Navarese et al’s meta-analysis of elective coronary revascularization and medical therapy versus medical therapy alone was recently published in the European Heart Journal.

This is an area of great controversy because there is evidence that revascularization does not improve prognosis in stable coronary artery disease. Naturally, the revascularization industry is disappointed by this. Revascularization did not reduce myocardial infarction and death in the COURAGE or ISCHEMIA trials. Navarese et al’s meta-analysis of 25 trials didn’t find a reduction in these outcomes with revascularization either. Nor did Chacko et al’s 2020 meta-analysis. If the case wasn’t closed after ISCHEMIA, surely it should be now.

Or should it? Navarese et al found, “In stable coronary artery disease patients, randomisation to elective coronary revascularisation plus medical therapy led to reduced cardiac mortality compared with medical therapy alone. The cardiac survival benefit after revascularisation improved with longer follow-up times and was associated with fewer spontaneous MIs.”

Why is the message so different? Is there truly a glimmer of hope that revascularization might have advantages over medical therapy alone?

The primary endpoint of the new meta-analysis was cardiac death. The finding of a reduction in cardiac death with a rate ratio of 0.79 (95% CI 0.67–0.93) was shown to be robust (i.e. consistent even when certain categories of trials were removed e.g. post-ACS). The meta-analysis appears to have been well-conducted.

However, it is more usual to consider all-cause mortality or all-cause mortality and myocardial infarction. To recommend a therapy as having prognostic value, the therapy needs to be shown to reduce all-cause mortality, not just disease-specific mortality. This distinction is important because it is easily possible to reduce disease-specific mortality whilst increasing mortality from other conditions, or having an overall null effect on mortality. Clinically, what matters is that death has happened. Disease-specific mortality is worth measuring but only as a secondary endpoint to explore how the therapy is reducing mortality.

Not all trials can assess for mortality benefit because the nature of the disease and the therapy might mean the trial would need to be too long to be feasible. In that situation the primary endpoint can be a composite of all-cause mortality and events that signify poor prognosis or disease progression. In the case of coronary artery disease the best event to look at is myocardial infarction because it is objective and clearly associated with worse outcomes. Admissions for unstable angina and heart failure are less objective (especially in unblinded trials) but are also fair game for inclusion in a composite primary endpoint as they are associated with worse outcomes and are indicative that the patient has started to experience complications of the disease. By itself cardiac death isn’t enough. The absence of an overall benefit for all-cause mortality and myocardial infarction strongly suggests that the therapy isn’t working as intended.

Navarese et al state that “Recommendations for medical therapy alone based on trials with limited follow-up have likely underestimated the benefits of revascularisation plus medical therapy.” This seems like wishful thinking to me. The benefits we need to be seeing (reduced death and MI) just aren’t there.

+ posts


  1. Good point, Alex. This meta-analysis has provoked a lot of attention on Twitter. Unsurprisingly, supporters of the aggressive use of revascularization strategies have tried to use it to support their position. I’d just like to briefly offer a few reasons why this meta-analysis should NOT be the basis for a more positive spin on revascularization than initially suggested by the ISCHEMIA trial.

    Unsurprisingly, the most positive position was taken by Gregg Stone on Twitter ( He wrote

    “Most important (to me) is that revascularization in stable CAD is clearly on net balance ‘safe’. No studies/meta-analyses show increased stroke, MI or death with PCI or CABG. So people should stop talking about the potential harms of revasc.”

    Stone is too quick to claim victory here. First of all, ALL medical interventions, and especially any procedures as serious as revascularization with PCI or CABG, may have harmful effects, and these should ALWAYS be carefully considered by both physicians and patients. It is the height of irresponsibility to recommend that “people should stop talking about the potential harms of revasc.”

    Stone also completely ignores the financial, economic, and opportunity costs related to a widespread aggressive approach.

    Additionally, there is the blithe indifference to the impact on an individual of having a major invasive procedure. Of course interventional cardiologists take this sort of thing for granted, but the vast majority of people don’t see PCI or CABG as quite so benign. For many it is a life-changing event, and not in a good way. They become permanent “heart” patients.

    One final point: it is the height of chutzpah to use the meta-analysis as a way to address the limitations of ISCHEMIA. Of course ISCHEMIA was designed to produce a robust answer to these questions. By failing to randomize patients in ISCHEMIA, the cardiology community in general, and the interventional community in particular, fatally weakened the trial. This problem was quite well known from the very beginning, yet the trial leaders and the cardiology community failed to respond adequately. Of course, no meta-analysis can possibly make up for the deficiencies of a trial with these limitations.

  2. This is an excellent article and comment from Larry. I add that PCI and CABG should not be pooled together. We are not offering both treatments to our patients. We are offering one or the other. Each with its advantages and disadvantages. Pooling has the net effect liming the outcomes of each individual treatment and does not show a realistic figure to base decisions

  3. I respectfully disagree with Alex and Larry.

    Alex states the primary endpoint of the meta-analysis should have been all-cause mortality. It is true that — 30 years ago — we designed trials with all-cause mortality with the primary endpoint. But in the last few decades, we have learned that that was not a good idea, and instead, we focused our efforts on the effect of drugs and interventions on cardiovascular mortality. We did so for good reasons. Cardiovascular interventions are not intended to reduce noncardiovascular deaths, so the inclusion of noncardiovascular events in an endpoint is unnecessarily dilutional. It is not true that, to recommend a therapy as having prognostic value, the therapy needs to be shown to reduce all-cause mortality.

    Interestingly, our shift in focus to cardiovascular death has been strongly encouraged by the US FDA. In fact, the Agency would prefer to grant a label for “reduction in cardiovascular mortality” even if a drug has been shown to reduce all-cause mortality, since such a label clearly defines what a cardiovascular drug can reasonably be expected to do.

    Alex believes that the optimal endpoint for coronary artery disease should be a composite of all-cause mortality and myocardial infarction. Why? She thinks that myocardial infarction is “objective”, but many myocardial infarctions are defined by a troponin exceeding some threshold value. These values are often rather arbitrary, and many such infarctions are of uncertain clinical relevance. The occurrence of hospitalizations for failure is actually far more objective than many myocardial infarction. But heart failure events comprise a small number of events in the absence of meaningful LV systolic dysfunction.

    Alex says that “cardiac death” is not enough., but I strongly disagree. Look at the data on implantable cardioverter defibrillators, which reduce cardiac death — without reducing myocardial infarction. Why is that not good enough?

    A researcher can rely on any single or composite endpoint as long as it is clinically meaningful to patients who have the disease being studied. An intervention that reduces cardiac death is very meaningful.

    I also disagree with Larry’s perspective. He thinks that having an invasive procedure for coronary artery disease turns a patient into a “heart” patient. But these patients were “heart” patients before their procedures, and their procedures were performed because they had cardiovascular disease. If the procedure were not performed, they would still have cardiovascular disease, i.e., they would still be “heart” patients. I know of no evidence that patients who know they have coronary artery disease become emotionally traumatized because they underwent a percutaneous revascularization procedure. I do know that — for many patients — concerns about their coronary artery disease are reduced following PCI, rightfully or wrongfully.

    None of my comments have anything to do with the merits or limitations of the Navarese meta-analysis. But in general, a meta-analysis whose upper bound of the 95% confidence interval approaches 1.0 should not be considered to be persuasive, especially when there are uncertainties about the completeness of follow-up for the primary outcome measure in the individual trials.

    1. I think Milton, like many physicians, may not fully appreciate the life-changing impact of a major medical procedure on people who are not fully immersed in the medical system all the time. Nevertheless, this is an interesting and worthwhile discussion to pursue, perhaps at some other point.

      More importantly, I find it highly interesting that nothing Milton writes contradicts the main message that both Alex and myself wanted to make, which is that this meta-analysis should NOT be used to overturn the negative findings of ISCHEMIA. Instead of focusing on the more technical issue of the endpoint I would love to learn what Milton really thinks about the relative value of revascularization versus medical therapy in patients with stable CAD.

    2. Very insightful comments by Dr Packer. I believe this discussion is the core of the subjects. Should CV or all cause mortality be the primary endpoint. From my perspective in open label trials all cause mortality should be the norm since differential mis classification bias probably be present. Furthermore, if an intervention lowers cardiovascular mortality and increases non-cardiovascular mortality it should not be used. A patient receives a treatment either to avoid symptoms or to decrease his/her probability lf death, not his probability of cardiovascular death. This applies also to screening procedures. An annual cardiac CT scan of may decrease CV mortality but will probably increase cancer mortality. Does that mean we should use it ? Overall mortality is a clean outcome and I can’t find strong arguments not to use it.

  4. I disagree with Larry once again — on two important points.

    Larry thinks that physicians may not fully appreciate the life-changing impact of a major medical procedure on patients. He thinks that percutaneous revascularization is emotionally turbulent for patients. Really? How does he know that? In his recent post, he wants to delay this debate until some future date, but why? He has raised the point that revascularization carries a major negative emotional risk, which should be considered in current decision-making. If he has evidence for that risk, he should share it.

    Larry also misses the whole point of the Navarese paper. The whole point of doing a meta-analysis is to go beyond the information that can be provided by a single trial. Meta-analyses are designed to look at endpoints that cannot be reasonably evaluated in a single trial, typically due to a lack of statistical power.

    Unfortunately, Larry insists on looking at the data in a binary way, i.e., ISCHEMIA was “negative”, and the meta-analysis is “positive”. Therefore, in his view, the meta-analysis contradicts the findings of ISCHEMIA. That way of thinking is entirely inconsistent with the whole purpose of doing a meta-analysis, and it is certainly not the way that we evaluate the ongoing accumulation of scientific evidence. There is absolutely nothing in the meta-analysis that is inconsistent with the findings of ISCHEMIA.

    The whole debate is whether the Navarese meta-analysis can teach us something valuable about the role of revascularization in stable coronary artery disease. Larry thinks that my personal opinion and choices matter, but he is wrong. The quality and the quantity of the evidence matters. And that is what Larry — and everyone — should be focused on.

    So I will ask Larry a simple question: What finding in the Navarese meta-analysis is inconsistent with the findings of the ISCHEMIA trial?

  5. 1. Milton, I promise to elaborate on my point about the important subjective effects of medical procedures in the near future. But I think that will require a separate and longer blog post, so let’s save that for the future.

    2. I am also happy to go further into the details of the meta-analysis, but I’d first like some clarification from Milton. In his original comment Milton hinted that he’s not entirely satisfied himself with the MA. So before getting into the weeds I’d like to know exactly where he stands on the value of the MA. I suspect, though I may be wrong, that we are actually far closer to agreeing that the MA is not terribly informative or useful.

  6. It’s a significant concern to me that patients are still being led to believe that by finding an area of stenosis and stenting it, their cardiologist has prevented a heart attack. It is not always the case that the person is already a heart patient — I’ve met people who underwent a stress test followed by angiography even though they were completely asymptomatic. The testing was ordered on the basis of them having risk factors. These patients believe that the cardiologist, by placing a stent, has prevented them from having a heart attack. Perhaps in some cases some good will occur in that they will then be more intensively treated with medications, but isn’t there a better way?? Can’t they receive appropriate treatment to control their risk factors without having that piece of metal inserted into their body?

    1. Marilyn, this is such an important point! And it so often is missed. In fact, despite the level of sophistication in the cardiology community, when you actually talk to most people who have been stented they almost always believe that they have undergone a life-saving procedure. I wonder where they got that idea from?

Leave a Reply