Response to Letter by Boden 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 Boden for his interest in our paper¹. As he mentions, our analysis demonstrated no benefit for PCI on the outcomes of death, cardiac death or myocardial infarction for stable coronary artery disease. Dr Boden approves of this conclusion.

Second, across all categories of unstable coronary artery disease (encompassing multivessel disease following STEMI; non-ST elevation acute coronary syndromes; and unrevascularized post-MI) we found a benefit for PCI on death, cardiac death and myocardial infarction. Dr Boden considers this invalid.

The VANQWISH trial² was not eligible for our meta-analysis. Patients in one randomized arm received coronary angiography, and any intervention that was considered clinically appropriate at time of angiography. Of the 462, 98 (21%) received PCI, which was balloon angioplasty without stenting (a further 95, 21%, received CABG.)

In the other arm, patients underwent, before discharge, a symptom-limited treadmill exercise test with planar thallium scintigraphy, or thallium scintigraphy with single-photon-emission computed tomography, and patients who had any positive test (or recurrent symptoms with ECG changes) then underwent angiography. Of the 458, 55 (12 %) received PCI. (A further 87, 19%, received CABG.)

Therefore, VANQWISH randomized patients with non-ST segment elevation myocardial infarction, but only gave differential treatment between arms to the subset who had no inducible ischemia on non-invasive testing; the arms otherwise received the same treatment. Our analysis was trying to examine the effect of invasive management in various classes of patient; in the VANQWISH trial the treatment was essentially the same in both arms for patients with postinfarct angina, ST depression during exercise, or ischemia on thallium testing. For this reason, VANQUISH was not eligible for our analysis.

However, VANQWISH does make an important point that invasive therapy may provide net harm, in the subpopulation of NSTEMI patients who subsequently have no angina and no inducible ischaemia. We have also re-run our analysis adding the VANQWISH trial. The relative risk for all-cause mortality across all categories of unstable CAD without VANQWISH was 0.84 (95% confidence interval 0.75 to 0.93, p=0.001), with no heterogeneity (I² 0.0%). If VANQWISH were to be added, the relative risk would be 0.84 (95% confidence interval 0.74 to 0.96 p=0.011), and heterogeneity would rise to 21.4% in the unstable CAD class.

For NSTEACS, without VANQWISH the relative risk was 0.84 (95% confidence interval 0.72 to 0.97, p=0.02), with no heterogeneity (I² = 0.0%); with VANQWISH, the relative risk would be 0.89 (95% CI 0.75 to 1.06, p=0.19). Again, adding VANQWISH alone would balloon heterogeneity to 31.4% for NSTEACS.

For the outcome of myocardial infarction, adding VANQWISH would not substantially change the results. Across all categories of unstable CAD, the relative risk without VANQWISH was RR 0.74 (95% CI 0.62 to 0.90, p=0.002, I²=57.6%). With VANQWISH the RR would be 0.75 (95% CI 0.63 to 0.90, p=0.002, I²=57.4%). For the outcome of myocardial infarction in NSTEACS trials, without VANQWISH the RR was 0.83 (95% CI 0.64 to 1.06, p=0.14, I²=66.9%). With VANQWISH the RR was 0.84 (95% CI 0.67 to 1.04, p=0.11, I²=57.4%).

Dr Boden is correct to suspect he is misreading Figure 3. Values are relative risks, not odds ratios. For the sake of clarity, only two NSTEACS trials reported cardiac death. Their pooled relative risk was 0.80 (95% CI 0.59 to 1.08, p=0.14) with no heterogeneity (I²=0.0%). When considered together with the other forms of unstable CAD in our pooled analysis, the point estimate for cardiac death was a relative risk of 0.69 (95% CI 0.53 to 0.90, p=0.007). There was moderate heterogeneity (I² = 39.4%).

We do not approve of excluding trials on the basis of size. The purpose of meta-analysis is to assemble the results from similar trials and look for an effect of therapy over the pooled results, where one might not be apparent from individual smaller trials. Meta-analysis calculation incorporates a lower weighting for information from smaller trials. Our analysis found zero heterogeneity for the outcome of all-cause mortality, both within the NSTEACS group and within the pooled results for all unstable CAD categories.

Dr Boden helpfully points out the review by Fanning et al³, which had a similar point estimate but wider confidence intervals than our analysis. Their RR for all-cause mortality was 0.87 (0.64 to 1.18); whereas ours was 0.84 (0.72 to 0.97). One key difference is the Fanning review was limited to trials in the stent era, whereas we did not impose such a restriction. The Fanning review also included trials that compared the timing of invasive management, as opposed to only comparisons of invasive therapy and medical therapy. Our analysis excluded such trials:NSTEACS trials were only eligible if they compared invasive versus conservative strategies, and not if they compared early versus late invasive strategies.

We are puzzled that half-way through the letter Dr Boden seems to have swapped sides, now presenting reasons why VANQWISH should not be included. He objects to interventional trials that had low rates of renin-angiotensin agents (22% in VANQWISH) and lipid lowering agents (13% in VANQWISH). Ultimately, there have been significant advances in both medical therapy and invasive therapy over the last two decades; in the VANQWISH trial, percutaneous coronary intervention was with balloon angioplasty only. In the intervening years there have been myriad advances in percutaneous coronary intervention technology (from balloons and into the stent era including bare metal and then drug-eluting stents, with further improvements in latter-era DES) and techniques (smaller sheath and catheter sizes, transradial access, intravascular imaging). We acknowledged this point in the Limitations section of our manuscript (“Other potential sources of heterogeneity include differences in length of follow-up, pharmacotherapy, invasive therapy (balloons, bare metal stents, drug-eluting stents), and study populations. Our analysis includes trials from 1992 and from 2019, and in that period of time there has been significant advancement in both the in the pharmacological and invasive management of CAD. This is a further source of heterogeneity in such an analysis”). We did not feel that there would be a systematic bias in favour of either the conservative or invasive arms by including both older and newer trials.

Ultimately, Dr Boden’s suggestion that in the modern era invasive therapy provides net harm to unstable CAD subsets is an exciting hypothesis, because we think the opposite might be the case. This degree of difference of opinion on a clinically important question makes it an ideal subject matter for a future randomized clinical trial.