We are concerned that the article by Dohn and colleagues [1] may have overstated the efficacy of oral atovaquone for the treatment of Pneumocystis carinii pneumonia in patients with the acquired immunodeficiency syndrome (AIDS). The study is described as an equivalency trial, that is, a test of the hypothesis that using oral atovaquone is not worse than using intravenous pentamidine for the treatment of AIDS-associated P. carinii pneumonia. The authors reported that the hypothesis was confirmed; in this small study, treatment failures were not significantly more frequent in the atovaquone group than in the pentamidine group. However, 29% of patients failed to respond to treatment with atovaquone compared with 17% of patients who received pentamidine. Therefore, patients in this study had a treatment failure rate that was 71% higher if they received atovaquone than if they received pentamidine. The study was designed to detect significant differences in treatment failure rates that were greater than 15%, and this difference was not achieved; moreover, study enrollment was terminated prematurely. If the study accrued fully and the failure rate remained the same, the difference would have closely approached significance (P = 0.10; two-tailed Fisher exact test). The authors concede that a larger study may have detected a statistically significant difference in treatment failure rate. Nonetheless, the conclusion that a failure rate of 29% is comparable to a failure rate of 17% in a small study that did not have sufficient power to detect major differences is surprising. Atovaquone should not be considered comparable to intravenous pentamidine for treatment of P. carinii pneumonia until the hypothesis is proven.