Genotypes of single nucleotide polymorphisms are subject to misclassification. If ignored, such misclassification can seriously distort the estimated genotype effects on the disease or outcome of interest. Validation data (gold standard or replicated surrogates) are required to assess the degree of misclassification and make adjustments. In practice, gold standard measurements may be unavailable or impractical. Collecting replicated surrogates is a reasonable option for validation data. In most practical applications, collecting replicated surrogates on all study subjects is not feasible; however, obtaining replicated surrogates on a subsample of the study population may be quite feasible. Generating duplicate data for a subsample of the study population is now common practice among genotyping laboratories. The authors propose a Bayesian method that can adjust for genotype misclassification using partial validation data. Simulation results show that the proposed method substantially reduces misclassification bias from the estimated genotype-disease association and provides appropriate uncertainty assessment, as well as improves other desirable properties of the estimated effects. The authors also provide an example showing the application of the proposed method to study data relating non-Hodgkin lymphoma to a single nucleotide polymorphism in the aryl hydrocarbon receptor gene.
- Bias (epidemiology)
- Polymorphism, single nucleotide
ASJC Scopus subject areas