When Evidence Doesn't Persuade

I had the evidence but lost the debate. To be fair, my opponent crushed me. This is a story about evidence and its inability to guide medical decisions.

My friend, a thoughtful interventional cardiologist, and I debated percutaneous coronary intervention (PCI) in patients with stable coronary artery disease (CAD).

The setting was the annual staff meeting at our hospital. The audience included experienced clinicians from all areas of medicine and surgery. I figured that since my colleagues were noncardiologists, they could serve as an impartial jury.

Using evidence, I made the case that PCI in patients with CAD is cardiology's biggest blind spot. I tried to convince my colleagues that the clogged-pipe theory^[1] of treating coronary disease is wrong.

My opponent and I agreed that the debate would not pit PCI vs medical therapy but rather whether stents in addition to medical therapy adds benefit. Everyone agrees that patients with coronary disease benefit from optimal medical therapy (aspirin, statins) and lifestyle changes.

The Evidence Against PCI in Stable CAD

My case began in the 1990s with the plain old balloon angioplasty trials. These showed angina improvement but no reduction in myocardial infarction (MI) or death.^[2,3]

Bare metal stents made PCI safer but did not reduce MI or death. I made special note of the MASS-II trial^[4] because it was a serious test, comparing PCI with stents to bypass or medical therapy in patients with multivessel disease, 90% of whom had the widow-maker lesion — in the proximal left anterior descending (LAD) artery. Overall survival did not differ among the three strategies, and, again, PCI did not reduce the rate of MI.^[5,6]

In the landmark COURAGE trial,^[7] investigators compared the strategy of optimal medical therapy vs optimal medical therapy plus PCI in more than 2200 patients. They found no difference in death or MI.

Since 2007, the authors of COURAGE have published numerous substudies looking for any group of patients that may benefit from PCI in addition to medical therapy. They found none: not those with positive nuclear studies,^[8] not those with many blockages or more myocardium in jeopardy,^[9] not those in the medical group that crossed over to the PCI group,^[10] and not even those with proximal LAD lesions or three-vessel disease and weak heart muscle.^[11]

COURAGE authors also published extended follow-up (median, 6.5 years) from the main trial;^[12] PCI still did not reduce MI or death compared with optimal medical therapy.

Critics of COURAGE could say it's just one trial and it used bare-metal stents. No. Other large studies and meta-analyses have shown no incremental benefit for PCI (even with drug-eluting stents) added to medical therapy in patients with stable CAD.^[13,14,15]

What about angina relief? Reducing a blockage from 90% to 0% has to help relieve chest pain. Indeed, until 2017, many studies showed that stents relieved chest pain.^[15 ] The problem, however, is that that relief wanes after 3 years.^[15,16]

The fatal flaw with all of these trials is that they were unblinded. Angina is subjective, so knowing your treatment assignment matters. Faith can improve subjective symptoms.^[17]

This is why the audacious ORBITA trial is so important.^[18] Primary investigator Rasha Al-Lamee, MD, and her team at Imperial College London, United Kingdom, randomly assigned 200 patients with single-vessel CAD to a real PCI or a placebo (sham) procedure. During the procedure, patients wore headphones and did not know if they received the stent or whether the blockage was left alone. After the patient left the lab, none of the doctors knew the treatment assignment. Such strict blinding equalizes the placebo effect.

The findings were shocking. PCI completely relieved ischemia when measured on stress testing, but this improvement in blood flow did not lead to the large differences in angina symptoms between the two treatment groups that PCI proponents had expected. There were also no significant differences in the change in exercise time (the primary endpoint), the physician rating of the degree of angina, or quality-of-life scores.

The Opposition

At this point, I felt that the evidence made a compelling case. I was wrong.

My opponent spent little time refuting the outcomes studies. Instead, he showed pictures of scary-looking blockages. "What if this was your LAD?" He told anecdotes of patients with unstented 90% blockages who came to need an urgent surgery, say a hot gall bladder. He spoke of the transition of 90% blockages to chronic total occlusions. Better to fix these things early, he warned.

While he did this, I looked at the audience. Their faces signaled my defeat.

As he told the scary anecdotes and proposed the possible benefits of fixing people before trouble occurred, my mind wandered to the superimposable Kaplan-Meier curves in the trials. I thought about the consistency of the results in the meta-analyses. I thought to myself: Why don't my colleagues realize that anecdotes even out in trials of thousands of patients? The whole reason for doing randomized controlled trials is so we are not fooled by our own biases.

The Verdict and Its Consequences for Evidence-Based Medicine

In the end, a show of hands confirmed my defeat. When presented with a picture of a stable proximal LAD lesion and asked whether they would want medical therapy alone or a stent, nearly the entire audience—my impartial jury—voted for the metal cage.

I tweeted news of my defeat, lamenting that evidence doesn't persuade—even physicians. A young doctor's response reflected the view of many senior leaders: "I think we need to be humble about our approach to stable CAD—BOTH pro- and anti-stent."

That comment reveals a core problem with evidence-based medicine: To wit, if a treatment—with risk and cost—does not improve outcomes and does not relieve symptoms more than placebo does, how can it have a pro side? What's more, if a clinician's gut feeling, emotion, and expertise can trump evidence, why bother doing trials?

The scientific method calls for testing a null hypothesis. Here the null hypothesis has been tested (in over 61 trials^[19]), and not a shred of evidence falsifies it.

I understand clinical judgment. No doubt there are outlier patients different from those included in the clinical trials. And ORBITA enrolled only patients with single-vessel disease; maybe fixing lesions in two arteries will improve exercise time by more than the nonsignificant 16 seconds they observed.

My point in telling the story of how evidence failed to persuade is that large swaths of doctors and patients still believe in the clogged-pipe theory of fixing CAD. When my thoughtful interventional partners refuse to "fix" a severe lesion because the patient has not had meaningful medical treatment, they face consternation from referring doctors and patients.

Other behaviors signal widespread disbelief in the evidence. One is the upgrading of symptoms to unstable angina. Atypical noncardiac pain at rest can easily become unstable angina in the presence of a 90% mid-circumflex lesion. Then there is the sedentary 380-pound man with new-onset shortness of breath walking up stairs. No, he's not breathless because of obesity; he's breathless because of the 90% distal right coronary artery lesion. We must fix that blockage.

These examples show the truth: Many people ignore the evidence and cling to the notion that stents are beneficial in patients with stable CAD. Anecdotes and angiograms are more believable than scientific evidence.

Finally, shared decision making and patient choice are overblown in the case of PCI. Shared decision making is necessary for preference-sensitive decisions: for example, live with persistent atrial fibrillation or take the risks inherent with rhythm control strategies (drugs and/or ablation). In this case, we seek patient choice because of uncertainty; we don't (really) know whether outcomes are better with rhythm control. Another example where patient choice is crucial is the primary prevention implantable cardioverter defibrillators (ICD). Here we know outcomes are better with the ICD, but the device comes with significant burdens.

PCI in (most) cases of stable CAD is different. We have no reason to offer it on the basis of reducing death or MI. And to offer PCI for angina relief means ignoring one of the most unassailable trials of our generation. And even if you doubt ORBITA, the angina relief from PCI observed in the unblinded trials wanes over time.

Before this debate, I knew evidence did not matter in politics. But I thought that since doctors were trained in science and reason they could be persuaded by evidence.

Maybe the German physicist Max Planck was correct when he implied science only advances one funeral at a time.

Comments

Commenting is limited to medical professionals. To comment please Log-in.

Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.