A Polygenic Response Predictor Indicates Survival Benefit of Beta-Blockade in Heart Failure Patients
Li J, She R, Gui H, Zeld N, Sabbah HN, Brunner-LaRocca HP, and Lanfear D. A Polygenic Response Predictor Indicates Survival Benefit of Beta-Blockade in Heart Failure Patients. J Card Fail 2019; 25(8):S111.
J Card Fail
Background: Beta-blockers (BB) are one of the most important therapeutic options for heart failure (HF). However, individual responses to BB treatment vary, which may be due, in part, to genetic variation. We sought to use an unbiased approach to look for a multi-genetic marker profile of BB effectiveness. No previous studies, to our knowledge, have provided insights about treatment-specific survival prediction by combining multiple genetic markers using a genome-wide (GW) approach. Method: Two studies were utilized; Henry Ford Pharmacogenomic Registry (HFPGR) and TIME-CHF. HFPGR patients who were of European ancestry (N=496), met Framingham criteria for HF and had a history of ejection fraction (EF) <50% were included. TIME-CHF patients with EF<50% (n=431) were included. HFPGR patients were randomly divided into derivation (N= 248) and validation (N= 248) sets, and TIME-CHF served as a test cohort. A statistical pipeline using five-fold cross-validation was implemented to identify beneficial gene-drug interactions and create a polygenic response predictor (PRP) using Cox regression for all-cause mortality with predictor variables including treatment, SNP and the SNP-by-treatment interaction in the HFPGR derivation set. Additional covariates included a clinical MAGGIC score and a propensity score for BB use. The PRP was reassessed for predictive validity in HFPGR validation and TIME-CHF cohorts. Results: After imputation and quality control, 7,055,710 SNPs, were available for analysis. Over a median follow up of 2.0 years there were 53 deaths in derivation, 36 in HFPGR validation and 165 in TIME. The optimized model identified 44 SNPs to make up the PRP with optimal cutoff at the 30th percentile (≤30% are BB responders and >30% are BB non-responsive). The results are shown in the figure. In the HFPGR validation set the HR of BB exposure for death in the BB responders vs. non-responders were 0.02 vs. 1.07 (p-value for interaction=0.208), while in TIME-CHF these were 0.12 vs. 0.98 (p-value for interaction=0.029). Conclusion: A BB-PRP among European ancestry patients (a 44-SNP genetic profile) replicated in two distinct cohorts and may identify patient subgroups that are BB responders vs. non-responders. Additional study is needed to prospectively validate this BB-PRP, illuminate the mechanisms underlying the PRP, and establish similar tools in other ancestral groups (e.g. African Americans).