Response

It is critical to validate genetic association results in independent data sets (1). We commend Wallis et al. for their timely validation genotyping of the five KLK6 germline variants we previously associated with aggressive prostate cancer (PCa) using imputation methods. They did not observe such association in 1907 men accrued from tertiary care centers in Toronto, Canada. This report highlights the complexity of genotype-phenotype association analyses in PCa. Our study was based on three cohorts including two large international PCa screening trials with well-documented patient characteristics, accrual methods, and disease definition (2, 3). In addition to single nucleotide polymorphism (SNP) analyses, we reported extensive results on haplotype analyses, adjusted analyses (for KLK2, KLK3, and other clinical factors), the correlation between the KLK6 variants with methylation and serum protein level, and further validation in the Canadian Prostate Cancer Genome Network (CPC-Gene) cohort, which links whole genome germline-somatic genotype information and clinical outcome (4). We reported an association between these variants and biochemical relapse. Since publication, we have validated two of five KLK6 SNPs in the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) consortium at the 5% significance level (data not shown, manuscript in preparation, courtesy PRACTICAL consortium). There are several potential sources for the differences in results between the two studies. First, while Wallis et al. suggest that their cohort (n = 1907) was twice the size of ours (n = 1858), the two are actually quite comparable in statistical power. However, family history and ethnic distribution differed across the two studies. Our cohort was restricted to Caucasian populations, which represented 78% of the multi-ethnic Wallis et al. cohort. While statistical methods were not detailed, population stratification adjustment is required for the association observed by Wallis et al. to be valid (5). These population differences may indicate that KLK6 SNPs have a clinical impact in the context of other germline SNPs elsewhere in the genotype. Second, while the study design and accrual method are somewhat unclear in Wallis et al., the clinical and pathological features of their PCa cases appear to be quite different. Our study included a majority of screen-detected cancers, accounting for the lower PSA levels and percentage of aggressive cancers compared with the Wallis cohort. These clinical differences may reflect a selective role for KLK6 SNPs in specific stages of disease. Third, we noticed some discrepancies in the Wallis data. For example, the first row of their table shows a total sum of variant frequencies (RR + RA + AA) of 91%, not 100% as one might anticipate. Presumably this is either a statistical or typographical error or reflects challenges in genotyping assays that may influence genotype-phenotype associations. We agree with Wallis et al. that the clinical utility of SNP-based biomarkers in the KLK6 gene remains uncertain. As we noted in our article, further studies are needed to identify the causal variants responsible for this association. Open and full sharing of genomic data and associated clinical information are key in this field, and we hope Wallis and colleagues will share their data to allow researchers to replicate and extend their results with those from other consortia.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.