Disease progression and treatment efficacy vary among individuals with multiple sclerosis. Reliable predictors of individual disease outcomes are lacking.

To examine the accuracy of the early prediction of 12-year disability outcomes using clinical and magnetic resonance imaging (MRI) parameters.

A total of 177 patients from the original Avonex-Steroids-Azathioprine study were included. Participants underwent 3-month clinical follow-ups. Cox models were used to model the associations between clinical and MRI markers at baseline or after 12 months with sustained disability progression (SDP) over the 12-year observation period.

At baseline, T2 lesion number, T1 and T2 lesion volumes, corpus callosum (CC), and thalamic fraction were the best predictors of SDP (hazard ratio (HR) = 1.7–4.6; p <= 0.001–0.012). At 12 months, Expanded Disability Status Scale (EDSS) and its change, number of new or enlarging T2 lesions, and CC volume % change were the best predictors of SDP over the follow-up (HR = 1.7–3.5; p <= 0.001–0.017). A composite score was generated from a subset of the best predictors of SDP. Scores of >=4 had greater specificity (90%–100%) and were associated with greater cumulative risk of SDP (HR = 3.2–21.6; p < 0.001) compared to the individual predictors.

The combination of established MRI and clinical indices with MRI volumetric predictors improves the prediction of SDP over long-term follow-up and may provide valuable information for therapeutic decisions.