External validation of a multiple sclerosis treatment decision score using data from the ProVal-MS cohort study.

Area of Science:

• Neurology
• Medical Informatics
• Biostatistics

Background:

• Predicting the course of relapsing-remitting multiple sclerosis (RRMS) and clinically isolated syndrome (CIS) is challenging due to disease variability.
• Prognostic algorithms are crucial for guiding treatment decisions in early MS, yet few have undergone external validation.
• External validation using independent data is essential for assessing the reliability of predictive models.

Purpose of the Study:

• To externally validate the multiple sclerosis treatment decision score (MS-TDS) using data from the ProVal-MS study.
• To assess the MS-TDS's ability to predict new or enlarging T2 lesions in patients with CIS or early RRMS.
• To demonstrate the feasibility of privacy-preserving federated analysis for multi-center data integration.

Main Methods:

• Prospective, multicentric, non-interventional cohort study (ProVal-MS) within the DIFUTURE consortium.
• External validation of the MS-TDS using area under the receiver operating characteristic curve (AUROC) and Brier score.
• Decision curve analysis (DCA) to compare MS-TDS guided decisions with those of treating neurologists.

Main Results:

• The study included 271 patients with CIS or early RRMS; 202 received platform treatment.
• The MS-TDS showed low predictive performance with AUROCs of 0.561 (pooled) and 0.567 (distributed).
• Decision curve analysis indicated a net benefit for MS-TDS comparable to experienced neurologists.

Conclusions:

• External validation revealed limited predictive accuracy for the MS-TDS in CIS/early RRMS.
• The MS-TDS may serve as a supplementary tool, especially for less experienced neurologists.
• Privacy-preserving federated analysis is a feasible and compliant method for external validation of predictive models.