Multiple sclerosis may have a non-progressive symptomatology for decades; however, it is not clear whether the disease activity may abate completely. We identified a cohort of patients, resident in Gothenburg at the time of disease onset, between the years 1950-64 (n = 307). These geographical and temporal restrictions, along with favourable conditions for a 'spider' epidemiological study, were optimal for an unbiased selection; this 15-year incidence cohort was essentially followed prospectively for 37-59 years after onset. The shortest follow-up time for patients without primary or secondary progression was 45 years. For patients with an initial relapsing-remitting course and multiple sclerosis diagnosis according to the Poser criteria (n = 202), the probability of non-progressive disease after 40 years was 22% (standard error 3.0%), and after 50 years it was 14% (standard error 3.2%). For attack onset including patients with possible multiple sclerosis, the corresponding probabilities after 40 and 50 years were 35% (standard error 3.3%) and 28% (standard error 3.5%), respectively. At the last follow-up in 2009-10, when patients reached the average age of the Swedish population life expectancy, only 13 patients from the multiple sclerosis diagnosis cohort, according to the Poser criteria, remained alive and non-progressive. Their annualized attack frequency diminished with time from 0.29 to 0.015. These patients had been functioning well socially. Nine patients had an Expanded Disability Status Scale score of 0-2.5, and four patients had a score of 3 or 3.5, with deficits dating back to attacks decades ago. Eight patients participated in a complete neuropsychological examination, which showed a slight difference (P < 0.01) concerning verbal memory and executive function compared to an age and socially matched reference group, whereas results for five other cognitive domains were within the normal range. Magnetic resonance images fulfilled the Barkhof-Tintoré criteria for multiple sclerosis in 10 of 11 patients, with conspicuously few subcortical lesions relative to extensive periventricular lesions and lesions extending from the inferior midline aspect of the corpus callosum. Prediction of the non-progressive stage was possible with moderate hazard ratios and low sensitivity. Early features that predicted a non-progressive course were complete remission of the onset attack, low or moderate initial relapse frequency and-when the patients with possible multiple sclerosis were included-dominating afferent symptoms. The clinical disease activity had abated in these 13 patients, with the caveat that transition to secondary progression continued to occur after four decades, albeit with decreasing risk.