<p>Amyotrophic lateral sclerosis (ALS) is a multifactorial neurodegenerative disease whose incidence increases with age. According to the gene–time–environment hypothesis, ALS onset occurs through the interaction between genes and environmental exposures during ageing, which may involve a continuous accumulation process. Alternatively, the multistep pathogenic hypothesis, based on the Armitage-Doll multistep model from cancer research, posits that a discrete number of specific sequential “hits” are necessary to trigger ALS. Here we analyzed three large population-based epidemiological datasets of ALS to formally test whether the ALS age-incidence curve is better described by a power law, as predicted by the Armitage-Doll model, or by an exponential function, which is generally associated to continuous accumulation of damage and is incompatible with the Armitage-Doll model. We obtained moderate-to-extreme Bayesian evidence in favor of the exponential function compared to the power law. Cancer data were instead better aligned, as expected, with the power law. These results suggest that the multistep pathogenesis hypothesis based on the Armitage-Doll model cannot be extended from cancer to ALS, because it is incompatible with the epidemiological data. This calls for a re-consideration of the current understanding of ALS pathogenesis. Our work also warns against extending the Armitage-Doll multistep model from cancer to other aging-related diseases solely based on age-incidence curves.</p>