Distribution of patients' paroxysmal atrial tachyarrhythmia episodes: Implications for detection of treatment efficacy

Distribution of Paroxysmal Atrial Tachyarrhythmia Episodes. Introduction: Clinical trials of treatments for paroxysmal atrial tachyarrhythmia (pAT) often compare different treatment groups using the time to first episode recurrence. This approach assumes that the time to the first recurrence is representative of all times between successive episodes in a given patient. We subjected this assumption to an empiric test. Methods and Results: Records of pAT onsets from a chronologic series of 134 patients with dual chamber implantable defibrillators were analyzed; 14 had experienced >10 pAT episodes, which is sufficient for meaningful statistical modeling of the time intervals between episodes. Episodes were independent and randomly distributed in 9 of 14 patients, but a fit of the data to an exponential distribution, required by the stated assumption, was rejected in 13 of 14, In contrast, a Weibull distribution yielded an adequate goodness of fit in 5 of the 9 cases with independent and randomly distributed data. Monte Carlo methods were used to determine the impact of violations of the exponential distribution assumption on clinical trials using time from cardioversion to first episode recurrence as the dependent measure. In a parallel groups design, substantial loss of power occurs with sample sizes <500 patients per group. In a cross-over design, there is insufficient power to detest a 30% reduction in episode frequency even with 300 patients. Conclusion: Clinical trials that rely on time to first episode recurrence may be considerably less able to detect efficacious treatments than may have been supposed. Analysis of multiple episode onsets recorded over time should be used to avoid this pitfall.