Follow-up data from a prognostic study done at L'Hôpital Neurologique de Lyon have been reanalyzed using a new methodologic approach. A stochastic survival model which allows the course of the disease to be described through the movement of patients from one disease state to another is presented. This approach extends previous modelling work by enabling the risk of transition between pairs of states to be written as a function of any number of discrete or continuous prognostic variables. Length of the first remission, sex, and mode of onset (mono- vs. polysymptomatic) were found to be significant prognostic indicators.