Establishing Criteria for Assessment of Efficacy of Antimicrobial Agents in Acute Otitis Media

In a recent article, Dagan et al. (1) report on the comparative bacteriologic efficacies of oral azithromycin and oral cefaclor in children with acute otitis media. The authors assert that based on a double-tympanocentesis method of evaluating bacteriologic eradication in otitis media susceptibility breakpoints for antibiotics can be determined and previous clinical efficacy studies are inadequate compared to this method. We would like to point out some of the pitfalls inherent in the bacteriologic eradication trials and the particular bias introduced into the trial described by Dagan et al.

In order to determine bacteriologic efficacy, the authors use a methodology first utilized by Howie and Ploussard in 1969 (5). Despite 30 years of experience, the double-tympanocentesis method has yet to be validated in clinical trials conducted to the most rigorous scientific standards in which all possible measures to exclude bias are employed. The diagnosis of acute otitis media is subjective in nature; therefore, the “gold standard” methodology for evaluation of anti-infective agents is the double-blind, double-dummy clinical trial using trained otoscopists. Open-label or single-blinded studies allow bias to influence both assessment by the investigator and actions of the patient. For example, the apparent failure of azithromycin to eradicate Haemophilus influenzae in the study reported by Dagan et al. may be due to the timing of the tympanocentesis in relation to the time-dependent bactericidal activity of the drug. In an open-label study, children on the shorter azithromycin regimen may return to the physician earlier compared with children on the 10-day beta-lactam regimen. This was, in fact, the case (Fig. 1). The median time to second tympanocentesis was 1 day earlier in the azithromycin arm compared with the cefaclor arm, and this trend, despite the small sample size, nearly reached statistical significance. A double-blind, double-dummy clinical trial design minimizes the potential for bias due to differences in length of therapy.

FIG. 1.

Median time to second tympanocentesis. For azithromycin (AZ) the median was 4 days (range, 3 to 6 days). For cefaclor (CFC) the median was 5 days (range, 4 to 6 days). P = 0.1044.

Dagan et al. cite their previous publication (2) as evidence that their model evaluating early bacteriologic eradication is a valid surrogate for eventual clinical response. The previous report uses the same data as the azithromycin-cefaclor study, combined with data from reportedly a third, randomized, amoxicillin-treated group of patients. The open-label design and the failure to follow the established randomization schedule preclude an unbiased assessment of drug-specific clinical response. Therefore, validation of a model in a three-arm study that uses the same data to report clinical and bacteriologic results from only two of the arms is a deviation from sound clinical-trial conduct and emphasizes how important it is to adhere to established methods in clinical-trial research in order to remove bias from scientific inference.

Dagan et al. also fail to provide a sound theoretical explanation for the observed results. The authors postulate that although azithromycin is highly concentrated in phagocytic leukocytes it is not available in the extracellular space to act against extracellular pathogens. This is inconsistent with previous clinical-trial experience and in vitro data demonstrating that 93.3% of the intracellular amount of azithromycin is released within 24 h under resting conditions (4) and immediately in bacteriologic stimulation (J. M. Hyatt, S. A. Mangione, and J. J. Schentag, Fifth Int. Conf. Macrolides Azalides Streptogramins Ketolides Oxazolidinones, abstr. 3.02, 2000).

It remains to be determined whether the timing of the tympanocentesis, investigator bias, and other factors can be removed from the evaluation of these double-tap trials. Given the potential pitfalls inherent in the methodology used in this study, as well as the lack of clinical validation, it would be premature to consider changes in susceptibility breakpoints or clinical-trial design based on these data as the authors have suggested. The peer review community must ask the difficult questions that might shed light on why a drug that has performed well in large, double-blinded clinical trials is perceived to be relatively inactive in children in a small bacteriologic study fraught with methodological problems. Until the double-tympanocentesis method is independently validated, clinical researchers should continue to use established designs for evaluating antimicrobials in otitis media by conducting large, double-blinded clinical trials in accordance with advice from experts in the field and the Food and Drug Administration (3).