Latent Class Analysis of Multivariate Longitudinal Data: Uncovering Response Patterns in a Stem Cell Study

• Activity Number: 424
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Qianqiu Li*+ and Xiaozhen Wang and Ian Harris and Bill Pikounis
• Companies: Johnson & Johnson Pharmaceutical R&D, LLC and Johnson & Johnson Pharmaceutical R&D, LLC and Johnson & Johnson Pharmaceutical R&D, LLC and Johnson & Johnson Pharmaceutical R&D, LLC
• Keywords: Latent class ; Mixture model ; Longitudinal data ; Small sample ; Response Pattern ; Stem Cell

Abstract

By applying latent class analysis to longitudinal data from a stem cell hCTC study with 3 time points and 5 treatment groups, we distinguished 3- and 4- subgroup mean response profiles and revealed their association with treatment and time. These results indicate efficacy of two dosed groups over 12 weeks, compared with only short-term (or 4-week) improvement for a comparator treatment group, and no improvement in the Vehicle group. In comparison with ANOVA and mixed model analyses, latent class analysis can be a powerful tool for inference on the mean response profile as well as posterior probabilities for membership of response subgroups. A maximum likelihood approach via the EM algorithm is used. The results demonstrate that the latent model is particularly effective in analysis of small sample data from animal models in presence of nonnegligible inter-individual variability.