Landmark Analysis

Landmark times were selected in advance for clinical reasons, but other intervals might capture symptom risk more precisely. The purpose of landmark analysis was to update symptoms, treatments, physiology and disease history over time rather than rely on a single baseline measurement. The study used landmark analysis with Royston-Parmar flexible parametric survival models to assess time-varying symptoms. Each symptom was identified by at least one relevant Read code in the 3 months before each landmark date.