No clear metrics for sensitive and reliable identification of the transition from relapsing-remitting multiple sclerosis (RRMS) to secondary progressive (SP)MS are available.
To compare diagnostic performances of two different data-driven Secondary Progressive Multiple Sclerosis definitions.
patient with RRMS with a follow-up ≥5 years, with a current age ≥18 years, and with ≥3 EDSS scores recorded were selected from the Italian MS Registry. Annual incidence of SPMS conversion was reported as number of patients converting to SP every 100 patients/year. Three different SPMS definitions have been used. Data-driven definitions based on the Lorscheider’s algorithm (LA) and on the EXPAND trial inclusion criteria were validated, using the neurologist’s definition as gold standard, in terms of calibration, discrimination and goodness of fit by calculating: sensitivity, specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), the Akaike information criterion (AIC), the Area Under the Curve (AUC). The overall calibration of the data-driven definitions was evaluated by the Calibration Slope test.
a cohort of 10,240 RRMS patients was extracted from the Italian MS Registry. According to the neurologist judgment, 880 (8.59%) patients were classified as SPMS in the dataset. By applying the LA and the EXPAND definition, 1,806 (17.64%) and 1,134 (11.07%) patients, respectively, were classified as SPMS. The annual rate of SP conversion during the follow-up was 0.74 every 100 patients/year based on the neurologist’s definition, 1.57 every 100 patients/year using the LA and 0.94 every 100 patients/year applying the EXPAND definition. Both the data-driven definitions were well calibrated, with a p-value of the Calibration Slope test higher than 0.05 (LA=0.55; EXPAND definition=0.57). The AIC (LA=4301; EXPAND definition=5510) and the R-Square (LA=0.15 vs EXPAND definition=0.05), were in favor of the LA. The LA showed a greater discrimination power (AUC: 0.83 vs 0.65) and a higher sensitivity (77.1% vs 38.0%) in comparison to the EXPAND definition. Both definitions showed similar specificity (88.0% vs 91.5%). The PPV and the NPV were both higher using the LA than those obtained by the EXPAND definition (37.5% vs 29.5%; 97.6% vs 94.0%, respectively).
An accurate definition of SP transition is needed for a timely and efficacious treatment of SPMS patients. Real-world data from the Italian MS Registry suggests that data-driven definitions had a greater ability to capture SP transition than neurologist’s definition and that the global accuracy of LA seems to be higher than a definition based on the EXPAND trial inclusion criteria.