RETRACTED ARTICLE: A Skew-Normal Bayesian Semi-parametric Latent Trait Linear Mixed Effect Model

Clinical trials have longitudinally collected the biomarkers that may be associated with the time-to-event endpoints via latent variables such as disease severity. These longitudinal data may consist of different types of measurements. The multilevel item response theory (MLIRT) model is widely used in several fields including public health and health sciences, for these longitudinal outcomes. However, the violation of the normality assumption may produce inaccurate inferences if skewness is significantly present for continuous outcomes. Furthermore, the trajectories of these biomarkers over time are often observed in a nonlinear manner. This implies that partial linear regression may be more appropriate in practice. Challenges remain in interpreting such complicated and long-term survival data due to data attributes, including a mix of longitudinal outcomes, measurement errors, and skewness. Ignoring these characteristics in the data could lead to biased conclusions. In this article, we relax the assumptions and extend the MLIRT model to accommodate mixed types of multivariate longitudinal data. This article presents a novel method for analyzing longitudinal clinical data from the ACCORD trial. The methods are evaluated through simulation studies.