Author + information
Given the extensive time, effort and expense involved in performing large-scale clinical trials, there is a natural tendency to perform subgroup analyses on the accumulated data subsequent to termination of the trial. Assmann et al. (1)recently stated: “Subgroup analyses are particularly prone to overinterpretation, and one is tempted to suggest “don’t do it” (or at least “don’t believe it”) for many trials, but this suggestion is probably contrary to human nature.” The report by Hallstrom et al. (2), a subgroup analysis of the Antiarrhythmics Versus Implantable Defibrillators (AVID) trial (3), falls into the category of “don’t believe it.”
The hypothesis of this substudy is reasonable—there may be subgroups of AVID patients who do not benefit from an implantable defibrillator as opposed to amiodarone therapy. Based on their data, the authors (2)conclude that the lowest-risk sextile is a subgroup in which implantable defibrillators may not offer benefit. However, this conclusion is faulty as the design of this substudy forces the outcome to be negative.
The AVID trial (3)evaluated whether the strategy of antiarrhythmic drug therapy versus implantable defibrillator therapy is better for treating patients with hemodynamically significant ventricular tachycardia or ventricular fibrillation. In designing the study, the investigators (4)concluded that a sample size of 1,200 patients was required to demonstrate the predicted benefit. The trial was terminated early, after only 1,096 patients were enrolled, because the analysis revealed that the difference in the primary outcome variable had crossed the statistical boundary for early termination. In the current report (3), the analysis of relative benefit between antiarrhythmic drugs and the implantable defibrillator relies on a sample size that is one-sixth the total sample size of the AVID trial. In addition to the small sample size, the subgroup being investigated is the group of patients in which the event rate and mortality is lowest. In subgroups with lower event rates, the difference between event rates in the two treatment groups typically decreases and a larger sample size is required to demonstrate that the difference is significant. It is misleading to suggest that one could demonstrate a difference in survival in the lowest-risk sextile with the sample size that is available.
Study sample size is typically chosen so that the power of the study or the probability of detecting the postulated difference is high, typically in the 80% to 90% range. The calculated power of this AVID substudy is 5.5% (calculated using a sample size of 166 patients, two year mortality of 11% [which was the observed mortality], a 25% reduction in mortality to 8.25% by one of the treatments), a value substantially less than the 80% to 90% value used in designing a trial. This means that there was only a minimal chance that this report could have demonstrated a difference between the two therapies even if implantable defibrillators or amiodarone reduced mortality by as much as 25%.
These substudy results should be interpreted as inadequate to answer the question posed by the investigators. The data are consistent with no benefit of implantable defibrillator over amiodarone therapy, with a benefit of implantable defibrillator therapy over amiodarone therapy and with a benefit of amiodarone therapy over implantable defibrillator therapy. In contrast, the clinical characteristics that identify the low risk sextile are interesting. A suggestion that the cost-effectiveness in this low risk sextile may be less favorable than in the other sextiles may also be valid. However, only properly designed clinical trials will be able to address the question of which therapy is appropriate or better in patients who have a low arrhythmia risk. Until such studies are performed, one can only conclude based upon the AVID trial that the strategy of implantable defibrillator therapy has a survival benefit compared to the strategy of antiarrhythmic drug therapy (amiodarone) in patients who have suffered either ventricular fibrillation or hemodynamically significant ventricular tachycardia.
- American College of Cardiology