Multidimensional latent trait linear mixed model: an application in clinical studies with multivariate longitudinal...

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Psychometrics

Background:

• Multilevel item response theory (MLIRT) models are common for longitudinal clinical data.
• These models often assume a single underlying latent trait, which may not capture disease heterogeneity.

Purpose of the Study:

• To propose a multidimensional latent trait linear mixed model (MLTLMM) that accommodates multiple latent variables.
• To address the limitations of the unidimensional assumption in analyzing complex diseases.

Main Methods:

• Developed a multidimensional latent trait linear mixed model (MLTLMM).
• Conducted extensive simulation studies comparing MLTLMM with unidimensional MLIRT models.
• Applied the MLTLMM to amyotrophic lateral sclerosis (ALS) clinical trial data.

Main Results:

• Simulation studies indicated MLTLMM outperforms unidimensional models for data with multiple latent variables.
• The proposed model demonstrated effectiveness in analyzing heterogeneous disease progression.

Conclusions:

• The MLTLMM offers a more flexible and accurate approach for analyzing multivariate longitudinal data in heterogeneous diseases.
• This model advances the analysis of complex conditions like ALS.