Assessing Complier Average Causal Effects from Longitudinal Trials with Multiple Endpoints and Treatment Noncompliance: An Application to a Study of Arthritis Health Journal

• Activity Number: 340 - SPEED: SPAAC SESSION III
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Lulu Guo* and Yi Qian and Hui Xie
• Companies: Department of Statistics & Actuarial Science, Simon Fraser University and Sauder School of Business, University of British Columbia and Simon Fraser University
• Keywords: Causal inference; Potential outcome model; Treatment effects estimation; Multi-level model; Hierarchical random effects model; Principal stratification

Abstract

Noncompliance often occurs in randomized controlled trials involving human subjects. To overcome the limitation of traditional approaches such as intention-to-treat and as-treated analysis, the complier-average causal effect (CACE) approach has been developed to estimate the intervention efficacy in the subgroup of compliers who would comply with their assignments. When evaluating multifaceted interventions for chronic diseases, such as arthritis, the endpoints often involve multiple outcomes to measure a complex trait. This raises the challenge of how to optimally pool treatment efficacy estimation across outcome measures. Motivated by Arthritis Health Journal Study, we develop multivariate mixed effects CACE (MCACE) model. The simulation shows, compared with univariate CACE model, MCACE model provides significant increase in the estimating efficiency, including up to 50% reduction in standard error and 1 fold increase in the power to reject the null. Finally, we apply the proposed MCACE model to Arthritis Health Journal. Our finding shows MCACE model yields significant effects on the endpoints while multiple univariate CACE models fail to detect treatment effect.