Reporting the results of randomized trials of empiric antibiotics in febrile neutropenic patients--a critical survey.

Twenty-one randomized trials of empiric antibiotic therapy in febrile neutropenic patients were reviewed and scored by methods suggested by Chalmers. In general, the scores were low, indicating that many of these articles failed to meet currently suggested methodologic standards for publication of results of clinical trials. Particularly striking was the lack of attention paid to statistical issues, such as power, exclusions, and repeated analyses. In addition, there appears to be a need for a better-developed system for assessing response in these circumstances.