We conducted a meta-analysis of randomized controlled clinical trials on steroid treatment for multiple sclerosis and optic neuritis. Of the 25 trials comparing steroids and controls without steroid treatment that we identified 12 were selected for this review. A meta-analysis was conducted to calculate the overall odds ratio across the studies for the numbers of patients without functional improvement and with new relapses. The trials included a total of 1714 patients: 998 with multiple sclerosis and 716 with optic neuritis. Any type of corticosteroids or adrenocorticotropic hormone (ACTH) treatment was considered, as was any dosage, route of administration, and length of treatment. Main outcome measures were: (a) number of multiple sclerosis patients who did not improve by at least one point on the EDSS or equivalent scale, or number of optic neuritis patients without complete recovery of visual acuity at 8 or 30 days and at longer follow-up; (b) number of multiple sclerosis patients with at least one new relapse, or number of optic neuritis patients in whom definite multiple sclerosis was diagnosed at longer follow-up. We found that corticosteroids or ACTH produced a significant improvement in disability or visual acuity at 30 days (odds ratio 0.49; 95% CI 0.37-0.64). The improvement was not statistically significant at longer follow-up (0.85; 95% CI 0.67-1.09). The treatment did not significantly reduce the number of patients with relapses (0.74; 95% CI 0.54-1.01). Both low and high doses were effective for 30-day improvement, but only high-dose and short-term therapy were factors that identified subgroups with some reduction in the risk of new relapse. However, the power of the statistical analysis to detect a reliable difference in the subgroups was low. Steroid treatment is therefore effective in accelerating short-term recovery in patients with multiple sclerosis or optic neuritis. Whether steroids are also effective in reducing the risk of relapse, and the optimal dose and length of treatment must still be determined.