A latent class mixed model for analysing biomarker trajectories with irregularly scheduled observations

This paper considers a latent class model to uncover subpopulation structur e for both biomarker trajectories and the probability of disease outcome in highly unbalanced longitudinal data. A specific pattern of trajectories ca n be viewed as a latent class in a finite mixture where membership in laten t classes is modelled with a polychotomous logistic regression. The biomark er trajectories within a latent class are described by a linear mixed model with possibly time-dependent covariates and the probabilities of disease o utcome are estimated via a class specific model. Thus the method characteri zes biomarker trajectory patterns to unveil the relationship between trajec tories and outcomes of disease. The coefficients for the model are estimate d via a generalized Ehl (GEM) algorithm, a natural tool to use when latent classes and random coefficients are present. Standard errors of the coeffic ients are calculated using a parametric bootstrap. The model fitting proced ure is illustrated with data from the Nutritional Prevention of Cancer tria ls; we use prostate specific antigen (PSA) as the biomarker for prostate ca ncer and the goal is to examine trajectories of PSA serial readings in indi vidual subjects in connection with incidence of prostate cancer. Copyright (C) 2000 John Wiley & Sons, Ltd.