The EARLY TAVR Trial Exposes the Main Tension in Medical Science

The Study of the Week explores yet another biased study design

We learn in high school that the purpose of science is to ask and answer questions about nature--through experiments. Scientists don’t know the results of an experiment beforehand, so, any result, negative or positive, adds knowledge.

Medical science should work the same way. Medical scientists would design and conduct experiments to answer important questions, say, whether a drug or surgery, improves an outcome. The outcome would not matter because knowledge is gained from either a negative or positive result.

Perhaps the biggest impediment to good medical science is that it is conducted by humans. Human nature holds that we respond to incentives.

The largest incentive in medical science is getting a “positive” result. Then, we can say a treatment worked. Everybody wins when this happens—the makers of the treatment, scientists, doctors and patients alike.

The tension in medical science comes between a) asking important questions and being happy with the knowledge that comes forth, or b) finding positive results.

These don’t have to be in opposition. In days of old, before there were established treatments, medical experiments often asked important questions and found positive results. The ACE-inhibitor trials in patients after myocardial infarction, for instance, were well conducted and positive.

Yet the more that we discover, the harder it is to align the matter of asking important questions and getting positive results. Progress increases the risk that conducting a proper experiment will yield a negative result.

---

A recent trial of two medical strategies demonstrates the growing problem of designing a medical experiment not to answer an important question but deliver a positive result.

The EARLY TAVR trial investigated two approaches to the common problem of managing patients who have severe aortic valve stenosis but no symptoms.

Quick background on aortic valve stenosis: The aortic valve lets blood out of the left ventricle to your body. If you are lucky enough not to die at a young age, you may develop thickening of the aortic valve leaflets, mostly due to calcium deposits.

Gradually, the valve leaflets thicken and lose their ability to open. The opening gets smaller; think putting your thumb on a garden hose. As the disease progresses, the opening can become like a pinhole. Patients then develop chest pain or breathlessness or fainting.

The good news is that surgeons can replace that valve. More recently, my colleagues in structural cardiology can put a new valve in through the leg. We call this transcatheter aortic valve replacement, or TAVR. There is a great debate about which patients do better with each procedure. But that is a question for future posts.

---

The question tested in EARLY TAVR is what to do with patients whose aortic valve looks terribly diseased on ultrasound, but they report no symptoms.

The old practice (control arm) was to wait until symptoms develop before recommending surgery. Clinical surveillance of asymptomatic patients was an easier sell when patients faced open heart surgery and months of recovery. TAVR now offers a far less invasive procedure and has changed the timing-of-intervention decision.

The other problem with clinical surveillance was that a (very) small number of patients with severe but asymptomatic disease died suddenly.

---

In recent years, two small surgical trials comparing surgery vs clinical surveillance in asymptomatic patients with severe aortic stenosis reported positive outcomes for early surgery. But these were small trials with small numbers of events.

The EARLY TAVR trial, which was sponsored by the makers of one of the two TAVR valves, asked the question of whether intervening on asymptomatic patients with a transcatheter approach would be superior to surveillance.

Recall that these are patients who have no symptoms. Trialists even checked their non-symptom status with a formal exercise test.

Before I tell you what happened in the trial, stop and think how you would answer the question of which is a better approach to patients who complain of nothing.

The two most worrisome outcomes of patients who have aortic valve stenosis or aortic valve surgery are death and stroke. A neutral medical scientist therefore might compare rates of stroke and death in the two strategies.

But that was not done in EARLY TAVR.

Instead, the authors chose to measure a composite primary outcome of stroke, death, and “unplanned hospitalization for cardiac reasons.” The latter component included hospitalization for having a TAVR.

Now think about the trial procedures. Patients who had no symptoms, but knew they had a severe valve problem were put in a trial. One arm got fixed. The other arm did not; instead, they were “clinically surveilled.”

EARLY TAVR delivered highly positive results. A primary end-point event occurred in 26.8% of patients in the TAVR group and in 45.3% of patients in the clinical surveillance group (hazard ratio, 0.50; 95% confidence interval, 0.40 to 0.63; P<0.001).

The components of the primary endpoint tell the story: I made two slides, which I showed in Brazil two weeks ago:

The upper left is the primary endpoint. Upper right is death—no difference. Bottom left is stroke—also not significantly different. Lower right is unplanned hospitalization—same as primary endpoint.

Now look at the slide with just the primary endpoint and unplanned hospitalization:

It is the same. There is immediate separation of the curves and then the curves are parallel.

Here is what happened: patients randomized to the surveillance arm had what Christopher Rajkumar and colleagues termed “subtraction anxiety.” The early surgery group were, as they write, “faith healed.”

Because the experiment was unblinded, patients in the clinical surveillance group knew they had severe disease, and knew they were not “fixed.”

This resulted in the rapid development of symptoms; strangely, right after passing a formal exercise test.

---

EARLY TAVR made news and was held up as a reason to intervene early in patients with severe asymptomatic aortic stenosis. Of course, this means, doing more procedures and selling more transcatheter aortic valves.

This takes me back to the tension of designing experiments to answer important questions or to get positive results.

EARLY TAVR looks to be in the latter category. They found no difference in hard endpoints. Instead, the main driver of “positive” results was the conversion of asymptomatic to symptomatic patients—largely, because they knew they had severe disease and were unfixed.

I wrote about this trial for doctors over at theHeart.org | Medscape Cardiology.

My final sentence:

Patients were experimented on, money and time were spent, and little was learned. Yet practice patterns will probably change. 

Somehow, we need to change the incentive structure for science.

One possible change is to consider the funding of science. I am not sure it is a great idea that those who make the device also design and conduct the trials to show its efficacy.

I could be wrong about this; but would it not be wiser that public funds go into testing device or drug efficacy? I am asking; I do not know.

Of course, another potential solution is that users of medical evidence equip themselves with enough skills in critical appraisal so that they are not easily bamboozled by such trials. (This, in fact, is one of the goals of Sensible Medicine). JMM

---

Thanks for your support. Recall that Sensible Medicine remains a place where critical appraisal can occur without the tension of an advertising model. This happens because of your support. Thank you.