Evaluation of Statistical Methods for Longitudinal Count Data with Dropouts

• Activity Number: 186
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Takayuki Abe*+ and Kazuhito Shiosakai and Yuji Sato and Manabu Iwasaki
• Companies: Keio University School of Medicine and Daiichi Sankyo Co. and Keio University School of Medicine and Seikei University
• Keywords: longitudinal clinical trials ; count data ; multiple imputation ; missing data ; GEE ; GLMM

Abstract

In clinical trials, handling missing data correctly is critical to draw unbiased inferences on treatment effect. We focused on clinical trials where longitudinal count data (e.g. number of episodes) are compared between two treatment groups (e.g. new treatment and control), in order to identify better analysis methods for incomplete count data. Statistical methods evaluated are complete-case analysis, last observation carried forward (LOCF), multiple imputation (MI) with some imputation models for count data and also with a calibration method, marginal models: generalized estimating equations (GEE), weighted GEE (WGEE), conditional models: generalized linear mixed-effects models (GLMM), and their combinations (e.g. MI + GEE). Integration techniques with respect to random-effects in the GLMM were also evaluated. An example motivated by an actual clinical trial was used in the evaluations. Simulation studies were performed under various settings in terms of missing mechanisms, sample sizes and degrees of missing information. The details of results are shown at the presentation. This work was supported by JSPS KAKENHI Grant Number 25240005.