How a “Positive” Trial Can Add Little to Clinical Care

The Study of the Week looks again at another positive trial. The DAPA-MI trial is a great example of why consumers of evidence need to look deeper than the top-line results.

Today I will tell you about a trial that delivered a positive result but will add little benefit to patients.

The teaching lesson surrounds the choice of endpoint to measure.

First some brief background.

When I started cardiology, much research focused on improving the care of people who had myocardial infarction (MI). Post-MI care was a target because injury to the heart from an occluded coronary led to bad things—like heart failure and arrhythmias.

Times have changed. A lot less research goes into post-MI care. That’s because my interventional cardiology colleagues open acutely closed coronary arteries in the cath lab. Doing so stops the heart attack and prevents damage to the heart muscle. We call this acute percutaneous coronary intervention or PCI.

Of course, not every patient who has an MI and acute PCI suffers zero heart damage. Sometimes PCI reduces the damage. In these patients, scientists continue to study ways to improve outcomes.

Swedish and UK researchers had the idea that the drug dapagliflozin might improve outcomes in patients who have an MI big enough to cause mild heart damage.  

Dapagliflozin is one of the drugs in the sodium-glucose cotransporter-2 inhibitor (SGLT2i) class. These drugs have shown benefit in three conditions—diabetes, heart failure due to reduced heart function and chronic kidney disease.

The trial is called the DAPA-MI trial. And it was simple in design, at least initially.

About 4000 patients who had MI and no other reason to take SGLT2i were randomized to dapagliflozin or placebo. This, of course, is in addition to the 3 or 4 other indicated medicines that post-MI patients take. So… the question was: does dapagliflozin—in addition to standard care—improve outcomes.

The focus of this column comes in how to measure these outcomes.

The standard in cardiology is to measure something called MACE—or major adverse cardiac events. These can vary but generally include MI, stroke, heart failure event, cardiovascular death (CVD) or death due to any cause.

That was the original plan for the DAPA-MI trial. But things changed. And this is the teaching point.

DAPA-MI investigators made estimates of how many events they would see. Their original events of interest included CVD and hospitalization for heart failure (HHF). The estimates determine the number of patients to include in the trial.

Soon into the trial, however, they learned that patients who have MI have a lot fewer events than they expected.

This is a huge problem for trialists. Because too few events makes it impossible to know if any observed differences are due to noise or actual signal.

The investigators now had three choices. They could a) forget the trial, b) enroll tons more patients (which is very pricy), or c) change the primary endpoint.

Changing the endpoint really means expanding the endpoint to include more events. More events to count helps sort out signal from noise.

They decided to expand the endpoints—of course, without looking at the data.

What they chose may be one of the bulkiest endpoints I have ever seen.

This now included a hierarchical composite endpoint of death, hospitalization for heart failure, MI, atrial fibrillation, new diagnosis of diabetes, the actual class of heart failure at last check, and a body weight decrease of 5% or greater at the last visit.

I have never seen such an endpoint. Obviously, these outcomes vary in importance. Weight loss is nice but it is a lot less important than death. Because of these differences in importance, the authors chose to analyze the outcome with something called a win-ratio method.

For the win-ratio analysis, matched pairs are created between two arms, then each strategy is assigned a win or a loss per pair, starting with the worst endpoint, death, then moving down the list sequentially if the previous event did not occur.

Win-ratio is a way to use all the data, but its downside is that it’s hard to apply clinically.

The Results

The analysis of the primary hierarchical composite outcome resulted in significantly more wins for dapagliflozin than for placebo (win ratio, 1.34; 95% confidence interval [CI], 1.20 to 1.50; P<0.001)

The problem was that hard clinical endpoints like death, CVD, HHF, MI, and even all-cause hospitalization were not different.

The primary endpoint was driven by softer endpoints like new diagnoses of diabetes and weight loss—both of which are way down the hierarchical scale.

Comments

DAPA-MI is emblematic of the main challenge for cardiology in the coming decades: it’s really hard to improve on what we have now, at least, in a cost-effective manner.

The patients in this trial were already on drugs that help reduce future outcomes. We call this guideline-directed medical therapy or GDMT. This cocktail of medicines plus the fact that most MIs are stopped quickly with PCI renders the rate of future bad things like heart failure or death very low. It’s hard for any new intervention to make low any lower.

The drug class of SGLT2i has been shown to reduce weight and improve glucose control. It is therefore not surprising that it reduced new diagnoses of diabetes and induced weight loss.

The teaching point is that DAPA MI is a positive trial. But it’s positive only because of very soft endpoints.

Given the drug’s high cost, and extra burden of another daily pill, we should not feel compelled to use the drug for this specific indication. For diabetes, yes. For heart failure, yes. For chronic kidney disease, yes. But not for patients after MI who have none of these three indications.

In sum, always look at the endpoints of a trial.

We want new therapies to make a substantial difference. New things should extend life or improve quality of life at a reasonable trade-off.

To know this, you have to look past the top-line results.