External validation of a clinical prediction model in multiple sclerosis


Moradi N., Sharmin S., Malpas C. B., Shaygannejad V., Terzi M., BOZ C., ...More

Multiple Sclerosis Journal, vol.29, no.2, pp.261-269, 2023 (SCI-Expanded) identifier identifier identifier

  • Publication Type: Article / Article
  • Volume: 29 Issue: 2
  • Publication Date: 2023
  • Doi Number: 10.1177/13524585221136036
  • Journal Name: Multiple Sclerosis Journal
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, BIOSIS, EMBASE, MEDLINE
  • Page Numbers: pp.261-269
  • Keywords: Outcome prediction, individual therapy response, external validation, multiple sclerosis, TREATMENT RESPONSE, INTERFERON-BETA, THERAPY, REVISIONS
  • Karadeniz Technical University Affiliated: Yes

Abstract

© The Author(s), 2022.Background: Timely initiation of disease modifying therapy is crucial for managing multiple sclerosis (MS). Objective: We aimed to validate a previously published predictive model of individual treatment response using a non-overlapping cohort from the Middle East. Methods: We interrogated the MSBase registry for patients who were not included in the initial model development. These patients had relapsing MS or clinically isolated syndrome, a recorded date of disease onset, disability and dates of disease modifying therapy, with sufficient follow-up pre- and post-baseline. Baseline was the visit at which a new disease modifying therapy was initiated, and which served as the start of the predicted period. The original models were used to translate clinical information into three principal components and to predict probability of relapses, disability worsening or improvement, conversion to secondary progressive MS and treatment discontinuation as well as changes in the area under disability-time curve (ΔAUC). Prediction accuracy was assessed using the criteria published previously. Results: The models performed well for predicting the risk of disability worsening and improvement (accuracy: 81%–96%) and performed moderately well for predicting the risk of relapses (accuracy: 73%–91%). The predictions for ΔAUC and risk of treatment discontinuation were suboptimal (accuracy < 44%). Accuracy for predicting the risk of conversion to secondary progressive MS ranged from 50% to 98%. Conclusion: The previously published models are generalisable to patients with a broad range of baseline characteristics in different geographic regions.