We observed 2 BRCA2 6174delT mutations and no BRCA1 mutations (Table ). One carrier, diagnosed at age 56 with a Gleason score of 7/10, had two uncles with prostate cancer; the second carrier, diagnosed at age 76 with a Gleason score of 9/10, had no relatives with prostate cancer but had an uncle and a daughter with breast cancer (Figure ).
BRCA1/2 founder mutation frequencies in AJ unselected prostate cancer cases and controls
Pedigrees of the 2 BRCA2:6174delT mutation carriers. Probands are identified with an arrow. PSU = primary site unkown; TAHBSO = total abdominal hysterectomy with bilateral salpingo-oophorectomy.
Recent studies reporting BRCA1/2
founder mutation frequencies in unselected AJ men with prostate cancer have also failed to find an excess of carriers among affected individuals [13
], with the exception of one study [16
] where a slight but significant excess of BRCA1
:5382insC carriers was observed. However, all studies were small, and lack of power may have prevented some differences from being detected. We therefore combined our results with observations in cases and, when available, controls from these studies. In addition, we compiled data from various reports which previously examined mutation frequencies in unselected AJ population controls [2
] and AJ volunteers [8
]. There were no statistically significant differences between the frequencies of the founder mutations in individuals affected with prostate cancer compared with their frequencies in AJ population controls (Table ).
Different recruitment methods were used by the various groups cited in Table as sources of controls. Hubert et al [14
] selected a group of 87 healthy elderly Israeli men (median age 71 years) with no history of cancer as controls, thereby attempting to compensate for possible age-related variations in mutation frequencies. The larger populations studies [2
] performed mutation testing on AJ individuals from either the US or Israel (or both) who were referred for unrelated genetic testing (e.g. Tay Sachs, Cystic Fibrosis, Fanconi Anemia, etc.), with no information available on gender, age or cancer history for these participants. Since genetic testing for the recessive conditions listed above would usually be undertaken prior to making the decision of having children, this larger group is more likely to have a lower median age than that observed in prostate cancer cases. This group should, however, otherwise provide representative population frequencies for the mutations. Finally, controls from Struewing et al [8
] were AJ volunteers from the Washington D.C. area who wished to participate in a study on breast and ovarian cancer. The authors acknowledge that such a recruitment scheme led to a higher than expected proportion of participants reporting a personal or family history of breast and ovarian cancer. This may conceivably result in an exaggerated mutation frequency in this control group compared to frequencies in the general population. In addition, participants in this control group included a number of siblings and relatives, including some who were mutation carriers, potentially further increasing the apparent frequency of BRCA1
mutations in this cohort. We therefore compared the mutation frequencies in volunteers from the Washington D.C. area [8
] to frequencies obtained from combining results from our other sources of controls [2
] to determine if frequencies from the volunteer group were different from those from the general population. Frequencies for BRCA1
:185delAG and BRCA2
:6174delT were slightly lower in the volunteer group, although the difference was not statistically significant. In contrast, the BRCA1
:5382insC mutation was significantly over-represented in volunteer controls compared to random population controls (OR = 3.3, 95%CI: 1.1–13; P
= 0.02). In consequence, we compared our combined cases to controls once again, this time excluding control data from the Washington D.C. study. While there was no change for BRCA1
:185delAG and BRCA2
:6174delT, we now observed a stronger association between the low frequency BRCA1
:5382insC mutation and prostate cancer (OR = 6.1, 95%CI: 0.54–42; P
= 0.07). However, as indicated by the wide confidence interval, this effect is driven entirely by 2 carriers observed by Vazina et al [16
], the only report of BRCA1
:5382insC mutations among prostate cancer cases across 3 studies. According to the authors, neither carrier has a family history of prostate cancer.
It is difficult to reconcile our observations with results from previous epidemiological studies suggesting an increased prostate cancer risk in relatives of mutation carriers [5
]. One possible explanation may be that a diagnostic bias exists in families where hereditary cancer cases are found. Specifically, having a relative affected with BRCA1/2
-related breast or ovarian cancer may encourage relatives to undergo testing and reveal the existence of prostate tumors which may otherwise have remained asymptomatic and undetected, thereby artificially increasing the incidence of prostate cancer cases in families bearing BRCA1/2
If there were a significant decline in the mutation frequency of founder BRCA1 and/or BRCA2 mutations in older Jewish males, then it is possible that using frequencies of 0.91% for BRCA1:185delAG, 0.27% for BRCA1:5382insC and 1.3% for BRCA2: 6174delT (Table ) from a potentially younger population will result in an underestimation of the odds ratios observed. There are no data published that address this possibility, but if we assume that age-matched controls would have overall allele frequencies of 0.64%, 0.19% and 0.91% (observed frequencies decreased by 30%) for the 185delAG, 5382insC and 6174delT alleles, respectively, then using the cases and controls in Table we still do not reach statistical significance for an association between founder mutations and prostate cancer: BRCA1:185delAG, OR = 1.7 (P = 0.23); BRCA1:5382insC, OR = 3.6 (P = 0.12); BRCA2:6174delT, OR = 1.4 (P = 0.33). Thirty percent is a generous reduction; these results suggest that a bias due to an age-related decline in mutation frequency in the control population is unlikely to be a major explanatory factor.
A third possibility is that only certain BRCA1/2
mutations are associated with an increased prostate cancer risk. In the initial report from the Breast Cancer Linkage Consortium [10
], there was a significant excess risk of prostate cancer for male BRCA2
mutation carriers (RR: 4.65, 95% CI: 3.48–6.22). In a further analysis, Thompson et al [17
] found that the risk of prostate cancer was lower in carriers of BRCA2
mutations located in the ovarian cancer cluster region (OCCR; nucleotides 3035–6629, including the AJ founder 6174delT) than in carriers of mutations clustering elsewhere in the gene (RR = 0.52; 95%CI = 0.24–1.00; P
= 0.05). This observation was recently indirectly supported by a large study examining 263 men with early-onset prostate cancer (55 years and less) where the authors sequenced the entire BRCA2
coding region and found 6 truncating mutations, all located outside the OCCR [18
]. However, this hypothesis cannot explain discrepancies in findings between epidemiological and direct mutation detection studies where an increased prostate cancer risk was also observed in relatives of AJ founder mutation carriers, including BRCA2:
]. It is notable, however, that in the study of Warner et al. [9
], a significant difference in cumulative incidence of prostate cancer to age 85 (33.6% vs. 12.6%, P
= 0.049) was observed when comparing relatives of women with and without founder BRCA1/2
mutations, who were themselves affected by breast cancer. Thus other factors may account for the differences in prostate cancer incidence observed.
With the combined results from nine studies, we have an 80% power to detect ORs of 2.7, 6.6 and 2.5 (185delAG, 5382insC and 6174 delT, respectively), while the values we observed range between 1.0 and 2.5. Therefore, we do not have the power to rule out small effects even with our combined sample size. For a mutation with a population frequency near 1%, we would need more than 3,200 cases and 3,200 controls to rule out an OR of 2.0 or greater; over 10,000 cases and 10,000 controls would be needed to exclude an OR of 1.5 or greater. These numbers indicate that larger studies will be needed to rule out a small effect by BRCA1/2 founder mutations on prostate cancer risk. However, using the observed mutation frequency in controls, we can estimate the population attributable risk per cent (PAR%) of these mutations as follows:
PAR% = Pe (RR-1)/ [1 + Pe(RR-1)] × 100
= proportion of exposure in controls and RR = observed relative risk between cases and controls. PAR% values will reflect the proportion of prostate cancer cases attributable to these AJ founder mutations based on their frequency in the population; from the combined data in Table , these values are 0.18% (BRCA1
:185delAG), 0.40% (BRCA1
:5382insC) and 0% (BRCA2
:6174delT). In comparison, based on previously published data on cases diagnosed after 55 years of age [19
], the PAR% of AJ BRCA1
(185delAG and 5382insC combined) and BRCA2
(6174delT) founder mutations are 3.8% and 2.5 %, respectively, in the case of breast cancer, and 23.8% and 16.7%, respectively, in the case of ovarian cancer.