PGX6847: Exploratory PGx investigation of dabrafenib efficacy in melanoma monotherapy studies BRF113710, BRF113929, and BRF113683

Trial description: Dabrafenib was recently approved as a single-agent therapy for treatment of advanced melanoma in patients with BRAF V600E mutations, and in combination with trametinib for patients with BRAF V600E/K-mutant metastatic melanoma. Inherited genetic variability may be one of several contributing factors that underlie individual treatment response to metastatic melanoma treatment. So far, there have been no previous pharmacogenetic (PGx) data generated around the genetics of treatment response to dabrafenib with regard to efficacy. The primary objective of this study is to identify germline genetic variants that predict differential response to dabrafenib, in metastatic melanoma subjects from clinical studies BRF113683, BRF113710 and BRF113929. These three studies will be assessed independently as well as meta-analyzed. The primary endpoint is Progression Free Survival (PFS), defined as the time from randomization until the first documented sign of disease progression or death due to any cause. Genome-wide association screen (GWAS) data from the Illumina Omni Express Exome (OEE) beadchip was previously generated for the three studies (BRF116604, n=390). These data include approximately 727,000 post-quality control (QC) genetic loci per patient distributed across the genome. The data will be stratified into 3 tiers for analysis: (1) Tier 1: functional variants in 9 candidate genes (n=10); (2) Tier 2: Other functional variants in 25 candidate genes (n=55) and Tier 3: GWAS (n=727,052). Prior to the PGx analysis, QC will be conducted on the genetic markers and on the subjects. No markers will be removed prior to the analysis based on their HWE P-values. However, markers showing substantial evidence of departure from HWE (P<1×10-8) will be investigated thoroughly for laboratory errors or other causes of departure from equilibrium and, if the cause remains undetermined, may be removed from the list of associated markers. Genetic ancestry estimates will be obtained by principal components analysis (PCA) in the EIGENSOFT version 2.0 BETA software package. Plots will be created using the principal components overlaid with self-reported ethnic groups to aid in visualization of the data clustering and to determine the number of principal components required to infer genetic ancestry. The PGx analysis population will consist of the patients enrolled in the clinical studies who provided written informed consent for PGx research, provided a blood sample for genotyping and were successfully genotyped for at least one of the genetic markers under study, have valid phenotype data and pass subject QC. Two PGx patient sub-groups will be analyzed: (1) subjects with BRAF V600E mutation from BRF113710, BRF113929 and BRF113683 respectively and (2) subjects with BRAF V600K mutation from BRF113710 and BRF113929, respectively. Patients with V600E mutation positive tumors demonstrated significantly better tumor responses compared to patients with V600K tumors in all dabrafenib studies. Therefore, the primary analysis will focus on subjects with V600E mutation. In the second sub-group analysis, only variants that show nominal association in V600E population will be analyzed in V600K population as their sample sizes are small. The 23 cross-over PGx subjects in BRF113683 will be analyzed in a separate sensitivity analysis along with the subjects only receiving dabrafenib in BRF113683 study. Genotype association tests will be performed assuming an additive genetic model. Cox proportional hazards model will be used to assess the association of genotype with time to event outcome (PFS). Covariates, which were listed in clinical RAPs as potential covariates or evaluated in sensitivity analyses or subgroup analysis of clinical studies, will be evaluated for association with PFS using Cox regression model and proportional hazard assumptions will be checked. Covariates for consideration will be individually evaluated for association with PFS by a univariate Cox model. All covariates that are significantly associated with PFS at p ≤ 0.05 will be included in a multivariate Cox model for further evaluation. Any covariate that is significant at p ≤ 0.05 in the multivariate Cox model will be included as covariate in PGx analysis model. If a covariate violated the proportional hazard assumptions, covariate by survival time interaction will be included for evaluation. Summary statistics [hazard ratio (HR)] from monotherapy arms of BRF113710, BRF113929 and BRF113683 will be meta-analyzed using the inverse variance method which allows for combing effect sizes and weighting by their variance. For declaring candidate gene and genome-wide statistical significance, Bonferroni-threshold criteria will be applied and weighted by 10 Tier I, 55 Tier II and 727,052 Tier III SNPs respectively. The p-value thresholds for significance are 5E-03, 9.09E-04 and 6.88E-08, respectively.