Clinically isolated syndromes (CIS), such as optic neuritis, brainstem or spinal cord syndromes are frequently the first clinical presentations of multiple sclerosis. However, not all CIS patients develop multiple sclerosis and in those who do, disability is highly variable. In previous follow-up studies, brain lesions on T2-weighted MRI are associated with increased risk of multiple sclerosis and to an extent disability. We evaluated the longitudinal relationships between the MRI lesions and clinical course over a period of 20 years. CIS patients were recruited between 1984 and 1987 and previously followed up after 1, 5, 10 and 14 years. Of the 140 subjects who were initially recruited with a CIS for a baseline MRI study, we followed up 107 patients after a mean of 20.2 years (range 18-27.7). Multiple sclerosis was diagnosed as clinically definite on clinical grounds only and disability determined using the Expanded Disability Status Scale (EDSS) and Multiple Sclerosis Functional Composite (MSFC) score. Clinically definite multiple sclerosis developed in 67 out of 107 (63%) overall: 60 out of 73 (82%) with abnormal and 7 out of 34 (21%) with normal baseline MRI. Multiple sclerosis was still relapsing-remitting in 39 (58%)--including 26 (39%) with a 'benign' course (EDSS < or = 3)--whilst 28 (42%) had developed secondary progression. T2 lesion volume at all time-points correlated moderately with 20-year EDSS (r(s) values 0.48 to 0.67; P < 0.001) and MSFC z-score [r(s) values (-0.50) to (-0.61); P < 0.001]. In those developing multiple sclerosis, a concurrent correlation of change in T2 lesion volume with change in EDSS was most evident in years 0-5 (r(s) = 0.69, P < 0.001). The estimated rate of lesion growth over 20 years was 0.80 cm3/year in those who retained a relapsing remitting multiple sclerosis course, and 2.89 cm3/year in those who developed secondary progressive multiple sclerosis, a difference of 2.09 cm3/year (95% CI: 0.77, 2.96; P < 0.001). This study extends previous follow-up of CIS patients and sheds new light on how the lesions evolve according to the natural history. Baseline MRI findings are predictive for development of clinically definite multiple sclerosis. Lesion volume and its change at earlier time points are correlated with disability after 20 years. Lesion volume increases for at least 20 years in relapse-onset multiple sclerosis and the rate of lesion growth is three times higher in those who develop secondary progressive than in those who remain relapsing remitting multiple sclerosis.