OBJECTIVE To evaluate the rate of bacteriologic failure of amoxicillin-clavulanate in the treatment of acute otitis media (AOM) and to identify the risk factors associated with failure. METHODS Ninety-nine subjects (mean age, 21.4 months) with AOM were treated with amoxicillin-clavulanate in two prospective study trials that compared efficacy of two experimental antibiotics with amoxicillin-clavulanate. Tympanocentesis for microbiologic studies was performed in all subjects at enrollment; at 3 to 6 days, during amoxicillin-clavulanate therapy; and at other times when clinically indicated. The subjects were followed up for 1 month. Clinical, bacteriologic, and virologic characteristics of the subjects were analyzed. RESULTS Bacteriologic failure of treatment occurred in none of 39 subjects (0%) with Streptococcus pneumoniae, two of 25 (8%) with Moraxella catarrhalis, and 11 of 29 (38%) with nontypeable Haemophilus influenzae (NTHi) infection. The failure rate for NTHi was higher than that for other pathogens (p = 0.0007) and was increased when compared with the preceding study period (p = 0.017). Bacteriologic failure was also associated with clinical failure (p = 0.041). In subjects with AOM caused by NTHi the rates of adequate drug compliance were comparable in both success and failure groups. Antimicrobial susceptibility testing by minimum inhibitory concentration and minimum bactericidal concentration (MIC/MBC) assays showed that amoxicillin-clavulanate resistance was not significantly associated with bacteriologic failure of treatment. However, in two subjects, MIC/MBC of the NTHi isolates during therapy were higher than MIC/MBC of the isolates before therapy; these strains of isolates pretherapy and during therapy were discordant as determined by outer membrane protein analysis. The bacteriologic failure rate was higher in nonwhite boys (p = 0.026) and in subjects with a history of three or more previous episodes of AOM (p = 0.008). Other factors such as age, bilaterality of disease, polymicrobial infection, and biotype pattern of NTHi were not associated with treatment failure. When children with adequate drug compliance were analyzed separately, only those with concomitant viral infection of the nasopharynx or middle ear were found to be at an increased risk of bacteriologic failure of treatment (p = 0.04). CONCLUSIONS The bacteriologic failure rate of amoxicillin-clavulanate therapy for AOM caused by NTHi was higher in the current study period than in the preceding period. Factors contributing to treatment failure were race, gender, proneness to otitis, and concomitant viral infection.