Article révisé par les pairs
Résumé : Severe sepsis and acute lung injury are challenging diagnoses as they relate to designing and reporting of clinical trials. The limited success in bringing forward new therapies in these areas is likely proof of that premise. The ability to use preclinical and phase I and II trial data to predict which patients and which dosing regimens are more likely to benefit is perhaps the greatest challenge. Animal models continue to be refined in attempts to more accurately reproduce human sepsis and acute lung injury. Oncology research should serve as a model for optimizing the integration of pharmacodynamics and pharmacogenetics into trial design. The European Organization for Research and Treatment of Cancer provides a valuable template for nonfunded multicenter clinical trial success. The marked heterogeneity of the patient population and small signal (tested therapy)-to-noise (comorbidities) ratio makes identification of treatment effect difficult. Dedicated investigators still enroll ineligible patients who are included in intent to treat analysis. High enrolling centers create less problems in an adequate test of a new therapy. Much has been learned from negative trials as to value of post hoc subgroup and interim analyses. Debate continues on fair and appropriate end point of trials. Extrapolation of adult positive trial results to children is problematic. Conflict of interest issues which rested dormantly for years are now at the forefront of discussion, and journal editorial board responsibility in this area is being recognized. Protocols may also help reduce heterogeneity of treatment across centers in clinical trials. This article reviews many of the problems encountered in clinical trial design and reporting and offers a perspective on dealing with them to the betterment of a clinical trial. © 2008 Elsevier Inc. All rights reserved.