# Multimodal predictors of disability progression and processing speed decline in relapsing–remitting multiple sclerosis

## TL;DR

This study uses a combination of data types to predict disability and cognitive decline in multiple sclerosis patients over 12 years.

## Contribution

The study introduces a multiverse approach to identify baseline predictors of disability and processing speed decline in multiple sclerosis.

## Key findings

- Baseline variables like treatment, EDSS, and vitamin levels predict disability progression.
- Disease-modifying treatment and baseline PASAT scores predict stable processing speed.
- Analytic flexibility significantly influences model variability and outcomes.

## Abstract

The underlying mechanisms for neurodegeneration in multiple sclerosis are complex and incompletely understood. Multivariate and multimodal investigations integrating demographic, clinical, multi-omics, and neuroimaging data provide opportunities for nuanced analyses, aimed to define disease progression markers. We used data from a 12-year longitudinal multicenter cohort of 88 people with multiple sclerosis, to test the predictive value of multi-omics, T1-weighted MRI (lesion count and volume, lesion-filled brain-predicted age), clinical examinations, self-reports on quality of life, demographics, and general health-related variables for future functional and cognitive disability. Systematic increases in Expanded Disability Status Scale (EDSS) scores were used to stratify a progressive disability group (PDG) from relatively stabile disability. A processing speed decline group (PSDG) was defined by a ≥ 20% decrease of Paced Auditory Serial Addition Test score from previous timepoints. We used a multiverse approach to identify which baseline variables were most predictive for PDG and PSDG memberships, considering multiple analysis paths. Future disability (median area under the curve: mAUC = 0.83 ± 0.04, median Brier score: mBS = 0.16 ± 0.02) and the loss of processing speed (mAUC = 0.89 ± 0.05, mBS = 0.10 ± 0.03) could be successfully classified across models. Varibles significantly (median p-values < 0.05) predicting stable disability included receiving disease modifying treatment at 12-year follow-up (median Odds Ratio: mORPDG = 7.44 ± 4.07, pmedian = 0.013, proportion of the OR’s directionality: PORSD = 100%), lower baseline EDSS for each 1-unit (mORPDG = 0.25 ± 0.11, pmedian = 0.013, PORSD = 100%), and counter-intuitively every year increase in baseline age (mORPDG = 1.12 ± 0.04, pmedian = 0.020, PORSD = 100%), and lower vitamin A per 1 umol/L (mORPDG = 0.10 ± 0.05, pmedian = 0.016, PORSD = 99.7%) and D levels per 1 nmol/L (mORPDG = 0.95 ± 0.02, pmedian = 0.025, PORSD = 100%). Variables significantly predicting stable processing speed were receiving disease modifying treatment at 12-year follow-up (mORPSDG = 0.10 ± 0.08, pmedian = 0.013, PORSD = 100%) and baseline PASAT score (mORPSDG = 0.86 ± 0.03, pmedian = 0.005, PORSD = 99.73%). These findings were supported by an additional simulation study. Concordant with the literature, disease modifying treatments influence disability progression, as well as a higher EDSS and PASAT scores at measurement start. Experimental and counterintuitive findings on vitamin A and D levels require further validation. The large variability across models suggests a strong influence of analytic flexibility, such as the selection of covariates.

The online version contains supplementary material available at 10.1038/s41598-025-17894-2.