In the recently proposed diagnostic criteria for multiple sclerosis (MS) by McDonald, the modified magnetic resonance imaging (MRI) Barkhof criteria have been incorporated. We examined the validity of this implementation in the Early Treatment of MS study, a randomized, double-blind, placebo-controlled study of 22 microg interferon beta1a given subcutaneously once weekly in 309 patients with a first episode consistent with demyelinating disease (and abnormal MRI). Conversion to clinically definite MS (CDMS) within 2 years of follow-up, as evidenced by a new clinical episode, occurred in 41% of patients (independent of treatment) with gadolinium enhancement or nine or more T2 lesions versus 11% of those without either finding (p = 0.017); similarly, proportions converting were 44% versus 31% for infratentorial lesions (p = 0.026), 40% versus 35% for juxtacortical lesions (p = 0.413), and 41% versus 17% for three or more periventricular lesions (p = 0.034). The rate of conversion to CDMS based on the number of modified Barkhof criteria was 22% for two or fewer positive criteria, increasing to 47% with four positive criteria. For a cutoff of three positive criteria, the hazard ratio for time to CDMS was 2.3 (95% confidence interval, 1.17-4.55; p = 0.016). Treatment effect seemed more evident as the number of positive criteria increased, and the number of patients needed to avoid one patient converting to CDMS decreased from 50 in patients with one or two positive criteria to 5.6 in patients with four positive criteria. However, the study was not powered to detect statistically significant treatment by variable interaction, and this remains an important issue for further study.