NHL and genomic variability in RAG1 and BRCA2

Response:

In the letter by Scott and colleagues in this issue of Blood, the authors failed to identify an association between RAG1 K820R or BRCA2 N372H variant alleles and risk of non-Hodgkin lymphoma (NHL), while we had observed a relationship in a previous study. Small differences in the ethnicity distribution between the 2 studies probably do not account for the discordant findings, as we did not detect risk estimate heterogeneity according to ethnicity. The 2 studies reported a similar gene variant prevalence among controls, which was also consistent with that seen in other populations.^1,2  The similar prevalence suggests another potential explanation for the discrepancy: that small differences in control genotype prevalence, coupled with slightly larger differences in case genotype distributions (as in our studies) can create apparently spurious positive (or null) results. This highlights the possibility that sampling variability may play a major role in seemingly discordant findings. If a slightly different sample of our respective eligible participants had been enrolled, we might have had concordant results. Resolution of this issue in genetic association studies will require very large datasets and thousands of study participants to develop large-scale evidence for genotype-disease associations, as several authors have noted.^3–5  Such initiatives are beyond the scope of most single investigations, and have led to the creation of consortia to address these questions, not only through standard meta-analyses, (which can be subject to publication bias), but by pooled analyses of individual-level data from both published and unpublished sources.⁶  Few results are yet available from these undertakings. The International Lymphoma Epidemiology Consortium (Interlymph), which includes related initiatives focused on other hematopoietic malignancies, is only one of many such consortia,⁷  and welcomes other investigators to join this collaborative effort.