Regression Modeling of Longitudinal Outcomes with Outcome-Dependent Observation Times and Discontinuous Risk Intervals, with Application to a Malaria Vaccine Study

• Activity Number: 187
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Kay See Tan* and Andrea Troxel and Benjamin French
• Companies: Memorial Sloan Kettering Cancer Center and University of Pennsylvania and University of Pennsylvania
• Keywords: Discontinuous risk intervals ; Informative observation times ; Joint models ; Outcome process ; Observation-time process ; Recurrent events

Abstract

In most randomized trials, the study protocol specifies the data-collection schedule. As-needed visits can also occur when participants seek additional care due to their characteristics and the occurrence of adverse outcomes. In addition, data-collection schedules can be interrupted by observation gaps, during which participants are not at risk for an event or physician visit; these gaps constitute discontinuous risk intervals. In a randomized trial that compared the efficacy of two anti-malaria vaccines, as-needed visits were common due to vaccine side effects and malaria symptoms, and, after a diagnosis of malaria, participants received treatment and were considered not at risk for a new malaria episode for 28 days. We introduce a generalized joint regression model for the longitudinal outcomes of interest and the observation times that can accommodate outcome-dependent observation times. Modified risk sets in the observation-time model accommodate discontinuous risk intervals. We illustrate the utility of our method using the vaccine trial and show that our method performs well under different patterns for outcome-observation dependence and discontinuous risk intervals.