I am torn on the PolyPill concept

Trials comparing the polypill to standard care need to be rigorous. The POLY-HF trial is not. But I am still not sure the polypill is a bad idea.

Many conditions in cardiology require multiple medications, for instance, high blood pressure, diabetes, and congestive heart failure.

Heart failure with reduced ejection fraction is a good example: clinical trials have shown that 4 drug classes improve important outcomes. Three of the four classes reduce mortality, and the fourth (SGLT2i) reduce a composite of cardiovascular death and hospitalizations for heart failure. It is clearly best if a patient can take all four drugs. (The other three HF drug classes: ACE, ARB or ARNI, beta-blockers, mineralocorticoid receptor blockers).

But 4-drug therapy is not easy. For starters, these are usually not the only medications these people take. Even if it were, getting four prescriptions filled, and remembering to take them every day, is work.

One of my rules of patient care is that I always seek the least disruptive way to care for people. Taking four separate drugs is a disruption. Yet trials show that if your patient is like those in the trials—youngish, ambulatory, solid blood pressure—then it is worth dealing with the hassle.

Enter the polypill: what if you could take one tablet that had multiple drugs in it? This reduces the work of being a patient and it may even increase adherence to beneficial drugs.

The downside of course is that you reduce the ability to dial in the best dose of the specific medications, and you may increase the risk of the most common side effect—low blood pressure. (Another Mandrola rule: low blood pressure is almost always worse than high blood pressure.)

When I was younger I shunned the polypill concept. Aging and taking a few daily medications has opened my mind to reducing the work of being a patient.

The Poly HF Trial

Investigators from two Texas hospitals studied a polypill in an RCT called the Poly HF trial. One group gets a polypill containing metoprolol succinate, (in four possible strengths), spironolactone 12.5, empagliflozin 10). The other group gets something called “enhanced” usual care. The enhanced part is important, and I will come back to it.

The authors screened nearly 5000 patients for eligibility in the EHR. They invited 474 to be in the trial and 212 were actually enrolled.

As it is in nearly all HF trials, enrolled patients were young (53 years), mostly male and had a baseline SBP of > 120 mmHg. The baseline LVEF was 25%.

Both groups of patients received sacubitril/valsartan—not part of the polypill.

The primary endpoint was LVEF at 6 months assessed by MRI. There were many secondary endpoints. They also calculated a win-ratio-based on many composite endpoints.

Results

• The primary endpoint was positive. LVEF was 40.4% vs 37.1% in the polypill vs control arm. The 3.3% difference had 95% confidence intervals of 0.2-6.4% and the p-value was 0.039.

• Secondary outcomes also favored the polypill. LVEF by echo (45.3% vs 42.0%). Win ratio = 1.72 (higher number favors polypill);

• HF hospitalizations or ED visits 37 per 100 patient years vs 86 per 100 patient years (rate ratio 0.40; 95% CI 0.40-0.88).

• KCCQ-OSS (quality of life) at 6 months was 71.8 vs 63.3. Diff: 8.5 pts (2.6–14.4).

• Medication adherence measured by blood levels favored the polypill.

• Adverse events were very low and not different in the two treatment arms.

The authors concluded

The polypill was well tolerated, with fewer adverse events with polypill treatment as compared to enhanced usual care. A polypill for heart failure was associated with a significant improvement in cardiac function as compared with enhanced usual care.

Comments

Let me say first that RCTs are the way to study polypills. Observational look-back studies are likely to be confounded by selection bias. You’d never know if the polypill users were healthier than non-polypill patients.

It’s also notable that the study enrolled a challenging population of people who were mostly uninsured, and nearly half were unemployed. These patients are underrepresented in typical clinical trials.

Yet there are serious limitations of this trial that preclude translating this data to polypill acceptance.

First and foremost, this is a highly selected small group of patients. The authors reviewed nearly 5000 health records—I assume of heart failure patients to enroll 100 patients per arm. This, I think, is one of the major limitations of polypill data: namely, that you have to be highly selected to be eligible for a polypill.

The second issue is the choice of endpoint. Left ventricular ejection fraction is both a weak estimation of cardiac function and a surrogate measure. Weak in the sense that—even with MRI—EF can vary a few percent based on many things, including the mood of the reader. Mostly, though, LVEF is a surrogate measure in that people care more to be alive and out of the hospital than a number on their chart. (Side note: a normal EF is not 100% but more than 50-55%).

What’s more. the difference in EF in this trial was only 3%—a number that reached statistical significance, but it’s not clinically significant.

The small difference may have been because the “enhanced” care part of the control arm did so well. Adherence in the standard care arm was shockingly high at 78% whereas it is more like 15% in real world data.

The teaching points here are that when you measure something like an ejection fraction you can enroll fewer patients, but then important clinical outcomes are underpowered—for instance, there was only 1 death in each arm.

The reason why a continuous outcome (LVEF) requires fewer patients is that each patient gives you a number on a continuous scale—which can then be compared with statistical tests. But doctors (and patients) care much more about hard outcomes like death or hospitalization. For these, most patients don’t have an event so you need more patients to discern differences in treatment effect.

As for the many secondary endpoints, I would discount most of them. The quality of life measures are marred by the open-label nature of the trial. The other secondary endpoints are few in number and the statistical testing was not corrected for multiple comparisons.

My summary

In this tiny, open-label trial, of super-selected patients, the polypill performed pretty well against an enhanced control arm.

It might do even better against standard every day care, but to know this, I would want larger multicenter trials that enroll a much less selected group of patients.

But I remain open to the concept of the polypill because it reduces the work of being a patient.

---

I am interested in your thoughts about the polypill concept. JMM