A Test for Constancy of Incremental Incidence Rates in a Long-Term Retrospective or Prospective Safety Study

• Activity Number: 679
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Girish Aras*+ and Jingyuan Yang
• Companies: Amgen and Amgen Inc.
• Keywords: Incremental incidence rates ; Long term safety study ; Survival Analysis ; Rare Adverse events

Abstract

In long-term safety studies, the incremental incidence rate is a useful concept in evaluating risks related to adverse events of medical interest. We propose estimates for the incremental (say annual) incidence rates in presence of independent random right censoring and describe their statistical properties. More importantly, we propose tests for the null hypothesis that the incremental incidence rates remain constant during the study against various alternative hypotheses, such as that they change over the years or have a specific trend, say monotonic increasing or decreasing. In contrast with the existing methods that test constant hazard rates, the proposed method allows non-constant hazard within each year, therefore it is especially useful for events of medical interest that have a seasonal occurrence pattern. This article explores and characterizes properties of these tests under various alternatives via theory and simulations. A numerical example based on a real long-term cohort safety study is also provided. It also illustrates limitations of eyeballing the rates, as opposed to formally testing them, to assert a causal link between exposure and