Describing complex disease progression using joint latent class models for multivariate longitudinal markers and...

Area of Science:

• Neuroscience
• Biostatistics
• Medical Statistics

Background:

• Neurodegenerative diseases present complex, multi-dimensional progression with various markers and clinical endpoints.
• Accurately describing disease progression is crucial for understanding natural history, staging patients, and predicting prognosis.
• Multiple system atrophy (MSA), a rare synucleinopathy, exemplifies these challenges due to its heterogeneous presentation and poor prognosis.

Purpose of the Study:

• To develop and validate a statistical approach for modeling complex disease progression and identifying patient subphenotypes.
• To apply this method to the specific case of multiple system atrophy (MSA) progression.
• To explore new pathological hypotheses by uncovering distinct disease trajectories.

Main Methods:

• A joint latent class modeling approach was employed, integrating multivariate mixed models for repeated markers and proportional hazard models for time-to-event data.
• The model accounts for multivariate repeated biomarkers, potential latent dimensions, and class-and-cause-specific risks.
• Maximum likelihood estimation was used, with the methodology available in the lcmm R package and validated through simulations.

Main Results:

• Five distinct subphenotypes of MSA were identified within a French cohort of 598 patients followed for up to 13 years.
• These subphenotypes differed significantly in the pattern and rate of biomarker degradation and their associated risk of mortality.
• Subphenotype membership was used to explore associations with external imaging and fluid biomarkers, accounting for membership uncertainty.

Conclusions:

• Latent class modeling provides a powerful tool for describing complex disease progression and identifying clinically relevant subphenotypes in neurodegenerative disorders.
• The identified MSA subphenotypes offer a refined understanding of disease heterogeneity and can guide future research into specific pathological mechanisms and therapeutic targets.
• This approach enhances the ability to predict prognosis and potentially personalize patient management strategies based on distinct disease trajectories.