Data-driven hypothesis discovery from disease trajectories in multiple sclerosis

Abstract

Introduction Multiple sclerosis (MS) is an incurable autoimmune disease marked by heterogeneous progression and a lack of reliable biomarkers, complicating prognosis and individualized care. This study introduces a novel trajectory-based statistical approach designed to identify patterns in patient histories within MS populations. Methods Using longitudinal clinical data from a real-world cohort of 1,025 MS patients (median follow-up: 6.75 years), two complementary analyses were conducted based on patient trajectory analysis. In the first analysis, the technique is applied to the complete dataset after removal of missing values (n = 985; 11,048 events) to uncover latent progressive trajectories. The second analysis evaluated the techniques’ performance on a smaller, limited-sample cohort (n = 83; 282 events). Results Across both analyses, the approach revealed previously unrecognized progression patterns, giving rise to new hypotheses, including an effect of Alemtuzumab on the bowel/bladder function (p < 0.01, RR = 2.83) and glatiramer acetate on the occurrence of relapses (p < 0.01, RR = 1.49). Known associations were also confirmed, such as the relationship between relapse activity and brain lesions (p < 0.01, RR = 1.20). Discussion The results demonstrate the method’s robustness across varying dataset sizes, highlight its methodological limitations, and show its potential to uncover previously unseen relationships among MS-specific diagnostic events. These findings provide a foundation for generating novel hypotheses relevant to biomarker discovery and therapeutic optimization.