Probabilistic Clustering Using Multivariate Growth Mixture Model in Clinical Settings-A Scleroderma Example.

Area of Science:

• Immunology
• Rheumatology
• Pulmonology

Background:

• Scleroderma (systemic sclerosis; SSc) is a heterogeneous autoimmune disease with variable progression across organ systems.
• Accurate patient stratification is crucial for guiding clinical care and managing SSc.
• Understanding disease trajectory aids in predicting outcomes and tailoring treatments.

Purpose of the Study:

• To classify patients with SSc into clinically meaningful subpopulations.
• To develop a real-time classification framework based on baseline characteristics and disease progression patterns.
• To guide clinical care by identifying patients at risk for rapid disease progression.

Main Methods:

• A Bayesian multivariate growth mixture model was employed to analyze lung function trajectories.
• Forced vital capacity (FVC) and diffusing capacity for carbon monoxide (DLCO) were jointly modeled in 289 SSc patients.
• A framework was developed to sequentially update patient subgroup probabilities using longitudinal data.

Main Results:

• Two distinct patient subgroups were identified: a
• stable
• group (n=150) with minimal lung function change over 10 years.
• A
• progressor
• group (n=139) exhibited significant decline in FVC and DLCO soon after disease onset.
• The algorithm calculates the probability of belonging to the progressor group using baseline data and longitudinal FVC/DLCO measurements.

Conclusions:

• The developed method enables baseline probability calculation for rapid progression, updated sequentially with accumulating patient data.
• This approach facilitates early identification of patients likely to experience rapid disease decline.
• Sequential data integration and classification hold potential for improving clinical decision-making and patient outcomes in SSc.