A case-control study of epilepsy in multiple sclerosis (MS) is presented using magnetic resonance (MR) imaging to semiquantitatively assess cortical-subcortical lesion load. In this sample of 13 pairs of cases with MS and epilepsy and controls with MS without epilepsy we found statistically higher cumulated cortical-subcortical lesion loads in the cases than in the controls (Wilcoxon, P = 0.036). Total lesion loads (cortical-subcortical plus deep white matter loads) did not differ significantly (P > 0.1) between cases and controls. The relative risk for seizures as determined by the odds ratio of a cortical-subcortical lesion load of > or = 20 was 8.8 (chi 2 = 5.23, P 0.025), the odds ratio of a large (> 1 cm) cortico-subcortical lesion was 4.7 (chi 2 = 4.9, P < 0.05), while the 2 MR criteria combined show an odds ratio of 19.2 (chi 2 = 8.0, P < 0.005). We conclude that: first, the presence of cortical-subcortical lesions in part accounts for the occurrence of seizures in MS patients; second, due to the substantial overlap of MR imaging scores between cases and controls the ultimate use of these MR imaging findings in the management of individual patients or in the organizations of trials should depend on the expected benefit of the treatment. If the benefit is only moderate or not known a cautious approach with exclusion of cases showing a substantial cortical-subcortical lesion load on MR imaging seems appropriate in trials with drugs, like 4-aminopyridine, that lower the epileptic threshold.