ABSTRACT
This chapter presents statistical methods for analyzing longitudinal data obtained from patient follow-ups in clinical trials over several time points. Generalized estimating equations (GEE) and generalized linear mixed models (GLMM) are introduced as statistical methods that account for dependence in repeatedly measured outcomes in longitudinal data. For clinical trials, score type tests can be derived under randomization, and these are presented for GEE and for GLMM under randomization. Since longitudinal data are more prone to missing data compared to studies that involve data from a single time point, the final section describes how missing data can be handled in GEE and in GLMM. A case study is presented based on a randomized double-blind trial in a study of isolated systolic hypertension in the elderly. This longitudinal study investigates the difference in treatment effect, illustrated using GEE and GLMM. Various working correlations are considered in GEE models and inverse probability weighted GEE models, and GLMM is also fitted and compared.
