Neuroimaging predictors of longitudinal disability and cognition outcomes in multiple sclerosis patients: A systematic review and meta-analysis

Background: Cross-sectional magnetic resonance imaging (MRI) studies have generated substantial evidence relating neuroimaging abnormalities to clinical and cognitive decline in multiple sclerosis (MS). Longitudinal neuroimaging studies may have additional value for predicting future cognitive deficits or clinical impairment, potentially leading to earlier interventions and better disease management. We conducted a meta-analysis of longitudinal studies using neuroimaging to predict cognitive decline (i.e. the Symbol Digits Modalities Test, SDMT) and disability outcomes (i.e. the Expanded Disability Status Scale, EDSS) in MS. Methods: Our systematic literature search yielded 64 relevant publications encompassing 105 distinct sub-analyses. We performed a multilevel random-effects meta-analysis to estimate overall effect size for neuroimaging' s ability to predict longitudinal cognitive and clinical decline, and a meta-regression to investigate the impact of distinct study factors on pooled effect size. Results: In the EDSS analyses, the meta-analysis yielded a medium overall pooled effect size (Pearson's correlation coefficient r = 0.42, 95% CI [0.37; 0.46]). The meta-regression further indicated that analyses exclusively evaluating gray matter tissue had significantly stronger effect sizes than analyses of white matter tissue or whole brain analyses (p < 0.05). No other study factors significantly influenced the pooled effect size (all p > 0.05). In the SDMT analyses, the meta-analysis yielded a medium overall pooled effect size (r = 0.47, 95% CI [0.32; 0.60]). The meta-regression found no significant study factors influencing the pooled effect size. Conclusion: The present findings indicate that brain imaging is a medium predictor of longitudinal change in both disability progression (EDSS) and cognitive decline (SDMT). These findings reinforce the need for further longitudinal studies standardizing methods, using multimodal approaches, creating data consortiums, and publishing more complete datasets investigating MRI modalities to predict longitudinal disability and cognitive decline.