Using Endpoint Data to Analyze Patients Rather Than Patient Data to Analyze the Endpoints

• Activity Number: 492
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Scott R. Evans*+
• Companies: Harvard School of Public Health
• Keywords: benefit:risk ; hospital chart analysis

Abstract

The standard approach to the analysis of clinical trials is to conduct between-arm comparisons with respect to each important endpoint (benefits and harms) and then weighing the results of these comparisons as part of a benefit:risk analyses. However this approach does not inform regarding the association between benefits and harms (i.e., whether the patients that benefit are the same or different than those that experience toxicity). This information is important for evaluating the utility of the intervention and for effectively treating patients. We propose the "hospital chart analysis" by creating an ordinal categorical outcome based on benefits and harms data, and then classifying each patient per this outcome. Interventions can then be compared with respect to this composite benefit:risk outcome. This approach not only incorporates the association between benefits and harms but can be constructed to integrate loss-to-follow up and other competing risk issues. An evolving example from an antimicrobial stewardship trial is discussed.