PGx6049: Pharmacogenetic Evaluation of Steroid Response in Fluticasone Furoate and Fluticasone Propionate Treated Asthma Subjects

Trial description: Asthma is a chronic disease of the lungs characterized by airway inflammation, bronchoconstriction and increased airway responsiveness. Inhaled corticosteroids (ICSs) are the most commonly used and the most efficacious therapy for asthma. However, considerable variation in response to corticosteroids among asthmatic subjects has been noted and there is increasing evidence for a genetic basis to this variation in response. The primary objective of this analysis is to use information gathered from ICS-treated asthma patients in three Phase 2B studies (FFA109684, FFA109685, and FFA109687) and four Phase 3A studies (FFA112059, HZA106827, HZA106829, and HZA106837) in an effort to identify genetic markers that are associated with steroid response. The primary endpoint measuring response will be change from baseline to 8 or 12 weeks of treatment (depending on study) in trough (PM pre-dose and pre-rescue bronchodilator) FEV1. In a related analysis, change from baseline to each scheduled visit in trough FEV1 will be used to determine if the linear trend of response during treatment is associated with marker genotypes. Secondary objectives include the identification of markers associated with change in the percentage of rescue-free 24-hour periods from baseline to the response assessment visit (8 or 12 weeks of treatment, depending on study) and to evaluate the ability of rs37973 (in gene GLCCI1) and rs1876828 (in gene CRHR1) to predict steroid response, as reported by McGeachie et al. (2012). An exploratory objective is to estimate the heritability of change in trough FEV1 using the mixed model method developed by Yang et al. (2010). A comprehensive set of genetic markers from across the genome will be analyzed. Genotype imputation will be conducted, with the 1000 Genomes Project whole genome sequence data as a reference set, using an in-house software pipeline developed by GSK Statistical Genetics. Bonferroni adjustments for multiple testing will be made, and significance criteria established. A combined p-value will be derived for each marker using a fixed effects, inverse variance meta-analysis. Ancestry principle components will be derived from genome wide markers and included in statistical analysis to correct for the potential effects of population stratification and potentially to refine self-reported geographic ancestry. QQ plots will be used to assess the extent to which the ancestry principal components have adjusted for population stratification. Variables which are known, or potential, covariates will be individually assessed for association with endpoint variables and with one another. These sets of independent variables will be included as covariates in statistical models or as stratification variables. Marker association will be tested in ICS-treated subjects using a general linear model that includes the identified covariates and the genetic marker, with genotypes coded to detect an additive genetic effect. The p-value for the genetic marker term will then be compared to the significance criterion established. If multiple genetic markers are significantly associated with response among steroid treated subjects, then the effect of each, conditional on the other(s), will be assessed. If one or more markers are discovered to be significantly associated with response, further analyses will be conducted to determine whether differences in response between steroid and placebo treated subjects differ among the genotypes in a useful manner. For the analysis of longitudinal data, mixed models for repeated measures (MMRMs) will be fit to change in trough FEV1 at specific visits after the first dose. Specifically, a random coefficients model will be applied to these data and the interaction between genotype and time will be assessed. To test the hypothesis that multiple rare variants within a gene are associated with steroid response, we will use the optimal unified test called SKAT-O (Lee et al., 2012), within the non-Hispanic white subset only, because existing methods don’t completely correct for rare variant population stratification (Mathieson, 2011).