Predicting survival of patients with amyotrophic lateral sclerosis at presentation: A 15-year experience

Objective: To describe the clinical features at first evaluation that best predict survival of the amyotrophic lateral sclerosis (ALS) population from the Salpêtrière Hospital between 1995 and 2009. Methods: Data are collected and entered into a clinical database from all patients seen at the Paris ALS Center. Variables analyzed were demographic and baseline information, strength testing (manual muscle testing; 1995-2009), the revised ALS Functional Rating Scale (ALSFRS-R; 2002-2009) and survival status. The χ² test and ANOVA assessed differences in variables by region and across time period. Univariate and multivariate Cox proportional hazards models determined which variables best predicted survival. Flexible modeling of continuous predictors (splines) assessed trends in survival for different variables. Results: 3,885 patients with ALS were seen in 1995-2009, of whom 2,037 had ALSFRS-R scores. Age, weight, strength, and site of onset varied by region of residence. The proportion of patients living outside Paris, the time to first visit, patient age, and motor function differed across time periods. In Cox models, site of onset, time to first visit greater than 18 months, strength and the year of visit after 2006 predicted survival (all p values <0.0001). Compared to patients first seen between 1999 and 2002, the hazard ratio of death was 1.04 (95% CI = 0.95-1.14) for 2003-2006, and 0.76 (95% CI = 0.66-0.87) after 2006, while adjusting for other predictors of survival. The use of noninvasive ventilation increased during 2004-2008 from 16 to 51% of patients. Conclusions: Older age, bulbar onset, shorter delay to first visit and poor motor function at first visit predicted shorter survival rates in this large center-based sample from France, showing marked consistency across time and region of residence. Survival improved after 2006, concurrent with increasing rates of noninvasive ventilation use. Clinicopathologic correlation could better define subgroups, but identification of etiologies may be needed to elucidate individual forms of ALS with unique survival patterns.