• breast cancer
• BRCA1
• BRCA2

• AJ, Ashkenazi Jewish
• YOD, year of diagnosis

There has been much discussion regarding the penetrance of breast cancer in BRCA1/2 mutation carriers (hereafter “carriers”). Both genetic and epigenetic factors could be influencing the reported penetrance estimates. We wanted to establish whether the penetrance of BRCA1/2 mutations is changing over time. To limit the genetic variability, we studied a cohort of 292 Ashkenazi Jewish (AJ) women diagnosed with first primary invasive breast cancer between 1 January 1980 and 1 November 1995 at a single Montreal Hospital, without regard to family history. All women were diagnosed at less than 65 years of age. Pathology blocks were identified from all women and the three AJ founder mutations in BRCA1/2 (185delAG, 5382insC (BRCA1) and 6174delT (BRCA2)) were identified in archival samples using PCR based techniques described previously.¹ We identified 41 (14%, 95% CI 10.2-18.6) BRCA1/2 carriers (31 in BRCA1 and 10 in BRCA2). The difference in mutation frequency between BRCA1 and BRCA2 carriers (10.6% v 3.4%) is statistically significant (Z=3.40, p=0.0007). Given that the population allele frequencies of AJ founder mutations in BRCA1 and BRCA2 are approximately equal (∼1%), this would suggest that BRCA1 has a higher penetrance than BRCA2 by age 65. This observation supports previous data from Canada.² We then divided the data into quartiles by year of diagnosis and determined whether the proportion of mutation carriers was changing over time. The number of founder BRCA1/2 mutations per quartile of year of diagnosis increased from eight (11.0%) to 15 (20.5%) over the 15 year period of the study, and the χ² p value for the trend in mean scores was 0.047 (table 1). This suggests that the penetrance of BRCA1/2 mutations to the age of 65 years is increasing. This is important, as previous studies of penetrance have not taken into account the year of diagnosis of the mutation carriers.³

The median age at diagnosis for all subjects was 53.6 years. Interestingly, BRCA2 carriers were diagnosed at a statistically significantly older median age than were BRCA1 carriers (59.8 years v 47.5 years, p=0.02, Mann-Whitney test). To study further the change in penetrance over time, we studied the effect of year of diagnosis (YOD) on age at diagnosis. The median YOD for all subjects was 1988.5. The median age at diagnosis for BRCA1 carriers diagnosed below the 50th centile was 50.7 years. For those diagnosed in the second half of the cohort study, the median age at diagnosis was 44.9 years (p=0.21). Although this difference does not reach statistical significance, in 251 non-carriers the equivalent ages were 54.5 and 53.3 years respectively, a much closer difference in age at diagnosis (p=0.98). For the 10 BRCA2 carriers, this difference was almost 10 years (60.4 years v 50.6 years, dichotomised at the median YOD, p=0.38). To show the change in age at diagnosis over time, we plotted the YOD against the age of diagnosis in BRCA1 carriers, in BRCA2 carriers, and in non-carriers (fig 1). The slopes for BRCA1 (point estimate of the slope −0.724, p=0.066) and BRCA2 (slope −0.303, p=0.68) clearly deviate from the non-carrier slope (−0.024, p=0.85), suggesting an interaction between mutation status and YOD. However, perhaps because of the small numbers of mutation carriers, these slopes are not statistically different, and their confidence intervals overlap (fig 1). The age at diagnosis of breast cancer among the non-carriers is virtually constant over time, whereas the age at diagnosis is falling for women diagnosed more recently with BRCA1 or BRCA2 related breast cancer. Taken together, these findings suggest that the age dependent penetrance of BRCA1/2 mutations could be increasing.

Some previous pedigree based studies have found that the age of diagnosis of familial and/or hereditary breast cancer is falling. Genetic anticipation,⁴ birth cohort effects,⁵ or ascertainment bias⁶ have all been suggested as possible underlying mechanisms. It is important to note that bias cannot be ruled out in any series of mutation carriers that were not ascertained at the time of diagnosis and without reference to family history.

Another possible reason why we might observe an increase in mutation frequency and a fall in age at diagnosis is that there were changes in referral patterns to the single hospital where all the subjects were diagnosed. For example, young women with a family history of breast cancer could have been preferentially referred in the latter years (1990-1995) of the study period to this hospital, whereas those without a family history or older women were referred to other hospitals. This would result in more BRCA1/2 mutation carriers being diagnosed at the study hospital without any change in penetrance. This is possible, but we cannot confirm or refute this suggestion as our study is anonymised and we have no family history information. It does, however, seem a somewhat implausible scenario. Notably, in fig 1, there is virtually no change in the age at diagnosis of non-carriers, so whatever is causing the fall in the overall median age at diagnosis per quartile (table 1) can only be acting on the carrier subgroup.

Another possible explanation for our observation is that from the mid-1980s onwards, some young high risk women decided to undergo preventive oophorectomy, which may have delayed or prevented their breast cancer diagnosis.⁷ This would result in a recent deficiency of late onset cases, and could resemble the pattern of an excess of early onset cases. This is a possible explanation for the decrease in the age of diagnosis of mutation carriers over time, but cannot explain the increase in the total number of mutation carriers in recent years.

In the general population, mammographic screening is likely to downstage those cancers identified. To consider this potential source of bias, we analysed the change over time in the size of the primary tumour and the probability of associated positive lymph nodes. Cancers in non-carriers diagnosed more recently were indeed smaller (median size 1.5 cm) when compared to those diagnosed earlier (1.8 cm, p=0.02, dichotomised at median YOD). In contrast, BRCA1/2 related cancers were not different in size (median size of 2.0 cm for those diagnosed before and after the median YOD, p=0.73). We noticed that among non-carriers, the proportion of women with node positive breast cancers was decreasing with time, whereas this effect was not seen in BRCA1/2 carriers. In a logistic regression model developed to predict nodal status over time, there was a noticeable interaction between mutation status and YOD. BRCA1/2 carriers diagnosed recently were twice as likely (odds ratio 2.1) to be node positive than were those diagnosed earlier, whereas in non-carriers the effect was reversed: non-carriers diagnosed earlier were twice as likely (odds ratio 1.9) as were those diagnosed recently to be node positive (p=0.07). Neither of these results suggest that mammography has resulted in a downstaging of BRCA1/2 related breast cancer. Mammography may also diagnose invasive breast cancer at younger ages than in non-screened subjects, and in recent years BRCA1/2 carriers may have used screening mammography at younger ages than non-carriers because of a positive family history of breast cancer. However, in another analysis of the same dataset, we showed that mammography is particularly ineffective in detecting small tumours in young BRCA1/2 mutation carriers,⁸ which is just the group that might be expected to be artificially skewing our results. It therefore seems unlikely that either potential biases are present in our dataset.

The importance of our observation is that the potential benefit of preventive interventions is intimately related to the incidence rate of cancer in those at risk. If the age dependent penetrance is increasing, then this should be taken into account when counselling unaffected women. Moreover, as our study is genetically restricted, it is likely that the change in penetrance we observed is the result of epigenetic (that is, reversible) factors. If such factors could be identified, then it would seem possible that the effect could be diminished, or even reversed. Several unmeasured factors could be contributing separately to the increase in number of mutations and the decrease in age at diagnosis with time, but it is important to note that these factors appear to be acting on the carriers only. This could suggest that the carriers are responding differently to pre-existing environmental or hormonal exposures. Nevertheless, we recognise that proving a change in penetrance of a cancer related gene over time is challenging and, therefore, these data can be regarded only as suggestive early evidence for such an effect. Large population based prospective cohort studies will be required to confirm or refute our preliminary observations.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Age at diagnosis (Y axis) is plotted against year of diagnosis (X axis). All 292 subjects recruited into the study are included. The key for the figure is shown in the box inset into the figure. The characteristics of the slopes are as follows: BRCA1, slope –0.72 (95% CI −1.56-0.12), r²=0.097, p=0.066; BRCA2, slope –0.30 (95% CI −1.98-1.37), r²=0.021, p=0.69; non-carriers, slope –0.02 (95% CI −0.28-0.23), p=0.85.