Our data suggest that breast cancer risk among mutationnegative women from BRCA1/2
mutation-positive families is similar to that observed in the general population. These results are similar to those reported by Domchek et al., who saw no increased risk of breast cancer among a similar cohort of 375 mutative negative women [6
]. While Domchek et al. did not report an aggregate SIR for invasive and non-invasive cancer, based on the total of four observed and 4.9 expected invasive plus in situ diagnosis, the risk does not appear to be elevated in this study. Conversely, Smith et al. reported a 5-fold increase in risk, although prospectively obtained data in a subset of their cohort suggested about a doubling of risk [3
]. Several other studies [4
] report risk estimates on the order of a 2- fold increase in risk as well. A comparison of pertinent study characteristics and findings is presented in .
Comparison of studies evaluating risk of breast cancer among mutation-negative women in BRCA mutation-positive families
There are several potential reasons for the wide variability in results from different studies. The Smith study had several fundamental limitations. One major criticism was the possibility of ascertainment bias—i.e., families that meet criteria for genetic testing will necessarily have multiple breast cancer cases, and thus families containing phenocopies are more likely to be tested than those without. In order to overcome this bias, Gronwald et al. looked at a Polish population of 261 sisters of 188 BRCA1
mutation carriers from an unselected population [4
]. They assumed that 50% of these women (130.5 individuals) would be mutation-negative. One breast cancer was reported among 72 non-carriers; based on these numbers they estimated that 2.5 cases would occur in the estimated 130.5 mutation-negative individuals—a number roughly double the expected number of cases in Poland. While this study design does account for selection bias, several other methodologic issues arise. First, the data are retrospective in nature, and the estimated risk ratio is based on a calculated probability of carrier or non-carrier status. Second, and more importantly, the estimate of risk, presented without confidence intervals, is extrapolated from only one case of breast cancer, and thus comes with a wide degree of uncertainty.
The Rowan series [5
] presents a prospective study of 104 mutation-negative women and includes three incident cases of breast cancer. Again the study is small, and the confidence limits surrounding the point estimate for risk is wide, although the lower boundary does not cross unity, suggesting at least a modest increase in risk. Interestingly, of the three women diagnosed with breast cancer, one had a maternal aunt with bilateral breast cancer who tested positive for a mutation in BRCA1
, and a mother who tested negative but who also developed bilateral breast cancer. This example illustrates the point that at least in some families, there is likely some component of risk that is familial, but that is not explained solely by mutations in BRCA1
Our study is the largest prospective study to date, with the longest follow-up, and mutation status in our participants was known rather than being assigned probabilistically. Similar methodology was used by Domchek et al., who also noted no increase in risk of invasive breast cancer among mutation-negative women [6
]. One strength of the current study was our ability to perform stratified analyses. We saw no difference by degree of relatedness to a known BRCA
-positive individual, but, interestingly, there was a slight though not statistically significant increase in risk among women with a first degree relative with breast cancer. This finding suggests the possibility that there exist genetic and/or environmental modifiers that may influence the risk of breast cancer among mutation carriers as well as non-carriers, and highlight the importance of individual assessment of family history when performing risk assessment for women who test negative for mutations in BRCA1/2
Our study takes into account the protective effect of oophorectomy, which no other analysis to date has done. We calculated adjusted O/E ratios using two methods—one which accounted for the breast cancer risk reduction seen in the general population after oophorectomy, and the second using estimates of risk reduction from a population of BRCA mutation carriers. Both methods have potential strengths and weaknesses. The general population estimate of the reduction in breast cancer risk from oophorectomy is probably most accurate for this population of mutation-negative women, but the study from which we derived this estimate did not take into account age at oophorectomy, which we know to be important. Thus, we also present the second model, which taken into account age at oophorectomy, but which may overestimate the effect of surgical menopause because it is derived from a population of BRCA mutation carriers. The adjusted O/E ratios using these two methods were 0.89 and 0.94, respectively. Thus, with optimal adjustment for oophorectomy, our point estimate would likely have been somewhere between these two figures.
An additional potential source of bias in this study is the phenomenon of “informative censoring,” which occurs when a given co-variate affects both the outcome of interest and the probability of being censored. In this case, if subjects who were most likely to develop breast cancer preferentially underwent prophylactic mastectomy, and thus did not develop breast cancer, the number of observed cancers would be lower than expected. In our study, only three participants underwent bilateral mastectomy during the study period (at ages 60, 46, and 44, contributing 16, 2, and 7 person-years of follow-up, respectively); if all three of these women were “destined” to develop breast cancer, which would have resulted in a total of 13 incident cases, the adjusted O/E ratio in this study may have been slightly greater than one.
It is also possible that there are women in our mutation-positive families on whom we did not have information who did develop breast cancer. The NCI Clinical Genetics Branch Hereditary Breast and Ovarian Cancer Registry sends periodic follow-up questionnaires to all study family members with known addresses; follow-up for the current analysis was truncated at the date of return of the most recent questionnaire. The questionnaires request updates on personal history as well as family cancer diagnoses, and we attempt to verify all reported cancer diagnoses in the family with pathology reports or physician records. In this analysis, we included any woman with known or inferable BRCA
mutation status on whom we had a recent cancer status update from either the individual or a first-, second-or third-degree relative. It is possible that the study participants who completed the questionnaires would not know about breast cancer diagnoses in more distant relatives. However, in our study, cancer status information came from third-degree relatives for only 13% of participants, and data suggest that family reporting of breast cancer diagnoses in the context of familial syndromes is quite accurate [10
Lastly, although the study had a mean of 17.7 years of follow-up, this cohort is still relatively young, and thus an excess risk of later-onset breast cancer may still emerge as the cohort ages.
In summary, we were unable to detect an excess risk of breast cancer among mutation-negative women in families with known BRCA mutations, with the possible exception of those with a family history of breast cancer in a first-degree relative. Even in that context, the O/E ratios and confidence intervals noted were in keeping with risk imparted by a first-degree family history in the general population. We saw no increase in risk after adjustment for the higher rate of prophylactic oophorectomy in our study population. Taken in concert with other prospective studies, these data suggest that current information is insufficient to mandate changes in screening practices for mutation-negative women from BRCA1/2 mutation-positive families. Additional research regarding both genetic modifiers and environmental factors that influence breast cancer risk may help to clarify individual breast cancer risk among women in this population.