An approach to the design and implementation of clinical trials of empirical antibiotic therapy in febrile and neutropenic cancer patients.

The results of many clinical trials on empirical therapy in febrile, neutropenic cancer patients cannot be readily transferred to the clinical practice, because the methodology is often flawed and definitions, study endpoints and eligibility criteria differ from trial to trial. This article critically reviews some issues related to the design and implementation of randomised clinical trials of empirical antibiotic therapy in cancer patients. Within the definition of phase III clinical trials, two approaches co-exist, based on the trial's specific aims: the "explanatory" approach and the "pragmatic" approach. The usual "explicit" aim of clinical trials of empirical therapy in febrile, neutropenic patients has been that of comparing the "efficacy" of two regimens. However, this term has been more often used with reference to the antibacterial activity of the regimen under study (explanatory aim) than to indicate the practical benefits it draws to the overall patient population treated for fever and neutropenia (pragmatic aim). These two meanings are often taken as perfectly interchangeable, while, conversely, they are completely distinct (though not independent) treatment effects. Most trials conducted in this patient population in recent years are explanatory trials, though not explicitly so, but their results have been widely applied to clinical practice, as they were pragmatic trials. In an explanatory trial the appropriate endpoint is success or failure (defined by clinical and laboratory data) among those patients affected with the specific infection for which the study drug is being given, while in pragmatic trials survival is probably the more appropriate outcome variable, since they are designed to assess the practical benefits that the overall population of febrile and neutropenic patients can obtain from the new empirical treatment. Unfortunately, survival is not a practical study endpoint for the difficulty in assessing the cause of death in this patient population and, especially, for the need for very large sample sizes, which might render the implementation problematic even for large, multicentre groups. Both types of trials need an intention to treat analysis, but this is especially crucial for pragmatic trials, which should not differentiate those cases in which success was obtained through multiple treatment modifications from those who did not require any treatment change. Obviously, this implies that no conclusion should be drawn about the antibacterial activity of the study drugs and that the number of treatment modifications should be taken into account in the interpretation of the results, especially for quality of life and cost evaluations. Information related to fever and signs of infection, age, underlying disease, neutropenia and concomitant administration of other antibiotics are crucial entry criteria that need to be clearly discussed and defined. Finally, the evaluation of toxicity is problematic in this patient population, due to the existence of a number of toxigenic factors, including the underlying disease, the type of infectious complication, the administration of chemotherapy and radiotherapy and the use of parental nutrition. All these effects tend to overlap, thus impairing the investigator's ability to detect specific drug-related side-effects.