COMPASS: Don't Pop the Champagne Yet

The positive results of the COMPASS trial,^[1] a comparison of two doses of low-dose rivaroxaban (Xarelto, Bayer/Janssen Pharmaceuticals) for secondary prevention in patients with vascular disease, made big news at the European Society of Cardiology 2017 Congress .

The makers of rivaroxaban and the clinical trialists won big. Patients, not so much.

For those who conduct clinical trials, victories hardly come bigger. They earned top billing in the world's biggest medical meeting, publication in the New England Journal of Medicine, and "sincere" congratulations^[2] from no less than Dr Eugene Braunwald (Brigham & Women's Hospital, Boston, MA), who was the discussant for the presentation.

Bayer/Janssen Pharmaceuticals, the makers of rivaroxaban, also reaped great treasure. A study of more than 27,000 patients carried out in 602 centers in 33 countries must have cost a lot. But now they have a new formulation of a drug that delivered impressive P values and could be used in 300 million patients.

But . . . even if this trial passes FDA muster, I would hold off the celebration. A more realistic view of the COMPASS results and their translation to clinical practice looks mundane.

COMPASS Results in Absolute Terms

When compared against aspirin alone in patients with vascular disease, the combination of rivaroxaban 2.5 mg twice daily plus aspirin reduced the rate of the composite primary outcome of cardiovascular death, stroke, or MI by 1.3%. The cost of this gain was a 1.2% increase in major bleeding. Rivaroxaban used alone at a dose of 5 mg twice daily did not significantly reduce events compared with aspirin alone.

Here are three other key messages attendees heard from the podium at the ESC meeting and from the COMPASS paper:

In addition to significant reductions in the composite end point, the rivaroxaban-plus-ASA combination reduced two other key composite secondary end points by 1.3% and 1.4% respectively, as well as lowering all-cause death by 0.7%.
Although major bleeding was higher with rivaroxaban/aspirin, fatal bleeding did not differ.
The rivaroxaban-plus-ASA combination reduced the net-benefit surrogate of cardiovascular death, stroke, MI, fatal bleeding, or symptomatic bleeding into a critical organ by 1.2%.

Summary

This is a classic case in which a trial with large numbers of patients produces extremely low P values but small differences in absolute risk reduction.

A 1.3% percent reduction in the composite end point translates to a number needed to treat of 76. Yes, if you multiply that gain to a population of 300 million people with vascular disease, you can show great benefit to populations. Regular doctors don't treat populations. We treat patients, and a gain of 1.3% means more than 98% of people who take this combination get the same result as those who took aspirin alone.

What's more, COMPASS was terminated early, and as the authors write in the discussion, "trials that are stopped early for efficacy may overestimate the treatment effect."

As it is with any clot-inhibiting drug, reducing vascular events comes with increased bleeding. Here, the 1.2% increase in major bleeding nearly exactly balances the gain. The authors downplay this because fatal bleeding was not increased.

That sort of interpretation approaches spin. In the supplement, the authors write that adjudication of fatal bleeding is "deemed by the investigator" (hundreds of them), and "bleeding leading to hospitalization or presentation to an acute-care facility with discharge on the same day relies on investigator judgment as to whether the bleed prompted the presentation." My take is that once one gets past the category of major bleeding, parsing out details in a trial with more than 600 centers is tough.

For example, spin can go both ways. If I wanted to spin the bleeding results to look worse, I could point to the subgroup analysis of bleeding (Figure S2 in the supplement), which showed major bleeding with the rivaroxaban-plus-aspirin combination was more than double in patients older than 75 years.

The net-benefit calculation also slants to favor the active arm. The calculation includes three ischemic events and only two bleeding events. Why, for instance, doesn’t bleeding requiring a hospital admission or transfusion count in a net-benefit calculation? These days, with our strict rules governing transfusions, bleeding that results in a blood transfusion is significant.

The final reason, the strongest reason, that I see these results as mundane comes from my point of view as a clinician.

The average patient in COMPASS was 68 years old with established vascular disease. In addition to aspirin, about 70% of patients in this trial took beta-blockers and ACE inhibitors; more than 90% took statins. That's a lot of pills, and it does not count the pills patients take for other conditions.

COMPASS proponents now want us to add another expensive (and you know it will be expensive) twice-daily drug to the concoction. This is not sustainable. Not for these benefits and harms.

This critical appraisal should not be taken as a knock against the trialists or the makers of rivaroxaban. Secondary prevention of heart disease in 2017 is hard. We are pretty darn good at it already. The really big gains in cardiology in the future will not likely come from fixing established vascular disease; it will come in preventing it.

Don't look for an expensive drug or device to do that.

JMM

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.