Much has been researched on the treatment of chronic coronary artery disease (CAD). 1-3 The ISCHEMIA (International Study of Comparative Health Effectiveness with Medical and Invasive Approaches) 4 trial had been long awaited, in the hope that it would answer many remaining questions. Our purpose is to prepare a critical analysis of this document, analyzing its methodology, its results and how it may impact the cardiologist's clinical practice. Initially, when examining the methodology, a

more »

... ence is noted between what was planned regarding sample size and what was actually accomplished. The sample size was initially calculated to detect an 18% difference in the event rate, when comparing the Invasive Strategy (INV Group) and the Conservative (CON Group) groups over a 4-year period, assuming an event rate of 20% in the second group. In other words, with an event rate of 16.4% in the INV Group, considering a test power of 90% and an alpha error of 0.05, approximately 2,500 patients would be required in each group. However, based on a data analysis performed in 2018, the rate of events was lower than previously estimated, and the researchers were forced to recalculate the sample size to achieve an 18.5% difference when comparing the CON and INV Groups, considering an aggregate event rate of 14% over a 4-year period in the CON Group, with a test power of 83% and a significance level of 0.05. Redoing these calculations in the R statistical software package, for the desired test power of the study, a total of 2,784 patients would be required in each group, totaling more than 5,500 patients. Furthermore, for the difference actually detected between the groups, i.e. 10%, the number of patients required for each group would be 7,500 patients, totaling 15,000. With this simple calculation, it is easy to conclude that the study did not have the statistical power for which it was proposed, since it presumed something that actually did not occur. The analysis of the differences between the two groups was estimated using Cox's proportional regression model. However, the statistical premises of this model were violated for the primary purpose of the study (due to treatment-time interaction, according to the researchers). The statistical analysis plan had specified that the presentation of the results would emphasize non-parametric estimates for cumulative event rate if the assumption of proportional risks were violated, and this actually happened. There were further problems in the study methodology, rendering it extremely complex to be analyzed. Examining the study results, it immediately stands out that 26% of patients in the CON Group underwent coronary angiography, and 21% of patients were revascularized. This means 544 patients out of 2,591, which is not a small number, and although these patients have undergone some type of revascularization, they were analyzed as if they were in the CON Group. Furthermore, the researchers reported that, in the INV Group, a total of 5,337 revascularization procedures were performed (including repeated