BRCA1 and BRCA2 Gene Mutations and Colorectal Cancer Risk

Systematic Review and Meta-analysis

Mok Oh; Ali McBride; Seongseok Yun; Sandipan Bhattacharjee; Marion Slack; Jennifer R. Martin; Joanne Jeter; Ivo Abraham


J Natl Cancer Inst. 2018;110(11):1178-1189. 

In This Article


The principal finding of our systematic review of 18 and meta-analysis of 14 studies is that the risk of colorectal cancer is moderately elevated in BRCA1 but not in BRCA2 mutation carriers. This increased risk was independent of study design, specific cancer type, ascertainment method, or age.

These findings are quite different from those reported by Sopik et al.[17] in a narrative review of, by their classification, four cohort studies, five case–control studies, and four pedigree studies. These authors concluded that there may be an increased risk for colorectal cancer in BRCA1 or BRCA2 mutation carriers from high-risk families, but that there is little evidence of such risk in unselected populations with founder mutations. Further, the Phelan et al.[22] prospective study of 7015 women with BRCA1 and BRCA2 mutations included in the Sopik et al.[17] review found no colorectal cancer risk signal in the sample at large, but a 4.76-fold increase in risk in women aged 30–49 years carrying the BRCA1 mutation (within a 95% CI margin indicating a 2.21-fold to ninefold incremental risk).[22]

Especially the findings from our meta-analyses provide additional insights into the association of BRCA1 and BRCA2 genotypes and colorectal cancer risk. We detected a statistically significant signal in colorectal risk when considering both BRCA1 and BRCA2 studies. However, our results showed that if there is an increased risk for colorectal cancer, it is among BRCA1 mutation carriers and not among BRCA2 mutation carriers. These findings were not a function of study design for BRCA2: whether examining through studies that reported age- and sex-adjusted estimates, focused on colon cancer, or ascertained BRCA2 carriers. BRCA1 showed a statistically significant risk signal for colorectal cancer with a higher estimates for subgroup analysis through age- and sex-adjusted estimates, and colon cancer. However, the results were not statistically significant for subgroup analysis when including only ascertained BRCA1 carriers. The conflicting results for BRCA1 among subgroups may be due to an inherent bias of pedigree-based studies, as these do not adjust for ascertainment bias. To reduce ascertainment bias, there are strategies to modify segregation analysis and condition on the ascertainment of the proband. However, the pedigree-based studies included in the meta-analysis did not use this statistical method. Note that, as explained in the Methods section, we did not perform statistical comparisons between summary effect measures because of the limited number of studies in the meta-analysis and the associated lack of statistical power. For instance, to compare BRCA1 vs BRCA2 studies on pedigree vs nonpedigree studies, the relative lack of statistical power (10 BRCA1 and 7 BRCA2 studies; 6 pedigree and 8 nonpedigree studies) would have increased the likelihood of type II errors and false-negative findings.

Both BRCA1 and BRCA2 work in a common pathway of genome protection known to function in homologous recombination. However, the two proteins work at different stages in the BRCA1- and BRCA2-mediated homologous recombination pathway. The BRCA1 functions upstream of BRCA2; therefore, the function of BRCA2 is dependent on BRCA1. Also, BRCA1 functions in both checkpoint activation and DNA repair, whereas BRCA2 is a mediator of the core mechanism of homologous recombination.[37] The links between the two proteins are not well understood, but the difference in mechanism might explain why BRCA1 but not BRCA2 is associated with an increased risk for colorectal cancer.

Our findings contribute to the evolving knowledge base about the association of BRCA1 and BRCA2 mutation status and the risk for certain cancers. Table 2 presents the results from previous meta-analyses for breast, ovarian, and prostate cancer, pancreatic cancer, as well as our results in colorectal cancer as reported herein. The association is well established for breast and ovarian cancer and examining for BRCAmutations has been routine practice. The association of BRCA mutations with the risk of developing cancer for less established cancer, such as prostate cancer, colorectal cancer, and pancreatic cancer, show relatively lower risk and with conflicting results. A meta-analysis of prostate cancer showed no association between BRCA1 mutation and cancer risk. Although many other studies showed statistical association for BRCA2,[11,21,25] no meta-analysis has been reported for the association between BRCA2 mutation and prostate cancer. For pancreatic cancer, there is no meta-analysis evidence of studies on the risk associated with BRCA1 and/or BRCA2.However, studies[25,40,41] have shown that BRCA mutations contribute to a higher risk of pancreatic cancer, with relative risk ranging from 2.11 to 5.79 for BRCA2 and 0.80 to 4.11 for BRCA1. Our meta-analyses provide evidence of increased risk of colorectal cancer for BRCA1 and can be further used in the polygenic risk for personalized colorectal cancer screening.

We could not meta-analytically validate the risk of early-onset colorectal cancer risk in BRCA1 carriers, which was perhaps the major conclusion for the previous Sopik et al.[17] study, as the only evidence is from the Phelan et al.[22] subanalysis. This remains a hypothesis to be further investigated, especially considering a clinical challenge of this finding by Evans et al..[42] Of the eight colorectal cancers in the BRCA1-positive women younger than 50 years, only four were confirmed in the Phelan et al.[22] analysis. Of these, two were squamous cell carcinomas of the anus, which Evans et al.[38] attribute to human papilloma virus. The two confirmed colorectal adenocarcinomas were similar to the expected number of 1.68 cases. Evans et al. also reviewed their own data on 890 women with a BRCA1 mutation, of whom 592 were less than age 50 years. One case of colorectal cancer occurred among those younger than age 50 years. There were no cases of colorectal cancer among women older than age 50 years, whereas 3.78 cancers would have been expected.

Further, our quantitative meta-analysis results from 14 studies differ from the qualitative interpretations by Sopik et al.[17] of 13 studies. Our data provide compelling evidence of an increased risk of colorectal cancer in carriers of the BRCA1mutation, but not carriers of the BRCA2 mutation. Reasoning by elimination, the finding of no elevated risk among BRCA2 mutation carriers and the dilution this likely had on the combined BRCA1 and BRCA2 analysis suggests that BRCA1 mutation carriers, in general, may be at greater risk of colorectal cancer. This quantitative finding is in contrast to the qualitative conclusion by Sopik et al.[17] of no increased risk in the population of BRCA1 mutation carriers at large, and the possible nearly fivefold risk in BRCA1-positive women younger than age 50 years.

There are some methodological differences between our study and the review by Sopik et al..[17] We complemented their narrative review with both a systematic review and a set of omnibus and stratified meta-analyses. We categorized our 18 studies as cohort studies with ascertained BRCA mutation carriers (n = 5), cohort studies involving pedigree analysis (n = 6), case–control studies (n = 5), cohort studies, and cohort studies involving kindred (n = 2), of which 14 were included in the meta-analysis. Further, both reviews covered the same time period of about 20 years. However, our review included one study[21] published after the conclusion of the Sopik et al.[17] evaluations, but also an additional four studies[24–27] released within the Sopik et al. 1994–2014 period. Of these five studies, four[21,25–27] were entered into the meta-analysis. They carried moderate weight in the overall BRCA1 or BRCA2 meta-analyses, low weight in the BRCA1 subanalysis, and high weight in the BRCA2 subanalysis. This may have strengthened the evidence against a BRCA2 risk association with colorectal cancer. The lack of a statistically significant BRCA2association may also have added to the finding of, at best, an equivocal association between general BRCA1 or BRCA2 mutation status and colorectal cancer risk. Importantly, the Mersch et al.[21] 2015 study in the BRCA1 subanalysis had a low relative weight in the meta-analysis that showed an increased in risk of colorectal cancer. This suggests that the studies included by Sopik et al.[17] in their review indeed would have yielded a statistically significant OR had these authors performed a meta-analysis.

Among the included studies, Brose et al.[28] and Thompson and Easton[11] had some duplication in terms of participants, but we included both the studies for three reasons. First of all, only 96 of 147 families included in Brose et al. are also used in the study by Thompson and Easton. Second, our study analyzed the presumed BRCA1 mutation carriers from the Brose et al. study. As to Thompson and Easton, our analyses included only 2245 tested BRCA1 mutation carriers, although the entire study included 11 847 participants. Therefore, all the presumed BRCA1 carriers in Thompson and Easton, duplicated in Brose et al., were not included in our analysis. Lastly, we conducted a sensitivity analysis by removing the Brose et al. study, and did not find a statistically significant difference. Therefore, we concluded that the overlap in the number of participants between Brose and Thompson in our analysis is relatively low, and included both of the studies.

Our meta-analysis has limitations and needs to be interpreted with some caution. Reviews hold the inherent risk of favoring some studies over others or some findings over others. First, heterogeneity in study designs must be acknowledged. Although most studies differentiated between BRCA1 and BRCA2 mutation carriers, some studies did not. Studies differed in how they ascertained BRCA mutation carriers with some studies defining mutation based on three well-known founder mutations, whereas other studies used whole-gene sequencing. Some studies did not confirm BRCA mutation by genetic testing but used pedigree instead, assuming that 50% of the study population would have BRCA mutations. Most studies focused on colorectal cancer, but three studies were limited to colon cancer. Due to the heterogeneity in study design, we performed subgroup analyses that, due to limited statistical power, did not permit stratification by BRCA mutation type or comparator. Only six studies were designed specifically to capture the risk of colorectal cancer in BRCA mutation carriers as their primary goal. The low incidence of colorectal cancers in some studies may be because these studies included several cancer types. Our stratified analyses carried with them the risk of limited statistical power, as the number of studies included in our analysis was small. Based on the NOS quality assessment results, all studies included in the meta-analysis were of moderate quality (NOS = 4 to 6). Therefore, it was not feasible to conduct further analyses to assess heterogeneity by study quality. Second, all the studies reviewed were noncontrolled observational investigations that, consequently, may have been unbalanced with regard to baseline clinical and demographic characteristics of the subjects. Although the real-world practice may be better reflected in observational studies, the relative lack of baseline characteristics including risk factors make their interpretation more difficult and may induce bias. Family history, preexisting cancer, diet, smoking status, and increasing age are known risk factors for colorectal cancer.[43] Studies varied in the effect measure used, including SIR, RR, and OR, which necessitated some effect measures to be converted into ORs to make them useable in the meta-analyses.

In conclusion, where a previous qualitative review suggested an increased risk of early-onset colorectal cancer only in women under the age of 50 years carrying a BRCA1 mutation, our systematic review and meta-analysis points at a 1.49-fold greater risk of colorectal cancer in BRCA1 mutation carriers, but not in BRCA2 mutation carriers.