Neurobehavioral phenotypes in multiple sclerosis: a data-driven classification using latent profile analysis

Cognitive impairment in multiple sclerosis (MS) is highly heterogeneous. Emerging evidence suggests that patients may cluster into distinct cognitive and affective phenotypes. Characterizing these phenotypes using clinically feasible tools is essential for improving stratification, prognostication, and the development of targeted monitoring and intervention strategies in routine MS care.

To identify neurobehavioral phenotypes in people with MS using a clinically feasible assessment battery and to evaluate the classification accuracy of individual measures for phenotype classification.

In this single-center cross-sectional study, 837 patients underwent clinical and neuropsychological assessments. Cognitive performance was evaluated using the Brief International Cognitive Assessment for Multiple Sclerosis (BICAMS). Depressive symptoms and levels of fatigue were also assessed. Latent profile analysis was performed. Post hoc analyses examined demographic and clinical differences between phenotypes. Receiver operating characteristic analyses assessed the ability of individual measures to discriminate among phenotypes.

The identified phenotypes were: Neuropsychologically Preserved (NP; 58.0%), mild Global Cognitive Dysfunction-Visuospatial Memory Predominant (mGCD-Vsp; 24.1%), significant Global Cognitive Dysfunction-Depression (sGCD-Depression; 5.1%), and Isolated Depression (I-Depression; 12.8%). More impaired phenotypes were associated with older age, longer disease duration, greater disability, progressive disease courses, and more frequent use of high-efficacy disease-modifying therapies. Visuospatial memory test demonstrated the highest discriminative accuracy for the mGCD-Vsp phenotype, whereas depression scores most accurately identified the sGCD-Depression and I-Depression phenotypes.

Distinct neurobehavioral phenotypes were identified in MS using brief cognitive and affective assessments routinely employed in clinical practice. The provision of classification accuracy metrics may facilitate clinical implementation and personalized monitoring and rehabilitation.