September 1st, 2014
Let’s Take a Close Look at PARADIGM-HF
Many have hailed the novel compound LCZ696 as the next great advance in heart failure (HF) treatment, on the heels of findings from the newly published PARADIGM-HF trial. I think this is too generous an interpretation.
LCZ696 is itself a combination of two drugs: valsartan (an angiotensin-receptor blocker) and sacubutil (an inhibitor of neprilysin). During a median follow-up of 27 months, LCZ696 showed superiority to enalapril with respect to the primary composite endpoint (death from cardiovascular causes and hospitalization for HF) and all-cause mortality — with a number needed to treat of 35. (These findings are likely exaggerated, given that the trial was stopped early, but that is not my primary concern with this study.)
Here are three reasons to doubt that PARADIGM-HF sets a new standard both for treatment of HF and for randomized comparative-effectiveness studies:
1. Drug Dosing
In PARADIGM-HF, oral enalapril was dosed up to 10 mg twice daily, whereas LCZ696 was dosed up to 200 mg twice daily (which includes a cumulative 320 mg of valsartan). The problem is that 320 mg is the maximum HF dose of valsartan per drug labeling, but enalapril can be dosed up to 40 mg daily (20 mg twice daily) — double the maximum dose proscribed per protocol.
The actual administered mean dose of the two drugs was 375±71 mg for LCZ696 and 18.9±3.4 mg for enalapril. Not surprisingly, given that the LCZ696 group received more angiotensin-pathway blockade, mean systolic blood pressure was lower in that group (3.2±0.4 mm Hg lower than in the enalapril group).
In effect, drug dosing in PARADIGM-HF was a “straw man” comparison. The reported outcomes may be entirely a consequence of more ARB versus less ACE inhibitor. That is reason enough to doubt the findings. Sacubutil, the novel drug, could have been a sugar pill, and the results may well have turned out the same. But there are two more good reasons to be skeptical.
2. Drug Run-In Periods
Drug run-in periods are a controversial study-design choice — one that I believe must unequivocally be abandoned. In PARADIGM HF, patients with systolic HF were asked to stop the ACE inhibitor or ARB they were already on and take oral enalapril 10 mg twice daily for 2 weeks. Of the 10,513 patients who entered the run-in phase, 1102 (10.5%) discontinued treatment at this point. Then the remaining 9412 patients took LCZ696 for more than twice as long (4 to 6 weeks), during which 977 (10.4%) patients discontinued treatment and 8442 underwent randomization.
The reason why drug run-in periods are problematic is discussed at length in the literature. In short, run-in periods exclude intolerant and nonadherent patients, foster spuriously large treatment effects, and (most troubling) create inclusion criteria that are irreproducible — i.e., that apply to no population we can clearly describe, as reasons for dropout are multifaceted and unique.
Even more concerning is that drug run-in periods test a different question than the one we think we are testing. In PARADIGM-HF, the run-in tested whether sticking with LCZ696 or switching to enalapril is better for HF patients who have taken and tolerated enalapril followed by LCZ696. It effectively turns the trial into a withdrawal study. If stopping LCZ696 is harmful, that counts against enalapril.
3. Stopping What Already Works
All PARADIGM-HF participants were already on an ACE inhibitor and an ARB, the standard of care for systolic dysfunction supported by several randomized trials — and they were asked to stop. Those of us who care for HF patients know quite well that there often is a very good reason why patients end up taking the ACE or ARB they are currently using. It makes little sense to stop what is working to enter this trial.
Here are some better options for how PARADIGM-HF could have been conducted:
- Option A: Compare the novel combo drug with valsartan alone (same dose of valsartan) among patients with newly diagnosed HF.
- Option B: Test sacubutil (the novel drug) against placebo as add-on therapy for patients with HF who are already on an ARB. This is the cleanest option.
- Option C: Randomize patients with HF to either continue their current therapy (and ramp up the dose to the maximal drug-label dose) or switch to the combination drug.
The Real Lesson of PARADIGM-HF
The response to PARADIGM-HF has been positive partly because it has been a long time any new drug has been shown to improve mortality in patients with HF. But our desire to find better options for these patients does not mean we should lower the standards for what counts as a good trial. Few providers would consider a randomized trial comparing valsartan at 320 mg with enalapril at 20 mg to be fair — and adding a novel drug to the valsartan does not make this comparison more palatable. Instead, PARADIGM HF shows that with a large number of patients and the “right” trial design, you can get any conclusion you want. As William Randolph Hearst said, “You furnish the pictures, and I’ll furnish the war.”
PARADIGM-HF does convince me that more angiotensin-pathway blockade is better, but it does not prove the benefit of a novel drug. A history of negative HF trials should teach us that the pretest probability that a novel drug works is very low in HF. The data from PARADIGM-HF do not greatly alter the probability that the drug works.
The concerns I have highlighted cannot be remedied, even with full access to the study data (which should nevertheless be made available) — instead, they are hard-wired into the trial design. For that reason, we must ask for another randomized trial — one that is truly done right. Heart failure patients deserve no less.
From the editors:
This post generated a lot of discussion. To facilitate reader navigation of the many issues addressed, we are presenting that discussion in a separate post (edited to better showcase the substantive points of members who offered their insightful comments). We hope you will help to continue this important conversation.
To view all of our coverage from the ESC meeting, go to our ESC.14 Headquarters page.