Response to Yang, Shi, Wang et al.

An underrecognized issue in meta-analysis is how to address heterogeneity, not of endpoints or outcomes across studies, but of study designs. Our meta-analysis on BRCA1 and BRCA2 in colorectal cancer (CRC) certainly faced the challenge of integrating findings from (clusters of) studies that differed in elements of design such as defining BRCA mutations, CRC patients, comparison groups, and mutation carriers, which we acknowledged in our paper but also benefitted from additional perspectives by Katona et al. (1) and now also Yang et al.

On the issue of including colon (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM]: 153) but not rectal cancer (ICD-9-CM: 154) in the studies by Ford et al. (2), Van Asperen et al. (3), and the Breast Cancer Linkage Consortium (4), we chose to err on the conservative side and included only pathologically diagnosed CRC in our analysis (3). This was based on a thoughtful observation by Evan et al. (5) that two of the four CRC cases in the Phelan et al. (6) study were cell carcinoma of the anus, which is likely attributable to human papillomavirus. The three papers cited above used the ICD-9-CM: 154 code for rectal cancer, which includes the five subcodes of 154.0, 154.1, 154.2, 154.3, and 154.8. Of these, only 154.0 and 154.1 are codes for neoplasm of the rectum, whereas the other 154 codes include neoplasm of the anus. Therefore, we used only the colon cancer portion reported in these three papers.

Type of comparison group is another issue of heterogeneity in designs, and studies varied in this regard. Yang and colleagues refer to Risch et al. (7), who used relatives of BRCA1/2-mutation carriers as the case group, which is different from the case group in other studies. However, although the Risch et al. study was included in our systematic review, it was one of the four (out of 18) studies not retained for the meta-analysis. It did not meet inclusion criteria because of the different control group used and was labeled a cohort study involving kindred. We share the correspondents’ desire for additional subgroup analyses by type of control group. However, our systematic review demonstrates the limited granularity of the studies with regard to case group stratification, as well as the limited statistical power.

Continued systematic synthesis of extant and future evidence will enhance the understanding of BRCA-mutation status and the risk of CRC. In the interim, our findings provide the level of synthesis afforded by the studies performed to date. With due caution, our study points to a 1.49-fold greater risk of CRC in BRCA1-mutation carriers relative to the risk in comparator populations, but not in BRCA2-mutation carriers.