a measure of deprivation (in QResearch only)), self-assigned ethnicity (based on practice and hospital data), family history of cancers and osteoporosis, history of other cancers, records of early and late menopause, oophorectomy or hysterectomy, uptake of mammography or scanning, menopausal symptoms, comorbidities, and use, or when possible, duration of use of other drugs. Comorbidities included benign breast disease, diabetes, and bipolar disorder or schizophrenia. 14 Other drugs included combined and progestogen only contraceptive drugs, aspirin, non-steroidal anti-inflammatory drugs, tamoxifen, and raloxifene. When numbers permitted, we categorised duration of use of other drugs up to one year before the index date as never, less than 1 year, 1-2 years, 3-4 years, and 5 years or more. Early menopause was estimated from records of menopausal symptoms or of oophorectomy or hysterectomy before age 45 years. Late menopause was considered if the first menopause related record was after 55 years for women older than 55 at the index date. For all other women we assumed onset of menopause was between age 50 and 55 years.

Statistical analysis

As data from QResearch and CPRD cannot be pooled, for all analyses we processed extracted datasets in parallel as similarly as possible. To calculate associations between breast cancer risk and different exposures to HRT, we used conditional logistic regression to estimate odds ratios with 95% confidence intervals. A small proportion of women had missing values for BMI, smoking status, and alcohol consumption, which we assumed to be missing at random. We imputed these separately for each dataset using chained equations over 10 imputed datasets, where the imputation model included all listed confounders, exposures, and case-control status indicators, and we combined the odds ratios obtained from the imputed datasets using Rubin’s rule. 17

We considered duration of exposure both in the form of defined categories of exposure and as a continuous variable. For ease of comparability with other studies and to simplify interpretation, our main results are presented using defined categories of exposure, with duration of HRT expressed in years. We used a meta-analytical technique to combine the obtained odds ratios from the separate analyses run on each database. 18 A fixed effect model with inverse variance weights was used for the main analysis and a random effect model as a sensitivity analysis. In the main tables and text we only include the combined results; the separate results for QResearch and CPRD are in the supplementary tables.

To model exposures as continuous variables, we ran separate analyses on each database using fractional polynomials to explore non-linear risk associations for durations of exposure, measured in days. 19 This was done for both recent and past exposures to all the types of HRT under investigation. Variables found to have non-linear associations were then transformed into the suggested powers, the separate analyses were rerun, and the resulting coefficients and standard errors were combined.

Additional and sensitivity analyses

To assess possible age related differences in risks associated with exposures to hormone, we performed additional analyses for different age categories at the index date: 50-59 years, 60-69 years, and 70-79 years. We ran another subgroup analysis for women in three different BMI groups: less than 25, 25 up to 30, and 30 or more. In this analysis, we included only controls in the same body mass category as their matched case.

For the main analysis, we considered women to have recently used HRT if they had a prescription between one and five years before the index date. The risk associated with HRT has been found to decrease rapidly after discontinuation, 20 so we needed a measure showing excess of risk for the most recently exposed women. To assess this, we repeated the analysis, defining recent use as exposure between one and two years before the index date.

It is possible that some women were classified as never exposed only because they were not registered with the practice at the time when they had used HRT. Although any systematic difference between cases and controls is unlikely, we addressed this possible misclassification of exposure by repeating the analysis in a subgroup of women with at least 10 years of medical records. Another sensitivity analysis dealt with unknown adherence to HRT, because it is possible that some women with apparent gaps between prescriptions had in fact spread their HRT supply over longer periods. In this analysis, we defined duration of HRT use as the period between the first HRT prescription and the last one prior to one year before the index date.

The main analysis was run on women aged 50 to 79, which may include some premenopausal and perimenopausal women who have a higher risk of breast cancer. 21 To deal with this and provide comparability of our results with those of a meta-analysis, 11 we also ran an additional analysis restricting our sample to women aged 55 to 79.

To check our assumption of missing at random for some confounders, we compared patterns of missingness in exposed and non-exposed women and repeated the analyses, including only cases and controls with recorded values. In the final sensitivity analysis, we dealt with problems that might have arisen from the different levels of linkage in QResearch and CPRD. All QResearch practices were linked to deprivation, hospital, mortality, and cancer registry data, so cases could be identified using all sources of data. For CPRD, however, only 60% of practices were linked to deprivation, hospital, and mortality data, so to include data from all usable practices, we were limited to identifying cases from all available data. To assess the possible effect on our results, we ran an analysis using only fully linked CPRD practices.

To estimate the excess of breast cancer cases associated with different HRT exposures we calculated the incidence rate in the unexposed female population for different age categories (50-59, 60-69, and 70-79) using the underlying cohort from CPRD. The rate in the exposed population was derived by multiplying the baseline rate by relevant odds ratios obtained from the combined analysis.

We used Stata v16 for all analyses. A 1% level of statistical significance was used to allow for multiple comparisons. To facilitate comparison with other studies, however, we present the results as odds ratios with 95% confidence intervals.

Patient and public involvement

This epidemiological study investigated a research question recommended by a National Institute for Health and Care Excellence committee, which included lay members. 5 It used routinely collected data and appropriate statistical techniques. The grant application process and the publication process of The BMJ both had lay involvement. No other lay people were involved in setting or extending the research question or the outcome measures in our study, nor were they involved in developing plans for the design or implementation of the study. However, to better understand motivations for starting HRT and possible adherence issues related to prescribed treatment, formal and informal conversations with some women taking HRT were also organised. In these, women generally reported high levels of adherence, regardless of whether they had sought treatment themselves or were recommended it by a doctor. Some of the women involved have also agreed to help further with interpretation and dissemination of the results through women’s menopausal forums.

Overall, 59 999 cases of breast cancer were identified in QResearch between 1 January 1998 and 31 July 2018, using general practice, hospital admissions, mortality, and cancer registry records. In total, 38 612 breast cancer cases were identified in CPRD between 1 January 1998 and 31 December 2018, using general practice records and, for linked practices, also using hospital admission records (until 31 December 2017) and mortality records (until 13 February 2018) ( fig 1 ).

Fig 1

Flow chart of included cases and controls

  • Download figure
  • Open in new tab
  • Download powerpoint

Table 1 shows the characteristics of the cases and matched controls from QResearch and CPRD. Cases were more likely than controls to be overweight or obese (53% v 50%), to be former smokers (29% v 27%), to have a record of benign breast disease (9% v 6%) or other cancers (3.1% v 2.6%), or to have a family history of breast cancer (4% v 2.5%).

Characteristics of women with breast cancer and matched controls one year before index date by database (QResearch and CPRD). Values are percentages (numbers) of participants unless stated otherwise

  • View inline

Across both databases, 33 703 (34%) cases and 142 391 (31%) controls had ever been exposed to HRT. Of those, 8860 (26%) cases and 42 799 (30%) controls had been exposed to oestrogen only therapy and 24 843 (74%) cases and 99 592 (70%) controls had been exposed to oestrogen-progestogen therapy (supplementary eTable 1). Women in the 60-69 age category were relatively more exposed to oestrogen only (47% in cases and controls) and oestrogen-progestogen therapies (48% in cases and controls). A high proportion of women using oestrogen only therapy had undergone oophorectomy or hysterectomy (89% cases and 90% controls), but, overall, users of oestrogen only and oestrogen-progestogen therapies had characteristics broadly similar to those of never users for most confounders. Some women switched between hormones during HRT exposure. About 20% of oestrogen only users had exposure to both oestrogens. About 57% of combined therapy users had only one oestrogen-progestogen combination recorded. About 0.6% of cases and controls started HRT in the year before the index date, but these are considered as never users in the analyses.

Overall exposure

Overall (or ever) exposure to HRT was associated with an increased risk of breast cancer (adjusted odds ratio 1.21, 95% confidence interval 1.19 to 1.23). The increased risk was mostly attributable to oestrogen-progestogen therapy (1.26, 1.24 to 1.29), with oestrogen only therapy showing a small increased risk (1.06, 1.03 to 1.10), both compared with never users (supplementary eTable 2). No increased risk was associated with oestrogen cream or vaginal preparations (supplementary eTable 3). The risks associated with HRT increased with duration of use, but the associations were less strong for oestrogen only therapy and for tibolone than for oestrogen-progestogen therapy, apart from estradiol-dydrogesterone preparations. Norethisterone, levonorgestrel, and medroxyprogesterone were associated with similar risks, increasing across all duration categories longer than one year. For all exposure durations, the combined treatment with the lowest associated risk increase was estradiol-dydrogesterone. No differences were found between low and high doses of oestrogens or between different application methods for estradiol, norethisterone, or levonorgestrel (supplementary eTable 4).

Associations between use of HRT and risk of breast cancer rapidly decreased with increasing years of discontinuation (supplementary eFigure 1 and eTable 5). For oestrogen only, estradiol combined with norethisterone and dydrogesterone, and tibolone, no significantly increased risk was found from two years after discontinuation. For medroxyprogesterone, the risk was reduced after two years but remained raised until after five years; for levonorgestrel until after 10 years.

Duration of recent and past exposures as categorical variables

Recent users of HRT (ie, those with prescriptions more than one year and less than five years before the index date) comprised 56% (18 879) of cases and 50% (70 931) of controls ever exposed to HRT. Figure 2 and figure 3 (supplementary eTable 6) show the associations between categorised durations of HRT and risks of breast cancer in women with recent and past exposures. The patterns of risks for recently exposed women were similar to those for overall exposures, but the risks were consistently higher and more pronounced, particularly for oestrogen-progestogen therapy. For women with past exposures, risks associated with longer durations of use of oestrogen-progestogen, particularly longer use of levonorgestrel (>3 years) and norethisterone (>5 years), remained high, but for other hormones the risks were not statistically significant. Findings for recent exposures to different doses and applications also had similar patterns to the overall exposure analysis, but with higher odds ratios (supplementary eTable 7).

Fig 2

Recent and past use of oestrogen, oestrogen-progestogen, and tibolone in association with breast cancer risk. Odds ratios are with reference to never users and adjusted for smoking status, alcohol consumption, Townsend fifth (QResearch only), body mass index, ethnicity, history of other cancers, oophorectomy or hysterectomy, records of early and late menopause, menopausal symptoms, mammography or scans, family history, comorbidities, other drugs, and years of data. Cases are matched to controls by age, general practice, and index date

Fig 3

Recent and past use of different hormones in association with breast cancer risk. Odds ratios are with reference to never users and adjusted for smoking status, alcohol consumption, Townsend fifth (QResearch only), body mass index, ethnicity, history of other cancers, oophorectomy or hysterectomy, records of early and late menopause, menopausal symptoms, mammography or scans, family history, comorbidities, other drugs, and years of data. Cases are matched to controls by age, general practice, and index date

A further restriction to recency (defined now as one prescription or more in the period 1-2 years before the index date) resulted in fewer women in each category of exposure, but the increased risks associated with longer exposures were even more pronounced (supplementary eFigure 2). In HRT users, 40% (13 463) of cases and 32% (44 972) of controls ever exposed to HRT had one or more prescriptions in the period 1-2 years before the index date. The patterns of risks for these recently exposed women were similar to those of overall exposures, but the risks were consistently higher and more pronounced for progestogens. For women with a last exposure more than two years before the index date, risks associated with long exposures to levonorgestrel (>3 years) remained high, but for other hormones the risks were not statistically significant (supplementary eFigure 3 and eTable 8).

Duration of recent and past exposures as continuous variables

Figure 4 (supplementary eTable 6) shows the associations between duration of different types of HRT and risks of breast cancer for recent (1-5 years before index date) and past users (prescriptions ≥5 years previously). A linear relation was found between duration of exposure as a continuous variable for most types of HRT, with risk increasing uniformly over time. However, for recent exposure to oestrogen-progestogen or to estradiol-norethisterone, and for past exposure to oestrogen-medroxyprogesterone, square root transformations gave the best fit for an association with breast cancer risk, showing that risk for these treatments increased faster earlier in the exposure. Additions of further fractional polynomial terms were not statistically significant.

Fig 4

Adjusted odds ratios for different durations of recent and past exposures to hormone replacement therapies in association with breast cancer risk. Odds ratios are with reference to never users and adjusted for smoking status, alcohol consumption, Townsend fifth (QResearch only), body mass index, ethnicity, history of other cancers, oophorectomy or hysterectomy, records of early and late menopause, menopausal symptoms, mammography or scans, family history, comorbidities, other drugs, and years of data. Cases are matched to controls by age, general practice, and index date. Model includes fractional polynomial terms for recent use of oestrogen-progestogen (power 0.5), estradiol-norethisterone (power 0.5), past use of oestrogen-levonorgestrel (power 0.5), and linear terms (1) for all other exposures

Risk increases for recent users were more pronounced than for past users, and different types of HRT showed different patterns of increase as the durations of exposure increased. Oestrogen-medroxyprogesterone and oestrogen-levonorgestrel formulations showed the greatest increases with duration. Oestrogen only (including separately conjugated equine oestrogen and estradiol), tibolone, and estradiol-dydrogesterone formulations showed the smallest increases with duration.

Subgroup analyses

The subgroup analyses for different age categories showed similar patterns in magnitudes of risk for recent and past exposures ( fig 5 , supplementary eFigure 4 and eTables 9 and 10). The oldest age group (70-79) had a smaller number of recent (1-5 years before the index date) users and, although odds ratios appeared to be higher than for the younger age groups, the confidence intervals were too wide to reach statistical significance for oestrogen only users. The younger age group (50-59) had the lowest odds ratios, which could reflect shorter durations of exposure, particularly in the category of five years or more (supplementary eTables 9 and 10). The mean duration for the category of 1-4 years, however, was only slightly lower for the younger group but similar between the older groups.

Fig 5

Use of oestrogen only, oestrogen-progestogen, and tibolone in women of different ages in association with breast cancer risk. Odds ratios are with reference to never users and adjusted for smoking status, alcohol consumption, Townsend fifth (QResearch only), body mass index, ethnicity, history of other cancers, oophorectomy or hysterectomy, records of early and late menopause, menopausal symptoms, mammography or scans, family history, comorbidities, other drugs, and years of data. Cases are matched to controls by age, general practice, and index date

Figure 6 presents the associations with breast cancer risk for recent and past exposures in different BMI categories (supplementary eFigure5 and eTables 11 and 12). Overall, the pattern of risks in the subgroups were similar to those of the main analyses. For women with a higher BMI (>30), however, the risks associated with HRT for recent users appeared slightly lower than in women with a lower BMI, both for oestrogen only and for oestrogen-progestogen therapies. For oestrogen only therapy and more than five years of use, the association with risk of breast cancer was statistically significant only in the lowest BMI group (1.24, 1.11 to 1.35) compared with never use. For oestrogen-progestogen, more than five years of use was associated with the highest adjusted odds ratio in the lowest BMI group and the lowest adjusted odds ratio in the highest BMI group (1.93, 1.80 to 2.05 for BMI <25; 1.71, 1.58 to 1.85 for BMI 25-30; and 1.38, 1.23 to 1.55 for BMI >30). For past use, no difference between BMI groups was observed.

Fig 6

Use of oestrogen-only, oestrogen-progestogen, and tibolone in women of different body mass index in association with breast cancer risk. Odds ratios are with reference to never users and adjusted for smoking status, alcohol consumption, Townsend fifth (QResearch only), body mass index, ethnicity, history of other cancers, oophorectomy or hysterectomy, records of early and late menopause, menopausal symptoms, mammography or scans, family history, comorbidities, other drugs, and years of data. Cases are matched to controls by age, general practice, and index date

Excess numbers in HRT users

The crude incidence rate of breast cancer in the underlying CPRD cohort was 33.0 (95% confidence interval 32.7 to 33.3) per 10 000 women years, whereas the crude incidence rate in women not exposed to HRT was 31.5 (31.1 to 31.7) per 10 000 women years. The rate for unexposed women varied with age, with the lowest rate in younger women (28.2, 27.6 to 28.7 in women aged 50-59; 34.1, 33.4 to 34.8 in women aged 60-69; and 33.3, 32.6 to 34.0 in women aged 70-79). The highest rate in the 60-69 years group was consistent with national data from cancer registration statistics in England. 22

Table 2 and fig 7 contain incidence rates and excess rates of breast cancer in users of HRT at different ages and for different durations. The number of extra cases is consistently larger for older women for all exposures. Compared with never users, the estimated number of excess cases per 10 000 women years in recent long term (≥5 years) users of oestrogen only treatment was three in women aged 50-59, four in women aged 60 to 69, and eight in women aged 70-79. Compared with never users, the number of excess cases per 10 000 women years in recent long term users of oestrogen-progestogen treatment was 15 in women aged 50-59, 26 in women aged 60-69, and 36 in women aged 70-79. For tibolone in recent long term users, the numbers exposed in younger women were too small to provide sufficient data, but within the older groups there are an estimated nine extra cases per 10 000 women years in women aged 60 to 69 and 15 extra cases per 10 000 women years in women aged 70 to 79.

Incidence rates and excess of cases of breast cancer compared with never use per 10 000 women years by different age categories and different durations and recency of hormone replacement therapy (HRT) use

Fig 7

Incident breast cancer rate per 10 000 women years for women unexposed and exposed for different durations to different hormone replacement therapies by age range. Rates were estimated using rates in unexposed populations multiplied by adjusted odds ratios derived from subgroup analyses for different age categories (see fig 5 )

Sensitivity analyses

The sensitivity analysis run on women with at least 10 years of recorded data showed similar patterns of risks associated with different durations of HRT use, but the risks appeared slightly higher, particularly for exposure to oestrogen-progestogen combinations of between five and 10 years (supplementary eFigure 6 and eTable 13). The sensitivity analysis on the subgroup of women aged 55 to 79 showed similar patterns of risks, with all values consistent with the subgroup analyses for different age groups (supplementary eFigure 6 and eTable 14). The sensitivity analysis with duration of exposure defined as from the first prescription of HRT to the end of the last prescription showed results similar to those of the main analysis (supplementary eTable 15). The results of analyses run on cases and controls without missing data for smoking status, alcohol consumption, and BMI were similar to those of the main analyses—as were the analysis restricted to CPRD cases and controls with linked data.

This large observational study found that exposure to most HRT drugs is associated with an increased risk of breast cancer. In comparison with a recent meta-analysis, however, our findings generally suggest lower increased risk associations between longer term HRT use and breast cancer, and we report a more noticeable decline in risks once HRT has stopped. Risk increases were mostly associated with oestrogen-progestogen treatments, but small increases were also associated with oestrogen only treatments. For all exposure durations, the combined treatment with the lowest associated risk increase was estradiol-dydrogesterone. Associations for all treatments depended on duration, with no increased risks for less than one year of treatment but increasing risks for longer exposures to medroxyprogesterone, norethisterone, and levonorgestrel. Associations were more pronounced for older women and less noticeable for obese women.

Strengths and weaknesses of this study

The main strengths of this original study are its size, consistent sources of primary care data, almost complete follow-up of diagnoses using linked data, consistent design, and resulting generalisability of the findings. Combining results from the two largest UK primary care research databases with national coverage has provided increased power, representiveness, and wide geographical coverage of included practices. The study relates to an important health problem and used only objective information on prescriptions for HRT in the UK, facilitating inclusion of the full range of preparations available within this national setting and presenting in detail the increased risk of breast cancer associated with usage patterns. The follow-up and validation of breast cancer diagnoses through linkages to hospital, mortality, and cancer registry (QResearch only) data reduced both ascertainment and recording bias. As our study was based on routinely collected data, it was also not susceptible to recall bias. Matching by general practice enabled consideration of possible differences in prescribing and recording patterns across practices. Differences in menopause onset and age at start of HRT were partially dealt with through matching by age. The results were also adjusted for information on lifestyle, comorbidities, and use of other drugs, and the study presents subgroup analyses for women in different age groups and in different BMI categories. Protopathic bias, from diagnostic and prescribing problems created by symptoms common to early breast cancer and onset of menopause, was minimised by excluding prescriptions issued in the year before the index date.

Some limitations of this study arise from inevitable shortfalls in completeness and accuracy within any routinely collected dataset. A small proportion of women had missing information on smoking status, alcohol consumption, and BMI, but these were dealt with by multiple imputation. As we did not have reliable data for age at onset of menopause for all women, we estimated onset from the first menopause specific record before the earliest HRT prescription. For women with no such record we assumed onset within the most common age range of 50 to 54 years. We did not investigate the differences between continuous and sequential HRT because these regimens are prescribed at different times after menopause. As our cases and controls were matched by age, they would likely have been prescribed similar regimens, making a comparison infeasible. Our primary focus, anyway, was recent long term exposure.

No reliable data were available for established risk factors for breast cancer, such as parity or time of the first pregnancy, but there is no evidence to show that these are related to HRT use. No data for physical activity were available, but a possible risk reduction for active women has been shown not to be influenced by menopausal status. 23 Some women might have joined their current practice after the onset of menopause, so records of past treatments might not have been available. The results from the subgroup restricted to women with at least 10 years of data, however, showed a similar pattern of risk associations to that of the main analysis. Use of HRT might have the side effect of increased breast density, possibly masking cancers and leading to diagnostic delays, 24 which could shift odds ratios for short duration towards unity. Also, although there was no information about adherence to HRT, any systematic differences between cases and controls seems unlikely because information was recorded prospectively before diagnosis. Conversations with lay women involved in this research also revealed a high adherence to HRT.

Strengths and weaknesses in relation to other studies

Our study used a nested case-control design, so it did not follow women prospectively from the start of HRT or assess average lifetime risks. Rather, it looked back at already recorded exposures to HRT for women with a diagnosis of breast cancer and matched controls in the age range 50 to 79 and produced comparisons of risks averaged across all time points at which diagnoses in the datasets occurred. The study is based on data derived from real world treatment settings, when women might not have had a constant supply of a preparation and might have needed to switch drugs during the study period. Including all exposures prescribed over time allowed us to present information for a wide range of common types of HRT.

Most trials produced results for a more restricted number of treatments. A meta-analysis of existing trials, 7 taken largely from the Women’s Health Initiative study, provided estimates only for the specific treatments of conjugated equine oestrogen with and without medroxyprogesterone. 6 In contrast to our estimates of a slightly increased risk for long term users of conjugated equine oestrogen (average duration for recent exposure of 5.6 years, odds ratio 1.07, 95% confidence interval 1.01 to 1.12), the meta-analysis found no difference in risk of breast cancer (relative risk 0.79, 95% confidence interval 0.61 to 1.02) after a mean duration of 7.2 years. The observed relative risk for the combined conjugated equine oestrogen with medroxyprogesterone therapy after a mean duration of 5.6 years (1.27, 1.03 to 1.56) was similar to our findings for recent exposure, with an average duration of 3.7 years (odds ratio 1.35, 1.30 to 1.41).

Our estimates were consistent with previous observational studies. 9 25 26 27 The Million Women Study 28 29 showed slightly higher risks than our study: for recent oestrogen only users a relative risk of 1.30 (95% confidence interval 1.21 to 1.40) compared with our odds ratio of 1.12 (1.08 to 1.16), and for recent oestrogen-progestogen users a relative risk of 2.00 (1.88 to 2.12) compared with our odds ratio of 1.51 (1.47 to 1.54). However, the Million Women Study only covered a selected population of women who had undergone mammography, and the initial study used just a single baseline questionnaire to collect information. 28 Taken together, the relatively high proportion of HRT users in the initial study (55% were ever users and 35% were current users) and the less than 65% response rate at three years of follow-up, which would be expected also to be skewed towards HRT users, would suggest that women who used HRT were more likely to have participated. 29

In general, some inconsistency was found between the proportions of women exposed to HRT in data used for our study and those used in the 2019 meta-analysis. 11 The predominant (40%) data source for the meta-analysis was from the Million Women Study, where, 51% of cases had ever been exposed to HRT and 18% of cases were current users (<5 years). The second largest data source, comprising 28% of the data used in the meta-analysis, was routinely collected CPRD data (one of the two data sources in our study), and here 40% of women with breast cancer had been exposed to HRT and 12% were current users. Both these exposure rates contrast with those in our study, which overall had 34% of cases ever exposed and 19% of cases with prescriptions within 1-5 years before diagnosis.

We cannot speculate on reasons for these differences in the CPRD data used because we do not have access to relevant information for the 2019 meta-analysis. That sample contained slightly older cases (mean age at diagnosis 66 v 63 in our sample) but no age range was reported. The estimations of risk for overall use of HRT in our CPRD analysis (odds ratio 1.21, 95% confidence interval 1.18, 1.25) were, however, similar to the CPRD specific estimations reported in the meta-analysis (relative risk 1.25, 95% confidence interval 1.20, 1.30). For recent use (prescriptions 1-2 years before the index date), when the proportions of HRT users in CPRD data used in our study and in the meta-analysis were closest, the estimates of risk were also similar. In our analysis of CPRD data on recent use, the odds ratios were 1.25 (1.17 to 1.34) for oestrogen only and 1.91 (1.83 to 1.99) for oestrogen-progestogen, whereas for the CPRD data used in the meta-analysis the corresponding findings were 1.36 (1.25 to 1.4) and 2.16 (2.02 to 2.31).

Comparative assessment of the findings from the 2019 meta-analysis is in general complicated by the heterogeneity of included studies and data sources. The meta-analysis included data from 24 differently designed prospective studies from around the world. Differences between findings from our large, consistently designed study and those from the meta-analysis might be related to the different periods covered by included studies or several problems relating to the different data sources. Some studies used routinely collected data with different definitions of exposure, 9 30 some used questionnaires with a single baseline assessment of exposure, 27 28 31 and others used repeated biennial questionnaires. 25 32 Some participants were recruited from different countries with ever exposure levels varying from 19% to 69%, 8 and some studies were from different profession related populations. 8 25

Overall, our results were broadly in line with those of the meta-analysis 11 but with slightly lower risks for long term exposures. This might partly be explained by almost half of the cases in the meta-analysis coming from the Million Women Study. For current use, however, the meta-analysis reported similar associations with risk of breast cancer, regardless of whether such use was restricted to HRT exposure within the past five years or within the past two years. By contrast, we found associations to be more pronounced for users with a prescription recency of 1-2 years before the index date, with higher odds ratios than for an exposure recency of 1-5 years. Our results with a recency definition of 1-2 years were broadly similar to those of the meta-analysis for either of their definitions of current use, whereas our findings for 1-5 years recency were lower. The difference in risk found by us seems to be more in line with previous expectations of declines in risk after cessation of HRT. 7

Our findings for oestrogen only users with recent (1-5 years) use of more than five years (odds ratio 1.15, 95% confidence interval 1.09 to 1.21) were lower than those from the meta-analysis (relative risk 1.33, 95% confidence interval 1.28 to 1.38). 11 Our study also found a marginally higher risk associated with estradiol than with conjugated equine oestrogen. For oestrogen-progestogen therapy, our finding for recent use of more than five years duration was also lower (1.79, 1.73 to 1.85) than the meta-analysis estimate (2.08, 2.02 to 2.15). Despite the similar average duration of exposures between our study and those in the meta-analysis, our findings for the different types of the most common progestogens and tibolone were consistently lower than those of the meta-analysis (supplementary eTable 16).

For dydrogesterone, our study found lower risks associated with more than five years of exposure than in the meta-analysis (1.24, 1.03 to 1.48 v 1.41, 1.17 to 1.71). 11 The risk from dydrogesterone was much lower than for any other progestogen, but one of our sensitivity analyses did show a statistically significant increased risk for a small subgroup of dydrogesterone users—those with a prescription 1-2 years before the index date and more than five years of use (112 cases, odds ratio 1.47, 1.19 to 1.83).

Our study showed differential risks associated with HRT use by age category. For recent exposures of more than five years’ duration, associated risks of breast cancer rose with increasing age category. This might partly be explained by generally longer usages in older age categories, although exposure of 1-4 years was similarly associated with increasing risk from younger to older age groups of women. Our findings for the 70-79 age group for oestrogen only use (1.25, 1.11 to 1.39) and for oestrogen-progestogen use (2.20, 2.02 to 2.39) are in line with findings from the meta-analysis 11 for women who started HRT at the age of 55-59 and continued treatment for 5-14 years : oestrogen only use of 1.26 (1.12 to 1.41) and oestrogen-progestogen use of 1.97 (1.81 to 2.15).

The adiposity of included women differed between our study and previous studies. Mean BMI in our study (27.7 in cases) was higher than in other observational studies (average 25) 33 but slightly lower than in the Women’s Health Initiative trial (28.5). 34 This could help to explain overall differences in associations between our findings and those of other studies, although the mean BMI in our study reflects the distribution within women with breast cancer diagnosed in the general UK population over the study period. Our findings for women matched by age and category of BMI are detailed and comprehensive estimations of duration dependent associations for HRT exposure and breast cancer risk. They are broadly similar to those from previous studies and the 2019 meta-analysis, 11 33 with the lowest associations between HRT use and risk of breast cancer in women in the highest BMI category. These concur with findings from the Million Women Study (which had relatively small numbers) and an earlier meta-analysis. 29 35 Some complex biological relation might exist between fat tissue and HRT, 36 although it might also be related to differences in timeliness of diagnoses between women with different body weights.

Implications for clinicians and policymakers

This study delivers more generalisable estimates of the different risks of breast cancer associated with specific progestogen components of HRT, while confirming no increased risks from short term use of oestrogen only, estradiol-dydrogesterone, and tibolone. Increasing duration of use was generally associated with increased risk, with tibolone and estradiol-dydrogesterone showing the smallest risks. The frequency of prescribing for treatments including dydrogesterone was, however, much lower than for those including norethisterone, medroxyprogesterone, or levonorgestrel.

Unanswered questions and future research

In our study protocol we did not prespecify analyses relating to cancer stage or tumour type because these lay outside the main question of interest. Although information on risk related to individual progestogens could be improved, previous studies have shown that the associated risks between HRT and tumour types might differ, with higher risks of developing oestrogen receptor positive tumours and lobular tumours. 11 Knowing the cancer stage could also address the question of risk differences between women of various body weights, to clarify whether systematic differences might exist in diagnostic delay. Other unknowns include questions about breast cancer survival rates and all cause mortality in women using HRT. 13

This large observational study of HRT and breast cancer risk based on two large primary care databases analysed in an identical manner has confirmed the excess risk to be attributable mostly to combined treatments, with the lowest risks associated with use of the least commonly prescribed dydrogesterone. Rarely prescribed tibolone also showed low increased risks.

Our findings of generally lower increased risks for combined HRT treatments and of more pronounced declines in risk once HRT has stopped, provide some counterbalance to the higher than expected risks reported in a recently published meta-analysis. 11 Our results add more evidence to the existing knowledge base and should help doctors and women to identify the most appropriate HRT formulation and treatment regimen, and provide more consistently derived information for women’s health experts, healthcare researchers, and treatment policy professionals.

What is already known on this topic

Long term systemic use of hormone replacement therapy (HRT) is associated with increased risks of breast cancer, mostly attributable to the progestogens medroxyprogesterone, norethisterone, and levonorgestrel

After discontinuation of treatment, the increased risks decline, but remain raised for some years

A recent large meta-analysis has reported higher than expected breast cancer risks associated with HRT

What this study adds

The study confirmed increased risks of breast cancer associated with long term use of oestrogen only therapy and combined oestrogen and progestogen therapy

The combined treatment associated with the lowest risk increase was estradiol-dydrogesterone

The findings suggest lower increased risks of breast cancer associated with longer term HRT use, and a more noticeable decline in risks once treatment is stopped compared with the meta-analysis

Acknowledgments

We acknowledge the contribution of EMIS practices who contribute to the QResearch database and EMIS and the University of Nottingham for expertise in establishing, developing, and supporting the QResearch database and the Chancellor masters and schools of the University of Oxford for continuing to develop and support the QResearch database. The Hospital Episode Statistics data used in this analysis are re-used by permission from the NHS Digital who retain the copyright. We thank the Office for National Statistics (ONS) for providing the mortality data. ONS and NHS Digital bear no responsibility for the analysis or interpretation of the data. This project involves data derived from patient level information collected by the NHS, as part of the care and support of patients with cancer. The data are collated, maintained, and quality assured by the National Cancer Registration and Analysis Service, which is part of Public Health England (PHE). Access to the data was facilitated by the PHE Office for Data Release. QResearch acknowledges funding from the National Institute for Health Research funded Nottingham Biomedical Research Centre until 31 January 2019 and the CRUK Cancer Centre and Wellcome Trust from 1 October 2019. Lauren Taylor (Division of Primary Care University of Nottingham) contributed clinical advice, in particular on the pharmacology of treatments and decisions made in the prescribing process, at the stage of interpreting the results and we should like to acknowledge these inputs with gratitude.

Contributors: YV contributed to the study protocol, reviewed the literature, designed the study, organised the extraction of CPRD data, did the analysis on both datasets and wrote the draft of the manuscript. JHC initiated the study, undertook the original literature review, drafted the study protocol, organised the extraction of the QResearch data, advised on the design and clinical aspects of the study and interpretation of the results and drafting of the paper. CC contributed to the development of the idea and the study design and advised on the analysis and interpretation of the results. JHC and CC critically reviewed the paper. YV is the guarantor of the study. All authors have approved the submitted version. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Funding: This work is partially funded by the National Institute for Health Research (NIHR) School for Primary Care Research (project reference 848619) and by Cancer Research UK (grant No C5255/A18085) through the cancer research UK Oxford Centre. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.

Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare support from the National Institute for Health Research School for Primary Care Research and by Cancer Research UK through the cancer research UK Oxford Centre; JHC is professor of clinical epidemiology at the University of Oxford and unpaid director of QResearch, a not-for-profit organisation which is a joint partnership between the University of Oxford and EMIS (commercial IT supplier for 60% of general practices in the UK). JHC was a paid director of ClinRisk until 2019, which produces open and closed source software to ensure the reliable and updatable implementation of clinical risk algorithms within clinical computer systems to help improve patient care; no other relationships or activities that could appear to have influenced the submitted work.

Ethical approval: The protocol for QResearch has been published in ePrints and was reviewed in accordance with the requirements for the East Midlands Derby Research Ethic Committee (ref 03/4/021). The protocol for CPRD has been approved by the Independent Scientific Advisory Committee for MHRA Database Research (N 16_282).

Data sharing: To guarantee the confidentiality of personal and health information only the authors have had access to the data during the study in accordance with the relevant licence agreements. Access to the QResearch data are according to the information on the QResearch website ( www.qresearch.org ). CPRD linked data were provided under a licence that does not permit sharing.

The lead author affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted and that any discrepancies from the study as planned have been explained.

Dissemination to participants and related patient and public communities: The results of the study will be sent to University of Nottingham press office, to related patients, and to the funders. A lay summary will be created and published at SPCR NIHR and QResearch website.

Provenance and peer review: Not commissioned; externally peer reviewed.

This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/ .

  • Marjoribanks J ,
  • Farquhar C ,
  • Roberts H ,
  • Lethaby A ,
  • ↵ Cancer Research UK. Risk factors for Breast Cancer. 2015. https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/breast-cancer/risk-factors .
  • ↵ National Collaborating Centre for Women’s and Children’s Health. Menopause. Full guideline. National Institute for Health and Care Excellence. 2015 (Version 1.5)
  • ↵ National Institute for Health and Care Excellence. Menopause: diagnosis and management. NICE guideline NG23. 2015.
  • Gartlehner G ,
  • Feltner C ,
  • Chlebowski RT ,
  • Manson JE ,
  • Fournier A ,
  • Graff-Iversen S ,
  • Weiderpass E ,
  • Kolonel L ,
  • Wilkens L ,
  • Henderson B ,
  • Collaborative Group on Hormonal Factors in Breast Cancer
  • Berrino F ,
  • Clavel-Chapelon F
  • ↵ Vinogradova Y, Coupland C, Hippisley-Cox J. Protocol to assess risk of breast cancer associated with use of hormone replacement therapy in real world settings: two nested case-control studies in primary care. 2019. https://nottingham-repository.worktribe.com/output/2309731 .
  • Delgado-Rodríguez M ,
  • Borenstein M ,
  • Hedges LV ,
  • Higgins JP ,
  • Rothstein HR
  • Royston P ,
  • Sauerbrei W
  • Stefanick ML ,
  • ↵ Office for National Statistics. Cancer registration statistics, England: 2017. 2019 www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/cancerregistrationstatisticsengland/2017 .
  • Colditz GA ,
  • Hankinson SE ,
  • Hunter DJ ,
  • Prentice RL ,
  • Henderson KD ,
  • Million Women Study Collaborators
  • Opatrny L ,
  • Dell’Aniello S ,
  • Assouline S ,
  • Alsaker E ,
  • Rossouw JE ,
  • Anderson GL ,
  • Writing Group for the Women’s Health Initiative Investigators
  • Feigelson HS ,

breast cancer case control study

ORIGINAL RESEARCH article

Machine learning algorithms to uncover risk factors of breast cancer: insights from a large case-control study.

Mostafa Dianati-Nasab,&#x;

  • 1 School of Medical and Life Sciences, Sunway University, Sunway City, Malaysia
  • 2 Department of Epidemiology, School of Public Health, Shiraz University of Medical Sciences, Shiraz, Iran
  • 3 Computational Intelligence & Intelligent Optimization Research Group, Business & Economics School, Persian Gulf University, Bushehr, Iran
  • 4 Department of Operation Management, Amsterdam Business School, University of Amsterdam, Amsterdam, Netherlands
  • 5 Department of Medicine, Iran University of Medical Sciences, Tehran, Iran
  • 6 Computer Science and Engineering, University of Westminster, London, United Kingdom

Introduction: This large case-control study explored the application of machine learning models to identify risk factors for primary invasive incident breast cancer (BC) in the Iranian population. This study serves as a bridge toward improved BC prevention, early detection, and management through the identification of modifiable and unmodifiable risk factors.

Methods: The dataset includes 1,009 cases and 1,009 controls, with comprehensive data on lifestyle, health-behavior, reproductive and sociodemographic factors. Different machine learning models, namely Random Forest (RF), Neural Networks (NN), Bootstrap Aggregating Classification and Regression Trees (Bagged CART), and Extreme Gradient Boosting Tree (XGBoost), were employed to analyze the data.

Results: The findings highlight the significance of a chest X-ray history, deliberate weight loss, abortion history, and post-menopausal status as predictors. Factors such as second-hand smoking, lower education, menarche age (>14), occupation (employed), first delivery age (18-23), and breastfeeding duration (>42 months) were also identified as important predictors in multiple models. The RF model exhibited the highest Area Under the Curve (AUC) value of 0.9, as indicated by the Receiver Operating Characteristic (ROC) curve. Following closely was the Bagged CART model with an AUC of 0.89, while the XGBoost model achieved a slightly lower AUC of 0.78. In contrast, the NN model demonstrated the lowest AUC of 0.74. On the other hand, the RF model achieved an accuracy of 83.9% and a Kappa coefficient of 67.8% and the XGBoost, achieved a lower accuracy of 82.5% and a lower Kappa coefficient of 0.6.

Conclusion: This study could be beneficial for targeted preventive measures according to the main risk factors for BC among high-risk women.

1 Introduction

Breast cancer (BC) stands as the foremost cause of cancer-related deaths among females and remains a significant global health concern, with over 2.3 million new cases and 685,000 deaths solely in 2020 ( 1 , 2 ). BC is anticipated to experience a considerable increase in cases by 2030, driven by significant lifestyle changes, as forecasted by the World Health Organization (WHO) ( 3 ).

BC is a prevalent cancer among Iranian women, accounting for nearly a third of all cancer occurrences ( 4 ) and it has been on an upward trajectory in recent years, reaching an age-standardized rate of prevalence of 47.1 per 100,000 Iranian women in 2018 ( 5 ). BC in Iranian women typically manifests at an earlier age and follows a more aggressive clinical course in comparison to Western populations ( 6 ). The concerning attributes of BC in Iran underscore the necessity for specific prevention and treatment strategies that take the population’s lifestyle and demographic characteristics into consideration.

While extensive research has been conducted to identify BC risk factors and preventive measures ( 7 – 9 ) the complex and multifactorial nature of BC necessitates innovative approaches for a comprehensive analysis. In recent years, advancements in machine learning (ML) techniques have showcased a promising future across various medical fields, including cancer research.

ML is known as a branch of artificial intelligence (AI) that relies upon a diverse set of statistical, optimization, and probabilistic techniques, facilitating computers in gathering insights from previous examples and identifying subtle patterns in complex datasets ( 10 ). These techniques have demonstrated high potential in identifying relevant factors and crafting personalized prevention strategies for different types of cancer ( 11 ). Consequently, ML models possess the capability to harness extensive datasets, extract invaluable insights from intricate patterns, and facilitate the identification of risk factors that may have been disregarded through traditional statistical methodologies ( 12 , 13 ).

The application of ML models in cancer research has shown promising results in improving risk prediction, prognosis estimation, and treatment selection ( 14 ). These models can integrate diverse sets of data, including clinical, genetic, lifestyle, and environmental factors, to generate accurate risk profiles for individuals. By employing the power of ML algorithms, researchers can analyze complex interactions among various risk factors and identify high-risk individuals who can benefit the most from tailored preventive interventions ( 15 , 16 ).

While ML models have been successfully employed in BC research globally ( 14 ), their application in the context of Iranian population-specific risk factors and preventive measures remains limited. Like numerous other nations, Iran exhibits distinct patterns concerning the incidence rate among younger age groups, the clinical attributes of this health issue, and the social and cultural surroundings of individuals dealing with BC ( 17 ). Therefore, exploring the application of ML models to identify risk factors and preventive measures specific to the Iranian population can provide valuable insights for tailored interventions and resource allocation in terms of disease control and prevention.

By integrating data from diverse sources and leveraging advanced ML algorithms, the complex etiology of BC is further understood through the advancements made by this research. The primary objective of this study was to investigate the potential of different ML models in identifying risk factors associated with primary invasive BC to develop more personalized preventive measures within the Iranian population. We aimed at training ML models that can accurately predict BC risk factors among women.

2 Materials and methods

The present study introduces a comprehensive framework comprising three distinct steps, as illustrated in Figure 1 . The initial phase delineates the construction of the database and elucidates a series of meticulous operations performed to preprocess the data, ensuring its suitability for subsequent modeling endeavors. These operations encompass the integration of disparate datasets, careful data cleansing, handling of missing values, and variable selection processes for the subsequent application of ML models.

www.frontiersin.org

Figure 1 Architecture of the framework of this study. All data analysis and modeling procedures were conducted utilizing R4.2.1 programming.

The output of this preliminary phase serves as a pivotal input for the subsequent two tiers of analysis. The second step involves the application of four diverse ML algorithms aimed at generating accurate predictions, which are subsequently evaluated using a range of statistical indices and receiver operating characteristic (ROC) curves. An ensemble approach is employed in the third and final step, where both a linear and a non-linear meta-learner algorithm forecast the anticipated outcomes.

Next, we conduct a comparative analysis between the results obtained from the second and third steps to identify the most optimal approach. In the subsequent sections, we provide a complete and detailed explanation of these processes, enabling a thorough understanding of the employed methodology. All data analyses and modeling procedures were conducted utilizing R4.2.1. programming.

2.1 Study population

Conducted at the Motahari Breast Clinic within Namazi Hospital, falling within the affiliation of Shiraz University of Medical Sciences in Iran, this large case-control study centered on women diagnosed with primary invasive BC. The clinic serves as the primary referral center for patients recently diagnosed with BC in the Fars province, with over 80% of these patients’ receiving treatment at this facility. The study included all eligible women with confirmed diagnosis of BC during the study period. Information about the study participants and methodology (including criteria for selecting cases and controls) has been detailed elsewhere ( 18 ).

Briefly, a total of 1,073 women were invited to participate in the study, of whom 64 were disqualified due to absent or insufficient information on histopathological reports, leaving a final sample of 1,009 cases. Written informed consent was obtained from patients who were literate, while verbal consent was obtained from illiterate patients. The study protocol was approved by the ethical committee of Shiraz University of Medical Sciences (no. 13748).

2.2 Case and control selection

In this case-control study, new patients with a confirmed diagnosis of BC and were admitted to the oncology and radiotherapy wards were included as incident cases. Between April 2014 and March 2017, control participants were chosen from among female attendees who were not previously diagnosed with BC and were visiting patients in different departments of the same hospital. Women in the control group were considered cancer-free if they verbally confirmed no current or history of cancer, without the need for a confirmatory examination or test. A final total of 1009 control participants were selected, frequency-matching to cases by age, using 5-year age-groups for matching ( 19 ).

2.3 Data collection

In a face-to-face context and within a timeframe of 2 to 8 weeks from the diagnosis of BC, interviews were carried out with the patients. Interviews took place in a private and quiet room in the hospital, facilitated by a trained female nurse. The questionnaire included questions related to education, occupation, family history of BC, smoking during adolescence and adulthood, history of oral contraceptive use, history of chest X-ray, history of benign breast disease, physical activity, body mass index (BMI), deliberate weight loss after 18 years of age, age at first delivery, total number of months of breastfeeding, history of miscarriage, menarche age, regular menstrual cycles, menopausal status, and history of type 2 diabetes. This questionnaire’s reliability has been discussed before ( 19 ).

2.4 Feature selection

In this research, we deliberately decided, after consulting with knowledgeable professionals in the area, to exclude the use of machine learning approaches for feature selection. This conclusion is based on recognizing the need to conduct a comprehensive examination and evaluation of all possible factors instead of just depending on automated selection techniques powered by ML algorithms.

2.5 Data description

Table 1 provides insights into the distribution of the study variables among individuals without and with BC, highlighting potential associations with the condition.

www.frontiersin.org

Table 1 Statistics of BC patients and individuals without cancer.

The provided table presents a comprehensive overview of various factors and their distribution among the two groups. A thorough analysis of the data reveals several noteworthy observations.

In terms of education, a relatively similar distribution is observed in the two groups, with a predominant presence of individuals with primary/illiterate or intermediate education. Both groups exhibit comparable proportions in this regard.

Regarding employment, most individuals in both groups were identified as housewives, constituting approximately 77% of the total population. The remaining individuals were classified as employed, comprising around 23%. This occupational distribution is consistent across the two groups. Exploring the family history of BC reveals a substantial distinction between the groups. Among BC patients, there is a notably higher percentage (17%) of individuals with a first relative affected by BC, compared to the “control” group (9%). This disparity suggests a potential association between familial history and the incidence of BC.

Considering smoking habits, there is an appreciable difference between the groups, as the prevalence of smoking is higher among the patients (15%) compared to the controls (7%). The use of oral contraceptive pills (OCP) demonstrates a slight difference between the two groups. BC patients exhibit a slightly higher proportion (47%) of OCPs use, in contrast to the “control” group (40%).

The presence of a history of benign breast disease presents a noteworthy contrast between the two groups. Among BC patients, a greater proportion (14%) had a history of benign breast conditions, while a smaller percentage (7%) is observed in the control group. This discrepancy suggests a potential link between prior benign breast conditions and an increased susceptibility to BC.

The remaining factors in the table, including chest X-ray history, physical activity, BMI, age at first delivery, breastfeeding duration, history of miscarriage, menarche age, regular menstruation, menopausal status, and type 2 diabetes, necessitate further s crutiny and analysis to determine their potential implications for BC risk.

3 Machine learning methods

The Caret package’s grid search method (Kuhn, 2008) in R was employed to optimize hyperparameters for all algorithms (RF, NN, XGBoosting Tree, and bagged CART) within the training set. The parameter values for each applied ML algorithm are presented in Table 1 .

3.1 Random forest

RF is an ensemble learning method that uses decision trees to predict classes in the case of classification or means for regression within the individual trees ( 20 ). The algorithm constructs a multitude of decision trees, and each tree is trained on a random subset of the training data and a random subset of the features. This helps to reduce overfitting by creating a diverse set of trees that are not highly correlated with each other. Overfitting is further prevented by randomly selecting a subset of features while constructing each tree. This is controlled by the hyperparameter mtry, which determines the number of variables randomly sampled at each split time. The optimal value of mtry is determined by using a grid search method ( 21 ) and was found to be 18 here with the Caret package.

3.2 Bagged cart

Bootstrap aggregating (bagging) is an ensemble meta-algorithm designed to enhance the stability and accuracy of ML algorithms using techniques such as classification and regression trees (CART). This involves generating multiple training sets through the process of resampling the original dataset with replacement. Subsequently, a model is trained on each of these newly created sets, leading to improved performance ( 22 ). The final prediction is then made by combining the predictions of all the generated models. Bagged CART, a variation of decision tree algorithm, employes bagging to reduce overfitting by randomly selecting a subset of features at each split to construct each tree ( 23 ). The performance of bagged CART depends on the values of hyperparameters such as the number of trees (B), the number of variables randomly sampled at each split time (mtry), and the minimum number of observations required to split an internal node (minsplit). Optimal values of these hyperparameters can be obtained through hyperparameter tuning.

The optimization process involved grid search, a method that systematically explores all possible combinations of hyperparameters within predefined ranges. The optimal values for hyperparameters were obtained by comparing the cross-validation performance of the hyperparameters mtry and minsplit control the complexity of each tree, while the hyperparameter B controls the number of trees in the ensemble. The optimal values of these hyperparameters depend on the characteristics of the dataset and the specific problem at hand. After the hyperparameter tuning was carried out, it was determined that the optimal values were as follows: mtry = [tuned value], minsplit = [tuned value], and B = [tuned value]. These tuned hyperparameter values play a crucial role in significantly enhancing the performance of Bagged CART, resulting in the creation of a ML model that is both more accurate and robust. The optimization of hyperparameters through grid search is a crucial step in developing effective ML models, and it ensures that our model can generalize well to unseen data and tackle real-world challenges effectively.

3.3 Neural networks

Neural Networks (NN) have captured significant attention in recent years due to their remarkable capacity to effectively address intricate challenges spanning a wide range of domains. Inspired by the structure of biological NN ( 24 ), this method employs a three-layered feedforward network. The key innovation lies in the notion of weights, which connects the hidden layers and facilitates learning between the output and input layers ( 25 ). By leveraging these weighted connections, NN excels at learning and adapting to complex patterns, making them a powerful tool in modern ML applications. One of the primary advantages of NN is their ability to learn from data without being explicitly programmed. This is achieved through a process known as backpropagation, where the network adjusts its parameters to minimize the difference between its predictions and the true outputs in the training data ( 26 ).

A multilayer perceptron (MLP) is a fully connected class of feedforward artificial neural network. To optimize the MLP, three hyperparameters were considered in this study: the number of neurons in the hidden layer, the learning rate, and the activation function. The number of neurons in the hidden layer determines the complexity of the model, with a higher number of neurons increasing model complexity ( 27 ). The learning rate governs the step size within the gradient descent optimization algorithm employed for training the network. Typically, a smaller learning rate leads to improved convergence and accuracy, optimizing the training process ( 28 ). The activation function applies a nonlinear transformation to the output of each neuron to introduce nonlinearity into the model ( 29 ).

3.4 Extreme gradient boosting tree

XGBoost is a popular gradient boosting algorithm that constructs an ensemble of decision trees sequentially, with each tree aiming to correct the errors of its predecessor ( 30 ). Among its many variants, XGBoost stands out as it uses decision trees as the base learner ( 31 ).

The primary objective of this study is to optimize the hyperparameters specifically for XGBoost. These hyperparameters encompass several key aspects, including the maximum depth of the trees, the learning rate (eta), the minimum child weight, the subsample ratio of columns when constructing each tree, and the gamma parameter. The parameter known as gamma, which is the focus here, holds significant importance in the context of this study. It plays a central role in establishing the minimum loss reduction required to initiate an additional partition on a leaf node, shaping the decision-making process within the algorithm ( 32 ).

The mathematical formulas for the hyperparameters are as follows:

- Maximum depth: The maximum depth of the decision trees, denoted as max_depth. A higher depth allows the model to capture more complex interactions but may also lead to overfitting.

- Eta : The learning rate, denoted as eta, controls the step size taken during the optimization process. A lower value results in slower learning but may improve generalization.

- Minimum child weight : The minimum sum of instance weight needed in a child, denoted as min_child_weight. It controls the minimum number of instances required in each leaf node, which helps prevent overfitting.

- Subsample : The subsample ratio of columns when constructing each tree, denoted as subsample. A lower value results in more conservative models.

- Gamma : The minimum loss reduction required to make a further partition on a leaf node, denoted as gamma. A higher value results in fewer splits and more conservative models.

Tuning these hyperparameters can lead to better performance and prevent overfitting of the model. By carefully selecting the optimal values for these parameters, XGBoost can achieve high accuracy and better generalization in real-world applications. These values are shown in Table 2 .

www.frontiersin.org

Table 2 Parameter values of the four applied ML algorithms.

The results section of this study provides an in-depth analysis and presentation of the findings derived from the conducted analysis. The dataset will be subject to comprehensive examination to uncover valuable insights and observations. Meticulously exploring the dataset, characteristics, patterns, trends, and relationships are identified.

The analysis involved a comprehensive evaluation of statistical metrics and a meticulous assessment of the models’ predictive capabilities. If the circumstances warrant, the integration of these models into an ensemble framework is explored, guided by both linear and non-linear meta-learner algorithms.

4.1 Variable importance

In this analysis, the interaction among demographic, medical history, and lifestyle factors linked to BC risk was explored using four distinct ML models. The results provide valuable insights into the varying degrees of significance exhibited by different factors. To determine the feature importance, a well-established technique called “permutation feature importance” was utilized. This approach assesses the contribution of each predictor in the models’ decision-making process. In a specific approach, the values of each predictor were shuffled throughout the dataset, and subsequently, the resulting reduction in performance metrics of the models—such as accuracy or area under the receiver operating characteristic curve (AUC-ROC)—was assessed. The degree of decrease observed directly reflects the predictor’s significance in influencing the model’s performance. Chest X-ray history, Deliberate weight loss, abortion history, and post-menopausal status ranked among the top five predictors in all models. Moreover, secondhand smoking, education (high school), menarche age (>14), occupation (employed), first delivery age (18-23), and duration of breast-feeding (>42 months) were important in at least two models.

Conversely, certain variables, such as OCP use and physical activity, were identified as crucial only in one model, emphasizing the significance of utilizing multiple models when predicting BC risk. The rankings of particular predictors between models also varied with education (high school) and OCP use demonstrating divergent results.

The variable importance analysis emphasizes the importance of selecting an appropriate ML model and utilizing multiple models to ensure the robustness and generalizability of the predictive model. These findings suggest that a combination of demographic, medical history, and lifestyle factors should be considered when assessing BC risk. This information can be leveraged to develop targeted interventions to prevent and manage BC, contributing to the ongoing efforts to improve the accuracy and effectiveness of BC risk prediction models. These algorithms will be employed to generate predictions, and their performance will be rigorously evaluated using statistical indices and receiver operating characteristic (ROC) curves.

Table 3 displays the importance of variables within the ML models.

www.frontiersin.org

Table 3 Variable importance.

4.2 Prediction models

In this study, the prediction of outcomes using a dataset was facilitated through the utilization of four distinct machine learning algorithms. To establish a well-balanced representation, the dataset underwent a randomized partition into training and testing sets, maintaining an 80:20 ratio. Specifically, the training set, encompassing 80% of the data, was employed for model training purposes, while the remaining 20% constituted the test set, serving as the evaluative benchmark to assess model performance.

To ensure the validity and reliability of our models, we included the widely adopted technique known as ten-fold cross-validation. This technique involved the systematic division of the training dataset into ten subsets, each of which was subsequently utilized for model training and evaluation in a carefully orchestrated manner. Through this iterative process, the models were trained on nine subsets while being meticulously evaluated on the remaining subset. By aggregating the results across these iterations, a comprehensive and robust estimation of the models’ predictive capabilities was derived.

Multiple measures were utilized to analyze each model’s performance, Accuracy,

95% Accuracy Confidence Interval (CI), Kappa, Sensitivity, and Specificity. The performance measures for the four ML methods are reported in Table 4 .

www.frontiersin.org

Table 4 Performance measures of four ML models.

The Accuracy evaluation computes the models’ ability to detect instances related to the diagnosis of breast cancer. To achieve this, we compute the proportion of accurately classified instances to all occurrences, accounting for True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN). Where True Positive (TP) is the number of correctly predicted positive instances, True Negative (TN) is the number of correctly predicted negative instances, False Positive (FP) is the number of incorrectly predicted positive instances, and False Negative (FN) is the number of incorrectly predicted negative instances.

Formula for calculating accuracy is as follows:

The RF and Bagged CART models achieved the highest accuracy scores of 0.8389 and 0.8152, respectively. The XGBoosting and NN models had relatively lower accuracy scores, obtaining 0.7133 and 0.6635, respectively.

The 95% Accuracy CI for the RF, Bagged CART, XGBoost, and NN models are (0.8002, 0.8726), (0.7748, 0.8511), (0.6675, 0.756), and (0.6162, 0.7085), respectively. The overlapping intervals suggest that there may not be statistically significant differences in the performance between the models. However, further analysis and consideration of other performance measures are necessary to make more robust conclusions about the model’s performance.

Kappa, a measure of inter-rater agreement in the diagnosis of breast cancer, the RF and Bagged CART models exhibited higher Kappa values of 0.6776 and 0.6306, respectively, indicating a substantial level of agreement beyond chance. Following widely accepted interpretation guidelines ( 33 ), these values are considered substantial. Conversely, the XGBoost and NN models displayed lower Kappa values of 0.426 and 0.3275, respectively, suggesting a moderate and slight level of agreement, respectively, which is comparatively lower for these models.

Sensitivity, which measures the model’s ability to correctly identify individuals with breast cancer, emphasizing the significance of true positive predictions, was determined. The RF model exhibited the highest sensitivity (0.8419), indicating its effectiveness in identifying true positive cases. Specificity, on the other hand, measures the model’s ability to correctly identify actual negative instances. The Bagged CART model had the highest specificity (0.8406), indicating its proficiency in identifying true negative cases.

The RF and Bagged CART models demonstrated superior performance in terms of accuracy, Kappa, sensitivity, and specificity compared to the XGBoost and NN models. However, the choice of the best model may ultimately depend on the specific application and the relative importance of sensitivity and specificity in the given context.

Finally, the evaluation of ROC/AUC revolves around the discriminating ability of the models in diagnosing breast cancer. The Receiver Operating Characteristic (ROC) curve serves as a valuable visual tool to evaluate the classification models’ performance, depicting the balance between correctly identifying positive cases and incorrectly classifying negative cases. The Area Under the Curve (AUC) metric provides a comprehensive measure of the models’ discriminative ability. Figure 2 represents the ROC curves of the four ML models. In the context of the presented results, it is noteworthy that the RF model showcased the highest AUC value of 0.900, indicating its remarkable proficiency in effectively distinguishing between the classes. Following closely, the Bagged CART model demonstrated a commendable AUC of 0.892. In contrast, the XGBoost model exhibited a relatively lower AUC of 0.783, while the NN model displayed the lowest AUC of 0.741. These findings substantiate the superiority of the RF model in accurately classifying the data, thereby establishing its significance and prominence within the analytical assessment.

www.frontiersin.org

Figure 2 The ROC curves of four ML algorithms. XGB, extreme gradient boosting tree; RF, random forest; b-CART, bagged CART; NN, neural networks.

4.3 Ensemble models

Ensemble methodologies encompass learning strategies that aim to create a robust and improved predictive model through the integration of multiple learning algorithms ( 34 ). The integration of weak learners in ensemble learning techniques can be achieved through two approaches: homogeneous and heterogeneous ensembles ( 35 ). Weak learners, recognized as base models within the realm of ML, represent algorithms that outperform random guesses with a noticeable yet moderate degree of effectiveness. Ensemble learning reveals the principle of homogeneity, which is embodied through the well-established methodologies of bagging and boosting. These influential techniques are highly regarded for their capability to aggregate multiple instances of similar weak learners, resulting in homogeneous ensembles that possess enhanced predictive capabilities. Conversely, heterogeneous ensembles adopt a distinct and advanced strategy known as stacking, which involves integrating diverse weak learners. Distinguished by its heterogeneity, stacking augments the ensemble’s predictive capabilities by harnessing a synergistic interplay of complementary algorithms ( 36 ).

In this study, the application of stacking ensemble learning for predicting the intended outcome was undertaken, followed by a comparison of the results with the best-performing machine learning (ML) algorithm. In this research, the stacked generalization, commonly referred to as stacking, algorithm was employed. Initially, a variety of learning algorithms were utilized to generate predictions, and subsequently, the outcomes were integrated through the application of a combiner algorithm, which represents an additional technique within the field of ML. This integration process allowed for a comprehensive assessment of the predictive performance.

Ensemble methods aim to enhance model performance by combining diverse base models, characterized by low correlations among them. In the evaluation of the association between these models, the correlation coefficients span from -1 to 1. A value of -1 signifies a perfect negative correlation, while a value of 1 indicates a perfect positive correlation. Nevertheless, within the framework of ensemble methods, the ideal correlation value is 0, signifying an absence of any relationship between the models. Table 5 displays the correlation coefficients among the base models used in ensemble learning.

www.frontiersin.org

Table 5 Base models correlation.

The low correlation among base models is crucial as it indicates that the models provide independent and distinct perspectives which means they make different types of errors or predictions. This diversity allows the ensemble model to capture a wider range of patterns and insights, ultimately improving prediction accuracy. In other words, this diversity is beneficial because when combined, the ensemble can compensate for individual model weaknesses and exploit the strengths of different models. By considering various perspectives and incorporating different types of information, the ensemble can produce more accurate and robust predictions than any individual model alone. Conversely, high correlations between base models can introduce redundancy and limit the ensemble’s ability to incorporate diverse insights. By leveraging a diverse set of base models with low correlations, ensemble methods offer a robust and accurate approach to prediction and generalization in ML. The collective wisdom of these models, integrated using a combiner algorithm, allows the ensemble model to harness the strengths of each base model and mitigate potential biases, leading to enhanced predictive capabilities and generalization performance.

Ensemble models, such as those incorporating the Generalized Linear Model and Boosted Classification Trees, provide a framework for combining the predictive abilities of multiple models, thereby mitigating individual model weaknesses, and capitalizing on their strengths. By blending diverse perspectives and capturing a wider range of patterns, ensemble models can potentially achieve higher accuracy and agreement than any individual model alone.

Table 6 presents the performance metrics of the best-performing previous model, RF, in comparison to two ensemble models: Generalized Linear Model and Boosted Classification Trees. These metrics provide insights into the models’ predictive capabilities and agreement with actual outcomes.

www.frontiersin.org

Table 6 Performance comparison of RF model with other ensemble models.

Regarding Accuracy and Kappa, the RF model achieved an accuracy of 83.89% and a Kappa coefficient of 67.76%. These results indicate that the RF model yielded correct predictions with an accuracy of 83.89% and demonstrated substantial agreement beyond what would be expected by chance. The Generalized Linear Model, when employed within the ensemble, achieved slightly lower accuracy and Kappa coefficient of 83.00% and 66.00%, respectively. Similarly, the Boosted Classification Trees, when incorporated in the ensemble, achieved a lower accuracy of 82.56% and lower Kappa coefficient of 0.65.

The RF model is known for its ability to handle complex relationships and interactions between variables. It combines multiple decision trees and aggregates their predictions to make accurate predictions. The results of Table 6 indicate that the RF model outperformed both the Generalized Linear Model and the Boosted Classification Trees in terms of accuracy and agreement. However, it is essential to acknowledge that ensemble models have their own merits, offering the potential for improved performance by incorporating diverse models so they can offer complementary insights and diversify predictions, potentially enhancing performance in specific contexts. Choosing the most suitable model or ensemble approach depends on the specific characteristics of the data and the objectives of the analysis.

5 Discussion

BC remains a significant public health concern globally, and understanding the risk factors associated with the disease is crucial for effective prevention and intervention strategies. This study aimed to investigate BC risk factors among women in Iran’s Fars province, contributing valuable insights to the understanding of disease etiology and informing targeted approaches to BC prevention and management. The rigorous methodology employed in this study, including data collection from a representative sample size and the use of four ML algorithms, ensures the reliability and robustness of the findings.

The analysis of variable importance across the ML models revealed variations in the rankings of risk factors, which underscores the inherent complexity and heterogeneity of BC etiology. However, in line with the existing body of literature certain variables including chest X-ray history, deliberate weight loss, and abortion history consistently emerged as influential across all four algorithms, suggesting their potential significance as BC risk factors ( 37 – 40 ). Additionally, variables such as post-menopausal status, second-hand smoking, OCP use, and education (high school) were consistently identified as significant risk factors by the majority of the ML models echoing earlier investigations ( 8 , 41 – 44 ).

In accordance with prior research findings, this study also indicates the factors that do not display significant involvement in the risk assessment of BC including occupation, physical activity, BMI, age at first delivery, breastfeeding, history of miscarriage, menarche age, regular menstruation, and menopausal status ( 19 , 45 – 51 ).

By combining the power of ML algorithms with comprehensive risk factor assessment, significant strides can be made in mitigating the burden of BC among women in both Iran and around the globe. The ML models showcase robust performance metrics. Notably, RF and Bagged CART stand out with their higher accuracy and Kappa values, reinforcing their potential for accurate BC risk prediction. However, the reasonable sensitivity and specificity values exhibited by XGBoost and NN highlight their ability to identify both true positive and true negative cases. With all ML models exhibiting good discriminatory power, these findings emphasize the effectiveness of ML algorithms in assessing BC risk.

This study significantly contributes to the expanding body of research focused on identifying predictors of BC risk. The findings underscore the importance of employing multiple ML models to enhance the accuracy of BC risk prediction. Furthermore, the study emphasizes the necessity of considering a wide array of risk factors during model development. By incorporating these approaches, the accuracy and effectiveness of risk prediction can be heightened, ultimately alleviating the burden of BC among women in Iran. Continued research advancements are crucial to further deepen our understanding and develop targeted interventions for BC prevention and management. Also, it is crucial to recognize that the fight against BC is an ongoing battle. Continued research advancements are vital for deepening our understanding of the complex interplay between risk factors and BC etiology. Moreover, these findings should be complemented with translational efforts to ensure that evidence-based strategies are effectively disseminated and implemented in clinical practice and public health policies.

By integrating the insights from this study into comprehensive risk prediction models, public health practitioners can enhance their ability to identify individuals at high risk of developing BC, enabling timely interventions and personalized prevention strategies.

This study is a significant step toward improving breast cancer prevention, early detection, and management. Continued research advancements and collaborative efforts are crucial to further deepen our understanding, develop targeted interventions, and ultimately reduce the burden of breast cancer on women’s health worldwide following interventions on the (modifiable) risk actors.

Although our study has several strengths, including the application of different methods, considering a vast variety of potential confounders and risk factors, and a large sample size, we would like to bring some limitations to the table. We collected data from a reference hospital in the southern part of the country but did not include data from other geographic areas of the country. Hence, in future research, it might be significant to consider a representative sample of different parts of the country to find any discrepancies between different ethnic groups. To ensure that ethnicity or other socio-demographic factors did not affect our results, we adjusted the results for potential confounders in the southern part of the country. Also, recruiting participants who visited the biggest referral center in the southern part of Iran makes the results generalizable to the city’s population.

Another limitation is recall bias from a case-control study, which is a common bias in this study design. To address this limitation, we could use some objective variables that participants can remember. Another potential confounding factor is alcohol consumption, which is not legal in our country, making it difficult to determine its true impact. However, we expect minimal influence on our results from alcohol consumption, as it is not prevalent in our country.

6 Conclusion

BC is a complex disease with many different risk factors that influence it, necessitating a thorough analysis. This study underscores the application of ML models to identify significant predictors of BC risk, thereby enhancing risk prediction accuracy. Within the context of the study population, this research highlights the pivotal role of demographic, medical history, and lifestyle factors in evaluating BC risk among women. The study’s outcomes indicated that a history of chest X-rays emerged as a noteworthy risk factor for BC. Furthermore, the presence of a family history of BC, smoking habits, and OCPs usage were identified as substantial predictors. These findings suggest that interventions targeting smoking cessation and promoting BC screening among women with a familial BC history could yield effective outcomes in reducing BC incidence.

Moreover, this research provides valuable insights into the intricate interplay of metabolic and hormonal factors contributing to BC development. The identification of deliberate weight loss, abortion history, and post-menopausal status as significant predictors underscores the significance of considering multiple factors when assessing BC risk. By expanding the existing knowledge base on BC risk factors, this study emphasizes the utilization of advanced ML techniques to elucidate complex interactions among various predictors. Subsequent studies can leverage these findings to develop more precise and efficacious BC risk prediction models, empowering clinicians, and patients to make informed decisions regarding BC prevention and management.

Data availability statement

The data analyzed in this study is subject to the following licenses/restrictions: The dataset was obtain from a healthcare center in Shiraz, Iran. Being obtained from different geographic can improve validity of the dataset. Requests to access these datasets should be directed to [email protected] .

Ethics statement

Ethical approval was not required for the study involving humans in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and the institutional requirements.

Author contributions

MD-N: Data curation, Software, Supervision, Writing – review & editing, Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Validation. KS: Data curation, Supervision, Writing – review & editing, Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Validation. RM: Formal analysis, Investigation, Methodology, Project administration, Resources, Validation, Writing – review & editing. SS: Investigation, Methodology, Software, Validation, Writing – review & editing. MF: Data curation, Supervision, Writing – review & editing, Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Validation. KH: Data curation, Writing – review & editing. BJ-S: Methodology, Validation, Resources, Writing – review & editing. TC: Data curation, Software, Supervision, Writing – review & editing, Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Validation. SD: Data curation, Software, Supervision, Writing – review & editing, Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Validation.

The author(s) declare that no financial support was received for the research, authorship, and/or publication of this article.

Acknowledgments

We appreciate all the healthcare professionals who agreed to participate in this study, and all staff for their great cooperation in data collection.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer KR declared a shared affiliation with the author KH to the handling editor at the time of review.

The author(s) declared that they were an editorial board member of Frontiers, at the time of submission. This had no impact on the peer review process and the final decision.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Abbreviations

BC, Breast cancer; ML, Machine learning; RF, Random Forest; NN, Neural Networks; Bagged CART, Bootstrap Aggregating Classification and Regression Trees; XGBoost, Extreme Gradient Boosting Tree; WHO, World Health Organization.

1. Breast cancer. Breast cancer (2021). Available at: https://www.who.int/news-room/fact-sheets/detail/breast-cancer .

Google Scholar

2. Torre LA, Islami F, Siegel RL, Ward EM, Jemal A. Global cancer in women: burden and trends. Cancer Epidemiol Biomarkers Prev (2017) 26(4):444–57. doi: 10.1158/1055-9965.EPI-16-0858

PubMed Abstract | CrossRef Full Text | Google Scholar

3. Gogate A, Wheeler SB, Reeder-Hayes KE, Ekwueme DU, Fairley TL, Drier S, et al. Projecting the prevalence and costs of metastatic breast cancer from 2015 through 2030. JNCI Cancer Spectr (2021) 5(4). doi: 10.1093/jncics/pkab063

4. Khoramdad M, et al. Breast cancer risk factors in Iranian women: a systematic review and meta-analysis of matched case-control studies. Eur J Med Res (2022) 27(1):311. doi: 10.1186/s40001-022-00952-0

5. Yang R, et al. RETRACTED ARTICLE: The circRNA circAGFG1 acts as a sponge of miR-195-5p to promote triple-negative breast cancer progression through regulating CCNE1 expression. Mol Cancer (2019) 18(1):4. doi: 10.1186/s12943-018-0933-7

6. GLOBOCAN. Estimated cancer incidence, mortality and prevalence worldwide in. Int. Agency Res. Cancer (2012) 2012:2015. Cancer, I.A.f.R.o. doi: 10.1200/JGO.17.00175

CrossRef Full Text | Google Scholar

7. Anstey EH, et al. Breastfeeding and breast cancer risk reduction: implications for black mothers. Am J Prev Med (2017) 53(3):S40–6. doi: 10.1016/j.amepre.2017.04.024

8. Chen WY, et al. Moderate alcohol consumption during adult life, drinking patterns, and breast cancer risk. Jama (2011) 306(17):1884–90. doi: 10.1001/jama.2011.1590

9. Martin LJ, et al. Family history, mammographic density, and risk of breast cancer. Cancer Epidemiol Biomarkers Prev (2010) 19(2):456–63. doi: 10.1158/1055-9965.EPI-09-0881

10. Dehdar S, et al. Applications of different machine learning approaches in prediction of breast cancer diagnosis delay. Front Oncol (2023) 13:1103369. doi: 10.3389/fonc.2023.1103369

11. Esteva A, et al. Dermatologist-level classification of skin cancer with deep neural networks. nature (2017) 542(7639):115–8. doi: 10.1038/nature21056

12. Ha R, et al. Convolutional neural network based breast cancer risk stratification using a mammographic dataset. Acad Radiol (2019) 26(4):544–9. doi: 10.1016/j.acra.2018.06.020

13. Wen W, et al. Prediction of breast cancer risk based on common genetic variants in women of East Asian ancestry. Breast Cancer Res (2016) 18(1):124. doi: 10.1186/s13058-016-0786-1

14. Wankhade Y, et al. (2023). Machine learning approach for breast cancer prediction: A review, in: 2023 2nd International Conference on Applied Artificial Intelligence and Computing (ICAAIC). Salem, India: IEEE.

15. Li K, et al. Risk prediction for estrogen receptor-specific breast cancers in two large prospective cohorts. Breast Cancer Res (2018) 20(1):1–16. doi: 10.1186/s13058-018-1073-0

16. Cuocolo R, et al. Machine learning in oncology: a clinical appraisal. Cancer Lett (2020) 481:55–62. doi: 10.1016/j.canlet.2020.03.032

17. Harirchi I, et al. Breast cancer in Iran: results of a multi-center study. Asian Pac J Cancer Prev (2004) 5(1):24–7.

PubMed Abstract | Google Scholar

18. Foroozani E, Akbari A, Amanat S, Rashidi N, Bastam D, Ataee S, et al. Adherence to a western dietary pattern and risk of invasive ductal and lobular breast carcinomas: a case–control study. Sci Rep (2022) 12(1):5859. doi: 10.1038/s41598-022-09725-5

19. Dianatinasab M, Fararouei M, Daneshi N, Rezaian S, Mohammadianpanah M, Chaman R, et al. Heterogeneity in risk factors for ductal and lobular breast carcinomas: a case–control study. Int J Cancer (2019) 145(11):2917–25. doi: 10.1002/ijc.32182

20. Breiman L. Random forests. Mach Learn (2001) 45:5–32. doi: 10.1023/A:1010933404324

21. Liaw A, Wiener M. Classification and regression by randomForest. R News (2002) 2(3):18–22. doi: 10.1021/ci034160g

22. Breiman L. Bagging predictors. Mach Learn (1996) 24:123–40. doi: 10.1007/BF00058655

23. Breiman L. Arcing the edge . Technical Report 486, University of California at Berkeley (1997) pp. 1–14.

24. McCulloch WS, Pitts W. A logical calculus of the ideas immanent in nervous activity. Bull Math Biophys (1943) 5:115–33. doi: 10.1007/BF02478259

25. Oja E. Simplified neuron model as a principal component analyzer. J Math Biol (1982) 15:267–73. doi: 10.1007/BF00275687

26. LeCun Y, Bengio Y, Hinton G. Deep learning. nature (2015) 521(7553):436–44. doi: 10.1038/nature14539

27. Bishop CM. Neural networks for pattern recognition . Oxford University Press (1995).

28. Bottou L. Stochastic gradient descent tricks. In: Neural networks: tricks of the trade (Berlin, Heidelberg: Springer), 2nd ed. (2012). p. 421–36.

29. Glorot X, Bordes A, Bengio Y. (2011). Deep sparse rectifier neural networks. In: Geoffrey G, David D, Miroslav D, editors. Proceedings of the fourteenth international conference on artificial intelligence and statistics. JMLR Workshop and Conference Proceedings (2011). p. 315–23.

30. Chen T, Guestrin C. (2016). Xgboost: A scalable tree boosting system, in: Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 785–94

31. Ridgeway G. Generalized Boosted Models: A guide to the gbm package. Update (2007) 1(1):2007.

32. Chen T, He T, Benesty M, Khotilovich V, Tang Y, Cho H, et al. Xgboost: extreme gradient boosting . R package version 0.4-2. (2015) 1 (4), 1–4.

33. McHugh ML. Interrater reliability: the kappa statistic. Biochem Med (2012) 22(3):276–82. doi: 10.11613/BM.2012.031

34. Ryu JW, Kantardzic M, Walgampaya C. Ensemble classifier based on misclassified streaming data, in: Proc. of the 10th IASTED int. Conf. on artificial intelligence and applications (2010) Austria. 347–54

35. Akour M, Alsmadi I, Alazzam I. Software fault proneness prediction: a comparative study between bagging, boosting, and stacking ensemble and base learner methods. Int J Data Anal Techniques Strategies (2017) 9(1):1–16. doi: 10.1504/IJDATS.2017.083058z

36. Saadatmand S, Salimifard K, Mohammadi R, Kuiper A, Marzban M, Farhadi A. Using machine learning in prediction of ICU admission, mortality, and length of stay in the early stage of admission of COVID-19 patients. Ann Operations Res (2023) 328:1043–71. doi: 10.1007/s10479-022-04984-x

37. Gao YT, Shu XO, Dai Q, Potter JD, Brinton LA, Wen W, et al. Association of menstrual and reproductive factors with breast cancer risk: results from the Shanghai Breast Cancer Study. Int J Cancer (2000) 87(2):295–300. doi: 10.1002/1097-0215(20000715)87:2<295::AID-IJC23>3.0.CO;2-7

38. Lipworth L, Bailey LR, Trichopoulos D. History of breast-feeding in relation to breast cancer risk: a review of the epidemiologic literature. J Natl Cancer Inst (2000) 92(4):302–12. doi: 10.1093/jnci/92.4.302

39. Sweeney C, Giuliano AR, Baumgartner KB, Byers T, Herrick JS, Edwards SL, et al. Oral, injected and implanted contraceptives and breast cancer risk among U.S. Hispanic and non-Hispanic white women. Int J Cancer (2007) 121(11):2517–23. doi: 10.1002/ijc.22970

40. Ursin G, Bernstein L, Wang Y, Lord SJ, Deapen D, Liff JM, et al. Reproductive factors and risk of breast carcinoma in a study of white and African-American women. Cancer (2004) 101(2):353–62. doi: 10.1002/cncr.20373

41. Ghoncheh M, Mahdavifar N, Darvishi E, Salehiniya H. Epidemiology, incidence and mortality of breast cancer in Asia. Asian Pac J Cancer Prev (2016) 17(S3):47–52. doi: 10.7314/APJCP.2016.17.S3.47

42. Jones ME, Schoemaker MJ, Wright LB, Ashworth A, Swerdlow AJ. Smoking and risk of breast cancer in the Generations Study cohort. Breast Cancer Res (2017) 19(1):118. doi: 10.1186/s13058-017-0908-4

43. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans Meeting. Hormonal contraception and post-menopausal hormonal therapy. In: IARC monographs on the identification of carcinogenic hazards to humans , vol. 91. World Health Organization (1999).

44. Johnson KC. Accumulating evidence on passive and active smoking and breast cancer risk. Int J Cancer (2005) 117(4):619–28. doi: 10.1002/ijc.21150

45. Vahdaninia M, Omidvari S, Montazeri A. What do predict anxiety and depression in breast cancer patients? A follow-up study. Soc Psychiatry Psychiatr Epidemiol (2010) 45(3):355–61. doi: 10.1007/s00127-009-0068-7

46. Clinton SK, Giovannucci EL, Hursting SD. The world cancer research fund/american institute for cancer research third expert report on diet, nutrition, physical activity, and cancer: impact and future directions. J Nutr (2020) 150(4):663–71. doi: 10.1093/jn/nxz268

47. Ganmaa D, Sato A. The possible role of female sex hormones in milk from pregnant cows in the development of breast, ovarian and corpus uteri cancers. Med Hypotheses (2005) 65(6):1028–37. doi: 10.1016/j.mehy.2005.06.026

48. Ziegler RG, Hoover RN, Pike MC, Hildesheim A, Nomura AM, West DW, et al. Migration patterns and breast cancer risk in Asian-American women. J Natl Cancer Inst (1993) 85(22):1819–27. doi: 10.1093/jnci/85.22.1819

49. Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, et al. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer (2015) 136(5):E359–86. doi: 10.1002/ijc.29210

50. Yala A, Lehman C, Schuster T, Portnoi T, Barzilay R. A deep learning mammography-based model for improved breast cancer risk prediction. Radiology (2019) 292(1):60–6. doi: 10.1148/radiol.2019182716

51. Liu X, Yuan P, Li R, Zhang D, An J, Ju J, et al. Predicting breast cancer recurrence and metastasis risk by integrating color and texture features of histopathological images and machine learning technologies. Comput Biol Med (2022) 146:105569. doi: 10.1016/j.compbiomed.2022.105569

Keywords: breast cancer, machine learning, risk factor, random forest, neural networks, bootstrap aggregating classification and regression tree, extreme gradient boosting

Citation: Dianati-Nasab M, Salimifard K, Mohammadi R, Saadatmand S, Fararouei M, Hosseini KS, Jiavid-Sharifi B, Chaussalet T and Dehdar S (2024) Machine learning algorithms to uncover risk factors of breast cancer: insights from a large case-control study. Front. Oncol. 13:1276232. doi: 10.3389/fonc.2023.1276232

Received: 30 August 2023; Accepted: 27 December 2023; Published: 15 February 2024.

Reviewed by:

Copyright © 2024 Dianati-Nasab, Salimifard, Mohammadi, Saadatmand, Fararouei, Hosseini, Jiavid-Sharifi, Chaussalet and Dehdar. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Samira Dehdar, [email protected]

† ORCID : Mostafa Dianati-Nasab, orcid.org/0000-0002-0185-5807 Khodakaram Salimifard, orcid.org/0000-0002-7059-724X

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

  • Open access
  • Published: 16 February 2024

High-protein diet scores, macronutrient substitution, and breast cancer risk: insights from substitution analysis

  • Mitra Kazemi Jahromi 1   na1 ,
  • Hamid Ahmadirad 2 ,
  • Hossein Farhadnejad 2   na1 ,
  • Mostafa Norouzzadeh 2 ,
  • Ebrahim Mokhtari 2 , 3 ,
  • Farshad Teymoori 2 , 4 ,
  • Niloufar Saber 2 ,
  • Zeinab Heidari 5 ,
  • Parvin Mirmiran 2 &
  • Bahram Rashidkhani 6  

BMC Women's Health volume  24 , Article number:  121 ( 2024 ) Cite this article

Metrics details

Evidence from recent studies suggested that variation in the quantity and quality of macronutrients in the diet may potentially play a role in predicting the risk of breast cancer (BC). In the current study, we aimed to assess the association of different high-protein diet scores and replacing fats and carbohydrate (CHO) with protein in the diet with the BC risk among Iranian women.

The current hospital-based case-control study was conducted on 401 participants, aged ≥ 30 years old, including 134 women in the case group who had been diagnosed with histologically confirmed BC and 267 women in the control group. Dietary intake data was collected using a validated food frequency questionnaire, and high protein diet scores were determined. Logistic regression models were used to determine the odds ratios (OR) and 95% confidence interval (CI) of BC across tertiles of high protein diet scores. Also, we assessed how substituting protein with other macronutrients affected BC odds while adjusting for the various confounding variables.

Participants’ mean ± SD of age and body mass index were 47.9 ± 10.3 years and 29.4 ± 5.5 kg/m 2 , respectively. The scores of high-protein-low-CHO and fat diet, high-protein and CHO-low-fat diet, and high-protein and fat-low-CHO diet in participants were 16.5 ± 3.8, 16.5 ± 6.7, and 16.4 ± 5.9, respectively. In the multivariable model, individuals in the highest tertile of high-protein-low-CHO and fat diet score (OR:0.71;95%CI:0.56–0.90) and high-protein and CHO-low-fat diet (OR:0.76;95%CI:0.60–0.97) had lower odds of BC compared to those in the lowest tertile ( P  < 0.05). However, no significant association was found between high-protein and fat-low-CHO diet and BC risk. Our results showed that replacing fat by protein (OR differences :-0.40;95%CI:-0.73,-0.07) and also replacing refined-CHO by plant protein (OR differences :-0.66;95%CI:-1.26,-0.07) in the diet are associated inversely with risk of BC( P  < 0.05).

Conclusions

The results of our study suggested that higher adherence to a high-protein-low-CHO and fat diet, characterized by a higher intake of plant proteins and a lower intake of refined grains and saturated fat can play a protective role against the odds of BC.

Peer Review reports

Breast cancer (BC) is a common malignancy affecting both sexes, with a higher incidence in women, causing a quarter of women’s cancers and affecting 1.5 million women annually [ 1 ]. In Iran, BC ranks as the third leading cause of death among women, with a ten-year survival rate of 58.1% [ 2 , 3 ]. BC impacts 8,000 Iranian women annually, and one-third of cases occur in women under 30 years old [ 4 ]. While early diagnosis of BC can improve the treatment process, its metastatic and multifactorial nature makes it difficult to treat effectively [ 5 ]. Consequently, healthcare systems bear a greater economic burden [ 6 ] and BC patients face premature death and reduced quality of life [ 7 ].

Genetic predisposition, sex, aging, unhealthy lifestyle, and poor diet are risk factors for predicting BC risk [ 8 , 9 ]. Crucial preventable causes of BC mortality include dysglycemia, obesity, alcohol, and red meat consumption. These factors are directly or indirectly linked to dietary choices [ 10 ]. Therefore, in the past decades, several studies focused on the role of diet in predicting the risk of BC at different levels, including food patterns, food groups, and nutrients [ 11 , 12 ].

Dietary protein intake can impact cancer risks depending on the type and amount of protein consumed [ 13 , 14 , 15 ]. The higher red meat consumption, as a source of dietary protein, is responsible for 3.21% of BC mortality [ 10 ]. It has been suggested that a high intake of protein could increase the risk of certain types of cancer, including prostate cancer, esophageal cell carcinoma, and colon cancer [ 13 , 14 , 15 ]. There are conflicting results from various studies. High-protein diets have been linked to an increased risk of respiratory tract and renal cell cancer [ 16 , 17 ], but a decreased risk of prostate cancer [ 18 ]. However, there is no clear link between high protein diet and the incidence or mortality of BC [ 19 ]. Studies have suggested that a higher intake of protein can improve the survival rate in individuals with BC [ 20 , 21 , 22 , 23 ]. However, a long observational study has highlighted that the source of protein consumed is a more crucial factor in determining the incidence of BC than the overall amount of protein consumed [ 24 ]. Notably, increased consumption of animal-based proteins may lead to a heightened risk of BC up to 20% [ 25 ].

Considering conflicting results in current research, the rising prevalence of BC in Iran, and the absence of conclusive findings on the association between the quantity and quality of dietary protein and BC risk in the Middle East and North Africa region, our objective was to explore the potential relationship between high protein diet scores and odds BC in Iranian adults. Also, we employed substitution models and compared different dietary protein sources with odds of BC.

Materials and methods

Study population.

We conducted a hospital-based case-control study using the Shohada and Imam Hossain hospitals as referral centers in Tehran, Iran from September 2015 to February 2016. The case group consisted of 136 newly diagnosed women with BC who were aged ≥ 30 years, had been diagnosed with BC within the previous 6 months, and were not undergoing any cancer treatments at the time of the interview. For the case group, we applied a set of exclusion criteria, including individuals who followed specific dietary habits, those with a history of hormone replacement therapy (HRT), and pregnant or lactating women. The control group included 272 women, aged ≥ 30 years, who were admitted to the same hospitals during the study period for non-neoplastic conditions, such as traumas and orthopedic disorders, disk disorders, acute surgical conditions, and eye, nose, ear, or skin disorders. Also, for the control group, we excluded individuals with a history of HRT or benign breast disease, physician-diagnosed cancer in any site, and those who were pregnant or lactating, as well as those who followed special dietary habits due to a particular disease or for weight loss purposes.

Of the subjects approached for participation in the study, less than 8% declined to be interviewed. Seven participants were excluded from the analysis due to reported energy intakes that deviated by more than ± 3 standard deviation (SD) from the mean energy intakes of the population, which included five subjects in the control group and two subjects in the case group. The exclusion of participants with extreme deviations in reported energy intakes was based on the assumption that these values were not representative of their actual intake and could introduce significant variability in the analysis. By excluding these outliers, we aimed to ensure the inclusion of participants with plausible energy intakes and minimize the potential bias caused by extreme values. Finally, a total of 401 participants (134 cases and 267 controls) remained for the final analysis.

Dietary assessment

To assess dietary intake during the year before diagnosis for cases or interviews for controls, a validated and reliable semi-quantitative 168-item food frequency questionnaire (FFQ) with standard serving sizes was used [ 26 ]. Participants were asked by skilled dietitians to report how frequently they consumed each food item on a daily, weekly, monthly, or yearly basis over the course of the previous year. The portion sizes of consumed foods were then converted into daily grams using household measures [ 27 ]. The energy and nutrient intake were computed using the United States Department of Agriculture (USDA) food composition table (FCT). The Iranian FCT was used for local food items not listed in the USDA FCT [ 28 ]. Adjustments were made to ensure compatibility and consistency between the two databases. By incorporating the Iranian Food Composition Table, we aimed to accurately estimate the nutrient composition of traditional Iranian foods in our analysis. It should be noted that due to religious considerations and legal restrictions in Iranian society, we could not collect data on alcohol consumption in participants, and therefore, it was not included in the analysis.

We calculated the main protein, carbohydrate (CHO), and fat subgroups based on their dietary sources as follows: Refined CHOs were defined as the total CHOs consumed from refined grains, sweets, and sweet snacks and drinks with added sugar. Carbohydrates obtained from other sources such as whole grains, dairy, fruits, and vegetables were considered to be non-refined CHOs. Moreover, protein and fat were classified into two categories: animal and plant sources. Animal protein was defined as dietary protein obtained from animal sources such as meat, poultry, fish, eggs, and dairy products, while plant protein was defined as protein obtained from plant sources such as legumes, nuts, seeds, and whole grains. Animal fat was determined as dietary fat obtained from animal sources such as meat, butter, cheese, and other dairy products, while plant fat was defined as fat obtained from plant sources such as nuts, seeds, and vegetable oils. In the current study, the breakdown of dietary subgroups, including refined CHOs, non-refined CHOs, animal protein, plant protein, and fat sources was made based on established nutritional classifications and their potential relevance to chronic diseases such as breast cancer risk [ 29 , 30 , 31 ]. These categorizations allow for a more detailed analysis of the impact of different types of CHOs, proteins, and fats on breast cancer outcomes.

High protein diet scores definition

To determine the dietary intake of protein, fat, and CHOs, the percentage of energy intake from each nutrient was calculated and categorized into deciles. A score of 1 to 10 was assigned to each decile for the high protein, high fat, and high CHO diets, respectively. Conversely, for the low-fat and low-CHO diets, a score of 10 to 1 was assigned to each decile, respectively. The scores for each nutrient were then summed to create three types of high-protein diets, including (1) high protein- low CHO and low-fat diet, (2) high protein- high CHO and low-fat diet, and (3) high protein- low CHO and high-fat diet.

Assessment of non-dietary exposures

Body weight was assessed using a digital scale (Seca, Germany) with a precision of 0.5 kg. Participants were instructed to wear light clothing and no shoes during the measurement. Height was measured to the nearest 0.5 cm using a tape meter that was fixed to a wall. This protocol was followed across all data collection sessions, and trained personnel conducted the measurements to minimize potential variations introduced by different individuals. The body mass index (BMI) was calculated as weight (kg) divided by height in square meters (m 2 ).

Trained dietitians conducted all other questionnaires and measurements. General questionnaires were used to collect participants’ socio-demographic, lifestyle, and clinical information, including age (years), age at menarche (years), age at first pregnancy (years), smoking status (yes, no), marital status (single, married, divorced, widowed), menopausal status (pre-menopause, post-menopause), education level (illiterate, less than a high school diploma, high school diploma and more), abortion history (yes, no), number of live births (number), breastfeeding history (month), history of HRT (yes, no), history of oral contraceptive pill (OCP) use, history of benign breast diseases (yes, no), family history of cancer (yes, no), family history of breast cancer (yes, no), bra-wearing habits (day (yes, no), night (yes, no)), vitamin D supplementation (yes, no), and use of anti-inflammatory medications (yes, no). The physical activity levels of participants were also assessed using a reliable and validated questionnaire [ 32 ], and the results were reported as Metabolic Equivalents hours per week (MET-h/week) [ 33 , 34 ].

To determine the socio-economic status (SES) score [ 35 ] of participants, three variables were used: family size (classified as ≤ 4 or > 4 people), education (categorized as academic or non-academic), and occupation (classified as employed or not employed). For each participant, a score of either 1 (if their family had ≤ 4 members, had an academic education, and were employed) or 0 (if their family had > 4 members, or had no academic education, and were unemployed) was assigned to each of the three variables. These scores were then summed to calculate the participant’s SES score. A score of 3 was considered high SES, a score of 2 was moderate, and a score of 1 or 0 was classified as low SES.

Statistical analysis

Statistical analysis was conducted using The Statistical Package for Social Sciences (Version 20.0; SPSS, Chicago, IL) (SPSS) software. The normality of the variables in both the case and control groups was assessed using a histogram chart and the Kolmogorov-Smirnoff test. All quantitative variables had normal distribution in the case and control groups because the results of the Kolmogorov-Smirnoff test were not statistically significant ( P  > 0.05), and also the histogram chart visually showed the normality of the distribution of the variables. We compared the mean values of continuous variables using the independent sample t-test. The chi-square test was also used to compare categorical variables. Logistic regression analysis was used to calculate the odds ratios (ORs) and 95% confidence intervals (CIs) for BC across tertiles of different high-protein diets. To account for potential confounding factors, we adjusted the logistic regression model for age, age at first pregnancy, menopausal status, family history of cancer, anti-inflammatory drug use, vitamin D supplementation, physical activity, BMI, SES, and energy intake. We also investigated the association between each of the major subgroups of CHOs, proteins, and fats, including refined and unrefined CHOs, animal and plant proteins, and animal and plant fats, with the risk of BC using logistic regression analysis with adjustment for potential confounding factors. Consequently, the data were analyzed by examining the odds of BC in individuals in the highest tertile compared to the lowest tertile and per increment of one SD of the aforementioned scores. Moreover, we examined how substituting protein with other macronutrients affected the BC risk after adjusting for the potential confounding variables mentioned above. We substituted 50 and 100 kcal of CHO, fat, animal protein, refined CHOs, and animal fat by total and plant protein intakes in the same multivariable logistic regression model. The difference in their coefficients plus their covariance was used to estimate the OR and 95% CI differences. All P -values are two-sided and P -values < 0.05 were considered statistically significant.

Participants’ mean ± SD of age and BMI were 47.9 ± 10.3 years and 29.4 ± 5.5 kg/m 2 , respectively. The scores of high-protein-low-CHO and fat diet, high-protein and CHO-low-fat diet, and high-protein and fat-low-CHO diet in participants were 16.5 ± 3.8, 16.5 ± 6.7, and 16.4 ± 5.9, respectively.

Table  1 indicates the baseline characteristics of subjects, including demographic and lifestyle variables, medical history, and dietary intakes in the case and control groups. The participants’ age, first pregnancy age, % of postmenopausal women, and cancer family history were higher in the case group compared to the control group, whereas the anti-inflammatory drug consumption, energy, plant protein, and vitamin D supplement intake in the control group were higher than case group ( P  < 0.05). Also, the high-protein-low-CHO and fat diet score in the control group was higher than the case group ( P  < 0.05). There were no significant differences between cases and controls in other variables.

Table  2 reported the ORs (95%CI) of BC based on the different high protein scores (including high-protein-low CHO and fat diet score, high-protein and CHO-low-fat diet, and high-protein and fat-low-CHO diet) in tertiles and a per increment of one SD among the study population. In the fully adjusted model after adjusting for age, first pregnancy age, menopausal status, family history of cancer, anti-inflammatory drug use, Vitamin D supplementation, physical activity, body mass index, socio-economic status, and energy intake, individuals in the third tertile of high-protein-low-CHO and fat diet score had lower odds of BC (OR: 0.48, 95%CI: 0.27–0.85, P-trend: 0.008) compared with those in the first tertile. However, based on the multivariable model, no significant association was found between the scores of high-protein and CHO-low-fat diet (OR: 0.58, 95%CI: 0.33–1.01, P-trend: 0.061) and high protein and fat-low-CHO diet with odds of BC (OR: 0.91, 95%CI: 0.52–1.59, P-trend: 0.665). Also, Table  2 showed that the odds of BC decreased by 29% (OR: 0.71, 95%CI: 0.56–0.90) with each SD increase in the high-protein-low-CHO and fat diet score. Furthermore, each SD increment in the high-protein and CHO-low-fat score was associated with 24% decreased odds of BC (OR: 0.76, 95%CI: 0.60–0.97) in the final model.

Table  3 expresses the substitute analysis for the association of replacing macronutrients together on the odds of BC, calculated using logistic regression models. Each 50 or 100 kcal replacement of fat by protein was associated with 20% (95%CI difference : -0.36, -0.03) and 40% (95%CI difference : -0.73, -0.07) lower odds of BC in participants, respectively. Also, the odds of BC were decreased by 33% (95%CI difference : -0.63, -0.03) with each 50 kcal and by 66% (95%CI difference : -1.26, -0.07) with each 100 kcal replacement of plant protein instead of refined CHO. Regarding the replacement of other macronutrients with each other, no significant was observed.

Figure  1 shows the association of 6 subgroups of macronutrient intake with the odds of BC in our study population. After adjusting for potential confounding variables, including age, first pregnancy age, menopausal status, family history of cancer, anti-inflammatory drug use, vitamin D supplementation, physical activity, body mass index, socio-economic status, and energy intake, participants in the highest tertile of plant protein intake had 50% (OR: 0.50, 95%CI: 0.29–0.89, P-trend: 0.018) decreased odds of BC than those in the lowest tertile. However, no significant association was observed between refined-CHO, non-refined-CHO, plant fats, animal fats, and animal proteins and the risk of BC.

figure 1

The odds ratio (95% confidence interval) of breast cancer according to tertiles of 6 subgroups of macronutrient intake using the multivariable model adjusted for age, first pregnancy age, menopausal status, family history of cancer, anti-inflammatory drug use, vitamin D supplementation, physical activity, body mass index, socio-economic status, and energy intake

The present study showed that higher adherence to diets with high-protein-low-CHO and fat score and high-protein and CHO-low-fat score were inversely associated with odds of BC. However, high-protein and fat-low CHO diet was not related to BC risk. In addition, our substitution analysis reported that substituting fats with protein or replacing refined CHO with plant protein in the diet of participants may contribute to decreasing the risk of BC.

To the best of our knowledge, our investigation is the first to examine the association between different high protein diet scores and the risk of BC in the framework of a case-control study among Iranian adults. In recent decades, many studies investigated the relationship between dietary intake of macronutrients and the risk of BC. In accordance with our findings, several randomized clinical trials (RCTs) reported that a low-fat dietary pattern significantly reduced the incidence of ovarian cancer [ 36 ] and the risk of BC mortality [ 37 ], however, other RCTs observed no significant association between a low-fat dietary pattern with risk of breast [ 38 ] and colorectal cancer [ 39 ]. In line with our study results, an observational study conducted in the US population shows that higher plant protein intake especially protein from vegetables was associated with lower BC incidence [ 24 ]. Also, the findings of the present, are in agreement with the results of a large prospective study that has reported higher intake of plant protein was associated with decreased cardiovascular and total mortality, and plant protein intake instead of red meat protein or processed meat protein was related to lower cancer-related, cardiovascular-related and total mortality [ 40 ]. Also, aligned with our findings, a meta-analysis of prospective studies reported higher total protein intake was related to a decreased risk of all-cause mortality, and plant protein intake was associated with decreased risk of all-cause and cardiovascular disease mortality. Also, this study shows that consumption of plant protein sources instead of animal protein could be associated with longevity [ 41 ]. Furthermore, an animal study shows that a diet with low-CHO, and high-protein decreases tumor growth and prevents cancer initiation [ 42 ]. However, contrary to our results, in the Nilsson. et al. study no significant association was observed between the higher low-CHO, and high-protein scores and the risk of BC [ 43 ]. Also, Chow et al. reported high protein intake has been related to the development of other chronic renal conditions that may increase renal cell cancer. Furthermore, two cohort studies indicated a low CHO-high protein diets were associated with increased risk of cardiovascular diseases [ 44 ] and total mortality [ 45 ] in Swedish women.

In the current study, we showed that replacing refined CHO with plant protein and replacing fats with protein in the diet can reduce the risk of BC. In other words, our findings support the idea that if a high-protein diet is based on plant-based food choices, it can be useful in preventing the occurrence of chronic diseases, such as BC; This is because a high-protein- low-fat diet based on plant-based food intakes emphasizes high consumption of legumes, whole grains, seeds, nuts, and soy products, and lower consumption of red and processed meat, refined grains, sweetened beverages, and high-fat foods. Adhering to this dietary pattern means high intakes of plant proteins, micronutrients with antioxidant properties, and fiber and less intakes of saturated fat, simple sugar, and salt. Such a dietary pattern can provide antioxidant [ 46 ], anti-inflammatory [ 47 ], and anti-insulin resistance properties [ 48 ] that can help protect individuals against the risk of BC [ 42 ]; therefore, it is expected that if individuals follow a high-protein-low-fat and CHO diet based on plant food choices, they can better protect themselves against the risk of BC. It should be noted that plant food sources containing high protein, such as whole grains, legumes, and seeds, are considered rich sources of complex CHOs. Therefore, in this case, a plant-based high-protein diet can be regarded as a high-CHO diet, all while still being known as a low-fat diet. The relationship between this type of diet and the risk of chronic diseases, such as BC, may also be inverse; our study confirms this claim, as we discovered that following a high-protein and CHO-low fat diet can be linked to a reduced risk of BC.

The mechanisms underlying the role of diet with different high protein scores and BC risk are not yet fully understood. It seems a diet with a high protein and low-fat score especially protein derived from plant sources, regardless of CHO intake, leads to a greater intake of fiber, vitamins, minerals, and polyphenols which can be effective in decreasing the risk of BC [ 10 , 49 , 50 , 51 ]. Simple CHOs as a possible factor in increasing the BC risk, make up the majority of CHO intake among the Iranian population and replace it with plant proteins such as legumes that have possible anti-carcinogenic effects [ 52 ]. The reason for this result may be that plant foods such as soy contain fibers and phytoestrogens that induce apoptosis [ 53 ]. Animal protein intake is usually associated with the intake of fat which may throughout carcinogenic heterocyclic amines lead to an increased risk of cancer [ 54 ]. Also, Taha et al. hypothesized that a diet with high casein might increase the progression of cancer cells in mice through the activation of the IGF/Akt/mTOR pathway [ 55 ]. Also, it seems proteins from animal sources throughout increased insulin-like growth factor-1 [ 56 ], and the expression of Ras homologous gene family member A and vascular endothelial growth factor receptor-2 [ 57 ] lead to tumor progression. However, plant protein intake was inversely associated with RhoA expression [ 57 ].

Our study has several strengths. To our knowledge, this is the first study on the association between different high-protein diet scores (high-protein-low-CHO and fat diet, high-protein and CHO-low-fat diet, and high-protein and fat-low-CHO diet) and the risk of BC in the Iranian population. Also, this study included a substitution analysis in which we examined the effects of substituting protein with other macronutrients on BC risk while controlling for potential confounding variables. This represents the first time this type of analysis has been conducted in relation to protein intake and BC. In addition, we used validated questionnaires to collect individual data on dietary intake and physical activity levels, which minimized the possibility of recall bias. We included the patients that newly diagnosed with BC (< 6 months) in the case group; therefore, the individuals included in the current study possibly had not changed their usual lifestyle (including diet and physical activity) due to their chronic illness. Finally, we tried to control the effect of various potential confounding variables in assessing the relationship between different high-protein diet scores and the risk of BC, as much as possible. Some limitations of the present study should be reported. First, recall bias and selection bias are difficult to avoid in case-control studies. Second, regarding the nature of case-control design, investigating the causality relationship in this study is impossible. Third, since alcoholic drinks such as wine and beer are not common or may be unreported in the Iranian population due to religious considerations and legal restrictions; therefore we could not determine participants’ data for alcohol consumption, which could have played a confounding role in the present study. Fourth, the effect of some variables that were unknown to us may not have been controlled in our statistical analyses. Finally, using questionnaires to collect dietary intake and physical activity information may cause measurement errors and recall bias, but to decline the errors we used validated and reliable questionnaires which were specially developed for the Iranian population.

In conclusion, our findings revealed that a high-protein-low-CHO and fat diet based on plant-based food choices can be associated with a reduced risk of BC. Also, the current study suggested that higher plant protein intake especially instead of fats and refined CHO in individuals’ diet can be considered for prevention of BC risk. Further dietary intervention trials, prospective and longitudinal studies are recommended to address the role of the different high protein diet scores in the prediction of BC risk and the mechanisms justifying this possible relationship. This finding is very important since it can help public health and be considered as a recommendation in dietary guidelines which a diet with high-protein-low-CHO and fat diet based on plant-based food choices, can easily prevent the occurrence of BC.

Availability of data and materials

The datasets analyzed in the current study are available from the corresponding author on reasonable request.

Abbreviations

  • Breast cancer

Body mass index

Carbohydrate

Confidence interval

Food composition table

Food frequency questionnaire

Hormone replacement therapy

Metabolic Equivalents

Oral contraceptive pill

Randomized clinical trial

Standard deviation

Socio-economic status

Statistical Package for Social Sciences

United States Department of Agriculture

Sun YS, Zhao Z, Yang ZN, Xu F, Lu HJ, Zhu ZY, et al. Risk factors and preventions of breast Cancer. Int J Biol Sci. 2017;13(11):1387–97.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Fouladi N, Amani F, Harghi AS, Nayebyazdi N. Five year survival of women with breast cancer in Ardabil, north-west of Iran. Asian Pac J Cancer Prev. 2011;12(7):1799–801.

PubMed   Google Scholar  

Abedi G, Janbabai G, Moosazadeh M, Farshidi F, Amiri M, Khosravi A. Survival rate of breast Cancer in Iran: a Meta-analysis. Asian Pac J Cancer Prev. 2016;17(10):4615–21.

PubMed   PubMed Central   Google Scholar  

Mousavi SM, Montazeri A, Mohagheghi MA, Jarrahi AM, Harirchi I, Najafi M, et al. Breast cancer in Iran: an epidemiological review. Breast J. 2007;13(4):383–91.

Article   PubMed   Google Scholar  

DeSantis CE, Fedewa SA, Goding Sauer A, Kramer JL, Smith RA, Jemal A. Breast cancer statistics, 2015: convergence of incidence rates between black and white women. CA Cancer J Clin. 2016;66(1):31–42.

Radice D, Redaelli A. Breast Cancer Manage PharmacoEconomics. 2003;21(6):383–96.

Arnold M, Morgan E, Rumgay H, Mafra A, Singh D, Laversanne M, et al. Current and future burden of breast cancer: global statistics for 2020 and 2040. Breast. 2022;66:15–23.

Article   PubMed   PubMed Central   Google Scholar  

Majeed W, Aslam B, Javed I, Khaliq T, Muhammad F, Ali A, et al. Breast cancer: major risk factors and recent developments in treatment. Asian Pac J Cancer Prev. 2014;15(8):3353–8.

Liu FC, Lin HT, Kuo CF, See LC, Chiou MJ, Yu HP. Epidemiology and survival outcome of breast cancer in a nationwide study. Oncotarget. 2017;8(10):16939–50.

Chen P, Li C, Li X, Li J, Chu R, Wang H. Higher dietary folate intake reduces the breast cancer risk: a systematic review and meta-analysis. Br J Cancer. 2014;110(9):2327–38.

Braakhuis AJ, Campion P, Bishop KS. Reducing breast Cancer recurrence: the role of Dietary Polyphenolics. Nutrients. 2016;8(9).

Aune D, Chan DS, Vieira AR, Rosenblatt DA, Vieira R, Greenwood DC, et al. Fruits, vegetables and breast cancer risk: a systematic review and meta-analysis of prospective studies. Breast Cancer Res Treat. 2012;134(2):479–93.

Article   CAS   PubMed   Google Scholar  

Alzahrani MA, Shakil Ahmad M, Alkhamees M, Aljuhayman A, Binsaleh S, Tiwari R, et al. Dietary protein intake and prostate cancer risk in adults: a systematic review and dose-response meta-analysis of prospective cohort studies. Complement Ther Med. 2022;70:102851.

Kong F, Geng E, Ning J, Liu Z, Wang A, Zhang S et al. The association between dietary protein intake and esophageal cancer risk: a meta-analysis. Biosci Rep. 2020;40(1).

Sieri S, Agnoli C, Pala V, Grioni S, Palli D, Bendinelli B, et al. Dietary intakes of animal and plant proteins and risk of Colorectal Cancer: the EPIC-Italy Cohort. Cancers. 2022;14(12):2917.

Nilsson LM, Winkvist A, Johansson I, Lindahl B, Hallmans G, Lenner P, et al. Low-carbohydrate, high-protein diet score and risk of incident cancer; a prospective cohort study. Nutr J. 2013;12:58.

Chow WH, Gridley G, McLaughlin JK, Mandel JS, Wacholder S, Blot WJ, et al. Protein intake and risk of renal cell cancer. J Natl Cancer Inst. 1994;86(15):1131–9.

Ax E, Garmo H, Grundmark B, Bill-Axelson A, Holmberg L, Becker W, et al. Dietary patterns and prostate cancer risk: report from the population based ULSAM cohort study of Swedish men. Nutr Cancer. 2014;66(1):77–87.

Nilsson LM, Winkvist A, Eliasson M, Jansson JH, Hallmans G, Johansson I, et al. Low-carbohydrate, high-protein score and mortality in a northern Swedish population-based cohort. Eur J Clin Nutr. 2012;66(6):694–700.

Rohan TE, Hiller JE, McMichael AJ. Dietary factors and survival from breast cancer. Nutr Cancer. 1993;20(2):167–77.

Borugian MJ, Sheps SB, Kim-Sing C, Van Patten C, Potter JD, Dunn B, et al. Insulin, macronutrient intake, and physical activity: are potential indicators of insulin resistance associated with mortality from breast cancer? Cancer Epidemiol Biomarkers Prev. 2004;13(7):1163–72.

Goodwin PJ, Ennis M, Pritchard KI, Koo J, Trudeau ME, Hood N. Diet and breast cancer: evidence that extremes in diet are associated with poor survival. J Clin Oncol. 2003;21(13):2500–7.

Holmes MD, Stampfer MJ, Colditz GA, Rosner B, Hunter DJ, Willett WC. Dietary factors and the survival of women with breast carcinoma. Cancer. 1999;86(5):826–35.

Pan K, Larson JC, Prentice RL, Mortimer JE, Neuhouser ML, Manson JE et al. Protein intake by source and breast Cancer incidence and mortality: the women’s Health Initiative. JNCI Cancer Spectr. 2020;4(6).

Taylor EF, Burley VJ, Greenwood DC, Cade JE. Meat consumption and risk of breast cancer in the UK women’s Cohort Study. Br J Cancer. 2007;96(7):1139–46.

Asghari G, Rezazadeh A, Hosseini-Esfahani F, Mehrabi Y, Mirmiran P, Azizi F. Reliability, comparative validity and stability of dietary patterns derived from an FFQ in the Tehran lipid and glucose study. Br J Nutr. 2012;108(6):1109–17.

Ghaffarpour M, Houshiar-Rad A, Kianfar H. The manual for household measures, cooking yields factors and edible portion of foods. Tehran: Nashre Olume Keshavarzy. 1999;7(213):42–58.

Google Scholar  

Azar M, Sarkisian E. Food composition table of Iran: National Nutrition and food research institute. Tehran: Shaheed Beheshti University; 1980.

Rigi S, Mousavi SM, Benisi-Kohansal S, Azadbakht L, Esmaillzadeh A. The association between plant-based dietary patterns and risk of breast cancer: a case–control study. Sci Rep. 2021;11(1):3391.

Romanos-Nanclares A, Gea A, Martínez-González MÁ, Zazpe I, Gardeazabal I, Fernandez-Lazaro CI, et al. Carbohydrate quality index and breast cancer risk in a Mediterranean cohort: the SUN project. Clin Nutr. 2021;40(1):137–45.

Sieri S, Krogh V, Muti P, Micheli A, Pala V, Crosignani P, et al. Fat and protein intake and subsequent breast Cancer Risk in Postmenopausal Women. Nutr Cancer. 2002;42(1):10–7.

Aadahl M, Jørgensen T. Validation of a new self-report instrument for measuring physical activity. Med Sci Sports Exerc. 2003;35(7):1196–202.

Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc. 2000;32(9):S498–S504.

Lee PH, Macfarlane DJ, Lam TH, Stewart SM. Validity of the international physical activity questionnaire short form (IPAQ-SF): a systematic review. Int J Behav Nutr Phys Activity. 2011;8(1):1–11.

Article   Google Scholar  

Garmaroudi GR, Moradi A. Socio-economic status in Iran: a study of measurement index. Payesh (Health Monitor). 2010;9(2):137–44.

Prentice RL, Thomson CA, Caan B, Hubbell FA, Anderson GL, Beresford SAA et al. Low-Fat Dietary Pattern and Cancer Incidence in the Women’s Health Initiative Dietary Modification Randomized Controlled Trial. JNCI: Journal of the National Cancer Institute. 2007;99(20):1534-43.

Chlebowski RT, Aragaki AK, Anderson GL, Thomson CA, Manson JE, Simon MS, et al. Low-fat dietary pattern and breast cancer mortality in the women’s Health Initiative randomized controlled trial. J Clin Oncol. 2017;35(25):2919.

Prentice RL, Caan B, Chlebowski RT, Patterson R, Kuller LH, Ockene JK, et al. Low-fat dietary pattern and risk of invasive breast cancer: the women’s Health Initiative Randomized Controlled Dietary Modification Trial. JAMA. 2006;295(6):629–42.

Beresford SAA, Johnson KC, Ritenbaugh C, Lasser NL, Snetselaar LG, Black HR, et al. Low-Fat Dietary Pattern and Risk of Colorectal CancerThe women’s Health Initiative Randomized Controlled Dietary Modification Trial. JAMA. 2006;295(6):643–54.

Budhathoki S, Sawada N, Iwasaki M, Yamaji T, Goto A, Kotemori A, et al. Association of animal and plant protein intake with all-cause and cause-specific mortality in a Japanese cohort. JAMA Intern Med. 2019;179(11):1509–18.

Naghshi S, Sadeghi O, Willett WC, Esmaillzadeh A. Dietary intake of total, animal, and plant proteins and risk of all cause, cardiovascular, and cancer mortality: systematic review and dose-response meta-analysis of prospective cohort studies. BMJ. 2020;370.

Ho VW, Leung K, Hsu A, Luk B, Lai J, Shen SY, et al. A low carbohydrate, high protein diet slows tumor growth and prevents cancer initiation. Cancer Res. 2011;71(13):4484–93.

Nilsson LM, Winkvist A, Johansson I, Lindahl B, Hallmans G, Lenner P, et al. Low-carbohydrate, high-protein diet score and risk of incident cancer; a prospective cohort study. Nutr J. 2013;12(1):58.

Lagiou P, Sandin S, Lof M, Trichopoulos D, Adami H-O, Weiderpass E. Low carbohydrate-high protein diet and incidence of cardiovascular diseases in Swedish women: prospective cohort study. BMJ. 2012;344.

Lagiou P, Sandin S, Weiderpass E, Lagiou A, Mucci L, Trichopoulos D, et al. Low carbohydrate–high protein diet and mortality in a cohort of Swedish women. J Intern Med. 2007;261(4):366–74.

Mullins AP, Arjmandi BH. Health benefits of Plant-Based Nutrition: Focus on beans in Cardiometabolic diseases. Nutrients. 2021;13(2):519.

Matsunaga M, Hurwitz EL, Li D. Adequate intake of plant protein foods and moderate intake of animal protein foods are inversely associated with C-reactive protein in US adults with diabetes: a cross-sectional study with National Health and Nutrition Examination Survey. Nutrition. 2021;89:111276.

Huang Y, Li X, Zhang T, Zeng X, Li M, Li H, et al. Healthy and unhealthy low-carbohydrate diets and plasma markers of cardiometabolic risk. Br J Nutr. 2023;130(1):137–46.

Naja F, Nasreddine L, Awada S, El Sayed Ahmad R, Hwalla N. Nutrition in the Prevention of breast Cancer: a Middle Eastern Perspective. Front Public Health. 2019;7:316.

Huang W-Q, Long W-Q, Mo X-F, Zhang N-Q, Luo H, Lin F-Y, et al. Direct and indirect associations between dietary magnesium intake and breast cancer risk. Sci Rep. 2019;9(1):5764.

Abdal Dayem A, Choi HY, Yang G-M, Kim K, Saha SK, Cho S-G. The anti-cancer effect of polyphenols against breast cancer and cancer stem cells: molecular mechanisms. Nutrients. 2016;8(9):581.

Sánchez-Chino X, Jiménez-Martínez C, Dávila-Ortiz G, Álvarez-González I, Madrigal-Bujaidar E. Nutrient and nonnutrient components of legumes, and its chemopreventive activity: a review. Nutr Cancer. 2015;67(3):401–10.

Mao Y, Tie Y, Du J. Association between dietary protein intake and prostate cancer risk: evidence from a meta-analysis. World J Surg Oncol. 2018;16(1):152.

Lumbreras B, Garte S, Overvad K, Tjonneland A, Clavel-Chapelon F, Linseisen JP, et al. Meat intake and bladder cancer in a prospective study: a role for heterocyclic aromatic amines? Cancer Causes Control. 2008;19(6):649–56.

Taha AAA, Koshiyama M, Matsumura N, Abiko K, Yamaguchi K, Hamanishi J, et al. The effect of the type of dietary protein on the development of ovarian cancer. Oncotarget. 2018;9(35):23987–99.

Klement RJ, Fink MK. Dietary and pharmacological modification of the insulin/IGF-1 system: exploiting the full repertoire against cancer. Oncogenesis. 2016;5(2):e193.

Shokri A, Pirouzpanah S, Foroutan-Ghaznavi M, Montazeri V, Fakhrjou A, Nozad-Charoudeh H, et al. Dietary protein sources and tumoral overexpression of RhoA, VEGF-A and VEGFR2 genes among breast cancer patients. Genes Nutr. 2019;14:22.

Download references

Acknowledgements

We appreciate the Shahid Beheshti University of Medical Sciences for their financial support of this study. The authors express their appreciation to all of the participants of this study.

This project was funded by a grant from the Shahid Beheshti University of Medical Sciences, Tehran, Iran. The funding body has no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

Author information

Mitra Kazemi Jahromi and Hossein Farhadnejad contributed equally to this work.

Authors and Affiliations

Endocrinology and Metabolism Research Center, Hormozgan University of Medical Sciences, Bandar Abbas, Iran

Mitra Kazemi Jahromi

Nutrition and Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran

Hamid Ahmadirad, Hossein Farhadnejad, Mostafa Norouzzadeh, Ebrahim Mokhtari, Farshad Teymoori, Niloufar Saber & Parvin Mirmiran

Department of Community Nutrition, School of Nutrition and Food Sciences, Shiraz University of Medical Sciences, Shiraz, Iran

Ebrahim Mokhtari

Department of Nutrition, School of Public Health, Iran University of Medical Sciences, Tehran, Iran

Farshad Teymoori

Student Research Committee, Ahvaz Jundishapur University of Medical Sciences, Ahvaz, Iran

Zeinab Heidari

Department of Community Nutrition, Faculty of Nutrition Sciences and Food Technology, National Nutrition and Food Technology Research Institute, Shahid Beheshti University of Medical Sciences, Tehran, Iran

Bahram Rashidkhani

You can also search for this author in PubMed   Google Scholar

Contributions

MKJ and FT contributed to conceptualizing and designing the current study. FT, ZH, and HF analyzed and interpreted the data. MKJ, HA, EM, MN, and NS drafted the initial manuscript. BR and PM supervised the project. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Farshad Teymoori or Bahram Rashidkhani .

Ethics declarations

Ethics approval and consent to participate.

Informed written consent was obtained from participants. All procedures performed in studies involving human participants adhered to the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The ethics research committee approved the study’s protocol of the Shahid Beheshti University of Medical Sciences, Tehran, Iran.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Jahromi, M.K., Ahmadirad, H., Farhadnejad, H. et al. High-protein diet scores, macronutrient substitution, and breast cancer risk: insights from substitution analysis. BMC Women's Health 24 , 121 (2024). https://doi.org/10.1186/s12905-024-02959-7

Download citation

Received : 30 October 2023

Accepted : 08 February 2024

Published : 16 February 2024

DOI : https://doi.org/10.1186/s12905-024-02959-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Dietary patterns
  • High-protein

BMC Women's Health

ISSN: 1472-6874

breast cancer case control study

Sex-steroid hormones and risk of postmenopausal estrogen receptor-positive breast cancer: a case–cohort analysis

  • Original Paper
  • Open access
  • Published: 16 February 2024

Cite this article

You have full access to this open access article

  • Frances E. M. Albers   ORCID: orcid.org/0000-0002-7319-5182 1 , 2 ,
  • Makayla W. C. Lou   ORCID: orcid.org/0000-0002-4518-5219 1 , 2 ,
  • S. Ghazaleh Dashti   ORCID: orcid.org/0000-0002-1399-7220 3 , 4 ,
  • Christopher T. V. Swain   ORCID: orcid.org/0000-0002-0158-2511 2 , 5 ,
  • Sabina Rinaldi   ORCID: orcid.org/0000-0002-6846-1204 6 ,
  • Vivian Viallon   ORCID: orcid.org/0000-0002-9799-4421 6 ,
  • Amalia Karahalios   ORCID: orcid.org/0000-0002-7497-1681 1 ,
  • Kristy A. Brown   ORCID: orcid.org/0000-0003-3382-5546 7 ,
  • Marc J. Gunter   ORCID: orcid.org/0000-0001-5472-6761 6 , 8 ,
  • Roger L. Milne   ORCID: orcid.org/0000-0001-5764-7268 1 , 2 , 9 ,
  • Dallas R. English   ORCID: orcid.org/0000-0001-7828-8188 1 , 2 &
  • Brigid M. Lynch   ORCID: orcid.org/0000-0001-8060-547X 1 , 2 , 10  

Sex-steroid hormones are associated with postmenopausal breast cancer but potential confounding from other biological pathways is rarely considered. We estimated risk ratios for sex-steroid hormone biomarkers in relation to postmenopausal estrogen receptor (ER)-positive breast cancer, while accounting for biomarkers from insulin/insulin-like growth factor-signaling and inflammatory pathways.

This analysis included 1208 women from a case–cohort study of postmenopausal breast cancer within the Melbourne Collaborative Cohort Study. Weighted Poisson regression with a robust variance estimator was used to estimate risk ratios (RRs) and 95% confidence intervals (CIs) of postmenopausal ER-positive breast cancer, per doubling plasma concentration of progesterone, estrogens, androgens, and sex-hormone binding globulin (SHBG). Analyses included sociodemographic and lifestyle confounders, and other biomarkers identified as potential confounders.

Increased risks of postmenopausal ER-positive breast cancer were observed per doubling plasma concentration of progesterone (RR: 1.22, 95% CI 1.03 to 1.44), androstenedione (RR 1.20, 95% CI 0.99 to 1.45), dehydroepiandrosterone (RR: 1.15, 95% CI 1.00 to 1.34), total testosterone (RR: 1.11, 95% CI 0.96 to 1.29), free testosterone (RR: 1.12, 95% CI 0.98 to 1.28), estrone (RR 1.21, 95% CI 0.99 to 1.48), total estradiol (RR 1.19, 95% CI 1.02 to 1.39) and free estradiol (RR 1.22, 95% CI 1.05 to 1.41). A possible decreased risk was observed for SHBG (RR 0.83, 95% CI 0.66 to 1.05).

Progesterone, estrogens and androgens likely increase postmenopausal ER-positive breast cancer risk, whereas SHBG may decrease risk. These findings strengthen the causal evidence surrounding the sex-hormone-driven nature of postmenopausal breast cancer.

Avoid common mistakes on your manuscript.

Breast cancer is a largely hormone-driven disease and the relationships between endogenous sex-steroid hormones – especially estrogens – and postmenopausal breast cancer are thought to be well established [ 1 , 2 , 3 ]. A recent systematic review and meta-analysis found moderate- to high-quality evidence that higher levels of estrogens and androgens, and lower levels of sex-hormone binding globulin (SHBG), were associated with increased risks of postmenopausal breast cancer [ 4 ]. The quality of the evidence in this review was largely determined by dose–response effects and large effect sizes [ 4 ]. No extracted result had adjusted for biomarkers from other biological pathways; namely, the insulin/insulin-like growth factor (IGF)-signaling and inflammatory pathways. These pathways may confound the effect of the sex-steroid hormone pathway (Fig.  1 ). For example, insulin and insulin-like growth factor-1 (IGF-1) can affect the bioavailability of estrogens and androgens via the regulation of aromatase and suppression of hepatic SHBG production [ 1 , 5 , 6 ]. They may also play a role in breast carcinogenesis: insulin and the IGF axis are proposed to have mitogenic and anti-apoptotic properties, and higher systemic concentrations of IGF-1 are associated with increased risks of breast cancer [ 1 , 2 , 5 , 6 , 7 , 8 ]. Further, a state of low-grade chronic inflammation – for example, in the context of physical inactivity and obesity – can foster a pro-carcinogenic environment via the overstimulation and dysregulation of immune cells, cytokines and adipokines [ 1 , 2 , 5 , 6 , 9 ]. Higher circulating levels of C-reactive protein (CRP) – a non-specific marker of chronic inflammation – are associated with increased risks of breast cancer, but the epidemiological evidence for other inflammatory markers remains uncertain [ 2 , 10 , 11 ]. Higher circulating levels of pro-inflammatory biomarkers including leptin, tumor necrosis factor-alpha (TNF-α) and interleukin-6 (IL-6) are also associated with enhanced aromatase activity and lower circulating levels of SHBG [ 1 , 2 , 5 , 6 , 9 , 12 ].

figure 1

Diagram of the assumed interrelationships between the inflammatory, insulin/insulin-like growth factor-signaling, and sex-steroid hormone pathways. SHBG sex-hormone binding globulin, CRP C-reactive protein, IGF insulin-like growth factor, IGF-1 Insulin-like growth factor-1

The aim of this study was to estimate risk ratios for sex-steroid hormone biomarkers in relation to postmenopausal breast cancer in a case-cohort of postmenopausal women within the Melbourne Collaborative Cohort Study (MCCS), while accounting for other biomarkers from the insulin/IGF-signaling and inflammatory pathways.

The Melbourne Collaborative Cohort Study

The MCCS includes 24,469 women aged 40–69 at recruitment from 1990 to 1994 [ 13 ]. At baseline and the second follow-up (F2, 2003–7), participants provided information about health status, lifestyle factors, sociodemographics and medical history via structured questionnaires [ 13 ]. Anthropometric and clinical measurements were performed at the study center, including the collection of blood samples [ 13 ]. At both times, plasma was stored in liquid nitrogen. Data linkages to national and state death and cancer registries—including the Victorian Cancer Registry and Australian Cancer Database—enabled vital status and cancer diagnoses to be determined prospectively [ 13 ]. The study protocol was approved by the Cancer Council Victoria Human Research Ethics Committee.

The case-cohort study

Initial eligibility criteria at second follow-up (2003–7).

This case-cohort study was restricted to women who attended F2. At F2, eligible women were postmenopausal, not known to be taking hormone replacement therapy (HRT), had provided a blood sample (within one year of the F2 questionnaire, if completed), had no prior invasive cancer diagnosis (except for keratinocyte cancers); at baseline, they had a body mass index (BMI) ≥ 18.5 kg/m 2 . Women were considered postmenopausal if they had had no menstrual periods in the past 12 months and met one of the following criteria: had experienced natural cessation of menses; had a bilateral oophorectomy; were age 55 years or older; or had had no periods in the 12 months prior to baseline and, for participants in a previous case-cohort study, measured estradiol concentration below 109 pmol/L at baseline (a threshold from that study [ 14 , 15 ]). The case-cohort comprised a random sample of the 10,669 eligible women and all eligible women diagnosed with estrogen receptor (ER)-positive postmenopausal breast cancer between blood collection at F2 and 31 October 2020.

An eligible tumor was defined as invasive adenocarcinoma of the breast (International Classification of Diseases, Tenth Revision [ICD-10] code C50) that was ER-positive. Tumors of unknown hormone receptor status were included as 88% of breast cancer diagnoses among eligible women of known ER status were ER-positive. ER-negative and progesterone receptor (PR)-positive cancers were also included as this tumor subtype may be misclassified and accounts for only 1–4% of diagnoses [ 16 , 17 , 18 , 19 ]. Unspecified adenocarcinomas and unspecified cancers were presumed to be adenocarcinomas as 99% of breast cancer diagnoses among eligible women of known morphology were adenocarcinomas.

In total, 1,412 women were selected for the case-cohort study, including 999 in the subcohort and 459 cases (46 from the subcohort) (Fig.  2 ). The subcohort was a random sample of eligible women (Online Resource 1).

figure 2

Selection of participants into the case-cohort study and analyses. N Number, F2 second follow-up wave, HRT hormone replacement therapy, BMI body mass index (kg/m. 2 )

Post-hoc criteria

Of the 1,412 selected women, 286 (20%) had unknown menopausal status and/or HRT use. Eligibility was confirmed for all selected women using the distribution of measured estradiol values at F2 for naturally postmenopausal women who were not taking HRT (806, 57% of selected women). Thirty-two women with estradiol values at or above the 99th percentile of this distribution (29.3 pg/mL, equivalent to 107.6 pmol/L) were excluded, regardless of age, menopausal status, or HRT use. Menopausal status and/or HRT use could not be determined for six women missing estradiol measurements. One woman was excluded as she did not participate in F2 despite providing a blood sample.

Four cases outside the subcohort were retrospectively disqualified as cases; three diagnoses were ascertained from death certificate only and one woman was diagnosed with non-adenocarcinoma breast cancer. To minimize the impact of death as a competing risk, follow-up was chosen to end on participants’ 86th birthday (Online Resource 2). Thus, 44 cases outside the subcohort were excluded and eight cases within the subcohort were analyzed as non-cases. Thirteen users of exogenous insulin were excluded so that measured insulin concentrations were of endogenous insulin.

The total study sample after post-hoc exclusions comprised 1,312 women, 969 in the subcohort and 378 cases (35 also in the subcohort) (Fig.  2 ).

Laboratory analysis of plasma biomarkers

Plasma samples of selected women were randomly ordered and allocated into 21 batches containing approximately equal numbers of cases. The samples were shipped on dry ice in two dispatches to the International Agency for Research on Cancer (IARC).

The plasma concentrations of all biomarkers were measured at the Nutrition Metabolism Branch, IARC. Plasma concentrations of sex-steroid hormones and SHBG were measured as previously described [ 20 ]. In brief, sex-steroid hormone concentrations were measured using a liquid chromatography-mass spectrometry system consisting of an ultra-high-performance liquid chromatograph (Agilent 1290, Agilent, Santa Clara, CA) and a QTRAP 5500 mass spectrometer (SCIEX, Framingham, MA). SHBG concentrations were measured by solid-phase “sandwich” enzyme-linked immunoassay (DRG International, Springfield, NJ). Interferon gamma (IFN-γ), IL-6, interleukin-8 (IL-8), interleukin-10 (IL-10), TNF-α, insulin, adiponectin, leptin, and CRP were measured by highly-sensitive and highly-specific electrochemiluminescent methods (Meso Scale Discovery, Rockville, MD). IGF-1 and insulin-like growth factor binding protein-3 (IGFBP-3) were measured by immunoassay methods by R&D Systems (Biotechne, Minneapolis, USA). C-peptide was measured by an enzyme-linked immunosorbent assay by ALPCO (Salem, USA). Further details are included in Online Resource 3. Three quality control samples at different concentration levels were measured in duplicate in each batch of analyses to assess the reliability of biomarker measurements. Reliability was assessed by calculating intra-assay and inter-assay coefficients of variation (CVs), as well as intra-batch and inter-batch intra-class correlation coefficients (ICCs), as described in Online Resource 4. Assay performance for estradiol and testosterone was evaluated by measuring samples created from reference standards with known concentrations. Measured values were compared with true values using validity coefficients and correlation plots, as described in Online Resource 5.

Normalization of biomarker values

Biomarker data were cleaned and normalized to correct for effects of batch, dispatch, and time since last meal (12% of study participants were not fasting at blood collection). The normalization technique was adapted from Viallon et al. [ 21 ]. Normalization models were used to estimate residual ICCs for the total proportion of variation attributable to batch for each biomarker. Methods for normalization and estimated ICCs are presented in Online Resource 6.

Calculation of free estradiol and free testosterone

Concentrations of free estradiol and free testosterone (i.e., not bound to SHBG) were calculated from normalized values of estradiol, testosterone and SHBG using the law of mass action assuming a fixed albumin concentration of 40 g/L (5.97 × 10 –4  mol/L) and the following association constants: 6 × 10 4 L/mol (binding of estradiol to albumin); 4 × 10 4 L/mol (binding of testosterone to albumin); 0.68 × 10 9 L/mol (binding of estradiol to SHBG); 1.6 × 10 9 L/mol (binding of testosterone to SHBG) [ 22 , 23 , 24 , 25 ].

Statistical analysis

Descriptive statistics were presented as medians and interquartile ranges (IQRs) or as frequencies and percentages, where appropriate. Weighted modified Poisson regression with a robust variance estimator was used to estimate risk ratios (RRs) and 95% confidence intervals (CIs) of postmenopausal ER-positive breast cancer, per doubling plasma concentration of progesterone, androstenedione, DHEA, total and calculated free testosterone, estrone, total and calculated free estradiol, and SHBG. Poisson regression models were weighted to account for the oversampling of cases, which can be considered a stratified form of sampling in which stratification depends on the outcome [ 26 ]. Case weights were one, and weights for non-cases were the inverse of the sampling probability for non-cases [ 26 ]. The latter was calculated as the number of non-cases in the eligible cohort divided by the number of non-cases in the subcohort ([10,669—459] /953).

Confounders including other biomarkers were identified a priori using causal diagrams informed by expert consensus and literature review. Sociodemographic and lifestyle confounders included: education; country of birth; socioeconomic disadvantage; diet at baseline (dietary intake of carotenoids and dietary intake of calcium); alcohol consumption at baseline; smoking status at baseline; adiposity at baseline; physical activity at F2; age at blood collection; and age at menopause. The identification, measurement and modelling of sociodemographic and lifestyle confounders are described in Online Resource 7. As age at menopause could only be measured for naturally postmenopausal women (821, 63% of the case-cohort after post-hoc exclusions), this variable was not included in the adjustment set for the primary analyses. Sensitivity analyses were conducted, restricting to naturally postmenopausal women with a recorded age at menopause to include this variable in adjustment sets. Biomarkers that were identified as potential confounders a priori but had correlations ≥ 0.50 with the biomarker of interest were not included in the primary analysis (Online Resource 8).

The primary analyses modelled all biomarker concentrations as continuous, normalized values on the log 2 -scale. A one unit increase on the log 2 -scale represents a doubling in biomarker concentration. Analyses were repeated without adjustment for other biomarkers (where applicable). Sensitivity analyses excluding ER-negative/PR-positive tumors and tumors of unknown hormone receptor status were conducted to test the assumption that these tumor subtypes were ER-positive. Additional sensitivity analyses excluded all cases diagnosed within one year of blood draw at F2 to assess the potential impact of reverse causation. These analyses also excluded deaths that occurred within this time frame to be consistent with the target trial approach [ 27 ]. In addition, analyses that modelled concentrations of each sex-steroid hormone biomarker as quartiles corresponding to the distribution of normalized biomarker values in the subcohort were performed without adjustment for other biomarkers.

All analyses were complete-case analyses. The linearity assumption was tested for the continuous, normalized biomarker values using restricted cubic splines and Wald-tests. All statistical analyses were performed using Stata 16 (StataCorp, College Station, TX).

Of the 1312 women eligible after post-hoc exclusions, 87 were excluded due to missing sociodemographic and lifestyle confounder data (Fig.  2 ). In addition, 17 women were excluded due to missing measurements for all sex-steroid hormone biomarkers. The characteristics of the remaining 1,208 women are summarized in Table  1 . Compared with non-cases, cases were more likely to be educated, have obesity, and experience the menopause at ≥ 53 years, and were less likely to be sufficiently active. The normalized concentrations of DHEA, total estradiol, free estradiol, leptin and CRP were higher, and the normalized concentration of SHBG was lower, for cases compared with non-cases. The characteristics of the 1,312 women eligible after post-hoc exclusions were not appreciably different from the 1,208 women analyzed (Online Resource 9).

Of the 342 cases eligible for complete-case analysis, 324 (95%) were ER-positive, 5 were ER-negative/PR-positive, and 13 were of unknown hormone receptor status. Twenty cases were diagnosed within one year of blood draw at F2, and 5 subcohort non-cases had died.

Reliability of biomarker measurements and assay performance

The calculated overall intra-assay and inter-assay CVs were below 10% and 15%, respectively, for most biomarkers (Online Resource Table 4.1). The estimated intra-batch and inter-batch reliability ICCs were above 80% and 70%, respectively, for most biomarkers (Online Resource Table 4.2). The validity coefficients for the true and measured values of estradiol and testosterone were 0.987 and 0.997, respectively. Correlation plots are presented in Online Resource 5.

Risk ratios per doubling of biomarker concentration

For the primary analyses, increased risks of postmenopausal ER-positive breast cancer were observed per doubling plasma concentration of progesterone (RR 1.22, 95% CI 1.03 to 1.44), androstenedione (RR 1.20, 95% CI 0.99 to 1.45), DHEA (RR 1.15, 95% CI 1.00 to 1.34), total testosterone (RR 1.11, 95% CI 0.96 to 1.29), calculated free testosterone (RR: 1.12, 95% CI: 0.98 to 1.28), estrone (RR: 1.21, 95% CI: 0.99 to 1.48), total estradiol (RR: 1.19, 95% CI: 1.02 to 1.39) and calculated free estradiol (RR: 1.22, 95% CI: 1.05 to 1.41) (Table  2 ). A decreased risk was suggested for SHBG (RR: 0.83, 95% CI: 0.66 to 1.05).

Results did not appreciably differ in analyses without adjustment for other biomarkers (Table  2 ), except that the inverse association for SHBG was somewhat weaker (RR: 0.90, 95% CI: 0.73 to 1.11). Point estimates for RR for all sex-steroid hormone biomarkers were slightly stronger when ER-negative/PR-positive tumors and tumors of unknown hormone receptor status were excluded (Online Resource 10). The results of the sensitivity analyses excluding cases and deaths that occurred within one year of blood draw at F2 were similar to the results of the primary analyses (Online Resource 11).

For the sensitivity analyses in the subset of naturally postmenopausal women with a recorded age at menopause (Online Resource 12), the point estimates for RR were closer to the null for progesterone (RR 1.11, 95% CI 0.90 to 1.36) and androstenedione (RR 1.08, 95% CI 0.85 to 1.39), and further away from the null for estrone (RR 1.30, 95% CI 0.99 to 1.69), total estradiol (RR 1.29, 95% CI 1.04 to 1.58) and calculated free estradiol (RR 1.31, 95% CI 1.08 to 1.60). Results with and without adjustment for age at menopause were similar, whereas the point estimates for RR without adjustment for other biomarkers were closer to the null for estrone, free estradiol and SHBG (Online Resource 12).

Risk ratios for quartiles of biomarker concentration

The highest versus lowest levels of biomarker concentrations were associated with increased risks of postmenopausal ER-positive breast cancer for progesterone (RR 1.56, 95% CI 1.09 to 2.24), androstenedione (RR 1.39, 95% CI 0.97 to 2.00), DHEA (RR 1.55, 95% CI 1.06 to 2.25), total estradiol (RR 1.49, 95% CI 1.01 to 2.19) and calculated free estradiol (RR 1.47, 95% CI 0.99 to 2.17) (Table  3 ). RRs were suggestive of monotonic increases for DHEA, estrone and total estradiol. In contrast, the positive relationship between calculated free estradiol and postmenopausal ER-positive breast cancer plateaued at the third-highest plasma concentration compared to the lowest.

Summary of principal findings

Higher plasma concentrations of progesterone, estrogens and androgens, and decreasing plasma concentration of SHBG, were associated with increased risks of postmenopausal ER-positive breast cancer in this case-cohort of postmenopausal women. Similar results were obtained with and without control for other biomarkers that were identified as potential confounders, suggesting that confounding by the insulin/IGF-signaling and inflammatory pathways was minimal. The exception was SHBG; a somewhat stronger inverse relationship was observed with adjustment for adiponectin, leptin, insulin and IGF-1. Slightly stronger associations were observed when cases that were ER-negative/PR-positive or of unknown hormone receptor status were excluded, indicating that some cases that were assumed to be ER-positive may not have truly been ER-positive. The impact of reverse causation was negligible. Results of the sensitivity analyses in the subset of naturally postmenopausal women with a recorded age at menopause were not sensitive to adjustment for age at menopause. Rather, the deviations observed from the primary analyses could be explained by reduced precision in the subsample, or differences between women who were naturally postmenopausal (with a known age at menopause) and women who were assumed to be postmenopausal for other reasons.

Strengths and limitations

A strength of our study was that careful consideration was given to biomarkers from the insulin/IGF-signaling and inflammatory pathways that may confound the associations between biomarkers of the sex-steroid hormone pathway and risk of postmenopausal ER-positive breast cancer. Biomarkers that may be potential confounders were identified a priori using a causal diagram that was informed by literature review and expert opinion. Causal diagrams can minimize the pitfalls of other confounder selection methods, including overadjustment bias [ 28 , 29 , 30 ]. However, residual confounding may remain if our assumptions are inaccurate or if important confounders have not been identified or correctly measured [ 28 , 30 ]. Depicting the true complexity of biomarker interrelationships and their role in breast carcinogenesis is challenging. The current body of causal knowledge is limited, and we could not account for bidirectional relationships as the biomarkers had only been measured at one point in time. Thus, we assumed what the net direction of the effects of the measured biomarkers would be in a relatively older cohort of postmenopausal women in our causal diagram. Our assumptions can be refined with the advancement of causal knowledge over time, ideally in studies that measure biomarkers at multiple points in time.

A potential limitation of our study was that selection bias may have been introduced if there were systematic differences between women who did and did not attend F2 and provide a blood sample. However, the sociodemographic and lifestyle confounders that we have adjusted for in our analyses are likely to have included the critical determinants of participation at F2 (e.g., age, country of birth, socioeconomic disadvantage, education, smoking, alcohol, adiposity) [ 13 ]. Restriction to women who had not been diagnosed with breast cancer before F2 could potentially bias the RR for hormones towards the null, since women with high hormone concentrations at baseline were more likely to develop breast cancer [ 15 ]. Bias may also arise due to the exclusion of participants with missing data, although in this study, the proportion of eligible women with missing data was small (8%). Considered collectively, we expect selection bias to have minimal impact on our study conclusions. We note that our findings are only generalizable to postmenopausal women who are not taking hormone replacement therapy or exogenous insulin and have no personal history of cancer (including breast cancer).

A major strength of our study was the use of a highly sensitive liquid chromatography-mass spectrometry method to measure the plasma concentrations of sex-steroid hormones in postmenopausal women with high precision and accuracy. We were able to demonstrate the validity of this method using reference standards for estradiol and testosterone. The measured and true values of estradiol and testosterone were highly correlated. Further, intra-assay and inter-assay CVs, as well as intra-batch and inter-batch ICCs, calculated from quality control samples indicated that biomarker measurements were reliable, with few exceptions that may be attributable to batch and dispatch effects (Online Resource 4). We adopted a novel analysis approach to correct for batch effects, dispatch effects and time since last meal, whilst retaining meaningful biological variation in the biomarker measurements [ 21 ]. Further, we measured the plasma concentrations of a breadth of biomarkers from the inflammation, insulin/IGF-signaling and sex-steroid hormone pathways, which were selected through expert consultation and literature review. However, plasma concentrations of biomarkers measured at only one point in time will not be perfect proxies of complex and time-varying biological processes that may operate at cellular and systemic levels.

Comparison with other studies

Previous studies that have adjusted for other biomarkers have typically compared results with and without adjustment for other sex-steroid hormones and/or SHBG [ 3 , 15 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 ]. These are often mutual or progressive adjustments to assess “independence” rather than confounding, on the basis that biomarkers share complex interrelationships and correlations. However, this practice can lead to overadjustment bias, which we have attempted to minimize by explicitly considering our assumed underlying causal structure a priori [ 29 ]. In addition, only a handful of other studies have measured and adjusted for biomarkers from other biological pathways that may be potential confounders. One study from the Women’s Health Initiative presented results for estradiol with and without adjustment for free IGF-1 and insulin; positive associations with postmenopausal breast cancer appeared stronger with adjustment for both IGF-1 and insulin [ 40 ]. In contrast, our results for estradiol were similar with and without adjustment for biomarkers that were identified as potential confounders (adiponectin, leptin, TNF-α, IL-6, IGF-1, insulin and SHBG). Another study from the UK Biobank presented results for total testosterone with and without adjustment for SHBG and IGF-1 that were not appreciably different [ 41 ]. Likewise, our results for total testosterone did not change with and without adjustment for SHBG, but IGF-1 was not identified as a potential confounder of this analysis.

Overall, our findings were generally consistent with previous studies, including a recent systematic review by Drummond et al. [ 4 ], a previous case-cohort study conducted at baseline (1990–1994) within the MCCS [ 15 ], and a pooled analysis of nine prospective studies examining the relationship between endogenous sex-steroid hormones and postmenopausal breast cancer [ 3 ]. A notable finding was the estimated risk ratio per doubling plasma concentration of progesterone; we observed the largest increased risk of postmenopausal ER-positive breast cancer for this biomarker (RR 1.22, 95% CI 1.03 to 1.44) compared to any other measured biomarker from the sex-steroid hormone pathway. Previous studies have either not measured endogenous progesterone or have drawn inconclusive results regarding its relationship with breast cancer after the menopause, largely due to insufficient assay sensitivity and low circulating levels in postmenopausal women [ 42 ]. Our result is in support of a recent study by Trabert et al. [ 43 ], which also used a highly sensitive liquid chromatography-mass spectrometry method and found increased risks of postmenopausal breast cancer per standard deviation increase in circulating endogenous progesterone levels (hazard ratio for invasive breast cancers: 1.24, 95% CI 1.07 to 1.43). Trabert et al. [ 43 ] also present evidence for effect modification: reduced risks of postmenopausal breast cancer were observed with higher levels of progesterone among women in the lowest quintile of circulating estradiol (< 6.30 pg/mL), while increased risks were observed among women in the higher quintiles (≥ 6.30 pg/mL). Collectively, these results may challenge the plausibility of our a priori assumption that progesterone does not have a direct effect on postmenopausal ER-positive breast cancer (depicted by no direct arrow from progesterone to postmenopausal breast cancer in our causal diagram, Online Resource Fig. 8.1). This assumption was based on the systematic review by Drummond et al. [ 4 ], which found moderate quality evidence of no association between progesterone and breast cancer risk (albeit in both pre- and postmenopausal women combined). The implication of this assumption is that we should interpret the risk ratio for progesterone as an indirect effect, possibly driven by its role as a precursor of androgens and estrogens in steroidogenesis. This finding—in addition to concerns over the sensitivity of progesterone measurements in early studies, as well as studies demonstrating paracrine effects of progesterone via neighboring PR-positive cells [ 42 ]—warrants future studies including mediation analyses to determine what dictates the effect of progesterone on postmenopausal ER-positive breast cancer.

Implications and future directions

Our study confirms the causal role that sex-steroid hormones and SHBG play in the etiology of postmenopausal ER-positive breast cancer. We strengthen the causal evidence by demonstrating that potential confounding from other biological pathways implicated in breast carcinogenesis is likely non-substantial. However, the limited state of causal knowledge about biomarker interrelationships and the potential for residual confounding should be considered when interpreting our findings. Of note, two recent systematic reviews found insufficient evidence to establish a causal link between the inflammation and insulin/IGF-signaling pathways and breast cancer [ 8 , 11 ]. Future research could examine whether adjustment for biomarkers from other biological pathways is more important for pre–menopausal breast cancer or ER-negative postmenopausal breast cancer. In addition, time-varying confounding could be examined in future studies that measure biomarkers at multiple points in time.

Data availability

The dataset generated for the current study is not publicly available due to compliance with participant informed consent and human research ethics committee approvals, but can be requested by contacting [email protected].

Gérard C, Brown KA (2018) Obesity and breast cancer - Role of estrogens and the molecular underpinnings of aromatase regulation in breast adipose tissue. Mol Cell Endocrinol 466:15–30. https://doi.org/10.1016/j.mce.2017.09.014

Article   CAS   PubMed   Google Scholar  

(2018) Absence of Excess Body Fatness. IARC Handbooks of Cancer Prevention. IARC, Lyon

Key T, Appleby P, Barnes I, Reeves G (2002) Endogenous sex hormones and breast cancer in postmenopausal women: reanalysis of nine prospective studies. J Natl Cancer Inst 94:606–616. https://doi.org/10.1093/jnci/94.8.606

Drummond AE, Swain CTV, Brown KA et al (2022) Linking physical activity to breast cancer via sex steroid hormones, Part 2: the effect of sex steroid hormones on breast cancer risk. Cancer Epidemiol Biomarkers Prev 31:28–37. https://doi.org/10.1158/1055-9965.Epi-21-0438

Friedenreich CM, Ryder-Burbidge C, McNeil J (2021) Physical activity, obesity and sedentary behavior in cancer etiology: epidemiologic evidence and biologic mechanisms. Mol Oncol 15:790–800. https://doi.org/10.1002/1878-0261.12772

Article   PubMed   Google Scholar  

Lynch BM, Leitzmann MF (2017) An evaluation of the evidence relating to physical inactivity, sedentary behavior, and cancer incidence and mortality. Current Epidemiol Rep 4:221–231. https://doi.org/10.1007/s40471-017-0119-7

Article   Google Scholar  

Key TJ, Appleby PN, Reeves GK, Roddam AW (2010) Insulin-like growth factor 1 (IGF1), IGF binding protein 3 (IGFBP3), and breast cancer risk: pooled individual data analysis of 17 prospective studies. Lancet Oncol 11:530–542. https://doi.org/10.1016/s1470-2045(10)70095-4

Drummond AE, Swain CTV, Milne RL et al (2022) Linking physical activity to breast cancer risk via the insulin/insulin-like growth factor signaling system, Part 2: the effect of insulin/insulin-like growth factor signaling on breast cancer risk. Cancer Epidemiol Biomarkers Prev 31:2116–2125. https://doi.org/10.1158/1055-9965.Epi-22-0505

Article   CAS   PubMed   PubMed Central   Google Scholar  

Sánchez-Jiménez F, Pérez-Pérez A, de la Cruz-Merino L, Sánchez-Margalet V (2019) Obesity and breast cancer: role of leptin. Front Oncol 9:596. https://doi.org/10.3389/fonc.2019.00596

Article   PubMed   PubMed Central   Google Scholar  

Chan DS, Bandera EV, Greenwood DC, Norat T (2015) Circulating C-reactive protein and breast cancer risk-systematic literature review and meta-analysis of prospective cohort studies. Cancer Epidemiol Biomarkers Prev 24:1439–1449. https://doi.org/10.1158/1055-9965.Epi-15-0324

Lou MWC, Drummond AE, Swain CTV et al (2023) Linking physical activity to breast cancer via inflammation, Part 2: the effect of inflammation on breast cancer risk. Cancer Epidemiol Biomarkers Prev 32:597–605. https://doi.org/10.1158/1055-9965.Epi-22-0929

Simó R, Sáez-López C, Barbosa-Desongles A, Hernández C, Selva DM (2015) Novel insights in SHBG regulation and clinical implications. Trends Endocrinol Metab 26:376–383. https://doi.org/10.1016/j.tem.2015.05.001

Milne RL, Fletcher AS, MacInnis RJ et al (2017) Cohort profile: the melbourne collaborative cohort study (Health 2020). Int J Epidemiol 46:1757–i. https://doi.org/10.1093/ije/dyx085

Baglietto L, English DR, Hopper JL et al (2009) Circulating steroid hormone concentrations in postmenopausal women in relation to body size and composition. Breast Cancer Res Treat 115:171–9. https://doi.org/10.1007/s10549-008-0069-3

Baglietto L, Severi G, English DR et al (2010) Circulating steroid hormone levels and risk of breast cancer for postmenopausal women. Cancer Epidemiol Biomarkers Prev 19:492–502. https://doi.org/10.1158/1055-9965.Epi-09-0532

Hefti MM, Hu R, Knoblauch NW et al (2013) Estrogen receptor negative/progesterone receptor positive breast cancer is not a reproducible subtype. Breast Cancer Res 15:R68. https://doi.org/10.1186/bcr3462

Olivotto IA, Truong PT, Speers CH et al (2004) Time to stop progesterone receptor testing in breast cancer management. J Clin Oncol 22:1769–70. https://doi.org/10.1200/jco.2004.99.251

De Maeyer L, Van Limbergen E, De Nys K et al (2008) Does estrogen receptor negative/progesterone receptor positive breast carcinoma exist? J Clin Oncol 26(335–6):5. https://doi.org/10.1200/jco.2007.14.8411

Nadji M, Gomez-Fernandez C, Ganjei-Azar P, Morales AR (2005) Immunohistochemistry of estrogen and progesterone receptors reconsidered: experience with 5993 breast cancers. Am J Clin Pathol 123:21–7. https://doi.org/10.1309/4wv79n2ghj3x1841

Mori N, Keski-Rahkonen P, Gicquiau A et al (2021) Endogenous circulating sex hormone concentrations and colon cancer risk in postmenopausal women: a prospective study and meta-analysis. JNCI Cancer Spectr. https://doi.org/10.1093/jncics/pkab084

Viallon V, His M, Rinaldi S et al (2021) A new pipeline for the normalization and pooling of metabolomics data. Metabolites 11:631. https://doi.org/10.3390/metabo11090631

Endogenous Hormones and Breast Cancer Collaborative Group (2003) Free estradiol and breast cancer risk in postmenopausal women: comparison of measured and calculated values. Cancer Epidemiol Biomark Prev 12:1457–61

Google Scholar  

Dunn JF, Nisula BC, Rodbard D (1981) Transport of steroid hormones: binding of 21 endogenous steroids to both testosterone-binding globulin and corticosteroid-binding globulin in human plasma. J Clin Endocrinol Metab 53:58–68. https://doi.org/10.1210/jcem-53-1-58

Rinaldi S, Geay A, Déchaud H et al (2002) Validity of free testosterone and free estradiol determinations in serum samples from postmenopausal women by theoretical calculations. Cancer Epidemiol Biomarkers Prev 11:1065–71

CAS   PubMed   Google Scholar  

Södergård R, Bäckström T, Shanbhag V, Carstensen H (1982) Calculation of free and bound fractions of testosterone and estradiol-17 beta to human plasma proteins at body temperature. J Steroid Biochem 16:801–10. https://doi.org/10.1016/0022-4731(82)90038-3

Borgan O, Langholz B, Samuelsen SO, Goldstein L, Pogoda J (2000) Exposure stratified case-cohort designs. Lifetime Data Anal 6:39–58. https://doi.org/10.1023/a:1009661900674

Article   MathSciNet   CAS   PubMed   Google Scholar  

Hernán MA, Sauer BC, Hernández-Díaz S, Platt R, Shrier I (2016) Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses. J Clin Epidemiol 79:70–5. https://doi.org/10.1016/j.jclinepi.2016.04.014

VanderWeele TJ (2019) Principles of confounder selection. Eur J Epidemiol 34:211–9. https://doi.org/10.1007/s10654-019-00494-6

Schisterman EF, Cole SR, Platt RW (2009) Overadjustment bias and unnecessary adjustment in epidemiologic studies. Epidemiology 20:488–95. https://doi.org/10.1097/EDE.0b013e3181a819a1

Hernán MA, Robins JM (2023) Causal Inference: What If. Chapman & Hall/CRC, Boca Raton

Kaaks R, Rinaldi S, Key TJ et al (2005) Postmenopausal serum androgens, oestrogens and breast cancer risk: the European prospective investigation into cancer and nutrition. Endocr Relat Cancer 12:1071–82. https://doi.org/10.1677/erc.1.01038

Tworoger SS, Zhang X, Eliassen AH et al (2014) Inclusion of endogenous hormone levels in risk prediction models of postmenopausal breast cancer. J Clin Oncol 32:3111–7. https://doi.org/10.1200/jco.2014.56.1068

Zhang X, Tworoger SS, Eliassen AH, Hankinson SE (2013) Postmenopausal plasma sex hormone levels and breast cancer risk over 20 years of follow-up. Breast Cancer Res Treat 137:883–92. https://doi.org/10.1007/s10549-012-2391-z

Zeleniuch-Jacquotte A, Bruning PF, Bonfrer JM et al (1997) Relation of serum levels of testosterone and dehydroepiandrosterone sulfate to risk of breast cancer in postmenopausal women. Am J Epidemiol 145:1030–8. https://doi.org/10.1093/oxfordjournals.aje.a009059

Zeleniuch-Jacquotte A, Shore RE, Koenig KL et al (2004) Postmenopausal levels of oestrogen, androgen, and SHBG and breast cancer: long-term results of a prospective study. Br J Cancer 90:153–9. https://doi.org/10.1038/sj.bjc.6601517

Sieri S, Krogh V, Bolelli G et al (2009) Sex hormone levels, breast cancer risk, and cancer receptor status in postmenopausal women: the ORDET cohort. Cancer Epidemiol Biomarkers Prev 18:169–76. https://doi.org/10.1158/1055-9965.Epi-08-0808

Berrino F, Muti P, Micheli A et al (1996) Serum sex hormone levels after menopause and subsequent breast cancer. J Natl Cancer Inst 88:291–6. https://doi.org/10.1093/jnci/88.5.291

Fourkala EO, Zaikin A, Burnell M et al (2012) Association of serum sex steroid receptor bioactivity and sex steroid hormones with breast cancer risk in postmenopausal women. Endocr Relat Cancer 19:137–47. https://doi.org/10.1530/erc-11-0310

Thomas HV, Key TJ, Allen DS et al (1997) A prospective study of endogenous serum hormone concentrations and breast cancer risk in post-menopausal women on the island of Guernsey. Br J Cancer 76:401–5. https://doi.org/10.1038/bjc.1997.398

Gunter MJ, Hoover DR, Yu H et al (2009) Insulin, insulin-like growth factor-I, and risk of breast cancer in postmenopausal women. J Natl Cancer Inst 101:48–60. https://doi.org/10.1093/jnci/djn415

Tin Tin S, Reeves GK, Key TJ (2021) Endogenous hormones and risk of invasive breast cancer in pre- and post-menopausal women: findings from the UK Biobank. Br J Cancer 125:126–34. https://doi.org/10.1038/s41416-021-01392-z

Trabert B, Sherman ME, Kannan N, Stanczyk FZ (2020) Progesterone and Breast Cancer. Endocr Rev 41:320–44. https://doi.org/10.1210/endrev/bnz001

Trabert B, Bauer DC, Buist DSM et al (2020) Association of circulating progesterone with breast cancer risk among postmenopausal women. JAMA Netw Open 3:e203645. https://doi.org/10.1001/jamanetworkopen.2020.3645

Download references

Acknowledgments

The authors thank Audrey Gicquiau for the measurement of the sex-steroid hormones and sex-hormone binding globulin, and Anne-Sophie Navionis for the measurement of the biomarkers of the insulin\insulin-like growth factor-signaling and inflammatory pathways. The authors also thank the participants of the Melbourne Collaborative Cohort Study. Cases and their vital status were ascertained through the Victorian Cancer Registry and the Australian Institute of Health and Welfare, including the National Death Index and the Australian Cancer Database

Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy or views of the International Agency for Research on Cancer/World Health Organization.

Open Access funding enabled and organized by CAUL and its Member Institutions. Melbourne Collaborative Cohort Study (MCCS) cohort recruitment was funded by VicHealth and Cancer Council Victoria. The MCCS was further augmented by Australian National Health and Medical Research Council grants 209057, 396414 and 1074383 and by infrastructure provided by Cancer Council Victoria. Funding for IIG_2018_1730 was obtained from World Cancer Research Fund (WCRF UK), as part of the World Cancer Research Fund International grant programme. The reference standards for estradiol and testosterone were purchased and analyzed using funds from a NIH grant (NIH R01 CA207369) held by Dr Sue Hankinson at University of Massachusetts. Frances EM Albers and Makayla WC Low are each supported by a Research Training Program Scholarship from the Australian Government and the University of Melbourne. Makayla WC Lou is further supported by a scholarship from the Macau Special Administrative Region Government Higher Education Fund (Governo da Região Administrativa Especial de Macau Fundo do Ensino Superior).

Author information

Authors and affiliations.

Centre for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, University of Melbourne, Melbourne, Australia

Frances E. M. Albers, Makayla W. C. Lou, Amalia Karahalios, Roger L. Milne, Dallas R. English & Brigid M. Lynch

Cancer Epidemiology Division, Cancer Council Victoria, Council Victoria, Level 8, 200 Victoria Parade, East Melbourne, Melbourne, VIC, 3002, Australia

Frances E. M. Albers, Makayla W. C. Lou, Christopher T. V. Swain, Roger L. Milne, Dallas R. English & Brigid M. Lynch

Clinical Epidemiology and Biostatistics Unit, Murdoch Children’s Research Institute, Melbourne, Australia

S. Ghazaleh Dashti

Department of Paediatrics, Melbourne Medical School, University of Melbourne, Melbourne, Australia

Department of Physiotherapy, Melbourne School of Health Sciences, University of Melbourne, Melbourne, Australia

Christopher T. V. Swain

Nutrition and Metabolism Branch, International Agency for Research on Cancer, Lyon, France

Sabina Rinaldi, Vivian Viallon & Marc J. Gunter

Department of Cell Biology and Physiology, University of Kansas Medical Center, Kansas City, USA

Kristy A. Brown

Cancer Epidemiology and Prevention Research Unit, Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, UK

Marc J. Gunter

Precision Medicine, School of Clinical Sciences at Monash Health, Monash University, Melbourne, Australia

Roger L. Milne

Physical Activity Laboratory, Baker Heart and Diabetes Institute, Melbourne, Australia

Brigid M. Lynch

You can also search for this author in PubMed   Google Scholar

Contributions

FEMA—Data curation, formal analysis, methodology, visualization, writing—original draft, writing—review & editing. MWCL—Data curation, methodology, writing—review & editing. SGD—Conceptualization, funding acquisition, methodology, supervision, writing—review & editing. CTVS—Methodology, supervision, writing—review & editing. SR—Conceptualization, data curation, funding acquisition, investigation, resources, writing—review & editing. VV—Methodology, writing—review & editing. AK—Conceptualization, funding acquisition, writing—review & editing. KAB—Conceptualization, funding acquisition, methodology, writing—review & editing. MJG—Conceptualization, funding acquisition, writing—review & editing. RLM—Conceptualization, funding acquisition, methodology, supervision, writing—review & editing. DRE—Conceptualization, data curation, funding acquisition, methodology, project administration, supervision, writing—review & editing. BML—Conceptualization, funding acquisition, methodology, supervision, writing—review & editing.

Corresponding author

Correspondence to Brigid M. Lynch .

Ethics declarations

Competing interests.

The authors have no relevant financial or non-financial interests to disclose.

Ethical approval

The study protocol was approved by the Cancer Council Victoria Human Research Ethics Committee (Approval Number: CCV IEC 9001).

Consent to participate

All participants of the MCCS gave written informed consent.

Consent to publication

Not applicable.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (PDF 959 KB)

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Albers, F.E.M., Lou, M.W.C., Dashti, S.G. et al. Sex-steroid hormones and risk of postmenopausal estrogen receptor-positive breast cancer: a case–cohort analysis. Cancer Causes Control (2024). https://doi.org/10.1007/s10552-024-01856-6

Download citation

Received : 03 October 2023

Accepted : 16 January 2024

Published : 16 February 2024

DOI : https://doi.org/10.1007/s10552-024-01856-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Breast cancer
  • Sex-steroid hormones
  • Progesterone
  • Sex-hormone binding globulin
  • Find a journal
  • Publish with us
  • Track your research

Log in using your username and password

  • Search More Search for this keyword Advanced search
  • Latest content
  • Publish with us
  • About the journal
  • Meet the editors
  • Specialist reviews
  • BMJ Journals More You are viewing from: Google Indexer

You are here

  • Volume 3, Issue 1
  • Association of vaginal oestradiol and the rate of breast cancer in Denmark: registry based, case-control study, nested in a nationwide cohort
  • Article Text
  • Article info
  • Citation Tools
  • Rapid Responses
  • Article metrics

Download PDF

  • http://orcid.org/0000-0003-4549-8271 Amani Meaidi 1 ,
  • Nelsan Pourhadi 1 ,
  • Ellen Christine Løkkegaard 2 , 3 ,
  • Christian Torp-Pedersen 4 , 5 and
  • http://orcid.org/0000-0001-6506-2569 Lina Steinrud Mørch 1
  • 1 Cancer Surveillance and Pharmacoepidemiology , Danish Cancer Institute , Copenhagen , Denmark
  • 2 Department of Gynecology and Obstetrics , Nordsjællands Hospital , Hillerod , Denmark
  • 3 Clinical Medicine , Copenhagen University , Copenhagen , Denmark
  • 4 Department of Cardiology , Nordsjaellands Hospital , Hillerod , Denmark
  • 5 Public Health , Copenhagen University , Copenhagen , Denmark
  • Correspondence to Dr Amani Meaidi, Cancer Surveillance and Pharmacoepidemiology, Danish Cancer Institute, Copenhagen, 2100, Denmark; amani-meaidi{at}live.dk

Objective To estimate the rate of breast cancer associated with use of vaginal oestradiol tablets according to duration and intensity of their use.

Design Registry based, case-control study, nested in a nationwide cohort.

Setting Based in Denmark using the civil registration system, the national registry of medicinal product statistics, the Danish cancer registry, the Danish birth registry, and statistics Denmark.

Participants Women aged 50-60 years in year 2000 or turning 50 years during the study period of 1 January 2000 to 31 December 2018 were included. Exclusions were a history of cancer, mastectomy, use of systemic hormone treatment, use of the levonorgestrel releasing intrauterine system, or use of vaginal oestrogen treatments other than oestradiol tablets. To each woman who developed breast cancer during follow-up (18 997), five women in the control group (94 985) were incidence density matched by birth year.

Main outcome measure The main outcome was pathology confirmed breast cancer diagnosis.

Results 2782 (14.6%) women with breast cancer (cases) and 14 999 (15.8%) women with no breast cancer diagnosis (controls) had been exposed to vaginal oestradiol tablets with 234 cases and 1232 controls having been in treatment for at least four years at a high intensity (>50 micrograms per week). Increasing durations and intensities of use (cumulative dose/cumulative duration) of vaginal oestradiol tablets was not associated with increasing rates of breast cancer. Compared with never-use, cumulative use of vaginal oestradiol for more than nine years was associated with an adjusted hazard ratio of 0.87 (95% confidence interval 0.69 to 1.11). Results were similar in women who had long term use (≥four years) and with high intensity of use (>50-70 micrograms per week) with an adjusted hazard ratio 0.93 (95% confidence interval 0.81 to 1.08).

Conclusions Use of vaginal oestradiol tablets was not associated with increased breast cancer rate compared with never-use. Increasing duration and intensity of use was not associated with increased rates of breast cancer.

  • Hormone replacement therapy
  • Breast neoplasms

Data availability statement

No data are available. Raw data used to conduct this study is only accessible through approval from the Danish Data Protection Agency and the Danish Health Data Board. Although anonymised, the data was available on an individual level, making data sharing restricted by the General Data Protection Regulation of EU law.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:  http://creativecommons.org/licenses/by-nc/4.0/ .

https://doi.org/10.1136/bmjmed-2023-000753

Statistics from Altmetric.com

Request permissions.

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

While systemic oestrogen treatment has been associated with increased risk of breast cancer development, use of vaginal oestrogen has been suggested to be risk-free

The effect of duration and intensity of vaginal oestrogen use on breast cancer risk is uncertain

WHAT THIS STUDY ADDS

In this nationwide study, use of vaginal oestradiol tablets was not associated with a significant increased risk of breast cancer

This finding remained even in women who were in treatment for >nine years and in women who used the drug long term at high intensity (>50 micrograms per week)

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY

The findings add reassurance to the breast cancer safety of vaginal oestrogen use in women

Introduction

Breast cancer is the most common cancer in women, affecting around 7.8 million women worldwide with 2.3 million incidences and 700 000 deaths annually. 1 Knowledge about external risk factors would potentially enable preventive actions.

Vaginally administrated oestrogen is the primary pharmaceutical treatment for the genitourinary syndrome of menopause, a condition caused by the physiological oestrogen deficiency following menopause. 2 Around 50% of postmenopausal women will have the syndrome with symptoms such as vaginal irritation, recurrent urogenital infections, and dyspareunia. 2 While systemic menopausal hormone treatment with oestrogen has been linked to an increased risk of breast cancer, use of vaginal oestrogen has been suggested to be risk-free. 3 However, studies have not been able to account for both duration and intensity of vaginal oestrogen use when assessing the association with breast cancer risk, which is highly relevant considering the variation in dosage and time of use among women in need of vaginal oestrogen. 3 4 Furthermore, vaginal oestrogen treatment has previously been associated with an increased risk of endometrial cancer. 4–6 This link to an oestrogen sensitive cancer further necessitates high quality evidence on the breast cancer safety of vaginal oestrogen treatment, especially considering the rise in use. 7

Nested in a Danish, nationwide, female population, we aimed to investigate the risk of breast cancer among women using vaginal oestradiol tablets.

Study population

In Denmark, access to healthcare is freely available to all Danish citizens. We conducted a nationwide, nested, case-control study using the following Danish national registries: (1) the civil registration system, which contains information about sex and vital status of all Danish citizens; (2) the national registry of medicinal product statistics, which includes information about all redeemed prescriptions at Danish pharmacies since 1 January 1995; (3) the Danish cancer registry, which includes all cancer cases since 1 January 1943; (4) the national registry of patients, which comprises information about discharge diagnoses and surgical codes on all somatic hospital admissions since 1 January 1976; (5) the Danish national birth registry, which holds information about all live and death births since 1 January 1973; and (6) statistics Denmark, which provides a yearly update on the education and income status of all Danish citizens. 8–12

We identified incident breast cancer cases and randomly chose controls with no breast cancer who had been matched by birth year. Women were chosen if they were recorded between 1 January 2000 and 31 December 2018 in a nationwide population of all Danish women aged 50-60 years on 1 January 2000 or turning 50 years throughout the study period. Women had no history of cancer (except non-melanoma skin cancer), mastectomy, or prior use of systemic hormone treatment, the hormone-releasing intrauterine system, other vaginal oestrogen treatments than oestradiol tablets, and anti-oestrogen medications.

Data sources and definitions of exclusion criteria are provided in online supplemental table S1 . The personal identification number given to all Danish citizens at birth or immigration allowed reliable linkage between data sources. The year of initiation of the study period and the age restriction of the study population were defined to ensure almost complete exposure history of local and systemic hormone treatment on all included women.

Supplemental material

Breast cancer.

The cancer registry contains records of all incidences of malignant neoplasms in the Danish population from 1943 and onwards. 10 A woman was considered a case with incident breast cancer from the date of a first time invasive breast cancer diagnosis in the Danish cancer registry (the International Classification of Diseases, 10th revision, codes C500-C509). 10 On the date of diagnosis, five women with no breast cancer (controls) were incidence density matched by birth year to each case of breast cancer. The national pathology registry provided information on oestrogen receptor positivity of the breast cancers (systemised nomenclature of medicine code F29521).

Vaginal oestradiol tablets

The exposure of interest was treatment with the vaginal oestradiol tablet as this form is by far the most commonly used type of vaginal oestrogen among Danish women. 7 During the study period, use of any vaginal oestradiol drug formulation required a prescription from a physician. Vaginal oestradiol tablets were available in doses of 10 µg and 25 µg.

Daily updated, individual level information about prescription redemption of vaginal oestradiol tablet was provided by the national registry of medicinal product statistics, which holds information on all prescriptions filled by the Danish population since 1995. 9 The registry receives its information electronically from the digital accounting systems of Danish pharmacies that primarily use the systems to secure reimbursement from the national health service. 9

A woman was considered to be using vaginal oestradiol tablets if she redeemed at least one prescription of the drug. Using the anatomical therapeutic chemical code (ATC) of oestradiol drug formulations (G03CA03) and conditioning on vaginal tablet administration, information on exposure status was obtained from the national registry of medicinal product statistics. 9 This national registry provided information on the date of redeemed prescription, size of drug unit, as well as size and number of redeemed packages. 9 This information was updated daily for each woman during the study period.

Vaginal oestradiol tablets are recommended to be taken once a day for the first two weeks of treatment followed by a maintenance dose of two tablets per week. However, the dosage may be regulated up or down according to the urogenital symptoms of the woman. The validated programme “medicinMacro”, accessible from Github ( www.github.com ) in the “tagteam/heaven” R package, was used to calculate the most likely daily dose. As such, also calculated was duration and time of use and non-use of vaginal oestradiol tablets according to information on date and amount of purchased drug, recommended default, minimum, and maximum dosages, and the prescription pattern of up to five most recent prescriptions. 13 14

Potential confounders

Potential confounding factors, such as educational level and yearly income, was obtained from statistics Denmark. Information about polycystic ovary syndrome, endometriosis, and chronic obstructive pulmonary disease (a surrogate measure for health threatening smoking) was identified from the national registry of patients, and data for redeemed prescriptions on bisphosphonates and diabetes mellitus medication was provided by the national registry of medicinal product statistics. 9 11 Information about parity was extracted from the national birth registry. 12

Statistical analysis

Conditional logistic regression models provided adjusted hazard rate ratios and corresponding 95% confidence intervals of breast cancer according to duration, intensity, and user status of vaginal oestradiol tablets at time of index date (date of diagnosis or matching). Duration was calculated as the cumulative duration of use at time of the index date without consideration to breaks in treatment. In a sensitivity analysis, duration was categorised according to the maximum length of continuous use without treatment breaks. Intensity of use was calculated as the cumulative dose of vaginal oestradiol tablets redeemed at time of index divided by the cumulative duration of use and categorised into low (≤25 μg/week), medium (>25-50 μg/week), high (>50-70 μg/week), and very high (>70 μg/week) intensity of use. User status was categorised into current use (use within 0-2 months prior to index date), recent use (use 2-24 months prior to index date), and previous use (>24 months prior to index date).

Women who had never used vaginal oestradiol tablets and other hormone treatments constituted the reference group in all analyses. The potential confounding factors described above were included in the statistical models.

Sensitivity analyses were conducted on the subpopulation of cases with breast cancers positive for oestrogen receptors and their matched controls as well as on the subpopulation of women with a Charlson comorbidity index score of zero (definition in online supplemental table S1 ).

All analyses were repeated with one year lag time. The level of statistical significance was set at P<0.05. Data was analysed using R statistical software (RStudio version 4.2.1). 15

Patient and public involvement

No patients or members of the public were involved in the design, analysis, or writing up of the study because the research project was undertaken by a small research group without funds or staff for patient and public involvement measures. The results of the study will, nevertheless, be disseminated to the public and health professionals by press releases and presentations at scientific conferences.

A total of 18 997 women with breast cancer and 94 985 women in the control group were identified ( figure 1 ). Characteristics of cases and controls are presented in table 1 . The overall prevalence of vaginal oestradiol use was 2782 (14.6%) among women with breast cancer and 14 999 (15.8%) among population controls. In the control group, 5261 (5.5%) of 94 985 women currently used, 3291 (3.5%) recently used, and 6447 (6.8%) previously used vaginal oestradiol tablets. The corresponding prevalence estimates were similar in women with breast cancer ( table 1 ). Median age at initiation of vaginal oestradiol tablets was 57 years (interquartile range 53-61) among women in the control group, median cumulative duration of use was 8.1 months (2.8-27.4), median cumulative dose was 1620 μg (750-5625), and the median intensity of use was 50.5 µg/week (49.5-55.0). The characteristics of vaginal oestradiol use were similar in breast cancer cases ( table 1 ).

  • Download figure
  • Open in new tab
  • Download powerpoint

Study flowchart

  • View inline

Characteristics of the study population and its use of vaginal oestradiol tablets

Use of vaginal oestradiol tablets was not associated with a significant increase in breast cancer rate compared with never-use ( figure 2 ). Increased cumulative duration of use did not imply increased rates of breast cancer ( figure 2 ), the adjusted hazard ratio of more than nine years of cumulative use was estimated to be 0.87 (95% confidence interval 0.69 to 1.11). The absence of association with duration of use persisted when only considering consistent, uninterrupted use ( online supplemental figure S1 ).

Rate of breast cancer according to duration of use of vaginal oestradiol tablets. Adjusted for educational level; yearly income; history of polycystic ovary syndrome, endometriosis, chronic obstructive pulmonary disease, use of bisphosphonates, or diabetes mellitus; and parity at index date

No consistent association was observed between intensity of use and rate of breast cancer ( figure 3 ). A total of 227 cases and 1186 controls were exposed to vaginal oestradiol tablets for more than four years with an intensity of more than 50-70 µg/week, which is above the recommended maintenance dose of 20-50 µg/week. In these users, the adjusted hazard ratio of breast cancer was found to be 0.93 (95% confidence interval 0.81 to 1.08) compared with never-use.

Rate of breast cancer according to duration and intensity of use of vaginal oestradiol tablets. Intensity categories were low (≤25 µg/week), medium (>25-50 µg/week), high (>50-70 µg/week), and very high (>70 µg/week). Adjusted for educational level, yearly income, polycystic ovary syndrome, endometriosis, chronic obstructive pulmonary disease, use of bisphosphonates, diabetes mellitus, and parity at index date

Results remained robust when stratifying according to user status at time of diagnosis or matching ( figure 4 ).

Rate of breast cancer according to timing, duration, and intensity of use of vaginal oestradiol tablets. Timing of use: current use (0-2 months within index date), recent use (2 months-2 years within index date), previous use (>2 years prior to index date). Adjusted for educational level, yearly income, polycystic ovary syndrome, endometriosis, chronic obstructive pulmonary disease, use of bisphosphonates, diabetes mellitus, and parity at index date

The lack of consistent association between duration and intensity of use of vaginal oestradiol tablets and breast cancer rate persisted in the subpopulation of oestrogen-receptor positive breast cancer cases ( online supplemental figure S2 ) and in healthy women with a Charlson comorbidity index score of zero ( online supplemental figure S3 ), respectively.

Sensitivity analyses including a one year lag time did not materially change the main estimates ( online supplemental figure S4 ).

In this real-world, nationwide, Danish population, increasing duration and intensity of use of vaginal oestradiol tablets was not found to be associated with an increased risk of breast cancer.

Several studies have found orally administrated oestrogen-only treatment to be associated with an increased risk of breast cancer. 3 A meta-analysis of individual participants worldwide reported a hazard ratio of 1.33 in current users of oral oestrogen-only treatment compared with no use. 3 The meta-analysis reports a hazard ratio of 1.09 with use of vaginal oestrogen without further consideration to intensity of use. 3 Similarly, a prospective cohort study by the Women’s Health Initiative of 45 663 postmenopausal women did not find any association between vaginal oestrogen use and breast cancer risk, but did not investigate the association according to duration or intensity of use. 4 A nationwide observational study from Finland reported that use of vaginal oestrogen for less than five years was not associated with an increased risk of breast cancer, but the study did not have sufficient power to study the effect of use of more than five years or the role of the intensity of use. 16

The apparent absence of association between use of vaginal oestradiol tablets and development of breast cancer have previously been explained by the low dose of oestradiol absorbed into the blood with vaginal application of low-dose oestrogen. 17 18 Our study suggests that use of >50-70 µg/week for more than four years is not associated with increased breast cancer risk (hazard ratio 0.93 (95% confidence interval 0.81-1.08)) compared with never-use. Use of >50-70 micrograms per week corresponds to 2.5-fold to 3.5-fold more than recommended weekly maintenance dose of the currently marketed low dose 10 µg vaginal oestradiol tablet.

To our knowledge, our study is the first to report on the breast cancer risk with vaginal oestradiol tablets according to duration and intensity of use. One strength of our study is its nationwide design with a large unselected study population. Additionally, a strength is the use of high quality registry data with accurate and continuously updated data for breast cancer diagnoses and vaginal oestradiol prescriptions as well as medical conditions, reproductive factors, and education. These data allow adjustment for several known risk factors for breast cancer and potential confounders. Use of registries covering the entire Danish population eliminated recall bias, minimised selection bias, provided a long study period, and resulted in no missing data for exposure, outcome, and covariates for all eligible study patients. Thus, no cases or controls were selected on missing data. Cancer diagnoses were from the cancer registry, in which all cancer diagnoses are histologically verified, further enhancing validity. 8 For all women included in the study, we had at least five years of prescription history (establishment of the National Prescription Registry in Denmark was in 1995).

Considering the observational nature of our study, the main limitation is potential existence of bias by unknown or unmeasured confounders. Women who were were adherent users of vaginal oestradiol tablets could potentially be healthier than women who did not use the tablets because adherence to a long term, expensive treatment for a physiological condition might be more likely in women prioritising health and having a favourable socioeconomic status. This potential healthy user bias may have biased our results towards the null. However, the results remained robust in a subpopulation of all healthy women with a Charlson comorbidity index score of zero. Furthermore, as in many other countries, in Denmark, high socioeconomic position has been associated with higher incidence of breast cancer, including in our study (data not shown). 19 Thus, if healthy user bias was present in our study, the direction of the bias would not necessarily cause an underestimation of the association. Finally, we did adjust for education and income in our study.

Despite controlling for several potential confounders, we cannot exclude the occurrence of residual confounding and unmeasured confounding. Obesity has been associated with an increased risk of breast cancer, and obesity is expected to be more common among women who do not use vaginal oestradiol tablets because their oestrogen production in lipid tissue likely decreases the need for exogenous oestrogen. 20 Thus, not adjusting for obesity might have caused an underestimation of the association between vaginal oestradiol and breast cancer. However, obesity is highly (and inversely) correlated with educational status in Denmark, and we did adjust for such. 21

Conclusions

Alongside recent studies suggesting that vaginal oestrogen treatment may safely be used by many women who have had breast cancer, this study adds reassurance to the breast cancer safety of vaginal oestrogen treatment. 22 However, considering the globally prevalent and often life-long indication and increasing use of vaginal oestrogen by postmenopausal women, further studies, especially in other populations, are warranted to confirm drug safety for all potential patients.

Ethics approval

This study was approved by the Danish Data Protection Agency and the Danish Health Data Board. Registry-based studies are not subject to ethics approval in Denmark.

  • World Cancer Research Fund/American Institute for Cancer Research
  • Portman DJ ,
  • Gass MLS , Vulvovaginal Atrophy Terminology Consensus Conference Panel
  • Collaborative Group on Hormonal Factors in Breast Cancer
  • Crandall CJ ,
  • Andrews CA , et al
  • Keiding N , et al
  • Bhupathiraju SN ,
  • Grodstein F ,
  • Stampfer MJ , et al
  • Goukasian I ,
  • Lidegaard O
  • Pedersen CB
  • Pottegård A ,
  • Schmidt SAJ ,
  • Wallach-Kildemoes H , et al
  • Gjerstorff ML
  • Schmidt M ,
  • Sandegaard JL , et al
  • Bliddal M ,
  • Pottegård A , et al
  • Torp-Pedersen C
  • Støvring H ,
  • Rostgaard K , et al
  • R Core Team
  • Lyytinen H ,
  • Pukkala E ,
  • Ylikorkala O
  • Notelovitz M ,
  • Nanavati N , et al
  • Sørdal T , et al
  • Dalton SO ,
  • Engholm G , et al
  • Berrino F ,
  • Abagnato CA , et al
  • Sundhedsstyrelsen
  • Jensen M-B , et al

Twitter @no

Contributors LSM and AM designed the study. AM conducted the data management, and NP and CT-P ran the analysis. The manuscript was drafted by AM and LSM, but all authors contributed to the final manuscript. All authors contributed to the interpretation of the findings. All authors approved the final version and made the decision to submit for publication. All authors had access to the statistically analysed data. AM and CT-P had access to and verified the raw data. AM is the guarantor of the overall content, takes responsibility for the work and conduct of the study, and controlled the decision to publish. The corresponding author attests that all listed authors meet the authorship criteria and that no others meeting the criteria have been omitted. Transparency: The lead author (the guarantor) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

Competing interests All authors have completed the ICMJE uniform disclosure form at www.icmje.org/disclosure-of-interest/ and declare: no support from any organisation for the submitted work; CTP has received grants from Novo Nordisk and Bayer outside of the current study; AM, EL, and LSM signed an authorship agreement with no financial benefit with Novo Nordisk after submission of this manuscript but before publication.

Provenance and peer review Not commissioned; externally peer reviewed.

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Read the full text or download the PDF:

  • Open access
  • Published: 06 February 2024

Experiences and perceptions of men following breast cancer diagnosis: a mixed method systematic review

  • Mary Abboah-Offei 1 ,
  • Jonathan Bayuo 2 ,
  • Yakubu Salifu 3 ,
  • Oladayo Afolabi 4 &
  • Theophilus N. Akudjedu 5  

BMC Cancer volume  24 , Article number:  179 ( 2024 ) Cite this article

560 Accesses

5 Altmetric

Metrics details

Men with breast cancer experience unique physical and emotional challenges. However, a thorough understanding of these experiences including the psychosocial effects and supportive care needs have received less attention. In some settings, men with breast cancer experience stigma within the healthcare system and their care needs are not prioritised. This influences the level of professional support offered, consequently worsening their health and well-being outcomes. This review explored the variabilities in the experiences and treatment modalities of male breast cancer (MBC) across different contexts.

All primary study designs including qualitative, quantitative, and mixed methods studies that reported on the experiences, treatment approaches and outcomes of MBC were included in this systematic review. Six databases (Embase, Medline, PsycINFO, Global Health, CINAHL and Web of Science) were searched for articles from January 2000 to September 2023. A results-based convergence synthesis was used for data analysis and reported using PRISMA guidelines.

Of the studies screened ( n  = 29,687), forty-four fulfilled the predetermined criteria and were included. Our findings relating to the experiences and treatment approaches of MBC are broadly themed into three parts. Theme 1—Navigating through a threat to masculinity: describes how males experienced the illness reflecting on detection, diagnosis, coming to terms with breast cancer, and disclosure. Theme 2- Navigating through treatment: captures the experiences of undergoing breast cancer treatment/ management following their diagnosis. Theme 3—Coping and support systems: describes how MBC patients coped with the disease, treatment process, aftercare/rehabilitative care, and the available support structures.

Conclusions

Men experience a myriad of issues following a breast cancer diagnosis, especially with their masculinity. Awareness creation efforts of MBC among the public and healthcare practitioners are urgently required, which could change the perception of men in promoting early diagnosis, adherence to treatments, post-treatment monitoring, oncological results and a better quality of life. Considerations for training, education and development of specialised guidelines for healthcare practitioners on MBC would provide the necessary knowledge and skills to enhance their practice through the adoption of person-centred and male-specific care strategies. Professional care intervention and support for MBC should not end after the diagnosis phase but should extend to the entire treatment continuum and aftercare including future research focusing on MBC specific clinical trials.

Trial registration

PROSPERO Registration No. CRD42021228778.

Peer Review reports

Male breast cancer (MBC) is a rare condition, accounting for less than 1% of all breast cancers. About 2,710 men are estimated to be diagnosed with breast cancer, with approximately 530 men projected to die from breast cancer in 2022 and have about 1 in 833 lifetime risk of being diagnosed with the disease in the United States [ 1 ]. Data from the Global Burden of Disease 2017 database indicate that the incidence of MBC increased from 8.5 thousand in 1990 to 23.1 thousand in 2017 with 123 countries showing a significant increasing trend in MBC incidence rates [ 2 ]. There are variations in the incidence of MBC among countries for instance, in Thailand MBC incidence was lower than that in Israel, and the rate of variability has been attributed to population-specific factors [ 3 ]. Additionally, disparities have been noted in the incidence, prevalence, mortality, and burden of cancer and related adverse health conditions in specific population groups [ 4 ]. Some of these disparities have been noted in the United States, where black men are reported to have higher incidence and mortality rates compared to white men in the context of all cancer [ 4 , 5 , 6 ].

Evidence suggests that MBC is mostly diagnosed late (49%) when the disease is more advanced compared to women (33%) leading to relatively worse prognosis [ 7 , 8 , 9 , 10 , 11 ]. This has been attributed to delayed presentation, lack of screening, reduced awareness by treating providers and a lack of awareness of the disease among men [ 12 , 13 , 14 , 15 ]. Consequently, MBCs are mainly diagnosed with more severe clinical manifestations with relatively complex tumour characteristics (i.e., larger sizes and extensive lymph node involvement) [ 16 ], associated with higher proportions of positive hormone receptors, which mostly results in prolonged treatment delay, and metastasis of the disease at diagnosis compared to female breast cancer [ 17 ]. This has been influenced by issues with lower socioeconomic status, barriers to accessing healthcare and insurance cover issues in the context of the United States, adherence to treatment, post-treatment follow-up, and stigma [ 7 , 18 , 19 , 20 ]. MBC patients suffer from a triple stigma including stigma by healthcare professionals, society, and especially by themselves as they struggle to accept the disease which has been labelled as a woman's disease [ 20 ].

Treatment for MBC has mainly been informed by available evidence for female breast cancer [ 21 ], and no randomised data exists for optimal management strategies for men including surgery, systemic therapy, and radiation [ 22 ]. Some guidelines have been published for the management of MBC [ 23 , 24 , 25 ]; however, these guidelines are rarely based on clinical trials leading to a paucity of literature on the evaluation of outcomes for MBC. According to Corrigan et al. [ 26 ], of the 131 breast cancer clinical trials conducted, there was only 0.087% of male patients represented among study participants.

Moreover, MBC being widely described as a 'woman’s disease' has psychosocially impacted the experience of men in terms of their body image and appearance as well as masculinity [ 27 , 28 ]. A critical psychosocial problem for MBC patients is concerns with body image [ 29 ], because both the disease and its treatment can lead to significant alterations to their looks and how the body functions [ 30 ]. With masculinity often associated with chest rather than breast [ 31 , 32 , 33 ], being linked to a “woman’s disease” attributed to the body part that men do not relate to is probably threatening their masculinity [ 34 ]. Men with breast cancer also face unique physical and emotional challenges however, there is inconclusive understanding of men’s experiences of the psychosocial implications of MBC as well as the supportive care needs [ 35 , 36 ]. Therefore, in this review, we explored the experiences of MBC patients and the management approaches across different demographic contexts.

Review question

What are the experiences and perceptions of MBC patients following diagnosis?

We conducted a mixed method systematic review with an interpretive and inductive stance [ 37 ] and reported in line with the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines [ 38 ].

Search strategy

We identified relevant studies through a search in six electronic databases: Global Health, CINAHL, Medline, PsycINFO, Embase, and Web of Science. Furthermore, we searched reference lists of included studies for additional studies. The search duration in these databases covered January 2000 to December 2023, and was updated in September 2023.

A combination of the following keywords was used for search strategy i) ‘Men’ OR ‘Male’ OR ‘Father’ OR ‘Husband’ AND ii) ‘Breast cancer’ OR ‘Breast carcinoma′ OR ‘Breast neoplasm’ OR ‘Breast tumour’ AND iii) ‘Experiences’ OR ‘Perceptions’ OR ‘Perspectives’ OR ‘Opinions’ AND iv) ‘Treatment’ OR ‘Approaches’ OR ‘Outcomes’. Multiple variations of the keywords were used including the truncations based on database requirements to broaden to capture all relevant studies.

Inclusion and exclusion criteria

This review included all primary studies of any design (qualitative, quantitative, or mixed methods) that report on MBC (included only men assigned male gender at birth); studies focussing on the experiences, perceptions, and treatment approaches for MBC; as well as studies conducted and reported in English (based on the resources available to the researchers). However, letters, editorials, commentaries, perspectives, case reports, opinion pieces, news reports and systematic reviews on MBC; studies reporting on cancers in men other than MBC; those that did not report on MBC experiences; as well as those reported in languages other than English were excluded.

Data extraction, quality assessment, synthesis and analysis

Search results were imported into Endnote reference manager (version 20) by the first reviewer (MA-O), duplicates removed and titles as well as abstracts were screened. The remaining studies were screened against the inclusion/ exclusion criteria, by three reviewers (MA-O, JB, OA), and any study for which inclusion was unclear was discussed and resolved by YS and TNA. Full texts studies were obtained if abstracts did not have enough information to determine the relevance of an article. Study variables such as authors, countries where studies were conducted, aims/objectives, study design, sample size and characteristics, experiences of MBC with verbatim quotes, MBC treatment approaches with outcomes and conclusions drawn were extracted to a common table (see Table  1 ).

We used a results-based convergent design [ 75 ] to guide data analysis, where we initially synthesised qualitative and quantitative findings separately, before integrating these findings from the two designs in the final analysis and synthesis (see Fig.  1 ). This allowed us to synthesise quantitative findings regarding treatment approaches of MBC and qualitative or mixed methods results on the experiences of MBC patients.

figure 1

A flow diagram on the results-based convergent design

Descriptive statistics was used in reporting the number of published studies and presented in a PRISMA flow diagram in Fig.  2 . We synthesised the descriptions of MBC experiences and treatment approaches reported across studies. All studies were analysed descriptively. To synthesise the data regarding the experiences of men with breast cancer, verbatim quotes reported in the qualitative studies were extracted by two authors (JB & TNA). An interpretive and inductive stance was employed [ 37 ] by reviewing verbatim quotes to generate codes (see Table  2 ). Similar codes were aggregated to generate sub-themes followed by formulation of higher order themes. For the quantitative data regarding the treatment modalities, we focused on describing the main reported treatment modalities rather than their frequencies. At the end of the analysis, both the qualitative findings and descriptions from the quantitative studies converged as one dataset. The themes generated from the initial process and the descriptions obtained from the quantitative studies formed the basis of undertaking a narrative synthesis.

figure 2

PRIMA flow chart of study search and selection process

The quality of included studies was assessed using the Quality Assessment Tool for Studies with Diverse Designs (QATSDD) tool [ 76 ], which is designed for use in mixed methods reviews and quality reporting in reviews that included qualitative, quantitative, mixed- and multi-methods research to ensure consistent and critical appraisal of relevant studies. In assessing study quality, studies were categorised as high quality if they achieved an aggregate score in excess of 70%, moderate quality were assigned to studies scoring between 50 and 70%, and those scoring less than 50% were assigned low quality (see Table  1 ). However, no study was excluded based on respective aggregate quality scores.

Study characteristics

Of the n  = 610 full-text articles assessed for eligibility. N  = 374 were excluded as these were letters, editorials, commentaries, perspectives, case reports, opinion pieces and news reports on MBC; including n  = 130 studies that did not report on MBC experiences and perceptions; and n  = 62 that were MBC related reviews (see Fig.  2 ). Following extensive search and screening, 44 studies were retained in the final synthesis and analysis, with publication years ranging from January 2000 to September 2023. Twenty-nine studies employed varied quantitative designs, 8 studies employed qualitative designs, and 6 studies employed mixed-method designs. Although most of the studies (n = 44) included only MBC, two retrospective studies compared males and females with breast cancer, and only the data reported on males were included in this review [ 58 , 68 ]. Study characteristics including quality assessment grading are reported in Table  1 .

Experiences and perceptions of males with breast cancer

As shown in Table  2 , three themes and nine sub themes emerged from the data which encapsulate the experiences of males with breast cancer.

Theme 1: Navigating through a threat to masculinity and one’s existence

This theme describes how males experienced the illness reflecting on detection, diagnosis, coming to terms with the disease, and disclosure. The subthemes are 1) emergence and awareness of a foreign illness and threat to one’s existence 2) coming to terms with a gendered disease and 3) opening up/ coming out of the illness closet. All included nine qualitative studies highlighted how the affected men perceived breast cancer as a threat to their sense of masculinity.

Emergence and awareness of a foreign illness and threat to one’s existence

Males generally perceived breast cancer as a feminine illness which cannot affect their bodies [ 31 , 34 ]. In fact, although all the men in the included studies had heard about breast cancer, most of them had not previously heard about breast cancer in males which made them rule out any possibility of ever living with it and may have contributed to delay in seeking healthcare [ 31 , 49 ]. This perception and the emerging non-specific symptoms often delayed early health seeking as the symptoms were interpreted as irrelevant or not requiring urgent attention [ 49 ]. It is worth highlighting that most of the affected men presented with palpable lump in the breast or discharge from the nipple of the affected breast. Some men had to be ‘pushed’ by their wives or partners to seek medical attention to rule out the possibility of breast cancer; a condition they felt was out of their scope [ 49 , 71 ]. A breast cancer diagnosis was met with varied emotions including being dumbfounded, shocked, surprised, debilitating stress, and a feeling of housing a feminised illness in a masculine body which threatened their sense of masculinity and personhood [ 13 , 31 , 34 , 49 ].

“…there is no reason why I shouldn’t have cancer, I’m only the same as anyone else. I’m just a bit disappointed really about where it got me. it’s not right on a man, is it? [ 31 ] (p.467). “From others at work, I always (hear) ‘admit it, you’re just trying to find excuses. You’re not a real man, or you wouldn’t have such an illness’. [ 34 ] (p.8). ‘I suppose the fact that it was breast cancer surprised me. The fact that it was cancer I suppose was a shock . . . So, I suppose a combination of both. You know the fact that it was breast cancer which I do not think I had heard of and the fact that it was cancer’’ [ 13 ] (p.336).

Receiving the diagnosis was challenging which some men kept to themselves or only informed family/ close friends [ 71 ]. The notion of breast cancer being a feminine illness made men view the disease as foreign or exotic to their bodies [ 49 ]. The growing awareness of the disease made the men feel a sense of oddity and shame for having a feminine illness alongside a feeling of losing one’s manhood to an illness not considered masculine [ 31 , 49 ]. Worry, anxiety, and uncertainty also marked their increasing awareness of the disease particularly regarding how the disease could distort the shape of their ‘masculine chest’ [ 13 ]. Despite the varied emotions, some males felt extremely lucky that the cancer was located at a site not considered ‘vital’ in terms of masculinity [ 67 ].

My biggest problem was how to tell my wife that I have a woman’s disease? Because I thought maybe you’re not a real man, perhaps half woman?” [ 34 ] (p.8). “Now when I first knew that I had it, I thought to myself …well how did Dickens get breast cancer? I’m not a woman. I’m a man. I was surprised more than anything… Women, it's an ever-present threat … Men – never occur to them. ‘‘When I first knew I did not want everyone knowing, because I did not want everyone coming round sympathising’’ . [ 13 ] (p.336).

Further to the above, the diagnosis of breast cancer forced the affected men to come face to face with their own mortality. This is because they felt a diagnosis of breast cancer threatened their existence and equated to a death sentence. The realisation of death lurking close by pushed the affected men to increase their efforts in attaining their dream before they died. This experience helped them to be more appreciative of their present lives, increased their consciousness about their health, and helped them to redefine their values and beliefs [ 60 ]:

“I appreciate life a lot more. Before my cancer, I didn’t take life seriously. I took life for granted. I didn’t appreciate the people in my life and the things I see. So, after the cancer, it was a good kick in the butt. Just how much you appreciate it, and also made me realise to go after my dreams, chase it, and achieve it. Go after it and every day is a gift” [ 60 ] (p.3).

Coming to terms with a gendered disease

Through the journey of receiving a breast cancer diagnosis and living with the illness, the affected men expressed the insights and perceptions they gained regarding living with an uncommon illness that is believed to affect mostly women [ 60 ]. Following the breast cancer diagnosis, males were faced with the reality of living with a condition they did not expect to have. Coming to terms with a feminised disease was gradual and a lonely journey for the affected men. In fact, some wished they could give their condition another name instead of breast cancer. The fear of being stigmatised made some men keep their diagnosis to themselves [ 13 , 32 ]. Others also felt a sense of awkwardness discussing such sensitive issues and would avoid [ 13 ]. Taken together, men with breast cancer often concealed or attempted to re-label their diagnosis to manage their sense of stigma, shame, and oddity as they navigated through coming to terms with living with a “feminine disease” in their masculine bodies [ 13 , 32 , 66 ]:

‘‘I told the guys I played golf with that I’d got cancer; I do not think so. I necessarily told them it was breast cancer’’. [ 13 ] (p.337) “…but if I did, I would talk about it as chest cancer. I wouldn’t use breast cancer. So that would be the term I would use, and, in the conversation, I would say that it is the same as breast cancer. It’s exactly the same thing; it’s just in my chest.” [ 66 ] (p. 964). “I think among old men they almost consider it to be a stigma, they almost don’t want to tell people, you know, it’s some kind of, I don’t know, a black mark, but I never looked at it that way…I think people younger would just view it a little differently, you know it’s cancer, it’s something they have to deal with, it doesn’t really matter what type of cancer it is.” [ 67 ] (p.37)

Opening up/ coming out of the illness closet

As the men gradually came to terms with living with the “foreign or exotic disease”, they were able to talk to their families and close friends about their diagnosis [ 13 ]. This required a lot of courage to navigate through such a sensitive issue. Interestingly, the men noted that the process of openly discussing their diagnosis in social spheres and coming out to others offered them an opportunity to reassert the meaning of masculinity, particularly as they recognize how fragile their masculine bodies are [ 31 ]:

“When I spoke to people about it, they thought I was telling fairy tales … that was really the worst thing about it.” [ 34 ] (p.8). “I want to prove to everybody that MBC is not a women’s disease and that a normal man can have MBC.” [ 31 ] (p.468).

In two studies, however, the authors described the phenomenon of selective disclosure in which the men only disclosed their illness to selected persons only [ 20 , 60 ]. For some men, the selective disclosure also meant revealing just the diagnosis, but not going further to reveal how they are experiencing the treatment process or the aftermath of the illness:

“The children know and our closest friends know, the very closest. Why? Because I disappeared for a while. I don’t talk about it within the family, not at all. Nobody talks with me about it, but they know. It is only information, and that’s it, not about the experience and not about the surgery, and not about the treatment” [ 20 ] (p.5).

Theme 2: Navigating through treatment

The theme captures the experiences of undergoing breast cancer treatment/ management following their diagnosis. The subthemes are 1) therapeutic interventions 2) navigating through feminised treatment pathways and 3) living with the effects of care/ ongoing treatment. All included qualitative, quantitative, and mixed method studies ( n  = 44) highlighted the treatment experiences and pathways respectively.

Therapeutic interventions

Several therapeutic interventions/ treatments were reported across the included studies. Five categories of treatments were ascertained across the included studies, and these are surgery, radiotherapy, chemotherapy, hormonal therapy, and palliative care. Surgical interventions included mastectomy with axillary dissection, mastectomy with sentinel node biopsy (both for men with late-stage breast cancer presentation), and lumpectomy [ 7 , 40 , 45 , 46 , 47 , 48 ]. Cronin et al., [ 46 ] noted that surgery and chemotherapy receipt were more likely among men up to age 65. In some studies, surgical interventions were the main forms of treatment with radiotherapy, chemotherapy, and hormone therapy playing adjuvant roles. For instance, in one study that included 37 men with breast cancer, radiotherapy (89.2%), hormonal therapy (56.7%), and chemotherapy (91.8%) were adjuvant therapies after surgery [ 48 ]. In one study, the authors reported several therapeutic regimens offered to men with breast cancer which included breast conserving surgeries, unilateral/ bilateral mastectomy, often with no reconstruction [ 44 ]. One third of the male breast cancer patients in the same study ( n  = 21) felt somewhat or very uncomfortable with their appearance after the surgery. Receipt of treatment was remarkably similar between blacks and whites in both age groups. Older black and white men had lower receipt of chemotherapy (39.2% and 42.0%, respectively) compared with younger patients (76.7% and 79.3%, respectively). Younger black men had a 76% higher risk of death than younger white men after adjustment for clinical factors only (HR, 1.76; 95% CI, 1.11 to 2.78), but this difference significantly diminished after subsequent adjustment for insurance and income (HR, 1.37; 95% CI, 0.83 to 2.24). In those age 65 years, the excess risk of death in blacks versus whites was nonsignificant and not affected by adjustment for covariates.

Navigating through feminised treatment pathways

Despite the reality of breast cancer among males, the care pathways and healthcare payment frameworks across various healthcare systems are significantly tailored to the needs of females which reinforces the notion of the disease as a feminine in nature [ 31 , 71 ]. A study from Germany highlighted the difficulty that these men experience in finding a physician as the practitioners felt their breast care specialty targeted women and would lose on reimbursement [ 34 ]. Even in facilities where they were given satisfactory care, the men felt the services and procedures still failed to consider their unique needs as men with breast cancer [ 31 , 42 , 71 ]. Some men were mistakenly addressed as females on the assumption that only females experienced breast cancer [ 34 ]. Male-specific psychosocial support and information were generally lacking across the studies. Information leaflets mostly contained pictures of female breast cancer patients which made the men feel excluded [ 34 ]. In fact, they felt the service was not designed for them:

“My GP said: ‘I don’t know what to do any more, it’s not my specialty area. I’ll have to refer you to someone else’. And the other doctor said, ‘This is a women’s practice (…) and we can’t get reimbursed for men, we don’t want men here.’” [ 34 ] (p.9). ‘‘. . . but I think as a male the information that I was given was female orientated and it could have been better presented for me and . . .I know that every case is different, but it was lacking in that respect’’. [ 13 ] (p.336).

Further to the above, some men had several challenges in scheduling for therapeutic regimen such as mammography [ 67 ]. Interactions with healthcare providers were often considered awkward as the providers often did not know what to say to the men with breast cancer. Subsequently, most men with breast cancer undergoing treatment often felt like outsiders, out of place, marginalised, and alone:

‘No information. Nothing at all. It was like men; you are on your own. I daresay women aren’t left like that . . .On leaving after the first operation the nurse gave me a leaflet, a piece of paper with women on it doing exercises you have to do and that was it’’. [ 13 ] (p.336). “I find that dealing with the mammograms and the technical staff to kind of tiptoe around you and put you in certain places because they don’t expect a male to be there, right, so they got women walking around in their gowns, so they don’t want you in those areas… they kind of shunt you into an isolated, a more isolated area so you’re not seeing the women walking by.” [ 66 ] (p.967).

Living with the effects of care/ ongoing treatment

Men undergoing treatment for breast cancer felt their lives, roles, and occupations were impacted adversely by the treatment regimen [ 60 ]. The clinical management process of the disease, in fact, further heightened the gendered essence of the disease. For men who underwent surgical intervention, the mastectomy scar served as a permanent reminder of the disease impacted on their masculinity [ 66 ]. Others felt their chest had deformed due to the scar [ 71 ]. The typical exposure of the male chest at leisure activities such as the beach was considered a no-go area to conceal the scar from public view. The scars also evoked a sense of perceived stigma among these men [ 32 ]:

“I’ve been abroad and sunbathed. People do look, they do look” [ 71 ] (p.1835). “I don’t feel like a complete person either because I’ve got something missing, haven’t I? ... My nipples are not there anymore. Sometimes I look in the mirror . . . I don’t like doing that. It’s gone. . . There’s a scar across there. . .Doctor said I look like a patchwork quilt. So, I don’t bother taking my shirt off now. And something else … yes you ought to have a tattoo as a nipple’’. [ 13 ] (p.337).

For men who underwent hormone therapy, it was observed that the side effects of the various medications threatened their notion of being a male. Experiencing erectile dysfunction and loss of libido were really challenging for these men as they felt they had lost their sense of masculinity or what made them men [ 34 , 77 ]. Hair loss from chemotherapy was also challenging and frustrating for them [ 43 ]. These men felt as though they had been transformed to ‘menopausal women’ [ 34 ].

“We’re candid and honest with one another … male sexual potency has gone.” [ 34 ] (p.9). “This has killed my sex life; I can no longer get an erection. I’m on this Tamoxifen which I’ve got to take for 5 years. You know it’s driving me mad. I get free Viagra but there is nothing there. There are no feelings or anything like that and it’s terrible. I couldn’t get an erection or nothing. I don’t know what it was, I just felt so no, no (silence) I just felt so embarrassed.” [ 31 ] (p.467).

Further to the above, some men felt they were a burden to others as they had to rely on others to have their needs met. Younger males felt their traditional roles as providers of the family was threatened as their dependence increased with a slow return to work and had to be supported by their spouses [ 54 ]:

“You start to receive only sickness benefits and when all of a sudden, you have over 500 euro less, you have to first see how you manage with that. And for me [...] it was even more because I only have a 60% part-time job and work as a freelancer on the side. And that I couldn't do any longer either.” [ 54 ] (p.6).

Theme 3: Coping and support systems

The theme describes how men with breast cancer coped with the disease, treatment process, aftercare/ rehabilitative care, and the available support and it was reported across qualitative ( n  = 9), quantitative ( n  = 5) and mixed methods ( n  = 4) studies. The subthemes are 1) active coping strategies 2) family support and 3) support from healthcare providers and other support groups.

Active coping strategies

Although the breast cancer diagnosis was considered threatening with intense emotional stress, some affected men remained optimistic and hopeful of improved outcomes. Affected men often worked towards accepting the disease which made the navigation process less challenging [ 47 ]. The treatment process and aftercare phase offered the affected men an opportunity to amend or reformulate their notion of masculinity [ 66 ]. Although dealing with the disease was difficult, the men reportedly gained new insights in life which helped to reshape their worldviews and life priorities [ 14 ]. In addition, previous experience with breast cancer in the family was associated with use of non-repressing coping styles (X 2 [1, N  = 26] r  = 5.60, p  < 0.05). There was also a higher use of mature defence patterns (superior healthy neurotic functioning) in patients who use non-repressive coping [ 70 ]. Despite the identified active coping mechanisms, one study reported that majority (70%) of men with breast cancer used immature and neurotic defensive functioning and 53.8% used a repressive approach to bottle up their emotions and concerns and [ 70 ]:

“I was kind of self-conscious the first year or so but um, I’m in pretty good shape, I’m relatively muscular, not super muscular, but I’m toned, I’m in shape, and I think a lot of times unless I’m really up close to people, I think a lot of times they don’t even see it… I’m not self-conscious. I go on vacation or go swimming at the beach, I don’t feel like people are staring at me.” [ 67 ] (p.38) “Breast cancer, for me, means a whole complex of experiences, of realisations. It’s like being in the military, you know. You meet somebody who’s been in the military, you don’t have to say anything. But if you meet someone who hasn’t, there’s not a way in the world to describe what it’s like.” [ 67 ] (p.38)

Family support

Studies found that majority of patients (61.3–80%) disclosed and discussed their diagnosis with their spouses and close families while 4–21% refused to disclose or discuss with anyone [ 7 , 13 , 61 ]. This might be because less stigmatization was reported from close families and friends compared to broader social settings [ 32 ]. Such disclosure might also be protective as availability of marital support was found to influence treatment choice and outcomes. Men who were not currently married received chemotherapy significantly less often [ 52 ] and had significantly higher (in some cases up to 21%) mortality than married ones [ 52 , 53 ].

This was corroborated by included qualitative studies which reported on the family support that men affected with breast cancer received. Spousal support was identified as a significant resource to seeking healthcare in the first instances as some wives had to push their partners to seek medical care [ 31 , 57 ]. Spousal and family support also helped men to navigate through the breast cancer diagnosis, coming to terms with the disease [ 49 , 57 ]. Family support was also an essential resource during the treatment and aftercare phase as family members offered emotional and practical support [ 47 ]:

“My wife was my support – she and I talked about everything. At the beginning we talked about it and agreed that I would have her as my support and she would have her family to support her through. It worked well and I also got support from her family . . . mine were useless’’. [ 13 ] (p. 338).

Support from healthcare providers and other support groups

Studies reported the dimensions, contents and timing of information needs demonstrated by the patients. Men with breast cancer acknowledged the support received from healthcare providers regarding diagnosis, information, treatment options, and aftercare support [ 49 , 57 ] with the most common source of information being verbal (92%), leaflets or booklets (53–71%) and internet (20%) [ 61 ]. Yet, 36–65% of participants felt their needs were not always met and wanted more information on various contents (particularly sexuality related information) at different times in their treatment (early/acute effects, late effects and ongoing quality of life) and in a more male specific manner [ 42 ].

Men with Breast cancer faced challenges in accessing needed support from healthcare facilities. Included studies reported experience of embarrassment and stigmatization within healthcare facilities where male breast cancer patients were meant to get support. 51.6% of patients experienced "extreme" or "very" severe embarrassment while waiting in the clinic among other female patients [ 13 ]. The experience of stigmatization was found to be higher within the cancer care system than other social surroundings with significantly higher stigmatization incidences reported in rehabilitation settings (mean = 1.50) and during hospitalisations (mean = 1.20) [ 53 ].

A mixed finding was observed regarding usage of peer supports. For one-to-one peer support, Iredale et al. (2006) reported low utilisation of formal support services with only 19% of participants speaking to other men who had breast cancer and only 1 in 4 indicating they would have liked that opportunity after their diagnosis. However, Midding et al. [ 53 ] found that more men (63.2%) had a one-on-one peer support from a female Breast Cancer Patient compared to 24.2% from another male breast cancer patient. This is consistent with the qualitative data which showed some men appreciated the opportunity to talk to other men with breast cancer on one-to-one basis [ 34 , 71 ], other men did not prefer this and were satisfied with the support offered by the healthcare providers and their families [ 13 ]:

‘‘…none of the guys wanted to have self-help groups ... I don’t think they need the psychological support that perhaps women do, and women tend to congregate and talk about these things anyway. I think this is, of course ... research I know ... but actually quite therapeutic in a way’’. [ 13 ] (p.338). “To be honest, I don’t know how I would be managing if I had never had (the support group). They gave me back the will to live and I will always be grateful for that.” [ 43 ] (p. 9).

In terms of group peer support, studies reported that only 15.3% of the participants were part of a peer support group and majority (96.3%) of participants who were not currently part of a support group did not wish to be part of a support group whether male only or mixed sex [ 53 , 61 ].

Breast cancer is generally perceived to be a disease common among women albeit incidence among men is slowly rising, creating a need for health systems to be responsive to their needs. To this end, this review sought to develop a comparative understanding of the experiences of men with breast cancer and the treatment options available to them across different demographic settings. The review findings highlight the embodiment of breast cancer as a ‘feminine’ disease which is incongruent with what it means to be a ‘man’ and hegemonic masculinity discourses. Throughout the trajectory of the disease (that is, from diagnosis to aftercare), the review findings underscore the gendered nature of the disease with a lack of health system preparedness to support men who develop a disease perceived to be ‘feminine’. Though the treatment pathways were similar to those observed in the management of female breast cancer patients, they do not necessarily meet the unique needs of MBC across the disease trajectory warranting urgent attention considering the increasing prevalence of the disease among men. Male-specific treatment pathways, ongoing education, and professional support are also required.

The breast is seen as a symbol of femininity, and as incongruent with being male, together with the significant public health emphasis on the prevention of breast cancer among females [ 78 , 79 ] have further championed the perception that breast cancer is a feminine illness [ 56 , 67 ]. Thus, it was not surprising that the finding regarding being out of sync with one’s body resonated across the included studies. The breast cancer diagnosis which commenced the illness trajectory was really challenging for the men and filled with varied emotions. Despite the difficulty, the professional support available was often gendered and unsuitable to their needs. Thus, they mostly had to rely on their spouses and close families/ friends if they were able to open up to them, which may take some time. Coupled with the hegemonic masculinity ideology that a man must always be in charge and not demonstrate any emotions which can be perceived as weakness, it is likely that men will navigate through these on their own which can make the journey very lonely for them. Agreeing with a previous study, depressive symptoms, anxiety, and traumatic stress symptoms were common occurrences following the breast cancer diagnosis [ 43 ]. The culture of silence around the issue can lead to utilising avoidant coping mechanisms which may delay support seeking among men. Taken together, the findings highlight a need for tailor-made, individualised counselling support service for men before, during, and after breast cancer diagnosis. The need for healthcare professionals to consider the impact of the MBC on men cannot, therefore, be overemphasised.

Commencing treatment and aftercare/ rehabilitative support is an equally challenging phase for men living with breast cancer. A previous study has observed that gender impacts on the experience with breast cancer treatment [ 15 ]. The review findings highlighted the ‘feminised’’ nature of the treatment pathways with some practitioners not even knowing how to support the affected men. Information leaflets and other educational materials were generally noted to be filled with images of females which made the men feel out of place. Overall, these can serve as structural barriers which potentially deter men from seeking help even when required [ 34 ]. Undoubtedly, breast cancer affects more females than males. However, healthcare service delivery should be tailored to the unique needs of men to overcome the feeling of marginalisation or being left out. The impact of the therapeutic regimen should also be highlighted particularly as they can lead to loss of libido or erectile dysfunction which further diminishes one’s sense of being a man in relation to societal norms. Surgical procedures can lead to scars which serve as permanent reminders of the illness which can have life-long impact on men. Professional support should therefore not end after the diagnosis phase but should extend to the entire treatment continuum and aftercare. There is also a need to raise awareness of male breast cancer among healthcare practitioners to improve their approach to individuals through person-centred and male-specific care strategies. It may be worth reiterating the recommendation by Nguyen et al., [ 34 ] suggesting a guideline targeting men with breast cancer to support healthcare practitioners in the health and social service delivery process.

The need for support was reiterated throughout the review, and this is corroborated in a previous study where family and spousal support was critically important for men with advanced prostate cancer [ 80 ]. Interestingly, mixed findings were observed regarding the need for male-specific support groups. Although this may be based on individual preferences, it may also emanate from the hegemonic masculinity ideology [ 80 , 81 ] or coping styles such as disengagement [ 20 ] as men may appear ‘stoic’ in the presence of such difficult moments and may not want to seek help [ 34 , 82 ]. A breast cancer diagnosis can profoundly impact masculinity, with men grappling with navigating a threat to masculinity which collectively challenges one's sense of self and traditional gender roles [ 82 , 83 , 84 ].

Recent research shows changing perceptions of breast cancer as a "feminine disease" due to awareness campaigns and shifts in societal attitudes [ 85 , 86 ]. Additionally, demographic factors like location of treatment, socioeconomic status, and age have been found to affect the quality of care and outcomes, while acknowledging the male breast cancer experience and its shared emotional aspects with women's experiences [ 87 , 88 ]. These highlights evolving healthcare practices and societal norms regarding breast cancer.

Despite this, it is still cogent to understand their lived experiences and advocate for men support groups, if they would like to join one, as they navigate through the diagnosis, treatment, and aftercare pathway. This study presents the synthesis of multicultural evidence to highlight the cross-cultural similarity in the reaction and lived experience of men when faced with the diagnosis of breast cancer.

Strengths and limitations

The strength of this mixed method is the inclusion of studies from different countries and settings in addition to including and synthesising studies on the experiences of patients with male breast cancer from diagnosis to aftercare. Notwithstanding, there are some limitations that need to be highlighted. Firstly, a real limitation of our review was including only studies published in English. Excluding studies that used a language other than English, potentially led to information loss that could come from relevant studies written in other languages and restricts this mixed methods review only to the views and perception of men living in English speaking countries or countries where practitioners write and publish in English. Secondly, we acknowledge that younger and older men may have unique experiences while navigating breast cancer diagnosis and treatment. These nuances were not captured in the current review and may be worth exploring in future studies.

Men experience a myriad of issues following a breast cancer diagnosis, underscored by their ideology of masculinity. Our findings suggest the need for healthcare professionals’ training and education on managing interactions with MBC patients in a way that does not propagate a sense of awkwardness and otherness in a feminised support structure. Additionally, policy must address the structural barriers to treatment access for MBC including healthcare finance reimbursements that limit access to gendered specialist breast cancer treatments. Awareness creation efforts of MBC among the public as well as healthcare practitioners are urgently required to explain to the public through television programmes and awareness meetings that breast cancer is a disease like any other that affects both men and women. Creating such awareness could lead to changing the perception of men and promote early diagnosis, adherence to treatments, post-treatment monitoring, oncological results, and a better quality of life. Professional care intervention and support for MBC should not end after the diagnosis phase but should extend to the entire treatment continuum and aftercare. Preserving sexual function is an important finding highlighted from this review. Research will be needed to develop and test testosterone-preserving treatment modalities or optimising existing therapies in a way that is relevant to the priorities of MBC. This will also require the development of specialised guidelines for healthcare practitioners on MBC to optimise care and treatment for MBCs in a person-centred manner as suggested by other studies. To develop such individualised support frameworks, it is imperative to understand the specific needs, priorities, and support preferences among MBC patients.

Availability of data and materials

All data generated or analysed during this study are included in this published article.

Society AC. Key Statistics for Breast Cancer in Men 2022 [Available from: https://www.cancer.org/cancer/breast-cancer-in-men/about/key-statistics.html .

Chen Z, Xu L, Shi W, Zeng F, Zhuo R, Hao X, et al. Trends of female and male breast cancer incidence at the global, regional, and national levels, 1990–2017. Breast Cancer Res Treat. 2020;180(2):481–90.

Article   PubMed   Google Scholar  

Health UDo, Services H. Making Cancer Health Disparities History: Report of the Trans-HHS Cancer Health Disparities Progress Review Group. Washington, DC: U.S. Department of Health and Human Services; 2004.

Hayat MJ, Howlader N, Reichman ME, Edwards BK. Cancer statistics, trends, and multiple primary cancer analyses from the Surveillance, Epidemiology, and End Results (SEER) Program. Oncologist. 2007;12(1):20–37.

Morris AM, Rhoads KF, Stain SC, Birkmeyer JD. Understanding racial disparities in cancer treatment and outcomes. J Am Coll Surg. 2010;211(1):105–13.

Rhoads KF, Cullen J, Ngo JV, Wren SM. Racial and ethnic differences in lymph node examination after colon cancer resection do not completely explain disparities in mortality. Cancer. 2012;118(2):469–77.

Co M, Lee A, Kwong A. Delayed presentation, diagnosis, and psychosocial aspects of male breast cancer. Cancer Med. 2020;9(10):3305–9.

Article   PubMed   PubMed Central   Google Scholar  

Ndom P, Um G, Bell EMD, Eloundou A, Hossain NM, Huo D. A meta-analysis of male breast cancer in Africa. The Breast. 2012;21(3):237–41.

Fiala L, Coufal O, Fait V, Foretova L. Male breast cancer-our experience. Rozhledy v chirurgii: mesicnik Ceskoslovenske chirurgicke spolecnosti. 2010;89(10):612–8.

PubMed   CAS   Google Scholar  

Rudlowski C. Male breast cancer. Breast care. 2008;3(3):183–9.

Robinson JD, Metoyer KP, Bhayani N. Breast cancer in men: a need for psychological intervention. J Clin Psychol Med Settings. 2008;15(2):134–9.

Thomas E. Men’s awareness and knowledge of male breast cancer. AJN The American Journal of Nursing. 2010;110(10):32–7.

Iredale R, Brain K, Williams B, France E, Gray J. The experiences of men with breast cancer in the United Kingdom. Eur J Cancer. 2006;42(3):334–41.

Pituskin E, Williams B, Au H-J, Martin-McDonald K. Experiences of men with breast cancer: A qualitative study. Journal of Men’s Health and Gender. 2007;4(1):44–51.

Article   Google Scholar  

Naymark P. Male breast cancer: incompatible and incomparable. Journal Of Men’s Health And Gender. 2006;3(2):160–5.

Ottini L, Palli D, Rizzo S, Federico M, Bazan V, Russo A. Male breast cancer. Crit Rev Oncol Hematol. 2010;73(2):141–55.

Rudan I, Rudan N, Basić N, Basić V, Rudan D. Differences between male and female breast cancer. II. Clinicopathologic features. Acta Medica Croatica. 1997;51(3):129–33.

Sineshaw HM, Freedman RA, Ward EM, Flanders WD, Jemal A. Black/white disparities in receipt of treatment and survival among men with early-stage breast cancer. J Clin Oncol. 2015;33(21):2337–44.

DeSantis CE, Miller KD, Goding Sauer A, Jemal A, Siegel RL. Cancer statistics for African Americans, 2019. CA Cancer J Clin. 2019;69(3):211–33.

Levin-Dagan N, Baum N. Passing as normal: Negotiating boundaries and coping with male breast cancer. Soc Sci Med. 2021;1(284):114239.

Sousa B, Moser E, Cardoso F. An update on male breast cancer and future directions for research and treatment. Eur J Pharmacol. 2013;717(1–3):71–83.

Article   PubMed   CAS   Google Scholar  

Fentiman IS. Surgical options for male breast cancer. Breast Cancer Res Treat. 2018;172:539–44.

Cardoso F, Bartlett J, Slaets L, Van Deurzen C, van Leeuwen-Stok E, Porter P, et al. Characterization of male breast cancer: results of the EORTC 10085/TBCRC/BIG/NABCG International Male Breast Cancer Program. Ann Oncol. 2018;29(2):405–17.

Korde LA, Zujewski JA, Kamin L, Giordano S, Domchek S, Anderson WF, et al. Multidisciplinary meeting on male breast cancer: summary and research recommendations. J Clin Oncol. 2010;28(12):2114.

Miao H, Verkooijen H, Chia K-S, Bouchardy Magnin C, Pukkala E, Larønningen S, et al. Incidence and outcome of male breast cancer: an international population-based study. J Clin Oncol. 2011;29(33):4381–6.

Corrigan KL, Mainwaring W, Miller AB, Lin TA, Jethanandani A, Espinoza AF, Piotrowski M, Fuller CD, Stauder MC, Shaitelman SF, Perkins GH. Exclusion of men from randomized phase III breast cancer clinical trials. Oncologist. 2020;25(6):e990–2.

Article   PubMed   PubMed Central   CAS   Google Scholar  

Manier KK, Rowe LS, Welsh J, Armstrong TS. The impact and incidence of altered body image in patients with head and neck tumors: a systematic review. Neuro-Oncology Practice. 2018;5(4):204–13.

Himmelstein MS, Sanchez DT. Masculinity impediments: Internalized masculinity contributes to healthcare avoidance in men and women. J Health Psychol. 2016;21(7):1283–92.

Paraskeva N, Clarke A, Harcourt D. Altered appearance from cancer. Body image care for cancer patients: Principles and practices. 2018;3:131–60.

Google Scholar  

Fingeret MC, Teo I, Epner DE. Managing body image difficulties of adult cancer patients: lessons from available research. Cancer. 2014;120(5):633–41.

Donovan T, Flynn M. What makes a man a man?: The lived experience of male breast cancer. Cancer Nurs. 2007;30(6):464–70.

Midding E, Halbach SM, Kowalski C, Weber R, Würstlein R, Ernstmann N. Men with a “woman’s disease”: Stigmatization of male breast cancer patients—A mixed methods analysis. Am J Mens Health. 2018;12(6):2194–207.

Pituskin E, Williams B, Au HJ, Martin-McDonald K. Experiences of men with breast cancer: A qualitative study. Journal of Men’s Health and Gender. 2007;4(1):44–51.

Nguyen TS, Bauer M, Maass N, Kaduszkiewicz H. Living with male breast cancer: a qualitative study of men’s experiences and care needs. Breast Care. 2020;15(1):6–13.

Younas A. Epistemic injustice in health care professionals and male breast cancer patients encounters. Ethics Behav. 2021;31(6):451–61.

Article   MathSciNet   Google Scholar  

Ly D, Forman D, Ferlay J, Brinton LA, Cook MB. An international comparison of male and female breast cancer incidence rates. Int J Cancer. 2013;132(8):1918–26.

Elbardan H, Kholeif AO, Elbardan H, Kholeif AO. An interpretive approach for data collection and analysis. Enterprise Resource Planning, Corporate Governance and Internal Auditing: An Institutional Perspective. 2017:111–65. 

Moher D, Liberati A, Tetzlaff J, Altman DG, Group* P. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Ann intern med. 2009;151(4):264–9.

Adekolujo OS, Tadisina S, Koduru U, Gernand J, Smith SJ, Kakarala RR. Impact of marital status on tumor stage at diagnosis and on survival in male breast cancer. Am J Mens Health. 2017;11(4):1190–9.

Ahmed A, Ukwenya Y, Abdullahi A, Muhammad I. Management and outcomes of male breast cancer in Zaria. Nigeria Int J Breast Cancer. 2012;2012:845143.

PubMed   Google Scholar  

Avila J, Herrick B, Attai DJ, Leone JP. Treatments for breast cancer in men: late effects and impact on quality of life. Breast Cancer Res Treat. 2023;201(3):489–98.

Bootsma TI, Duijveman P, Pijpe A, Scheelings PC, Witkamp AJ, Bleiker EM. Unmet information needs of men with breast cancer and health professionals. Psychooncology. 2020;29(5):851–60.

Brain K, Williams B, Iredale R, France L, Gray J. Psychological distress in men with breast cancer. J Clin Oncol. 2006;24(1):95–101.

Chichura A, Attai DJ, Kuchta K, Nicholson K, Kopkash K, Pesce C, Yao K. Male breast cancer patient and surgeon experience: the male WhySurg study. Ann Surg Oncol. 2022;29(10):6115–31.

Crew KD, Neugut AI, Wang X, Jacobson JS, Grann VR, Raptis G, et al. Racial disparities in treatment and survival of male breast cancer. J Clin Oncol. 2007;25(9):1089–98.

Cronin PA, Romanoff A, Zabor EC, Stempel M, Eaton A, Smyth LM, et al. Influence of age on the clinical outcome of breast cancer for men and the development of second primary cancers. Ann Surg Oncol. 2018;25(13):3858–66.

Duarte do Amaral DE, Manfrim Muniz R, Habekost Cardoso D, Tuerlinckx Noguez P, Ferreira Fagundes R, Costa Viegas A. Male breast cancer: the survivor's context. J Nurs UFPE/Revista de Enfermagem UFPE. 2017;11(5).

El-Beshbeshi W, Abo-Elnaga EM. Male breast cancer: 10-year experience at mansoura university hospital in Egypt. Cancer Biol Med. 2012;9(1):23.

PubMed   PubMed Central   Google Scholar  

France L, Michie S, Barrett-Lee P, Brain K, Harper P, Gray J. Male cancer: a qualitative study of male breast cancer. The Breast. 2000;9(6):343–8.

Giordano SH, Perkins GH, Broglio K, Garcia SG, Middleton LP, Buzdar AU, et al. Adjuvant systemic therapy for male breast carcinoma. Cancer. 2005;104(11):2359–64.

Halbach SM, Midding E, Ernstmann N, Würstlein R, Weber R, Christmann S, et al. Male breast cancer patients’ perspectives on their health care situation: a mixed-methods study. Breast Care. 2020;15(1):22–9.

Harlan LC, Zujewski JA, Goodman MT, Stevens JL. Breast cancer in men in the United States: a population-based study of diagnosis, treatment, and survival. Cancer. 2010;116(15):3558–68.

Hill TD, Khamis HJ, Tyczynski JE. Comparison of male and female breast cancer incidence trends, tumor characteristics, and survival. Ann Epidemiol. 2005;15(10):773–80.

Hiltrop K, Heidkamp P, Halbach S, Brock-Midding E, Kowalski C, Holmberg C, et al. Occupational rehabilitation of male breast cancer patients: Return patterns, motives, experiences, and implications—A qualitative study. Eur J Cancer Care. 2021;30(4): e13402.

Hoffman A, Horesh N, Shabtai M, Forschmidt E, Rosin D, Gutman M, et al. Breast Cancer in Men: A Single Center Experience Over a Period of 22 years. The Israel Medical Association Journal: IMAJ. 2020;22(3):160–3.

Kowalski C, Steffen P, Ernstmann N, Wuerstlein R, Harbeck N, Pfaff H. Health-related quality of life in male breast cancer patients. Breast Cancer Res Treat. 2012;133(2):753–7.

Midding E, Halbach SM, Kowalski C, Weber R, Würstlein R, Ernstmann N. Social support of male breast cancer patients—a mixed-methods analysis. Am J Mens Health. 2019;13(4):1557988319870001.

Nahleh ZA, Srikantiah R, Safa M, Jazieh AR, Muhleman A, Komrokji R. Male breast cancer in the veterans affairs population: a comparative analysis. Cancer. 2007;109(8):1471–7.

Özkurt E, Tükenmez M, Yılmaz R, Cabioğlu N, Müslümanoğlu M, Dinççağ AS, et al. Favorable long-term outcome in male breast cancer. European Journal of Breast Health. 2018;14(3):180.

Potter AM, Bentz B, Crue L, Leiby S, Bashi S, Maguire K, Meyers J, Mieczkowski K. Men’s Lived Experiences of Breast Cancer and Changes in Occupation. Occup Ther Int. 2023;13:2023.

Rayne S, Schnippel K, Thomson J, Reid J, Benn C. Male breast cancer has limited effect on survivor’s perceptions of their own masculinity: a record review and telephone survey of patients in Johannesburg, South Africa. Am J Mens Health. 2017;11(2):246–52.

Sanguinetti A, Polistena A, Lucchini R, Monacelli M, Galasse S, Avenia S, et al. Male breast cancer, clinical presentation, diagnosis and treatment: Twenty years of experience in our Breast Unit. Int J Surg Case Rep. 2016;20:8–11.

Article   PubMed Central   Google Scholar  

Sarmiento S, McColl M, Musavi L, Gani F, Canner JK, Jacobs L, et al. Male breast cancer: a closer look at patient and tumor characteristics and factors that affect survival using the National Cancer Database. Breast Cancer Res Treat. 2020;180(2):471–9.

Shah S, Bhattacharyya S, Gupta A, Ghosh A, Basak S. Male breast cancer: a clinicopathologic study of 42 patients in eastern India. Indian J Surg Oncol. 2012;3(3):245–9.

Shin JY, Kachnic LA, Hirsch AE. The impact of race in male breast cancer treatment and outcome in the United States: a population-based analysis of 4,279 patients. Int J Breast Cancer. 2014;2014:685842.

Skop M, Lorentz J, Jassi M, Vesprini D, Einstein G. “Guys don’t have breasts”: The lived experience of men who have BRCA gene mutations and are at risk for male breast cancer. Am J Mens Health. 2018;12(4):961–72.

Thompson EH Jr, Haydock AS. Men’s lived experiences with breast cancer: The double consciousness of marginal men. Sex Roles. 2020;82(1–2):28–43.

Visram H, Kanji F, Dent S. Endocrine therapy for male breast cancer: rates of toxicity and adherence. Curr Oncol. 2010;17(5):17–21.

Wang Y, Chen K, Yang Y, Tan L, Chen L, Zhu L, et al. Incidence and survival outcomes of early male breast cancer: a population-based comparison with early female breast cancer. Ann transl med. 2019;7(20):536.

Weber R, Ehrenthal JC, Brock-Midding E, Halbach S, Würstlein R, Kowalski C, Ernstmann N. Defense mechanisms and repressive coping among male breast cancer patients. Front Psych. 2021;10(12): 718076.

Williams B, Iredale R, Brain K, France E, Barrett-Lee P, Gray J. Experiences of men with breast cancer: an exploratory focus group study. Br J Cancer. 2003;89(10):1834–6.

Yadav BS, Sharma SC, Singh R, Dahiya D, Ghoshal S. Male breast cancer: Outcome with adjuvant treatment. J Cancer Res Ther. 2020;16(6):1287.

Yoney A, Kucuk A, Unsal M. Male breast cancer: a retrospective analysis. Cancer/Radiothérapie. 2009;13(2):103–7.

Zongo N, Ouédraogo S, Korsaga-Somé N, Somé OR, Go N, Ouangré E, et al. Male breast cancer: diagnosis stages, treatment and survival in a country with limited resources (Burkina Faso). World journal of surgical oncology. 2018;16(1):1–7.

Hong QN, Pluye P, Bujold M, Wassef M. Convergent and sequential synthesis designs: implications for conducting and reporting systematic reviews of qualitative and quantitative evidence. Syst Rev. 2017;6(1):1–14.

Sirriyeh R, Lawton R, Gardner P, Armitage G. Reviewing studies with diverse designs: the development and evaluation of a new tool. J Eval Clin Pract. 2012;18(4):746–52.

Ruddy KJ, Giobbie-Hurder A, Giordano SH, Goldfarb S, Kereakoglow S, Winer EP, et al. Quality of life and symptoms in male breast cancer survivors. The Breast. 2013;22(2):197–9.

Dozier R. Beards, breasts, and bodies: Doing sex in a gendered world. Gend Soc. 2005;19(3):297–316.

Sulik GA. Pink ribbon blues: How breast cancer culture undermines women’s health. Walton Street, Oxford: Oxford University Press; 2010.

Salifu Y, Almack K, Caswell G. ‘My wife is my doctor at home’: A qualitative study exploring the challenges of home-based palliative care in a resource-poor setting. Palliat Med. 2021;35(1):97–108.

Connell RW, Messerschmidt JW. Hegemonic masculinity: Rethinking the concept. Gend Soc. 2005;19(6):829–59.

Salifu Y, Almack K, Caswell G. ‘Out of the frying pan into the fire’: a qualitative study of the impact on masculinity for men living with advanced prostate cancer. Palliative Care and Social Practice. 2023;17:26323524231176828.

Quincey K, Williamson I, Wildbur D. Men with breast cancer and their encounters with masculinity: An interpretative phenomenological analysis using photography. Psychology of Men & Masculinities. 2021;22(4):690.

Quincey K, Williamson I, Winstanley S. ‘Marginalised malignancies’: A qualitative synthesis of men’s accounts of living with breast cancer. Soc Sci Med. 2016;149:17–25.

Yadav S, Karam D, Bin Riaz I, Xie H, Durani U, Duma N, Ruddy KJ. Male breast cancer in the United States: treatment patterns and prognostic factors in the 21st century. Cancer. 2020;126(1):26–36.

Wang F, Shu X, Meszoely I, Pal T, Mayer IA, Yu Z, Shu XO. Overall mortality after diagnosis of breast cancer in men vs women. JAMA oncol. 2019;5(11):1589–96.

Download references

Acknowledgements

Not applicable.

Author information

Authors and affiliations.

School of Health and Social Care, Edinburgh Napier University, Sighthill Court, Sighthill Campus, Edinburgh, UK

Mary Abboah-Offei

School of Nursing, The Hong Kong Polytechnic University, Hong Kong Special Administrative Region, Hongkong, China

Jonathan Bayuo

International Observatory On End of Life Care (IOELC), Faculty of Health and Medicine, Division of Health Research, Lancaster University, Lancaster, LA1 4AT, UK

Yakubu Salifu

Florence Nightingale Faculty of Nursing, Midwifery and Palliative Care, King’s College London, London, WC2R 2LS, UK

Oladayo Afolabi

Institute of Medical Imaging & Visualisation, Department of Medical Science & Public Health, Faculty of Health & Social Science, Bournemouth University, Bournemouth, UK

Theophilus N. Akudjedu

You can also search for this author in PubMed   Google Scholar

Contributions

This review was conceived and designed by JB, MA-O, YS, OA, and TNA. The first reviewer imported all search results to Endnote reference manager version X9, de-duplicated, then all authors screened titles and abstracts of all identified studies, any article for which inclusion was unclear were discussed and if necessary adjudicated by YS and TNA. All authors critically appraised and contributed to the manuscript.

Corresponding author

Correspondence to Yakubu Salifu .

Ethics declarations

Ethics approval and consent to participate, consent for publication, competing interests.

The authors declare no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Abboah-Offei, M., Bayuo, J., Salifu, Y. et al. Experiences and perceptions of men following breast cancer diagnosis: a mixed method systematic review. BMC Cancer 24 , 179 (2024). https://doi.org/10.1186/s12885-024-11911-9

Download citation

Received : 28 September 2022

Accepted : 22 January 2024

Published : 06 February 2024

DOI : https://doi.org/10.1186/s12885-024-11911-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Male breast cancer
  • Experiences
  • Perceptions
  • Treatment approaches
  • Systematic review
  • Masculinity

ISSN: 1471-2407

breast cancer case control study

A case-control study of breast cancer risk factors in 7,663 women in Malaysia

Affiliations.

  • 1 Department of Applied Mathematics, Faculty of Engineering, University of Nottingham Malaysia Campus, Semenyih, Selangor, Malaysia.
  • 2 Cancer Research Malaysia, Subang Jaya, Selangor, Malaysia.
  • 3 Department of Surgery, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia.
  • 4 Biomedical Imaging Department, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia.
  • 5 Sime Darby Medical Centre, Subang Jaya, Selangor, Malaysia.
  • 6 Jeffrey Cheah School of Medicine and Health Sciences, Monash University Malaysia Campus, Subang Jaya, Selangor, Malaysia.
  • 7 Department of Paediatrics, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia.
  • PMID: 30216346
  • PMCID: PMC6138391
  • DOI: 10.1371/journal.pone.0203469

Background: Breast cancer risk factors have been examined extensively in Western setting and more developed Asian cities/countries. However, there are limited data on developing Asian countries. The purpose of this study was to examine breast cancer risk factors and the change of selected risk factors across birth cohorts in Malaysian women.

Methods: An unmatched hospital based case-control study was conducted from October 2002 to December 2016 in Selangor, Malaysia. A total of 3,683 cases and 3,980 controls were included in this study. Unconditional logistic regressions, adjusted for potential confounding factors, were conducted. The breast cancer risk factors were compared across four birth cohorts by ethnicity.

Results: Ever breastfed, longer breastfeeding duration, a higher soymilk and soy product intake, and a higher level of physical activity were associated with lower risk of breast cancer. Chinese had the lowest breastfeeding rate, shortest breastfeeding duration, lowest parity and highest age of first full term pregnancy.

Conclusions: Our study shows that breastfeeding, soy intake and physical activity are modifiable risk factors for breast cancer. With the increasing incidence of breast cancer there is an urgent need to educate the women about lifestyle intervention they can take to reduce their breast cancer risk.

Publication types

  • Clinical Trial
  • Multicenter Study
  • Research Support, Non-U.S. Gov't
  • Breast Feeding
  • Breast Neoplasms / epidemiology*
  • Case-Control Studies
  • Malaysia / epidemiology
  • Middle Aged
  • Risk Factors
  • Soybean Proteins / adverse effects
  • Soybean Proteins

Grants and funding

  • 203477/Z/16/Z/Wellcome Trust/United Kingdom
  • MR/P012930/1/MRC_/Medical Research Council/United Kingdom
  • V203477/Z/16/Z/WT_/Wellcome Trust/United Kingdom

ScienceDaily

Strongest contender in decades in fight against breast cancer

For decades, hormonal treatment of breast cancer has been going in one direction -- blocking estrogen. Now a global study involving researchers at the University of Adelaide has discovered there may be another, less toxic way to defeat the most common form of breast cancer.

The study found the drug enobosarm stimulates the androgen receptor (AR), making it effective against estrogen receptor-positive (ER+) breast cancer, which constitutes up to 80 per cent of all breast cancer cases.

"The effectiveness of enobosarm lies in its ability to activate the AR and trigger a natural defence mechanism in breast tissue, thereby slowing the growth of ER+ breast cancer, which relies on the hormone estrogen to grow and spread," said senior co-author Professor Wayne Tilley, Director of the Dame Roma Mitchell Cancer Research Laboratories at the University of Adelaide.

"This clinical study is supported by our pre-clinical research, previously published in Nature Medicine , which established that the AR is a tumour suppressor in both normal breast tissue and ER+ breast cancer."

Along with investigators from the University of Adelaide and Dana-Farber Cancer Institute (DFCI) in Boston, USA, the international study also included researchers from the University of Liverpool in the UK and other experts around the world.

The team assessed enobosarm's efficacy and safety in 136 postmenopausal women with advanced or metastatic ER-positive, HER2-negative breast cancer. Enobosarm showed significant anti-tumour activity and was well-tolerated by patients, without adversely affecting their quality of life or causing masculinising symptoms.

This discovery represents the first advancement in hormonal treatment of ER+ breast cancer in decades and offers a promising new oral treatment strategy for the most prevalent form of breast cancer.

The new hormonal strategy, published in The Lancet Oncology , differs from the existing standard-of-care hormonal treatments, which have been around for decades and involve suppressing estrogen activity in the body or inhibiting the ER.

Although successful initially, treatments targeting ER can cause severe side effects and treatment-resistant progression of the disease is common.

"Our findings are very promising. They demonstrate that stimulating the androgen receptor pathway with enobosarm can be beneficial," said senior co-author and study Principal Investigator Dr Beth Overmoyer from DFCI.

"This is the first time a non-estrogen receptor hormonal treatment approach has been shown to be clinically advantageous in ER+ breast cancer. The study supports further investigation of enobosarm in earlier stages of breast cancer as well as in combination with targeted therapies, such as ribociclib, a CDK 4/6 inhibitor."

Around 57 Australians are diagnosed with breast cancer every day, with more than 2.3 million cases identified globally each year.

"The data strongly encourages more clinical trials for AR-stimulating drugs in treating AR-positive and ER-positive breast cancer. The fact that this drug is well-tolerated also opens possibilities for its use in breast cancer prevention," said co-author Dr Stephen Birrell, a clinical affiliate of the University of Adelaide.

  • Breast Cancer
  • Women's Health
  • Colon Cancer
  • Lung Cancer
  • Diseases and Conditions
  • Ovarian Cancer
  • Breast cancer
  • Mammography
  • Breast reconstruction
  • Breast implant
  • Colorectal cancer
  • Cervical cancer
  • Monoclonal antibody therapy

Story Source:

Materials provided by University of Adelaide . Original written by Jessica Stanley. Note: Content may be edited for style and length.

Journal Reference :

  • Carlo Palmieri, Hannah Linden, Stephen N Birrell, Sally Wheelwright, Elgene Lim, Lee S Schwartzberg, Amy R Dwyer, Theresa E Hickey, Hope S Rugo, Patrick Cobb, Joyce A O'Shaughnessy, Stephen Johnston, Adam Brufsky, Wayne D Tilley, Beth Overmoyer. Activity and safety of enobosarm, a novel, oral, selective androgen receptor modulator, in androgen receptor-positive, oestrogen receptor-positive, and HER2-negative advanced breast cancer (Study G200802): a randomised, open-label, multicentre, multinatio . The Lancet Oncology , 2024; DOI: 10.1016/S1470-2045(24)00004-4

Cite This Page :

  • Gravastar Inside Another Gravastar?
  • Root Microbes: Secret to a Tastier Cup of Tea
  • Ancient Retroviruses and Our Big Brains
  • Diverse Ancient Volcanoes On Mars
  • Brain 'Programmed' to Learn from Those We Like
  • Targeting the Microenvironment to Heal Hearts
  • Visual Cues Inform Decision to Cooperate
  • Early Stone Plaza Found in the Andes
  • Sustainable Chemically Synthesized Sugars
  • Cultured Beef-Rice: Animal Cells in the Grains

IMAGES

  1. (PDF) Risk of breast cancer in symptomatic women in primary care: A

    breast cancer case control study

  2. Characteristics of the Study Subjects in a Breast Cancer Case-control

    breast cancer case control study

  3. PPT

    breast cancer case control study

  4. Pathophysiology

    breast cancer case control study

  5. breast cancer care pathway case study

    breast cancer case control study

  6. (PDF) Breast cancer case control study: Conceptual framework and

    breast cancer case control study

VIDEO

  1. Breast Cancer Prevention

  2. Team Breast Cancer Treatment

COMMENTS

  1. A case-control study to evaluate the impact of the breast screening programme on mortality in England

    An initial pilot case-control study in London indicated that attending mammography screening led to a mortality reduction of 39%. Methods Based on the same study protocol, an England-wide study was set up. Women aged 47-89 years who died of primary breast cancer in 2010 or 2011 were selected as cases (8288 cases).

  2. A case-control study to evaluate the impact of the breast screening

    An initial pilot case-control study in London indicated that attending mammography screening led to a mortality reduction of 39%. Methods Based on the same study protocol, an England-wide...

  3. Association Between Change in Mammographic Density and Breast Cancer Risk

    The breast cancer risk factors assessed at entry for the women in this study, stratified by case and control status, are presented in Table 1. Most women in both the case and control groups were postmenopausal (209 and 482 women, respectively) and parous (228 and 508 women, respectively).

  4. Biomarkers of inflammation and breast cancer risk: a case-control study

    Breast cancer is the commonest cancer and leading cause of cancer death in women worldwide, with an estimated 1.7 million cases and over 520,000 deaths in 2012, accounting for 25% of all female ...

  5. A case-control study of breast cancer risk factors in 7,663 ...

    Published: September 14, 2018 https://doi.org/10.1371/journal.pone.0203469 Article Authors Metrics Comments Media Coverage Abstract Background Figures Abstract Background Breast cancer risk factors have been examined extensively in Western setting and more developed Asian cities/countries.

  6. Case-control Studies on the Effectiveness of Breast Cancer

    Observational studies, and particularly case-control studies, are an important monitoring tool for breast cancer screening programs. The focus should be on diminishing bias in observational studies and gaining a better understanding of the influence of study design on estimates of mortality reduction.

  7. A case-control study to evaluate the impact of the breast screening

    A case-control study to evaluate the impact of the breast screening programme on mortality in England Br J Cancer. 2021 Feb;124 (4):736-743. doi: 10.1038/s41416-020-01163-2. Epub 2020 Nov 23. Authors Roberta Maroni # 1 , Nathalie J Massat # 1 , Dharmishta Parmar 1 , Amanda Dibden 1 , Jack Cuzick 1 , Peter D Sasieni 2 , Stephen W Duffy 3

  8. Use of Underarm Cosmetic Products in Relation to Risk of Breast Cancer

    We conducted a 1:1 age-matched hospital-based case-control study aiming to investigate the risk for breast cancer in relation to self-reported UCP use. We included measurements of aluminum concentrations in breast tissue from a large series of breast cancer patients and healthy individuals in a controlled epidemiologic study.

  9. Breast cancer screening case-control study design: impact on breast

    Breast cancer screening case-control study design: impact on breast cancer mortality Abstract Background Recent case-control studies on the effectiveness of population-based breast cancer screening show differences in the magnitude of breast cancer mortality reduction.

  10. Possible role of lead in breast cancer

    Essential characteristics of the study population are represented in our previous work (Anđelković et al. 2021).Briefly, women diagnosed with BC were significantly older with higher BMI values than the control group (both P < 0.001). The average age of the control group was 39.5 ± 10.8 years (BMI 22.8 kg/m 2, IQR 20.3-25.7), while the average age of women diagnosed with cancer was 61.2 ± ...

  11. Statins to reduce breast cancer risk: A case control study in U.S

    514 Background: Statins (HMG CoA reductase inhibitors) are commonly used cholesterol-lowering agents that are noted to suppress tumor growth in several animal models, however clinical data for chemoprotective role of statins in breast cancer is conflicting. While some studies report reduced risk of breast cancer with lipid lowering drugs, others report an increased risk. We investigated the ...

  12. Are better AI algorithms for breast cancer detection also better at

    To evaluate our hypothesis, we designed a case-control study using paired mammograms at diagnosis and at the previous screening visit. The study included n = 3386 women from the OPTIMAM registry, that includes mammograms from women diagnosed with breast cancer in the English breast screening program 2010-2019.

  13. Use of hormone replacement therapy and risk of breast cancer: nested

    Abstract Objective To assess the risks of breast cancer associated with different types and durations of hormone replacement therapy (HRT). Design Two nested case-control studies.

  14. Dietary quality index and the risk of breast cancer: a case-control study

    Diet quality is a significant determinant in the etiology of breast cancer (BrCa), but further studies are required to explore this relationship. Therefore, we tried to assess if diet quality, assessed using the Diet Quality Index-International (DQI-I), was related to BrCa among the Iranian population. In the present case-control research, 134 women with a recent diagnosis of BrCa and 267 ...

  15. A case-control study of breast cancer risk and ambient

    We conducted a case-control study of breast cancer risk from exposure to pesticides using a Geographical Information Systems (GIS)-based method that combines geocoded residential and occupational histories with state pesticide use reports and land use data 24 in California's highest-ranking counties (Fresno, Tulare, and Kern) for agricultural de...

  16. Machine learning algorithms to uncover risk factors of breast cancer

    Introduction: This large case-control study explored the application of machine learning models to identify risk factors for primary invasive incident breast cancer (BC) in the Iranian population. This study serves as a bridge toward improved BC prevention, early detection, and management through the identification of modifiable and ...

  17. High-protein diet scores, macronutrient substitution, and breast cancer

    Background Evidence from recent studies suggested that variation in the quantity and quality of macronutrients in the diet may potentially play a role in predicting the risk of breast cancer (BC). In the current study, we aimed to assess the association of different high-protein diet scores and replacing fats and carbohydrate (CHO) with protein in the diet with the BC risk among Iranian women ...

  18. A case-control study to evaluate the impact of the breast screening

    Our estimation of overdiagnosis was a multi-stage process, as follows: Conduct a case-control study where cases were women with breast cancer and controls were women with no diagnosis of breast cancer prior to the age of their individually matched case (although in principle, they could develop breast cancer thereafter and potentially be a case), with aspects of screening history as the ...

  19. Sex-steroid hormones and risk of postmenopausal estrogen ...

    Cancer Causes & Control - Sex-steroid hormones are associated with postmenopausal breast cancer but potential confounding from other biological pathways is rarely considered. ... This analysis included 1208 women from a case-cohort study of postmenopausal breast cancer within the Melbourne Collaborative Cohort Study. Weighted Poisson ...

  20. Association of vaginal oestradiol and the rate of breast cancer in

    Objective To estimate the rate of breast cancer associated with use of vaginal oestradiol tablets according to duration and intensity of their use. Design Registry based, case-control study, nested in a nationwide cohort. Setting Based in Denmark using the civil registration system, the national registry of medicinal product statistics, the Danish cancer registry, the Danish birth registry ...

  21. Experiences and perceptions of men following breast cancer diagnosis: a

    Male breast cancer (MBC) is a rare condition, accounting for less than 1% of all breast cancers. About 2,710 men are estimated to be diagnosed with breast cancer, with approximately 530 men projected to die from breast cancer in 2022 and have about 1 in 833 lifetime risk of being diagnosed with the disease in the United States [].Data from the Global Burden of Disease 2017 database indicate ...

  22. A case-control study of breast cancer risk factors in 7,663 women in

    A case-control study of breast cancer risk factors in 7,663 women in Malaysia PLoS One. 2018 Sep 14;13 (9):e0203469. doi: 10.1371/journal.pone.0203469. eCollection 2018. Authors

  23. Strongest contender in decades in fight against breast cancer

    The study found the drug enobosarm stimulates the androgen receptor (AR), making it effective against estrogen receptor-positive (ER+) breast cancer, which constitutes up to 80 per cent of all ...