Autoimmune comorbidities on cognitive decline in multiple sclerosis

Background: Cognitive impairment is a frequent and clinically important feature of multiple sclerosis (MS), affecting information processing speed and memory and contributing to functional limitations. Autoimmune comorbidities occur in a meaningful subset of people with MS and plausibly influence cognitive trajectories through systemic inflammation, endocrine dysfunction (notably thyroid disease), symptom burden (fatigue, sleep disturbance, depression), and therapy-associated secondary autoimmunity.Objective: To present a complete, publication-style analytic workflow evaluating associations between autoimmune comorbidity and longitudinal cognitive change in MS using a cohort of 182 participants and established cognitive measurement approaches.Methods: A dataset of 182 adults with MS phenotype categories (RRMS/SPMS/PPMS), disability (EDSS), demographics, mood (BDI-II), fatigue (MFIS), sleep quality (PSQI), disease-modifying therapy (DMT) category, and autoimmune comorbidities (autoimmune thyroid disease, psoriasis, inflammatory bowel disease, type 1 diabetes, celiac disease, uveitis) was generated. The primary cognitive outcome was Symbol Digit Modalities Test (SDMT) at baseline, 12 months, and 24 months. Linear mixed-effects models with random intercepts tested autoimmune status-by-time interactions adjusting for age, education, EDSS, mood, fatigue, sex, and MS subtype.Results: In the cohort, 24/182 (13.2%) had ≥1 autoimmune comorbidity, with psoriasis (8.2%) and thyroid disease (6.0%) most common. Baseline SDMT was ~52 points. Time was associated with SDMT decline (~0.92 points/year). The autoimmune-by-time interaction suggested additional decline among those with autoimmune comorbidity but did not reach statistical significance in this dataset.Conclusions: This analysis illustrates an end-to-end approach to evaluate whether autoimmune comorbidities modify cognitive decline in MS using SDMT/BICAMS-consistent outcomes, standardized exposure definitions, and mixed-effects modeling with key confounders. The manuscript and data schema can be adapted to real-world cohorts for genuine inference.