Multilevel random effect and marginal models for longitudinal data

Abstract

In many clinical trials, in order to characterize the safety profile of a subject with a given treatment, multiple measurements are taken over time. Mostly, measurements taken from the same subject are not independent. Thus, in cases where the dependent variable is categorical, the use of logistic regression models assuming independence between observations taken from the same subject is not appropriate. In this paper, marginal and random effect models that take the correlation among measurements of the same subject into account were fitted and extensions on the existing models also proposed. The models were applied to data obtained from a phase-III clinical trial on a new meningococcal vaccine. The goal is to investigate whether children injected by the candidate vaccine have a lower or higher risk for the occurrence of specific adverse events than children injected with licensed vaccine, and if so, to quantify the difference. Moreover, in the paper, extensions for the random intercept partial proportional odds model and generalized ordered logit model which assumes identical variability for different category levels were extended by introducing category specific random terms. This is very appealing to study the association between different category levels. Instead of using the classical logistic regression, Generalized Estimating Equations (GEEs) and random effect models are appropriate when measurements taken from the same subject are not independent. The result reveals that, in both marginal and random effects model, significant difference between the two vaccines were found for pain and redness adverse event.

Keywords: Generalized estimating equations, generalized linear mixed models, generalized ordered logit models, meningococcal vaccine, partial proportional odds models