New printed evidence reignited the controversy over the EXCEL affair, exposing more important flaws, both in the trial’s statistical analysis and the revelation that the previously concealed myocardial infarction data… Click to show full abstract
New printed evidence reignited the controversy over the EXCEL affair, exposing more important flaws, both in the trial’s statistical analysis and the revelation that the previously concealed myocardial infarction data has now been made available. Further to the extensive disclosure of flagrant bias introduced into the design, analysis, and publication of the 5-year follow-up of the EXCEL trial[1], the June edition of the Journal of American Medical Association (JAMA) Internal Medicine revealed and highlighted that the formal hypothesis testing, which in the EXCEL trial was prespecified as a demonstration of noninferiority at 3 years, was without discussion or explanation switched to a superiority test in the 5-year analysis[2]. If as prespecified, the same noninferiority analysis had been conducted at 5 years it would not have met the criteria for statistical significance, and the printed conclusion would not have been possible. This is a classic case of interpretation bias, otherwise known as spin. The newly identified violation follows in the wake of the controversy over data manipulation in the EXCEL trial. The article authored by Professor James M. Brophy, who is a professor of Medicine, Epidemiology, and Biostatistics at McGill University in Montreal, Canada, re-examines the design and results of the EXCEL trial and other recent randomized clinical trials related to the left main coronary artery disease (LMCAD) using Bayesian methods[2]. The article shows furtive aspects and details of the statistical analysis, not previously acknowledged, that in effect concealed and circumvented the real results of the EXCEL trial to attain the desired conclusion. The method of statistical analysis used to interpret the 5-year data was different from that used in the 3-year publication, and in direct contrast to the original protocol, without explanation or justification. If EXCEL’s 5-year results had been interpreted in
               
Click one of the above tabs to view related content.