Design and analysis issues in cardiac arrhythmia trials: insights from the Detect Supraventricular Tachycardia Trial

BACKGROUND: The Detect Supraventricular Tachycardia (Detect SVT) trial was a randomized single blind study in subjects receiving a dual chamber implantable cardioverter defibrillator (ICD). The primary goal was to evaluate whether dual chamber enhancements result in greater discrimination between supraventricular tachycardia (SVT) and ventricular episodes when compared to ventricular-only enhancements. PURPOSE: The purpose of this report is to describe our experiences from an analytic perspective, including overall study design, the value of blinded episode review, and the statistical analysis plan. METHODS: The Detect SVT study was a parallel group study, with blinded central review of all arrhythmia episodes occurring over a six-month period. The primary response was the proportion of inappropriate detections (SVT episodes that are inappropriately classified as ventricular by the device). RESULTS: A parallel group design was essential as unplanned crossovers were very differential. The blinded review of episodes was a major use of resources, with over 2300 episodes observed. This review did not materially alter the findings, however, the avoidance of any perceived bias was a major strength. The value of collecting large numbers of episodes on a single subject was found to be minimal. A Kaplan-Meier analysis of the number of SVT episodes until first inappropriate detection was found to add additional information. LIMITATIONS: The recommendations reported here relate to cardiac arrhythmia trials, or other studies with very similar endpoints. These observations are based on the results of a single trial and should be confirmed in other studies. CONCLUSIONS: The parallel group design was essential, as was accounting for within-subject correlation in the analysis. The blinded review process was also important, but significant resources could be saved by closing-out episode collection after a fixed number of episodes. Lessons learned from the Detect SVT trial may prove useful to others designing trials in this active research area.