We strongly agree that cost should always be considered when prescribing a medication. Our study, however, like most randomized, controlled trials, was designed to evaluate the safety and efficacy of two agents, fluconazole and ketoconazole, for candida esophagitis. We did not conclude that either agent was the "treatment of choice." We feel that the reader can use the results of our study in combination with other considerations such as local drug costs and individual patient characteristics (for example, previous antifungal therapy) to make decisions about treatment.

A formal decision analysis incorporating cost could be extremely helpful in the treatment of patients with HIV infection and esophageal symptoms. Such an analysis, however, would represent a separate study. Unfortunately, a simple statement of the costs of fluconazole and ketoconazole is of minimal value by itself because it ignores the numerous variables that have to be included in choosing a medication. Even the more ambitious assessment of Drs. Thio and Rutledge fails to take into account several important factors.

Because the etiologic agent in most HIV-positive patients with esophageal symptoms is Candida, I recommend giving empiric antifungal therapy to most patients and reserving endoscopy (the gold standard for diagnosis) for those who show no response in 1 to 2 weeks. The total cost of an upper endoscopy with the attendant histologic and microbiologic tests is in the range of $700 to $1200—an expense that far exceeds the cost of a course of either fluconazole or ketoconazole. The 20% greater rate of clinical resolution seen with fluconazole might avoid the expense of endoscopy in 20% of patients. In addition, patients treated with ketoconazole in our trial required doubling of their dose twice as often as those given fluconazole (at our institution, we generally begin ketoconazole at 400 mg/day for candida esophagitis)—again decreasing the cost differential. Finally, patients in our study had a more rapid response to fluconazole, which might allow for a shorter course of therapy. Determining whether these considerations outweigh the cost differential requires formal analysis.

The letters also included several specific questions. Dr. Neely must have missed our Figure 1, which shows the clinical response of patients to the study medications. No patients in the trial were taking ddI, and no significant difference was observed between the two treatment groups in the proportion of patients taking ketoconazole before entry into the study. Resistance is a concern with the frequent and long-term use of any anti-infective agent. The incidence of fluconazole-resistant Candida albicans is unknown. Because in-vitro testing of fungal susceptibilities often does not parallel results in in-vivo animal models, the reproducibility and clinical predictability of in-vitro testing have been questioned [1]. Further, clinical resistance to oral antifungal therapy probably is frequently due to host factors (for example, compliance, absorption, degree of immune deficiency).