A Spline-Based Flexible Model for Comparing the Cumulative Effects of Time-Dependent Exposures on Survival Outcomes

• Activity Number: 648
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Chenkun Wang*+
• Companies: Indiana University
• Keywords: Survival analysis ; Cox's proportional hazard model ; Regression spline ; Pharmacoepidemiology ; Time-dependent covariates ; Likelihood ratio test

Abstract

In pharmacoepidemiologic studies, comparing the risk of adverse effects between two medications presents a unique challenge in terms of capturing the complex time-dependent drug exposure and modeling the varying effects of the medication over time. Conventional models often rely on a prior chosen cumulative effect, such as the current dose or a total cumulative dose. We present a flexible modeling approach for comparing the cumulative effects of two time-dependent exposures. The weight functions that assign weights to dose in the past are modeled using cubic regression splines and estimated in the Cox proportional hazards model. Likelihood ratio test is used to compare the equality of two cumulative medication exposures on survival outcomes. We evaluate the proposed method in a simulation study and also apply it to an example data set comparing the risk of cardiovascular diseases between two types of antidepressants, i.e. selective serotonin reuptake inhibitor (SSRI) and tricyclic antidepressants (TCA).