Case Study 28.2 — PARADIGM-HF: Why Preserving the Peptide Beat Supplying It

And why the combination was chemistry, not marketing

The setup

Read this case study after Case Study 28.1, because it only makes sense against that failure.

By the time the nesiritide question was settled, the position looked bleak for the natriuretic peptide system as a therapeutic target. The physiology was as compelling as it had been in 1981. The attempt to exploit it directly had produced an expensive drug that improved every measurement and changed no outcome.

The obvious conclusion — the system does not work as a drug target — was wrong. But you could not know that yet, and several people drew it.

What was available was a second idea, which had been circulating since the 1990s and had already failed once in a form nobody wanted to repeat.

Idea: stop throwing the peptide away

Neprilysin is a membrane-bound zinc peptidase, abundant on the kidney tubule brush border, that degrades natriuretic peptides among many other substrates. Inhibit it and the peptides the body already makes survive longer.

The logic is architecturally identical to the DPP-4 inhibitors of Chapter 7 §7.6: rather than injecting a hormone, block the enzyme that destroys the hormone and let endogenous production go further. Preserve rather than supply.

It has a built-in ceiling, and the ceiling is worth stating up front. You can only preserve what the body actually makes. A degradation inhibitor cannot push a receptor past physiological concentrations; a receptor agonist can. That is precisely why DPP-4 inhibitors produce a real but modest glycemic effect while GLP-1 receptor agonists produce a transformative one.

So the reasonable expectation was: preserving will work, and it will work less well than supplying. Hold on to that expectation. It is about to be inverted.

The first attempt, and why it failed

Neprilysin inhibition alone has a problem, and the problem is that neprilysin is not selective.

Its substrate list includes ANP, CNP, BNP, adrenomedullin, substance P, bradykinin — and angiotensin II. That last one is fatal to the simple version of the strategy. Angiotensin II is the peptide at the business end of the renin-angiotensin-aldosterone system: vasoconstriction, aldosterone release, sodium retention, remodeling. It is the thing heart failure treatment spends its whole effort suppressing.

Inhibit neprilysin on its own and you raise the peptides that help and the peptide that harms. You lift the brake and the accelerator at the same time, in a disease where the accelerator is already the problem.

The first-generation solution was to combine neprilysin inhibition with ACE inhibition in a single molecule — a "vasopeptidase inhibitor." Block the angiotensin-producing enzyme and block the natriuretic-peptide-degrading enzyme together. On paper this is elegant.

In practice it produced angioedema — swelling of lips, tongue, and airway, which is uncommon but can be life-threatening — substantially more often than an ACE inhibitor comparator, in a very large hypertension program. The mechanism is bradykinin. Both neprilysin and ACE degrade bradykinin. Block both and bradykinin accumulates from two directions at once. The compound was not approved.

That failure is the reason the successful drug is built the way it is, and it is why the label of the drug that eventually worked carries a contraindication and a washout requirement to this day. History is encoded in labels.

The redesign

Pair neprilysin inhibition not with an ACE inhibitor but with an angiotensin receptor blocker.

Now the arithmetic works. Inhibiting neprilysin raises natriuretic peptides, adrenomedullin, and other beneficial substrates. It also raises angiotensin II — but the receptor angiotensin II acts through is blocked, so the elevation has nowhere to go. And because an ARB does not touch ACE, bradykinin has one of its two disposal routes intact, so accumulation is not compounded.

Sacubitril/valsartan is that pairing: a neprilysin inhibitor prodrug and an angiotensin receptor blocker, co-crystallized into a single product. The class acquired a name, ARNI — angiotensin receptor–neprilysin inhibitor.

Say the important part plainly, because it is routinely misunderstood as a commercial decision: the combination is not two drugs bundled for convenience. Blocking the angiotensin receptor is what makes neprilysin inhibition usable at all, and it has to be a receptor blocker rather than an ACE inhibitor because of bradykinin. Every element of the design is a scar from a previous failure.

Note also, in a book about peptides: neither component is a peptide. Both are small molecules. The peptides in this story are the endogenous ones the drug protects.

The trial

PARADIGM-HF. 8,442 patients with chronic heart failure with reduced ejection fraction, symptomatic, with elevated natriuretic peptide levels, on background therapy including a beta-blocker. Randomized, double-blind.

The comparator was enalapril, not placebo. This is the detail that makes the result carry the weight it does. Enalapril is an ACE inhibitor with an established mortality benefit in this exact population. The trial did not ask whether the new drug beat nothing. It asked whether it beat a treatment already known to save lives.

Design features worth noticing. Before randomization, patients passed through sequential single-blind run-in periods on enalapril and then on the study drug; those who tolerated neither were not randomized. A run-in enriches a trial for people who tolerate the drugs, so the tolerability observed inside the trial is likely better than tolerability in unselected practice. Because both arms were run in, it is less clear that this inflates the efficacy comparison.

The result. The trial was stopped early, at a median follow-up of about 27 months, on the recommendation of its data monitoring committee, because the boundary for overwhelming benefit on cardiovascular mortality had been crossed.

• Primary composite of cardiovascular death or heart failure hospitalization: 21.8 percent versus 26.5 percent, hazard ratio 0.80 (95% CI 0.73–0.87)
• Cardiovascular death alone and heart failure hospitalization alone both moved in the same direction
• All-cause mortality 17.0 versus 19.8 percent, hazard ratio 0.84
• More symptomatic hypotension; less cough, less hyperkalemia, less renal impairment than enalapril
• Angioedema uncommon in both arms, numerically more frequent with sacubitril/valsartan, without a significant excess of serious cases

Guidelines were revised. The ARNI class moved into the position of preferred renin-angiotensin system inhibitor for heart failure with reduced ejection fraction.

The inversion, and four candidate explanations

Return to the expectation: preserving should work less well than supplying, because a degradation inhibitor is capped by endogenous production.

The opposite happened. Supplying failed outright. Preserving produced one of the more consequential heart failure results of the last twenty years. Four explanations are on the table, and no one can prove which is doing the work.

Timing and setting. Nesiritide was tested over hours to days, in acutely ill patients already receiving treatments that produce the same physiological effects. Sacubitril/valsartan was tested over years, in chronic disease, where remodeling — the slow structural deterioration natriuretic peptides oppose — has time to be affected. A disease-modifying effect cannot appear in 48 hours.

Physiological pattern. This is Chapter 3, and it may be the deepest of the four. A preserved endogenous peptide is still released by the body, on the body's schedule, in the tissues where it is made, in proportion to actual load. An infusion is flat, systemic, and indifferent to what the heart is doing minute to minute. In a system that already shows receptor downregulation and resistance (§28.2), that difference is plausibly decisive.

The one-peptide framing may simply be wrong. BNP is a relatively poor neprilysin substrate compared with ANP and CNP. Neprilysin inhibition raises several peptides at once, and attributing the benefit to BNP was never well founded. Possibly the trial succeeded because it changed a network rather than a molecule.

Different questions were asked. PARADIGM-HF beat an active comparator on a hard endpoint over years; the nesiritide trial added a drug to standard care and counted 30-day events. Part of the difference in answer is a difference in what was asked.

Only the second of those is really about peptides. The rest are about disease biology and trial design — which is the usual shape of these explanations, and a reason to be suspicious of anyone who explains a trial result purely in molecular terms.

Two things the trial cannot tell you

It cannot separate the two components. There was no arm receiving the ARB alone; the comparator was a different RAAS drug. The benefit is attributable to the combination. The mechanistic account above — that neprilysin inhibition raises natriuretic peptides and the ARB neutralizes the angiotensin II problem — is a well-supported inference, not a trial result.

Trials stopped early for benefit tend, as a class, to overestimate effect size, because stopping rules are triggered by favorable random fluctuation as well as by real effect. The direction of the result is not in doubt. The magnitude is plausibly a little generous.

And a third, which is about scope rather than the trial: a later study of the same drug in heart failure with preserved ejection fraction did not reach statistical significance on its primary endpoint. The ✅ belongs to the HFrEF claim. It does not transfer.

Discussion questions

1. The pairing of a neprilysin inhibitor with an angiotensin receptor blocker is presented as mechanistically required. Suppose a competitor marketed a neprilysin inhibitor alone, arguing that patients could add whichever RAAS drug their clinician preferred. Construct the strongest case for that position, then rebut it using the substrate list. What evidence would settle the argument?

2. PARADIGM-HF beat an active comparator with a proven mortality benefit. Explain precisely why that is harder than beating placebo, and why the resulting effect estimate is in one sense an underestimate of the drug's total benefit and in another sense not.

3. The run-in period and the early stop both affect how the result should be read, in different directions and on different quantities. Work through each: what does it bias, in which direction, and does it threaten the direction of the finding or only its magnitude? Which of the two would worry you more if you were writing a guideline?

4. The most successful therapy in this chapter contains no peptide. Discuss what that implies for the framing of "peptide therapeutics" as a coherent field. Is the useful category the drug's chemistry or the target's chemistry? Give one example from elsewhere in this book that supports your answer and one that complicates it.

5. The expectation going in was that preserving would work less well than supplying, and the opposite happened. Rank the four candidate explanations by how persuasive you find them, and justify your top choice. Then design — in outline, not in detail — a study that could discriminate between your top two.

6. A patient stable on an ACE inhibitor is switched to sacubitril/valsartan. Explain to a non-specialist why a gap between the two is required, what would happen without it, and why the requirement traces back to a drug that was never approved. Then explain, to the same person, why their BNP result may go up while they are getting better — and which test should be used instead.