Case-Only Trees and Random Forests for Exploring Genotype-Specific Treatment Effects in Randomized Clinical Trials with Dichotomous End Points
TLDR Case-only trees and random forests improve predictions of treatment effects in clinical trials.
The study introduced case-only trees and random forests to identify gene–treatment interactions and estimate marker-specific treatment effects in clinical trials with dichotomous end points. Using a prostate cancer prevention trial as an example, the researchers demonstrated that these methods, which focus solely on cases, provided more accurate predictions of treatment effects and better measures of variable importance compared to the interaction tree method that includes both cases and controls. This approach showed potential in discovering genotypes that may affect the efficacy of finasteride in preventing prostate cancer.
