Abstract 17486: Non-Proportional Hazards in Randomized Cardiology Clinical Trials: A Standardized Universal Secondary Analysis (SUSAN) to Enhance Clinically Meaningful Appraisal of Results

Abstract

Introduction: In patients with CAD, more invasive interventions are expected to have more early complications in anticipation of improved longer-term outcomes. This often leads to violation of the proportional hazard (PH) hypothesis, thereby complicating interpretation of results. Hypothesis: Analyzing the early and late post-procedural periods separately may provide important additional information for clinicians beyond the cumulative outcome. Methods: We identified 36 RCTs comparing more versus less invasive strategies for the management of CAD. Of those, 16 exhibited crossover of the Kaplan Meier curves, indicating violation of PH. Hereafter, we focus on 4 landmark trials exhibiting such findings: STICH, EXCEL, FREEDOM, and ISCHEMIA. Complete results in all trials will be presented. Results: Initial evaluation of outcomes violating PH includes assessment using algorithms testing proportionality. Thereafter, focus should be shifted not to the time of crossing of curves, but to identifying the time of trend switch: this signifies transition from the earlier to the later disease states (acute/peri procedure or longer term/chronic phase) and may be accomplished by calculation of instantaneous hazards. Following this, the two clinically relevant periods should be studied individually to quantify the impact of different approaches in each of the periods (Figure). As a final step, the cumulative effect of the intervention on hard endpoints should be calculated using statistical designs not assuming proportionality, such as Restricted Free Survival Time or the Com-Nougue. Conclusions: In addition to the cumulative assessment of the primary endpoint, SUSAN allows analysis of the two post-procedural periods separately, as identified by the point of trend switch. This would be particularly informative for clinicians and patients in reviewing early risks and late benefits; adhering to a standardized method may allow a universal presentation of RCT results.