Background: Few studies have evaluated the role of host (germline) genetics in predicting outcome in follicular lymphoma (FL) patients. Using patients treated with rituximab (R) based immunochemotherapy (IC) from 3 clinical trials and one prospective observational study, we conducted the first genome-wide association study (GWAS) of prognosis in FL patients to identify germline genetic predictors of event-free survival (EFS) overall as well as failure within the first 24 months after treatment initiation (EFS24). Methods: We conducted a GWAS study of FL patients (n=1,054) treated in first-line therapy with R-CHOP (n=726, 69%), R-CVP (n=255, 24%) and R-FM (fludarabine and mitoxantrone) (n=73, 7%) from the following studies: PRIMA (n=391) and RELEVANCE (n=205) of the Lymphoma Study Association; FOL05 (n=199) of the Fondazione Italiana Linfomi; and the University of Iowa/Mayo Clinic Molecular Epidemiology Resource (MER) (n=259). After IC, 437 patients (41%) received R-maintenance. Genotyping was performed using Human Core exome (PRIMA, RELEVANCE, FOL05) and 660 Quad (MER) BeadChips. Minimac4/1000G was used to impute additional SNPs. After stringent quality control measures, 1,287,537 SNPs across the 4 cohorts with a minor allele frequency (MAF) >0.05 were evaluated. Association testing was conducted assuming a log-additive genetic model. We calculated the hazard ratios (HR) by Cox regression analysis for EFS. Results were adjusted for FLIPI score, sex, cohort, principal component analysis of population stratification, treatment (R-CHOP/R-FM vs. R-CVP), and use of R-maintenance (modeled as a time-dependent variable). To identify early clinical failures, we conducted a GWAS using a piecewise Cox model to find SNPs associated with events within the first 24 months (EFS24) vs after 24 months. Because of the heterogeneity of our prognostic cohorts, we utilized a cross-validation statistical technique to identify robust findings that replicate across cohorts. The data were iteratively analyzed in discovery and validation blocks using a "leave-one-cohort-out" (LOCO) strategy: in each round, one cohort was left for validation while the other three cohorts were combined for discovery. Top SNPs were selected from each discovery block of three cohorts with an inclusive p<0.0001 threshold to move to validation in the remaining cohort. SNPs were considered validated with a p<0.1 threshold and a consistent HR direction between discovery and validation blocks. Validated SNPs were then combined into a final pooled analysis and were ranked based on p-value and number of cross validations in the context of standard pooled GWAS results. Results: In the final analysis for EFS prediction, the top SNP was rs72625024 at 3q27.3 (MAF=0.11; HR=0.62, p=9.37x10-8), which was discovered and validated in all 4 LOCO rounds. Figure 1 shows this and other top SNPs from the LOCO analysis mapped onto traditional pooled GWAS results. This SNP also predicted overall survival (OS) in a model independent of FLIPI and other prognostic factors (HR=0.69, p=0.013). rs72625024 is near FETUB (fetuin B), which is involved in the inactivation of the PI3K/AKT signaling pathway. The other top SNPs for EFS were at 3q21.1 and 8p23.2, but they did not reach statistical significance and showed weak validation in the LOCO design. In the analysis of EFS24, rs114695031 at 14q32.13 predicted events during the first 24 months and was discovered and validated in all 4 LOCO rounds (MAF=0.20; HR=0.58, p=1.93x10-8) (Figure 2). This variant is an intronic SNP localized in T Cell Leukemia/Lymphoma 6 gene, a long non-coding RNA and putative tumor suppressor. Suggestive associations for early failure were observed for rs12824955 at 12q24.11 (MAF=0.27; HR=0.60, p=1.46x10-7) and validated in 3 of 4 LOCO rounds, while rs333302 at 3q21.1 (MAF=0.42; HR=1.61, p=1.15x10-6) did not reach statistical significance. For events after 24 months, no SNPs were statistically significant. Conclusions: In this large GWAS study, a locus at 3q27.3 (near FETUB) was associated with EFS after adjustment for other prognostic factors and treatment. Furthermore, a genome-wide significant locus (p<5x10-8) at 14q32.13 (near TCL6) predicted EFS24, suggesting a specific role in early failures. These results, requiring additional external validation, support a role for inherited susceptibility impacting outcomes in IC-treated FL and that may vary by timing of failures. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal
Read full abstract