We randomly assigned children with otitis media with effusion to receive either erythromycin-sulfisoxazole, cefaclor, amoxicillin or placebo for a 2-week period, primarily to determine whether either erythromycin-sulfisoxazole or cefaclor would have greater short term efficacy than that found previously for amoxicillin, and secondarily to supplement earlier data on outcomes in placebo-treated subjects. Interim analyses showed no statistically significant (P less than 0.05) differences between the three antimicrobial treatment groups in the primary outcome measures, i.e. the prevalence of middle-ear effusion 2 and 4 weeks after entry, and indicated that postulated differences favoring the erythromycin-sulfisoxazole and cefaclor groups over the amoxicillin group were unlikely to be found even if the originally calculated sample size were attained. Subject accrual was therefore terminated. Final analysis showed no significant between-group differences in other outcome measures as well. In antimicrobial vs. placebo comparisons neither erythromycin-sulfisoxazole nor cefaclor gave more favorable outcomes than placebo, whereas more children were effusion-free in the amoxicillin group than in the placebo group at 2 weeks (31.6% vs. 14.1%, P = 0.007), but not at 4 weeks. We conclude that when antimicrobial treatment for otitis media with effusion is deemed advisable, neither erythromycin-sulfisoxazole nor cefaclor should replace amoxicillin as first line treatment.