BRCA1 and BRCA2 gene mutations and risk of breast cancer. Public health perspectives

Validation of a digital identification tool for individuals at risk for hereditary cancer syndromes

Background

The number of individuals meeting criteria for genetic counseling and testing for hereditary cancer syndromes (HCS) is far less than the number that actually receive it. To facilitate identification of patients at risk for HCS, Counsyl developed a digital identification tool (digital ID tool) to match personal and family cancer history to National Comprehensive Cancer Network (NCCN) BRCA-related Hereditary Breast and Ovarian Cancer (HBOC), Lynch syndrome, and polyposis testing criteria in one-to-one, automated fashion. The purpose of this study was to validate the ability of the digital ID tool to accurately identify histories that do and do not meet NCCN testing criteria.

Methods

Third-party recorded three-generation pedigrees were retrospectively reviewed by a certified genetic counselor (CGC) to determine if independent events included in pedigree histories met NCCN guidelines, and were then sorted into groups: high risk events (meets criteria) and low risk events (does not meet criteria). Events were entered into the digital ID tool to determine the extent of its concordance with events sorted by CGC review. Statistical tests of accuracy were calculated at a 95% confidence interval (CI).

Results

One hundred ninety-seven pedigrees were reviewed consecutively representing 765 independent events for analysis across groups. 382/382 (100%) high risk events identified by the digital ID tool and 381/383 (99.47%) low risk events identified by the digital ID tool were concordant with CGC sorting. The digital ID tool had a sensitivity of 100% (99.04–100% CI) and specificity of 99.48% (98.13–99.94% CI). The overall accuracy of the digital ID tool was estimated to be 99.74% (99.06–99.97% CI), reflecting the rate at which the digital ID tool reached the same conclusion as that of CGC review of pedigree events for the recommendation of genetic testing for individuals at risk for HCS.

Conclusions

The digital ID tool accurately matches NCCN criteria in one-to-one fashion to identify at-risk individuals for HCS and may be useful in clinical practice, specifically for BRCA-related HBOC and Lynch Syndrome.

Electronic supplementary material

The online version of this article (10.1186/s13053-018-0099-8) contains supplementary material, which is available to authorized users.

Epidemiology of Breast Cancer in Women

Epidemiologic studies have contributed importantly to current knowledge of environmental and genetic risk factors for breast cancer. Worldwide, breast cancer is an important cause of human suffering and premature mortality among women. In the United States, breast cancer accounts for more cancer deaths in women than any site other than lung cancer. A variety of risk factors for breast cancer have been well-established by epidemiologic studies including race, ethnicity, family history of cancer, and genetic traits, as well as modifiable exposures such as increased alcohol consumption, physical inactivity, exogenous hormones, and certain female reproductive factors. Younger age at menarche, parity, and older age at first full-term pregnancy may influence breast cancer risk through long-term effects on sex hormone levels or by other biological mechanisms. Recent studies have suggested that triple negative breast cancers may have a distinct etiology. Genetic variants and mutations in genes that code for proteins having a role in DNA repair pathways and the homologous recombination of DNA double stranded breaks (APEX1, BRCA1, BRCA2, XRCC2, XRCC3, ATM, CHEK2, PALB2, RAD51, XPD), have been implicated in some cases of breast cancer.

Biomarkers and the genetics of early neoplastic lesions

It has become increasingly evident that the study of DNA is inadequate to explain many, if not most, aspects of the development and progression of neoplastic lesions from pre-invasive lesions to metastasis. Thus, the term "genetic" can no longer refer to just the study of the genome. Much of the action in genetic research now shifts to the methods by which the pre-mRNA from one gene is processed to yield multiple different proteins, different quantities of the same protein as well as other forms of regulating RNA. Thus, the age of post-transcriptional processing and epigenetic control of the transfer of information from the genome has arrived. The mechanisms of post-transcriptional processing and epigenetic control that must be characterized in greater detail including alternate splicing, regulation of mRNA degradation, RNA regulatory factors including those factors which extensively edit mRNAs, control of translation, and control of protein stability and degradation. This chapter reviews many of the processes that control information from the genome to proteins and how these factors lead from less than 40,000 genes to more than an order of magnitude increase more proteins which actually control the phenotypes of cells - normal or neoplastic. It is usually the products of genes (e.g., mRNA, microRNA and proteins) that are the molecular markers that will control translational research and ultimately, individualized (personal) medical approaches to disease. This chapter emphasizes how the process of neoplasia "hijacks" the normal processes of cellular operations, especially those processes that are important in the normal development of the organisms - including proliferation, cellular death, angiogenesis, cellular mobility and invasion, and immunoregulation to ensure neoplastic development, survival and progression. This chapter reviews the wide range of processes controlling the information that flows from the genome to proteins and emphasizes how molecular steps in pure processes can be used as biomarkers to study prevention, treatment and/or management of diseases.

Risk Communication During Screening for Genomic Breast Cancer Susceptibility

Screening for genomic predisposition to diseases of adulthood is emerging in standard medical practice. Breast cancer predisposition testing provides a useful example for understanding the potential implications of risk-related screening in practice. Risk is an important and complex psychological construct that is understood differently both between scientific disciplines and between medical and lay persons. The process of establishing thresholds for classifying whether one's risk reaches levels warranting departure from standard preventive strategies is an important predictor of both physician feedback and patient interpretation of their risk. Messages about genomic breast cancer risk arising during the pedigree screening context are conveyed via explicit and implicit modes of communication. Best practices for communicating about genomic risk and the implications for health disparities related to screening for genomic susceptibility remain unknown. More work is urgently needed given the societal forces that are catalyzing population-based screening for genomic predisposition into the routine practice of preventive medicine.

Glutathione S-transferase polymorphisms associated with risk of breast cancer in southern Taiwan

In this study, the genetic polymorphisms associated with breast cancer in southern Taiwan were investigated. Two categories of genes were analyzed: (1) BRCA1, BRCA2, and Rad51, the DNA repair factors involved in homologous recombinational repair; and (2) CYP1A1, COMT, GST, and NAT2, the xenobiotic-metabolizing enzymes (XME) involved in estrogen metabolism. We found that the number of deletions and/or mutations in the GST genes was highly correlated with the occurrence of breast cancer. These data suggest that the GST enzymes, which detoxify the catechol estrogen quinones, are important target molecules for screening in populations at high risk of breast cancer.

Analysis of the mononuclear inflammatory cell infiltrate in the normal breast, benign proliferative breast disease, in situ and infiltrating ductal breast carcinomas: preliminary observations

BACKGROUND

Mammary carcinogenesis is a multistep process entailing the transition from normal breast to benign proliferative breast disease (ductal hyperplasia) to ductal carcinoma in situ to infiltrating ductal carcinoma.

HYPOTHESIS

These transitions are associated with changes in the mononuclear inflammatory cell infiltrate.

MATERIALS AND METHODS

A total of 53 mastectomy specimens of normal breast, benign proliferative breast disease, ductal carcinoma in situ and infiltrating ductal carcinoma were evaluated for mononuclear inflammatory cell infiltrate by using immunohistological methods and monoclonal antibodies including CD20, CD68, CD3 and granzyme B, histiocytes, T cells and cytotoxic T cells.

RESULTS

Transitions from normal breast to the subsequent tissue with lesions (normal skin v benign proliferative breast disease v ductal carcinoma in situ v infiltrating ductal carcinoma) were associated with significantly (p<0.01) increased mean (SD) density of mononuclear inflammatory cell infiltrate at the parenchyma (3.2 (1.0) v 26.4 (7.8) v 33.6 (7.9) v 39.1 (4.7) for CD20+ B cells; 2.8 (1.0) v 81.5 (14.0) v 84.0 (14.9) v103.7 (3.9) for CD3; 1.3 (2.0) v 3.8 (4.0) v 12.7 (23) v 22.1 (25.0) for CD68+ macrophages; 2.0 (1.0) v 58.3 (5.0) v 60.0 (10.0) v 74.1 (28.0) for granzyme B+ cytotoxic T cells) and at the stroma (0.7 (1.0) v 3.0 (5.0) v 13.3 (20) v 16.7 (30.0) for CD20+ B cells; 1.0 (2.06) v 4.0 (2.5) v 16.7 (5.0) v 21.7 (15) for CD68+ macrophages; 1.4 (0.6) v 4.2 (1.2) v 46.6 (16.7) v 77.0 (5.0) for CD3+ cells and 0 (0) v 0.5 (1.0) v 0.7 (1.0) v 0.7 (1.0) for granzyme B+ cytotoxic T cells).

CONCLUSIONS

The increased mononuclear inflammatory cell infiltrate during mammary carcinogenesis may reflect non-specific or specific immunological processes.

Genetic and histopathologic evaluation of BRCA1 and BRCA2 DNA sequence variants of unknown clinical significance

Classification of rare missense variants as neutral or disease causing is a challenge and has important implications for genetic counseling. A multifactorial likelihood model for classification of unclassified variants in BRCA1 and BRCA2 has previously been developed, which uses data on co-occurrence of the unclassified variant with pathogenic mutations in the same gene, cosegregation of the unclassified variant with affected status, and Grantham analysis of the fit between the missense substitution and the evolutionary range of variation observed at its position in the protein. We have further developed this model to take into account relevant features of BRCA1- and BRCA2-associated tumors, such as the characteristic histopathology and immunochemical profiles associated with pathogenic mutations in BRCA1, and the fact that approximately 80% of tumors from BRCA1 and BRCA2 carriers undergo inactivation of the wild-type allele by loss of heterozygosity. We examined 10 BRCA1 and 15 BRCA2 unclassified variants identified in Australian, multiple-case breast cancer families. By a combination of genetic, in silico, and histopathologic analyses, we were able to classify one BRCA1 variant as pathogenic and six BRCA1 and seven BRCA2 variants as neutral. Five of these neutral variants were also found in at least 1 of 180 healthy controls, suggesting that screening a large number of appropriate controls might be a useful adjunct to other methods for evaluation of unclassified variants.

DNA repair gene polymorphisms and risk of second primary neoplasms and mortality in oral cancer patients

OBJECTIVES/HYPOTHESIS

We tested the hypothesis that polymorphisms in genes involved in DNA repair pathways are associated with the development of second primary neoplasms of the upper aerodigestive tract (UADT), as well as mortality, in patients previously diagnosed with oral squamous cell cancer (OSCC).

METHODS

DNA specimens from 279 OSCC patients who had participated in two previous population-based case-control studies were assayed for the following polymorphisms: X-ray repair cross-complementing (XRCC) 1 Arg399Gln, XRCC3 Thr241Met, xeroderma pigmentosum complementing group D (XPD) Lys751Gln, and O-methylguanine- DNA methyltransferase (MGMT) Leu84Phe and Val143Ile. Baseline demographic information was obtained from personal interviews and tumor characteristics and treatment were obtained from cancer registry files. Cox proportional hazards models were used to calculate hazards ratio (HR) estimates for each polymorphism in relation to the risk of developing second primary neoplasms at any site, UADT, and head and neck. HRs were also determined for associations with all-cause mortality and oral cancer specific mortality.

RESULTS

A significant increased risk of second neoplasms (all sites combined, as well as for UADT sites and for head and neck squamous cell cancers) was observed among XRCC3 241Met allele homozygotes (HR 2.65-3.44, P < .02). No significant association with the development of second neoplasms was observed for the XRCC1 399Gln, XPD 751Gln, or MGMT 84Phe or 143Ile alleles. Although no associations with oral cancer-specific mortality were observed, we found a significant inverse association between all-cause mortality and possessing at least one copy of the XRCC1 399Gln allele (HR 0.68, 95% confidence interval [CI] 0.47-0.97, P = .03), as well as a suggestion of a direct association between all-cause mortality and having one copy of the XRCC3 241Met allele (HR 1.39, 95% CI 0.95-2.03, P = .09).

CONCLUSIONS

Polymorphisms in the DNA repair enzyme gene XRCC3 241Met was associated with an increased risk of second neoplasms, and polymorphisms of the XRCC1 399Gln gene were associated with a decreased risk of all-cause mortality in patients with primary OSCC. These findings require confirmation in other populations before the clinical implications can be considered.

Application of embryonic lethal or other obvious phenotypes to characterize the clinical significance of genetic variants found in trans with known deleterious mutations

This work describes an approach to characterize the clinical significance of genetic variants detected during the genetic testing of BRCA1 in patients from hereditary breast/ovarian cancer families. Results from transgenic mice and extensive clinical testing support the hypothesis that biallelic BRCA1 mutations result in embryonic lethality. Therefore, it is reasonable to conclude that variants of uncertain clinical significance found to reside in trans with known deleterious mutations impart reduced risk for cancer. This approach was applied to a large data set of 55,630 patients who underwent clinical BRCA1 screening by whole gene direct DNA sequencing. Fourteen common single nucleotide polymorphisms (SNPs) were used to assign 10 previously defined common, recurrent, or canonical haplotypes in 99% of these cases. From a total of 1,477 genetic variants detected in these patients, excluding haplotype-tagging SNPs, 877 (59%) could be unambiguously assigned to one or more haplotypes. In 41 instances, variants previously classified as being of uncertain clinical significance, mostly missense variants, were excluded as fully penetrant mutations due to their coincidence in trans with known deleterious mutations. From a total of 1,150 patients that harbored these 41 variants, 956 carried one as the sole variant of uncertain clinical significance reported. This approach could have widespread application to other disease genes where compound heterozygous mutations are incompatible with life or result in obvious phenotypes. This largely computational technique is advantageous because it relies upon existing clinical data and is likely to prove informative for prevalent genetic variants in large data sets.

Risks and benefits of population-based genetic testing for Mendelian subsets of common diseases were examined using the example of colorectal cancer risk

OBJECTIVE

Genetic testing for adult-onset, common diseases is becoming more commonplace in clinical medicine. We modeled the proportions of hypothetic populations that would potentially benefit or suffer harm from widespread predisposition testing.

METHODS

Using the traditional two-by-two table from the discipline of epidemiology, we modeled three hypothetic populations using the example of genetic testing for hereditary colorectal cancer in three groups: the general population, a genetically increased-risk population, and a population at increased risk due to nongenetic factors.

RESULTS

We demonstrate that the potential benefits are increased and risks are reduced when testing is limited to those at increased genetic risk when compared with testing in the general population. Where disease incidence is increased due to nongenetic factors, genetic testing has the potential to detract from the detection and reduction of other potentially important risk factors.

CONCLUSION

While targeted testing can benefit those truly at increased risk, broadly applied genetic testing can do more harm than good.

FISH analysis of BRCA1 copy number in paraffin-embedded ovarian cancer tissue samples

We investigated the BRCA1 gene copy number in unselected ovarian malignancies. Both additional genes (amplification) as well as deletion (loss of heterozygosity, LOH) are often thought to have a role in the initiation or progression of cancer. In addition, if there were little change, deletion studies might help identify BRCA1 mutation carriers. Forty-seven paraffin-embedded ovarian tissue blocks obtained between 1984 and 1997 were used for this study. A sample was "deletion-positive" when BRCA1-deleted cells in the tumor area were significantly different from the benign area. Twenty-five (53%) cases were found to be "deletion-positive". The average age of onset of "deletion-positive" patients was 50.8 years and of "deletion-negative" 57.8 years (P < 0.05). There was no statistical difference between groups in the staging, histology, or prognosis. A Kaplan-Meier study did show a trend towards poorer survival for "deletion-positive" patients. FISH permits unique molecular characterization of malignancies at a cellular level. Double amplification of HER-2 and c-myc predicts poor ovarian cancer survival. There appears to be a definite role for BRCA1 deletion in reducing the age of ovarian cancer onset and possibly in overall survival. Further FISH studies of this and other patient sets using additional molecular markers are needed.

[Collective health, the new genetics, and market eugenics]

Due to constant advances in genetic manipulation, the field of public health is faced with the possible emergence of a reductionist genomic emphasis, beyond socio-cultural influences. Under such circumstances, the main focus of public health interventions would shift towards a genomic approach to "individuals" and their "families", to the detriment of their consecrated object - "populations", thereby moving away from the current central priority of efforts (and resources) aimed at reducing ill-health due to socioeconomic inequalities. Even admitting the benefits of genetic knowledge, a new eugenic practice may emerge from the availability of genetic tests on the marketplace aimed at individuals that can afford to consume them.

Polymorphisms of XRCC1 and XRCC3 genes and susceptibility to breast cancer

Mammalian cells are constantly exposed to a wide variety of genotoxic agents from both endogenous and exogenous sources. Genetic variability in DNA repair may contribute to human cancer risk. We used a case-control study design (162 cases and 302 controls) to test the association between three amino acid substitution variants of DNA repair genes (XRCC1 Arg194Trp, XRCC1 Arg399Gln, and XRCC3 Thr241Met) and breast cancer susceptibility. We found a weak association between the XRCC1 194Trp allele and breast cancer risk (adjusted odds ratio (OR)=1.98; 95% confidence interval (CI)=0.85-4.63). We also found a potential gene-gene interaction between the XRCC1 194Trp allele and XRCC3 241Met allele and breast cancer risk (adjusted OR=8.74; 95% CI=1.13-67.53). Although larger studies are needed to validate the study results, our data suggest that amino acid substitution variants of XRCC1 and XRCC3 genes may contribute to breast cancer susceptibility.

Application of population screening principles to genetic screening for adult-onset conditions

Recent advances in molecular genetics have highlighted the potential use of genetic testing to screen for adult-onset chronic diseases. Several issues must be addressed, however, before such tests can be recommended for population-based prevention programs. These issues include the adequacy of the scientific evidence, the balance of risks and benefits, the need for counseling and informed consent, and the costs and resources required. Ongoing assessment of the screening program and quality assurance of laboratory testing are also needed. This paper considers the application of general principles for mass screening to genetic testing for susceptibility to adult-onset chronic diseases. Evaluation of proposals for genetic screening in context of these principles reveals that needed evidence is often absent, particularly with respect to the predictive value of tests, efficacy of interventions, and social consequences of testing. The principles of population screening are developed into a framework for public health policy on genetic screening that has three stages: assessment of the screening test and interventions for those who test positive, including assessment of risks and costs, policy development, and program evaluation. Essential elements are identified, including evaluation of evidence and processes for consensus development and program evaluation. The proposed framework for public health policymaking outlined in this commentary, when combined with future efforts that involve an authoritative consensus process, may be useful for the evaluation and planning of genetic screening programs aimed at reducing morbidity and mortality from adult-onset chronic diseases.

Who's liable for breast cancer prevention? Your patient can sue--and win--if preventive options aren't made clear

Strategies to reduce the likelihood of breast cancer in high-risk patients present complex medical-legal issues. An accurate and thorough medical history, objective calculation of risk, thorough discussion of preventive options and side effects, meticulous documentation of physician-patient interactions, and meticulous follow-up are essential. These elements form the foundation of a sound approach to breast cancer prevention in high-risk patients and should reduce physician liability if cancer occurs.

Genetic regulation of ionizing radiation sensitivity and breast cancer risk

Genetic variability in DNA repair may contribute to hypersensitivity to ionizing radiation (IR) and susceptibility to breast cancer. We used samples collected from a clinic-based breast cancer case-control study to test the working hypothesis that amino acid substitution variants of DNA repair genes may contribute to prolonged cell-cycle delay following IR and breast cancer risk. Fluorescence-activated cell sorter (FACS) analysis was used to measure cell-cycle delay. PCR-restriction fragment length polymorphism (RFLP) assays were used to determine four genotypes of three DNA repair genes: XRCC1, 194 Arg/Trp and 399 Arg/Gln; XRCC3, 241 Thr/Met; and APE1, 148 Asp/Glu. The data showed that breast cancer patients had a significantly higher delay index than that of controls (P < 0.001); the means +/- SD for cases and controls were 36.0 +/- 13.1 (n = 118) and 31.4 +/- 11.5 (n = 225), respectively. There was a significant dose-response relationship between delay index, categorized into quartiles, and an increasing risk of breast cancer (crude odds ratios: 1.00, 1.00, 1.27, and 2.46, respectively; P(trend) = 0.002). In controls, prolonged cell-cycle delay was significantly associated with the number of variant alleles in APE1 Asp148Glu and XRCC1 Arg399Gln genotypes (P(trend) = 0.001). Although larger studies are needed to validate the results, our data suggest that an inherited hypersensitivity to IR may contribute to human breast carcinogenesis.

Lavage and nipple aspiration of breast ductal fluids: a source of biomarkers for environmental mutagenesis

The acquisition of breast ductal fluid by nipple aspiration and ductal lavage are simple noninvasive procedures to sample breast tissue. Nipple aspiration fluid (NAF) obtained with gentle suction and a simple syringe-adapted apparatus may evaluate the secretory components that bathe the ductal epithelial cells. Evaluations have included the quantification of soluble markers (carcinoembryonic antigen and prostatic-specific antigen), DNA amplification, protein gel electrophoresis, and mutagenesis assays. It has been suggested that environmental mutagens in the breast ductal system may contribute to carcinogenesis. The feasibility of mutagenesis assays on NAF has been limited by the small size of the samples obtained. Three small clinical studies detected mutagens in 6-14% of the samples using the Salmonella Ames assay. Ductal lavage collects more of a cellular aspirate from the ductal system utilizing a microcatheter. Early studies on ductal lavage fluid have included cytology and methylation-specific PCR. Ductal lavage in a high-risk group has identified cellular atypia in 21% of those sampled. Samples obtained through the nipple, by aspiration or lavage, are the proteinaceous secretions from the ductal system and ductal epithelial cells. The fluid represents the cellular events and the dynamic secretory process of the breast and may include potential initiators of the carcinogenesis process in the cellular microenvironment. Fluid obtained by ductal lavage may allow for more detailed studies of the role of mutagens in breast cancer.

A longitudinal study of physical activity and breast cancer prediction

Physical activity was analyzed as a predictor of breast cancer in the Aerobic Center Longitudinal Study. The study sample included 4,520 healthy women, with 150 incident cases of breast cancer. The study sample was limited to women who did not have a diagnosis of breast cancer at initial entry to the Aerobic Center Longitudinal Study. Women in the study were predominantly from an upper-middle class, college-educated, Caucasian American group being evaluated at a world-famous fitness center. Women with and without breast cancer had similar reports for diet history, alcohol and tobacco use, body composition, and breast health history. Women with breast cancer were older and had an earlier menarche and later menopause. Logistic regression analysis of physical activity for women with and without breast cancer was performed controlling for age. Intensity and duration of physical activity were not predictors of breast cancer. Less frequency of jogging was a significant predictor of breast cancer among women in the sample, O.R.=1.14, P =.04. Women who reported jogging as a type of activity were less likely to develop breast cancer than those who did not report jogging, O.R.=2.32, P=0.05.

Impact of educational print materials on knowledge, attitudes, and interest in BRCA1/BRCA2: testing among Ashkenazi Jewish women

BACKGROUND

The recent identification of several BRCA1/BRCA2 founder mutations among Ashkenazi Jewish individuals has led to increased salience of BRCA1/BRCA2 testing for Jewish individuals. Little is known about interest in BRCA1/BRCA2 testing among Ashkenazi Jews from the general population. Furthermore, previous research has not generally evaluated the impact of education on interest in testing among individuals from the general population. The goal of the current study was to examine whether a brief educational booklet regarding BRCA1/BRCA2 testing would influence knowledge, attitudes, and interest in testing among Ashkenazi Jewish women from the general population.

METHODS

After a baseline telephone interview, participants were randomized to receive either genetic testing educational print materials (n = 195 women) or general breast cancer education control materials (n = 196 women). One month after receiving these materials, the authors reassessed knowledge, attitudes, and interest in BRCA1/BRCA2 gene testing.

RESULTS

Relative to the breast cancer education control materials, the genetic testing education materials led to increased knowledge, increased perception of the risks and limitations of testing, and decreased interest in obtaining a BRCA1/BRCA2 mutation test.

CONCLUSIONS

These data indicate that preliminary print education can be used to educate low-risk individuals about BRCA1/BRCA2 genetic testing. This approach may be used to educate low-risk individuals about the benefits and risks/limitations of BRCA1/BRCA2 testing, so that they can make informed decisions about whether to pursue genetic counseling.

Ethical considerations in clinical pharmacogenomics research

In recent years there have been unprecedented advances in our understanding of the involvement of genetic polymorphisms in the response to drug therapies. Polymorphisms have been identified that lead to variable patient responses to several medications including cardiovascular, psychiatric, anti-infective and analgesic therapies. The potential for the development of customized, genotype-based therapies is scientifically and clinically attractive. However, these developments, although bearing scientific promise, raise ethical concerns for the conduct of research with human subjects, particularly with respect to confidentiality, risk-benefit analysis, DNA-banking and pharmacoeconomic issues. This article discusses some of the ethical considerations that are related to the use of pharmacogenomics in clinical research protocols.

A meta-analysis of estrogen replacement therapy and risk of epithelial ovarian cancer

Estrogen replacement therapy (ERT) has not been associated with epithelial ovarian cancer in most reported epidemiologic studies that have looked for an association. Some studies may have found weak statistically nonsignificant associations because the number of cases or number of women who reported estrogen use was small. We performed a meta-analysis of data from 15 case-control studies that provided data on ERT and risk of epithelial ovarian cancer. The 15 combined studies were statistically heterogeneous (chi(2) (14) = 26. 3, P < 0.05) in terms of the effect they found. When we combined these studies using a random effects model, we did not find a significant association of ERT with ovarian cancer (odds ratio = 1.1, 95% confidence interval = 0.9-1.3). There was no clear evidence of a dose-response relation with increasing duration of estrogen use in a subset of five studies that reported estrogen use by duration (overall slope = 0.0012, 95% confidence interval = -0.0055 to 0. 0080). The influences of statistical outliers, study design (hospital or clinic controls vs. community controls), and location (U.S. and Canada vs. Europe and Australia) were examined. The odds ratio was 1.3 (95% confidence interval = 1.0-1.6) in the relatively homogeneous subset of four U.S. case-control studies with community controls, but we cannot rule out the possibility of uncontrolled confounding. The odds ratios for estrogen use for other subgroups defined by geographic location and type of control group were not significantly different from one.