PARAGON-HF Trial: Sacubitril/Valsartan in HFpEF

PARAGON-HF was a landmark randomized trial that tested sacubitril/valsartan against valsartan alone in patients with heart failure and preserved ejection fraction, a population with almost no proven drug therapies at the time. The trial narrowly missed its primary endpoint, with a p-value of 0.06, placing it in a statistical gray zone that has generated years of debate about who actually benefits from the drug and how the result should influence practice.¹ The subgroup findings, pooled cross-trial analyses, and subsequent regulatory decisions tell a more layered story than the headline result alone suggests.

Why Heart Failure With Preserved Ejection Fraction Is So Hard to Treat

Heart failure does not always mean a weak heart pump. In roughly half of heart failure cases, the heart squeezes normally but stiffens and fills poorly during relaxation. This is heart failure with preserved ejection fraction, or HFpEF. Patients are typically older, more often female, and carry multiple other conditions like diabetes, high blood pressure, obesity, and kidney disease. Despite being just as common as heart failure with reduced pumping, HFpEF had no drugs proven to reduce hospitalizations or death before the late 2010s. Trial after trial failed.

Part of the difficulty is that HFpEF is not one disease. It is a syndrome driven by different mechanisms in different patients. The same neurohormonal stress pathways that drive classic heart failure are activated in HFpEF, but to a lesser degree, which may explain why drugs targeting those pathways tend to produce smaller and more inconsistent benefits.² PARAGON-HF was designed as a serious attempt to move the needle in this frustrating space.

How Sacubitril/Valsartan Works

Sacubitril/valsartan combines two mechanisms in one pill. The valsartan component blocks the receptor for angiotensin II, a hormone that promotes fluid retention, blood vessel constriction, and harmful remodeling of the heart muscle. The sacubitril component inhibits neprilysin, an enzyme that breaks down beneficial hormones called natriuretic peptides. Blocking neprilysin raises natriuretic peptide levels, which in turn promotes salt and water excretion, lowers blood pressure, and counteracts fibrosis and stiffening of the heart.³

This dual action had already proven dramatically effective in heart failure with reduced ejection fraction, where the PARADIGM-HF trial showed clear reductions in death and hospitalization. The hope was that the same mechanism would help in HFpEF, even if the underlying biology was different.

Who Was Enrolled

PARAGON-HF recruited 4,796 patients from across the globe, spanning Asia-Pacific, Central and Western Europe, North America, and Latin America. The average age was about 73 years, and 52% were women, making this one of the most gender-balanced major heart failure trials. The average ejection fraction was about 58%, well into the preserved range. Most patients had mild to moderate symptoms, and roughly four in ten had diabetes while nearly half had chronic kidney disease.⁴

The trial used an active comparator rather than placebo. Every patient received valsartan, which is itself an effective blood pressure drug with some heart failure benefit. This matters when interpreting the results: the question was not whether sacubitril/valsartan works at all, but whether adding neprilysin inhibition on top of angiotensin blockade provides extra benefit.

The Primary Result and Why It Fell Just Short

The primary endpoint was a composite of total hospitalizations for heart failure and death from cardiovascular causes. Over a median follow-up of about 35 months, there were 894 primary events among patients assigned to sacubitril/valsartan and 1,009 among those assigned to valsartan. This translated to a rate ratio of 0.87, meaning a 13% lower event rate with sacubitril/valsartan, but the 95% confidence interval crossed 1.0 (0.75 to 1.01) and the p-value was 0.06.⁵

In conventional statistical terms, that is a miss. The trial did not cross the pre-specified threshold for declaring a positive result. But the margin was razor thin, and many cardiologists felt the data still pointed toward a real treatment effect that the trial was simply underpowered to confirm conclusively in the full population. Cardiovascular death rates were nearly identical between groups (about 8.5% versus 8.9%), so whatever signal existed was driven mostly by fewer heart failure hospitalizations (690 versus 797).⁶

Subgroups That Showed Clear Benefit

Where PARAGON-HF became genuinely interesting was in its pre-specified subgroup analyses. Two stood out as particularly striking and have shaped how the drug is used in practice.

The first was ejection fraction. Among patients whose ejection fraction fell below the median of 57%, the treatment effect was substantially larger, with a rate ratio of 0.78, a roughly 22% reduction that was statistically significant. This mirrors the known benefit of sacubitril/valsartan in reduced ejection fraction and suggests a continuum of response as the pumping function declines.⁷

The second was sex. Women assigned to sacubitril/valsartan had a 27% lower rate of the primary endpoint compared with valsartan (rate ratio 0.73), while men showed essentially no difference (rate ratio 1.03). The interaction was statistically significant, with a p-value for interaction of 0.017.⁸ Why women benefited more remains an open question, though it may relate to sex differences in cardiac remodeling, natriuretic peptide biology, or the fact that women with HFpEF tend to have slightly lower ejection fractions than men at any given symptom level.

Kidney Function and Treatment Response

A separate analysis looked at how baseline kidney function influenced treatment response. Among patients with the most impaired kidney function (estimated filtration rate at or below 45), sacubitril/valsartan reduced the primary cardiovascular outcome by about 31% compared with valsartan (rate ratio 0.69) and reduced cardiovascular death specifically by 35% (hazard ratio 0.65).⁹ This is a particularly important finding because kidney disease is extremely common in HFpEF patients and worsening kidney function is one of the biggest drivers of hospitalization and death in this population.

Patients already taking a mineralocorticoid receptor antagonist (a type of potassium-sparing diuretic used in heart failure) also appeared to get more benefit from sacubitril/valsartan. In MRA users, the rate ratio for the primary outcome was 0.73, compared with 0.94 in non-users. Additionally, the combination of an MRA and sacubitril/valsartan was associated with a slower annual decline in kidney filtration rate.¹⁰ Though the interaction did not reach formal statistical significance, the pattern reinforced the idea that combination neurohormonal blockade might work better than any single agent alone.

Quality of Life

Beyond hard clinical events, PARAGON-HF also measured how patients felt using the Kansas City Cardiomyopathy Questionnaire, a validated tool that captures symptoms, physical limitations, and social functioning. At 8 months, patients on sacubitril/valsartan had a borderline higher score, with an adjusted difference of about 1 point (p = 0.051). Over the full 36-month follow-up, the difference was modest but statistically significant. More telling was that patients on sacubitril/valsartan had roughly 31% higher odds of experiencing what clinicians consider a meaningful improvement in quality of life, defined as a 5-point or greater increase in the questionnaire score.¹¹

A 1-point average difference sounds small, and on its own it would not move practice. But the proportion of patients achieving meaningful improvement was meaningfully different, suggesting the average masks a real benefit in a subset of responders.

Biomarker Dynamics and a Puzzling Convergence

One of the more curious findings involved NT-proBNP, a blood marker of heart wall stress that clinicians use to track heart failure severity. Sacubitril/valsartan reduced NT-proBNP quickly, with a noticeable drop at 4 weeks and a statistically significant difference at 12 weeks. But by 36 weeks, patients in the valsartan-only group had also experienced a delayed decline in NT-proBNP, and the gap between the two arms disappeared.¹²

This convergence is hard to interpret. It may reflect the biological reality that valsartan alone is a reasonably effective neurohormonal blocker, and the advantage of adding neprilysin inhibition narrows over time. Or it may be an artifact of trial dynamics: sicker patients who would have had higher NT-proBNP values may have been hospitalized or died, selectively removing them from later measurements and pulling the valsartan group’s average down.

The Hypotension Problem

Safety in PARAGON-HF was broadly acceptable, but one issue stood out: low blood pressure. About 13% of all patients experienced documented hypotension, and it occurred more frequently in the sacubitril/valsartan arm. That matters because hypotension in this analysis was not just an inconvenience. Following a hypotension episode, patients had roughly 63% higher rates of the composite of cardiovascular death and heart failure hospitalization and a similar increase in all-cause death.¹³

This finding creates a practical tension. The patients most likely to benefit from sacubitril/valsartan, particularly those with lower ejection fractions and worse kidney function, may also be those most vulnerable to blood pressure drops. Careful dose titration and monitoring are essential, especially in older patients already on multiple blood-pressure-lowering medications.

Did the Run-In Period Hide a Positive Result?

PARAGON-HF used a run-in design, meaning all patients took both sacubitril/valsartan and valsartan sequentially before randomization, and those who could not tolerate either drug were excluded. About 16% of subjects dropped out during this phase. A later analysis found that these dropouts tended to be sicker: they had lower blood pressure, worse kidney function, higher heart failure biomarkers, and more prior hospitalizations.¹⁴

When researchers reweighted the analysis to give more representation to patients resembling those who dropped out, the treatment effect looked slightly better (rate ratio 0.86, with a confidence interval of 0.74 to 1.00). This is a methodological point, not a clinical one, but it reinforces the suspicion that PARAGON-HF may have selectively enrolled patients who were less likely to show a big treatment difference, essentially making it harder for the drug to prove itself.

Pooled Analyses Across the Ejection Fraction Spectrum

Because PARADIGM-HF (in reduced ejection fraction) and PARAGON-HF (in preserved ejection fraction) used similar designs and endpoints, investigators pooled the two trials in a pre-specified combined analysis of over 13,000 patients.¹⁵ These pooled analyses have been influential in shaping the narrative around sacubitril/valsartan as a drug with benefit across a broad range of ejection fractions, not just below 40%.

A post hoc pooled analysis found that sacubitril/valsartan significantly reduced all-cause hospitalization across both trials, with an 8% lower hazard over roughly two and a half years of follow-up.¹⁶ Another pooled analysis examined both cardiac and non-cardiac disease burden and reinforced the treatment signal across the combined population.¹⁷ These results helped support the idea that the benefit of sacubitril/valsartan extends into the mildly reduced and lower-preserved ejection fraction range, even if the PARAGON-HF result on its own did not reach statistical significance.

Regulatory and Guideline Impact

Despite missing its primary endpoint, PARAGON-HF influenced regulatory decisions. The U.S. FDA expanded the indication for sacubitril/valsartan to include heart failure patients with ejection fraction “below normal,” a deliberately vague wording that encompasses much of the PARAGON-HF population without endorsing the drug for the very highest ejection fractions. The FDA’s logic leaned heavily on the pooled analyses and the consistent subgroup signals rather than the standalone trial result.

However, no drug approved for HFpEF has been shown to reduce mortality. In the last decade, the FDA approved three drugs for morbidity reduction in HFpEF, but none of them, including sacubitril/valsartan, demonstrated a mortality benefit in this population.¹⁸ This is an important reality check. The best available treatments for HFpEF reduce hospitalizations and improve symptoms, but the mortality problem remains unsolved.

Cost-Effectiveness and Real-World Tradeoffs

Sacubitril/valsartan is an expensive branded medication, which raises practical questions about value. One economic modeling study projected that for patients with ejection fractions of 60% or less, sacubitril/valsartan adds roughly half a quality-adjusted life year compared with standard angiotensin blockade, at an incremental lifetime cost of about $41,000. That translates to an incremental cost-effectiveness ratio of about $77,000 per quality-adjusted life year, which falls within the range that most health systems consider acceptable.¹⁹

A separate analysis took a more skeptical view. When evaluating sacubitril/valsartan as an add-on to other therapies like SGLT2 inhibitors (the newer class of drugs also approved for HFpEF), the incremental cost-effectiveness ratio exceeded $250,000 per quality-adjusted life year in both the mildly reduced and preserved ejection fraction groups. Only at generic pricing did the numbers become favorable.²⁰ This tension between moderate clinical benefit and high drug cost is likely to sharpen further as SGLT2 inhibitors, which now have strong standalone evidence in HFpEF, become generic first.

Phenotypic Clustering and the Search for Responders

One lingering question after PARAGON-HF was whether the drug works better in particular clinical phenotypes of HFpEF, beyond the broad subgroup analyses. Researchers applied phenotypic clustering, a statistical technique that groups patients by shared clinical features rather than by a single variable like ejection fraction, to both the PARAGON-HF and DELIVER (a trial of dapagliflozin in HFpEF) populations. In PARAGON-HF, the effect of sacubitril/valsartan did not differ across the identified phenotype clusters.²¹

That null finding is itself informative. It means that beyond ejection fraction and sex, researchers have not yet found a reliable clinical fingerprint that predicts who will respond well to neprilysin inhibition in HFpEF. The biology of this syndrome remains stubbornly heterogeneous, and the precision-medicine promise of tailoring therapy to individual HFpEF subtypes is still early-stage work. For now, the strongest predictors of benefit remain the relatively simple ones: being female, having a lower ejection fraction within the preserved range, having worse kidney function, and already being on a mineralocorticoid receptor antagonist.