Rate and Shape of Functional Decline in Amyotrophic Lateral Sclerosis Identify Clinically Distinct Amyotrophic...

The ALS Functional Rating Scale-Revised decline rate compresses trajectory shape into a single number, potentially missing distinct phenotypes. We tested whether shape identifies phenotypes that rate does not distinguish, and whether post-hoc analysis recovers a hidden ultrarapid subgroup.

From the Piemonte and Valle d'Aosta register (2000-2022), we applied 2 models to 1,686 amyotrophic lateral sclerosis patients: a Gaussian mixture model (GMM; rate axis) and a latent-class mixed model (LCMM) with cubic splines (shape axis). Concordance, prognostic performance, calibration, and clinical utility were assessed at 24 months. External validation used the Emilia-Romagna register.

The rate model identified 4 classes (GMM-Slow 31%, GMM-Inter-slow 46%, GMM-Inter-fast 17%, GMM-Catastrophic 6%; slopes -0.44 to -3.52 points/month); the shape model 4 (LCMM-SlowLinear 46%, LCMM-PlateauFall 23%, LCMM-LateDecline 15%, LCMM-RapidEarly 16%). Concordance was modest (kappa = 0.273). Adding shape to rate changed discrimination negligibly (C-index 0.761 vs 0.755), but separated subgroups within rate classes (within GMM-Catastrophic, 30.1 vs 12.0 months). LCMM-LateDecline combined moderate early decline with reduced respiratory reserve (median forced vital capacity 83.5%) and 22.7-month median survival. The combined model gave the greatest net benefit for the composite outcome. Of 945 patients excluded for sparse sampling, 367 were ultrarapid progressors with shorter survival than LCMM-RapidEarly; findings replicated in the Emilia-Romagna register.

Rate and shape describe partially overlapping aspects of the disease. LCMM-LateDecline, with moderate early decline yet occult respiratory compromise, is clinically challenging, and the 3-visit criterion hides over half of ultrarapid patients. Rate predicts the overall burden of progression, whereas shape predicts the timing of deterioration. ANN NEUROL 2026.