For women with a BRCA1 mutation, the risk of breast cancer to age 70 has been estimated to be between 64 and 84%.1–3 Among women with a mutation, it is generally assumed that cases of breast cancer are attributable primarily to the mutation; it is not known to what extent the local environment or genetic background modifies the risk of cancer. At present, most women with a BRCA1 mutation are provided with a cancer risk estimate that is derived from one of a few international collaborative studies1–3 based largely on subjects from western Europe or North America. The risks of breast cancer in the general population vary widely from country to country.4 Canada and the United States have relatively high incidence rates of breast cancer, whereas in Poland the rate is much lower. The annual incidence of breast cancer in Canadian women is 83 per 100,000 per year (GLOBOCAN), in the USA it is 126 per 100,000 per year (SEER white women) and in Poland it is 49 per 100,000 per year (GLOBOCAN). The sources of the variation are not known, but it is likely that genetic and environmental factors both play a role. Currently, it is not clear if the regional variation in incidence extends to the subgroup of women who carry a BRCA1 mutation. In this prospective study, we compare the risk of breast cancer for women with a BRCA1 mutation from Poland with that of women from Canada or the United States.
Women with a BRCA1 mutation face a high lifetime risk of breast cancer. It is unknown to what extent environmental factors modify the inherent genetic risk. If women from different countries, but with similar mutations, experience different levels of cancer risk, nongenetic risk modifiers are likely to be present. Study subjects were a cohort of 1477 women with a BRCA1 mutation, from Canada (n = 358), the United States (n = 256) and Poland (n = 863). The women were followed for a mean of 4.3 years and 130 incident cases of breast cancer were recorded. Annual cancer incidence rates were calculated, and based on these, penetrance curves were constructed for women from North America and Poland. In a Cox proportional hazards model, residence in Poland, versus North America, was associated with an adjusted hazard ratio of 0.54 (95% CI 0.34-0.86; p = 0.01). The risk of breast cancer to age 70 was estimated to be 49% for women from Poland and 72% for women from North America. Among women with BRCA1 mutations, the risk of breast cancer in women who reside in Poland is less than that of women who reside in North America. The reasons for the difference are unknown, but this observation suggests that environmental factors or genetic modifiers are important in determining risk.
Subjects and Methods
Study population and data collection
Eligible study subjects were identified from a cohort of women with a BRCA1 or BRCA2 mutation. Subjects were drawn from 26 contributing study centers in Canada, the United States and Poland. The cohort study was initiated in 1995, Poland joined the study in 1999. All study subjects received genetic counseling (with the exception of a small number of subjects) and all provided written informed consent for study participation. A woman was eligible for the cohort study when molecular analysis established that she was a carrier of a deleterious mutation in BRCA1. Each woman completed a standardized questionnaire at the time of genetic testing or study enrollment. The questionnaire was the same in each study center; the Polish questionnaire was a translation of the English questionnaire. Variables of interest included information about medical and reproductive histories. A family history was also collected from each woman at the time of genetic testing. Information was recorded on first- and second-degree relatives, including cancer site and age at diagnosis. The women were asked to report new diagnoses of cancer in themselves (site and date of diagnosis) through the completion of a mailed questionnaire every 2 years. The woman was asked how the breast cancer was first detected (mammogram, Ultrasound, MRI, self-detected, physician examination). We attempted to obtain a pathology report for each incident case of cancer reported in the cohort and have obtained a report for over 70% of these. We included only invasive cancers in the analysis.
A woman was eligible for the study if she was between the ages 25 and 65 years at baseline and if she did not have a prior mastectomy or a known diagnosis of breast or ovarian cancer. The subjects were followed from the date of completion of the baseline questionnaire until the first of: development of breast cancer, prophylactic mastectomy, death, or date of last follow-up. 2111 women were identified as potentially eligible. 634 women were excluded, including 528 women for whom follow-up information was not available and 106 women who were lost to follow-up. Therefore, 1477 women were eligible and were included in the analysis (70%). Study subjects were divided into those born in Poland and those born in Canada or the United States. The proportion of eligible subjects who were in the cohort was 70% in Poland and 70% in North America. Survival curves were constructed using the Kaplan-Meier method. Statistical significance of the differences between the curves was tested using the log-rank test. The two curves were consistent for proportionality using the SAS statistical software and a multivariable survival analysis was conducted. The hazard ratio for breast cancer was estimated for women who resided in Poland, compared to women from North America or Canada. The Cox proportional hazards model was implemented in SAS, version 9.1.3. The hazard ratio was adjusted for oophorectomy, age at study entry, parity, age at menarche, oral contraceptive use and hormone replacement therapy. Oophorectomy was treated as a time-dependent covariate.
Age-specific cancer rates (and their 95% confidence intervals) were calculated, based on 5-year intervals. For each 5-year interval, the number of observed cancers was divided by the total number of person-years at risk contributed by members of the cohort to that interval. Age-specific incidence rates were estimated for the entire cohort and separately for the two subgroups of women (Poland versus North America). Based on the calculated age-specific cancer rates, theoretical penetrance curves were constructed for the two subgroups. These curves were generated by applying the observed cancer rates annually to theoretical cohorts of healthy women from age 25 to age 70.
Risk of breast cancer
One thousand four hundred and seventy-seven unaffected women with a BRCA1 mutation were followed for new cases of breast cancer. After a mean follow-up of 4.3 years, 130 new cases of breast cancer were diagnosed in the cohort. The subjects from Poland and North America are compared in Table 1. Over the entire follow-up period, cumulative incidence was 15.9% for women from North America (95% CI = 12.0–19.8%) and was 12.1% for women from Poland (95% CI = 8.0–16.2%) (Figs. 1a and 1b). The average annual risk of breast cancer was 2.4% (95% CI 1.8–2.9%) for women from North America and was 1.7% (95% CI 1.2–2.1%) for women from Poland. The difference in risk was greatest for young women (<age 40) (Table 2). Notably, the risk in North America was highest for women between the ages of 25 and 40 and declined thereafter. By contrast, in Poland, the annual risk was relatively stable from age 25 to 60. Using the age-specific cancer rates for breast cancer calculated in Table 2, we estimated the penetrance to age 70 to be 48.6% for BRCA1 carriers from Poland and 71.7% for carriers from North America.
Oophorectomy was associated with a strong reduction in risk (Table 3) and was much more commonly performed in North America than in Poland (Table 1). To eliminate the effect of oophorectomy (i.e., to estimate the risks for women with intact ovaries) the analysis was repeated, but excluding women with a history of oophorectomy at baseline and censoring the follow-up at date of oophorectomy. In this analysis, the estimated penetrance was 57.5% to age 70 for women from Poland and was 76.3% for women from North America.
A multivariate survival analysis was conducted to estimate the extent to which the hazard was influenced by country of origin. The survival analysis was adjusted for oophorectomy, age at study entry, age of menarche, parity (0,1,2,3,4+), oral contraceptive use (ever/never), tamoxifen use (ever/never) and hormone replacement therapy (ever/never), smoking (ever/never), regular alcohol use (ever/never) and family history (numbers of first- and second-degree relatives with breast cancer). Based on this adjusted analysis, Polish women experienced roughly 50% of the risk of breast cancer of their North American counterparts (HR = 0.54; 95% CI 0.34–0.86; p = 0.01).
It is possible that the different penetrance estimates from the two regions might be due to different distributions of BRCA1 mutations (with different penetrances associated with each mutation). To test this hypothesis, the analysis was restricted to the 604 carriers with the most common BRCA1 mutation (5382insC). The results were essentially the same (Figures 1a and 1b). It is also possible that the higher incidence rates observed in North America were due to more intense surveillance. However, the proportion of screen-detected cancers was not higher in North America than in Poland (Table 4).
In our study, we show that the annual incidence rates for breast cancer are lower for young women with a BRCA1 mutation in Poland than in North America (Canada). We sought to identify factors that might account for the group differences from among the known risk factors. In several respects, the Polish women were at a priori higher risk from breast cancer; they had a lower prevalence of oophorectomy5 and less frequent use of tamoxifen (which was rarely used in either group). However, Polish women had a later age of menarche by an average of 0.9 years. The number of family members with breast cancer were higher for North American women than for Polish women, but adjustment for family history did not affected the hazard ratio. To a large extent, the difference in the strength of the family histories for the Polish and North American women reflects the fact that the great majority of BRCA1 mutations in Polish carriers are three founder mutations6 and as a result, criteria for genetic testing are less stringent than in North America. North American women were also more likely to take oral contraceptives and hormone replacement therapy, but neither of these was a significant risk factor for breast cancer in the present study (on the contrary, hormone replacement therapy was associated with a reduced risk of breast cancer, in keeping with the result of our recent case-control study7). By adjusting for the hormonal and reproductive factors, the effect of Polish residence on breast cancer risk was not attenuated. Furthermore, when we removed women with an oophorectomy from the analysis, the difference was similar.
We did not study dietary factors in this analysis; it is possible that certain dietary components or micronutrients are protective and are more common in Poland. The risk difference is unlikely to be due to obesity or caloric intake per se because the mean weights and BMIs were similar in the two groups (Table 1). In the future, we hope to include a diet questionnaire in this cohort study and to collect and analyze samples for specific nutrients. If dietary patterns contribute to the risk, it may be possible to identify important foods by studying differences between women both within and between countries.
The risk difference observed here does not appear to be due to differences in the intensity of screening; 44% of the breast cancers in Poland were detected by screening versus 32% of the cancers discovered in North America (Table 4).
It is possible that the observed difference represents a chance finding, although the difference was robust at a conventional level of statistical significance. A total of 130 incident cases of breast cancer was diagnosed in the cohort and within each age group, the number of incident cancers was relatively few. Also, the point estimates of lifetime penetrance for the two cohorts are derived directly from the age-specific cancer rates presented in Table 2 and do not take into account the variance in these underlying rates.
It is also possible that the observed effect of residence is based on differences in genetic background. Poland is a genetically-homogeneous country whereas the North American population is ethnically-mixed. In theory, it is possible that there are variant alleles of one or more modifying genes that increase the risk of breast cancer in BRCA1 carriers, and which are relatively rare in Poland, but to date, the variant alleles that have been discovered through GWAS studies of BRCA1 carriers are associated with hazard ratios that are too modest to explain the large difference in risk observed here. For example, an allele of the rs2363956 SNP is associated with a hazard ratio of 0.84 (95% CI: 0.80–0.89) for breast cancer in BRCA1 carriers.8 Even if the allele was present in 100% of North American carriers and 0% of Polish carriers then the baseline risk in Poland would be reduced by only 16%, far less than the 46% difference observed here. Of course, it is theoretically possible that as yet unidentified modifiers have larger impacts than those reported to date.
The impact of residence on the risk of breast cancer before age 40 (Table 2) attests to the importance of risk factors that act early in life. Of the known reproductive risk factors for breast cancer in BRCA1 carriers, age of menarche is perhaps the most profound.9 In our study, we adjusted for age of menarche in the multivariable analysis and therefore the differences in risk could not be accounted for by the small group difference in the age of menarche. It will be of interest to consider other early exposures; future studies should address potential risk factors in the intrauterine period, early childhood and puberty. In this respect, it is notable that early diagnostic chest X-rays (i.e., below age 18) have been implicated as a possible risk factor for BRCA1 carriers in two studies10, 11 but screening mammograms (initiated after age 25) have not been associated with risk.12 Also, oral contraceptives have been suggested to increase the risk for breast cancer in BRCA1 carriers if used before age 25.13
We did not study BRCA2 carriers because few have been identified in Poland. The Polish carriers are represented by three founder mutations,6 and the most common of these, 5382insC is also seen in North America, predominantly among Jewish and Eastern European women. The difference in the incidence rates of cancers among mutation carriers in the two regions mirrors the difference in rates in the underlying populations, suggesting that the modifying factors may have an impact beyond the small proportion of cases which is due to the BRCA1 gene.
It is not clear if the rates of cancer in the Polish carriers are increasing with calendar time. The data from the present study is based on a relatively short period of follow-up - all the incident breast cancers which contribute to the rates were diagnosed between 1996 and 2010 and this period is too short to study trends in cancer incidence. We restricted our study to three countries; there were too few subjects in our international study from other countries to generate meaningful estimates. Outside of Poland and North America there are several other cohorts of BRCA1 or BRCA2 carriers and it is desirable that studies of this type are conducted in different populations.
This work was supported by the Canadian Breast Cancer Research Alliance. S.L.N. was supported by the Morris and Horowitz Families Endowment.