We previously compared the diagnostic capabilities of MRI of the head with CT, evoked potentials, and CSF oligoclonal banding (OB) analysis in a prospective evaluation of 200 patients with suspected multiple sclerosis (MS). To examine the ability of MRI and other paraclinical tests to predict the diagnosis of clinically definite MS (CDMS), we did a systematic clinical follow-up of 200 patients who were previously reported. In that study, 85 of 200 could be diagnosed as having laboratory-supported definite MS (LSDMS). In follow-up, we excluded one patient diagnosed as LSDMS who in retrospect was considered to have had CDMS at entry and 15 patients who were eventually diagnosed as having other diseases. After a mean follow-up of 2.1 years, 55 of the remaining 184 patients (30%) had developed CDMS. Thirty-eight of 84 patients with an original diagnosis of LSDMS (45%) and 17 of the remaining 100 patients with suspected MS (17%) had become CDMS. Forty-six of the 55 patients who developed CDMS in follow-up (84%) had an initial MRI that was strongly suggestive of MS. Fifty-two of those 55 CDMS patients (95%) had at least one MS-like abnormality on MRI when originally studied. In contrast, 38 of 55 (69%) had CSF OB, 38 of 55 (69%) had an abnormal VEP, 35 of 55 (64%) had an abnormal SEP, and 21 of 55 (38%) had an abnormal CT when first studied. MRI was the most sensitive single paraclinical test for predicting CDMS. CDMS developed during follow-up in 46 of the 94 patients (49%) whose initial MRI was strongly suggestive of MS.(ABSTRACT TRUNCATED AT 250 WORDS)