Abstract
Severe weight loss in amyotrophic lateral sclerosis (ALS) is common, multifactorial, and associated with shortened survival. Using longitudinal weight data from over 6000 patients with ALS across three cohorts, we built an accelerated failure time model to predict the risk of future severe (≥ 10%) weight loss using five single-timepoint clinical predictors: symptom duration, revised ALS Functional Rating Scale, site of onset, forced vital capacity, and age. Model performance and generalisability were evaluated using internal-external cross-validation and random-effects meta-analysis. The overall concordance statistic was 0.71 (95% CI 0.63-0.79), and the calibration slope and intercept were 0.91 (0.69-1.13) and 0.05 (-0.11-0.21). This study highlights clinical factors most associated with severe weight loss in ALS and provides the basis for a stratification tool.