Objectives To generate a prognostic classification method for Amyotrophic Lateral Sclerosis (ALS) from a prognostic model built using clinical variables from a population register.

Materials and Methods We carried out a retrospective multivariate analysis of 713 patients with ALS over a 20 year period from the South-East England Amyotrophic Lateral Sclerosis (SEALS) population register. Patients were randomly allocated to ‘discovery’ or ‘test’ cohorts. A prognostic score was calculated using the discovery cohort and then used to predict survival in the test cohort. This score was used as a predictor variable in subsequent survival analyses, either as a raw value for a Cox regression or split into four prognostic categories (good, moderate, average, poor).

Results A prognostic score generated from one cohort of patients predicted survival for a second cohort of patients (r²=0.72). Six variables were included in the survival model: age at onset, diagnostic delay, El Escorial category, use of riluzole, gender and site of onset. Cox regression demonstrated a strong relationship between these variables and survival (χ² 80.8, df 1, p<0.0001, n=343) in the test cohort. Kaplan-Meier analysis demonstrated a significant difference in survival between clinical categories (log rank 161.932, df 3, p<0.001).

Conclusion It is possible to correctly classify patients into prognostic categories using clinical data easily available at time of diagnosis.