Sample Size Calculation in Clinical Trials with Mixed Binary and Continuous Co-Primary Endpoints Modeled by Gaussian Copulas

• Activity Number: 260
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Beilei Wu* and Alexander de Leon and Daniel Bonzo
• Companies: PAREXEL International and University of Calgary and PAREXEL International
• Keywords: Latent means test ; maximum likelihood estimation ; Monte Carlo approximation ; multivariate Gaussian ; power ; polychoric and polyserial correlations

Abstract

The paper introduces a Gaussian copula-based methodology for calculating sample size in clinical trials with multiple mixed binary and continuous co-primary endpoints. Our Gaussian copula joint model permits adoption of flexible non-Gaussian marginal distributions for mixed endpoints, and includes the conditional grouped continuous model (CGCM) as a special case. The proposed method employs a latent variable description of binary endpoints and tests latent mean differences in the binary proportions. Tests based on latent means result in a simple and streamlined methodology akin to that for multiple continuous co-primary endpoints studied in Sozu et al. (2011); Wu and de Leon (2013) showed them to be more powerful than Sozu et al.'s (2012) approach. Our simulations show that standard asymptotic calculations of sample size are not reliable in practice when the continuous endpoints and the latent variables underlying the binary endpoints are non-Gaussian, as they may lead to considerably smaller or bigger samples than necessary; a method based on Monte Carlo approximation is proposed. Numerical examples based on the PREMIER Study and the VALOR Trial are used to illustrate our method.