Recurrent Events Analysis Using Landmark Andersen-Gill Model with Time-Varying Covariates

• Activity Number: 351 - Statistical Issues Specific the Therapeutic Areas- 3
• Type: Contributed
• Date/Time: Tuesday, July 31, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Zheyu Liu* and Vivian Lanius and Dejian Lai
• Companies: Bayer Pharmaceuticals and Bayer AG and The University of Texas Health Science Center at Houston
• Keywords: Recurrent events; Andersen-Gill model; landmark analysis; time-varying covariates; stroke

Abstract

We propose an extension of the landmark Andersen-Gill model on recurrent events analysis with time-varying covariates in cardiovascular trials. In the proposed model, covariates effects can be estimated at any specific time interval during the study period, adapted to the changing at-risk population, and while applying the most recent longitudinal measurements. We pre-define a set of landmark time points within the follow-up study interval. At each landmark time datasets, an Andersen-Gill intensity model was fitted using individual longitudinal data who were at risk at pre-define landmark period. In addition, we assume the time varying covariates effects on the risk of recurrent events are smooth functions over time. Simulation studies on recurrent events reveal that this proposed approach has consistent performance compare to standard model but is nevertheless robust against model misspecification. We applied the landmark approach to both neurological stroke study and a cardiovascular stroke trial by assessing time varying covariates effects on time to event outcomes.