Response by Ahmad et al Regarding Article “Effects of percutaneous coronary intervention on death and myocardial infarction stratified by stable and unstable coronary artery disease: a meta-analysis of randomized controlled trials”

We thank Dr Shah for his interest in our article¹. The results of the ISCHEMIA trial were already public knowledge, having been presented at the American Heart Association Scientific Sessions as Kaplan-Meier curves, and extensively discussed in the medical and lay media. Since the trial investigators had been confident enough in the correctness of their results to present them publicly and discuss it with lay media, it would have been scientifically inexcusable to pretend these data did not exist.

We used the curves provided by the investigators and the quoted numbers at risk to derive event counts. This was explained in the Methods section of our manuscript: ‘If studies did not provide the event counts, data were extracted from Kaplan-Meier curves by digitization of the survival curves which were combined with the numbers at risk to derive the number of events, using the R package reconstructKM’.. We therefore committed to updating our analysis in the event there were significant changes between the conference presentation and the journal publication: ‘The ISCHEMIA trial has not yet been published in full. If the full published data differ from the presentation, we will update this analysis accordingly.’

The NEJM publication² does not quote event counts for cardiac death, so the event rates in the conference presentation has not been displaced as the best source for this information. It did, however, report event counts for all-cause death and myocardial infarction, so we have now re-run our analyses using these values for these outcomes.

The effect size for all-cause mortality in stable CAD changed from RR 0.98 (95% CI 0.87-1.12, p=0.79) using the conference data to RR 0.98 (95% CI 0.86-1.11, p=0.71) using the NEJM data. The effect size for myocardial infarction in stable CAD changed from RR 0.98 (95% CI 0.86-1.11, p=0.72) using the conference data to RR 0.97 (95% CI 0.86-1.11, p=0.69) using the NEJM data. We have included these new Forest plots here as Figure 1 and 2. We can see why the investigators chose to present these data promptly at the conference, since the results are important.

Figure 1. The effect of PCI on all-cause mortality.

Figure 2. The effect of PCI on myocardial infarction.

Dr Shah asks why we did not include three trials: (1) the VANQWISH trial, (2) the ORBITA trial (of which we are authors), and (3) an unreferenced study by Estevez-Loureiro et al.

First, we have explained in detail why the VANQWISH trial was not eligible, in our response to the letter from Dr Boden³. Second, the ORBITA trial was not eligible because itwas measuring the placebo-subtracted effect of PCI on exercise time, and deliberately only lasted 6 weeks, to avoid leaving enough time for events such as death. It was not a trial comparing long term effects of PCI with medical therapy for the outcomes of death and myocardial infarction, and was therefore ineligible.

Third, the study by Estevez-Loureiro et al is not referenced in Dr Shah’s letter because it has never been published. We cannot be sure, but we assume Dr Shah is referring to an abstract presented at the ESC Congress in Barcelona in 2014. We were aware of this study. This unpublished trial randomized patients with STEMI and multivessel disease to either complete revascluarization or ischemia-guided revascularization following a stress echocardiogram. The inclusion criteria for this category in our analysis was: ‘Multivessel disease following STEMI: patients who underwent successful primary PCI for STEMI and had residual coronary lesions, and who were randomized to PCI versus no PCI for those residual lesions.’ The unpublished study by Estevez-Loureiro et al therefore clearly does not fit the inclusion criteria for our analysis.

We agree that unrevascularized post-MI patients are unusual in contemporary clinical practice. Prior meta-analytic work⁴ had classified these studies as ‘stable CAD’; we wanted to make the distinction that these were not truly stable patients.

We are, once again, puzzled that a commenter has criticised us for including ‘older and lower quality’ studies but in the same breath insist that we include a trial published in 1998. Ultimately, any assertion that in the modern era medical therapy would be superior to invasive therapy for unstable CAD subsets would only be credible if it was the finding of a contemporary randomized clinical trial.