Digoxin’s Second Chance Just Ran Out: What a Failed Phase 3 RCT Reveals About Cardiovascular Drug Development

Picture a data monitoring committee convening in late 2023 to review the final unblinded results from a trial that took eight years and 1,212 patients to complete. The drug in question has been used for over two centuries. The investigators had done everything right — pre-specified endpoints, rigorous randomization, extended recruitment from a planned 36 months all the way to 84 months to ensure adequate power. And yet when the hazard ratio landed on the table, it told a story the cardiology community had quietly feared: low-dose digoxin, administered with precision, had reduced hospitalizations for worsening heart failure — but had failed to move the composite primary endpoint with statistical significance. The study published in Nature Medicine was rigorous. The result was real. And the implications cut deeper than one drug’s fate.

The trial’s design was deliberate and, by contemporary standards, exemplary. Investigators enrolled patients with heart failure with reduced or mildly reduced ejection fraction, targeting serum digoxin concentrations in the low therapeutic range — between 0.5 and 0.9 ng/mL — the window that post-hoc analyses of older data had identified as potentially beneficial while avoiding toxicity. That precision mattered because digoxin’s history is inseparable from its toxicity profile: the drug carries a narrow therapeutic index, and concentrations above 1.0 ng/mL offer no additional benefit while substantially increasing arrhythmia risk. The trial wasn’t testing digoxin as clinicians used it in 1985. It was testing a calibrated, pharmacokinetically disciplined version of the molecule — arguably the fairest test the drug has ever received.

But the negative primary endpoint now sits in the medical record, and the cardiovascular drug development community has to decide what to do with it.

A 27-Year Shadow Over the Data

To understand why this result matters beyond the headline, you have to go back to 1997. The original DIG trial, published in the New England Journal of Medicine, randomized 6,800 patients with heart failure and a left ventricular ejection fraction of 0.45 or less to digoxin or placebo on top of diuretics and ACE inhibitors. The trial’s primary finding — no effect on all-cause mortality, but a reduction in heart failure hospitalizations — has haunted digoxin ever since. The drug reduced hospitalizations. It didn’t extend lives. In the therapeutic hierarchy of cardiology, that result translated into a slow, decades-long retreat from first-line use, accelerated by the arrival of beta-blockers, mineralocorticoid receptor antagonists, and eventually SGLT2 inhibitors.

What the 1997 DIG trial could not answer was whether the harm signals embedded in its data — particularly excess mortality at higher serum concentrations — had contaminated its efficacy signal. That is the question the low-dose hypothesis was designed to resolve. If you strip out the toxic tail of the concentration curve and study only patients maintained at 0.5–0.9 ng/mL, does the benefit-risk calculus change? The new trial, with its 1,212 evaluable patients and its 84-month recruitment arc, was the answer to that question. The answer, on the primary composite endpoint, was no — or at least, not convincingly enough.

That logic holds — until you look at the hospitalization data in isolation, which is where the story gets operationally interesting for sponsors thinking about their own cardiovascular programs.

The FDA’s guidance on endpoints for heart failure drug development is explicit on a point that many sponsors misread: an effect on symptoms or physical function, even without a favorable effect on survival or risk of hospitalization, can serve as a basis for drug approval. The guidance explicitly contemplates scenarios where symptom benefit alone supports approval. What it does not do is tell sponsors how to weight a drug that reduces hospitalizations — a costly, clinically meaningful outcome — without achieving statistical significance on a composite that bundles hospitalizations with mortality. That ambiguity is not new, but the digoxin result sharpens it.

The Trial Design the New Era Demands

Consider what happened in parallel therapeutic areas during the eight years this trial was running. In heart failure with reduced ejection fraction, the DAPA-HF trial demonstrated that dapagliflozin reduced the primary composite outcome of worsening heart failure or cardiovascular death to 16.3% versus 21.2% in the placebo group — a clean, statistically robust separation that translated directly into approval and guideline incorporation. EMPEROR-Reduced followed with empagliflozin. Two SGLT2 inhibitors reshaped the standard of care while the digoxin trial was still enrolling its final patients.

This is not a coincidence of pharmacology. The SGLT2 inhibitor trials were powered to detect mortality signals in well-characterized, contemporary populations receiving optimized background therapy. They enrolled faster because the regulatory hypothesis was sharp: reduce cardiovascular death or worsening heart failure, and the composite will move. The digoxin trial faced a structurally harder problem — it was rehabilitating a molecule whose benefit had always been hospitalization-centric, in a population now receiving background therapy that includes the very drugs that had proven mortality benefit. Designing around that confounder required either a larger sample, an adaptive design that could shift the primary endpoint pre-specified conditions, or a primary endpoint built from the start around hospitalization rate rather than a composite including mortality.

None of those choices are simple. But the operational lesson is precise.

When a drug’s pharmacological mechanism is dissociated from mortality — when the honest hypothesis is “this reduces hospitalizations and improves functional status, not that it extends life” — the primary endpoint architecture must reflect that hypothesis from Day One of protocol development. A composite that includes all-cause mortality will dilute a genuine hospitalization signal if the mortality curves are flat, which is exactly what happened here. The trial, which extended recruitment to 84 months and ultimately enrolled 1,240 patients before the analysis set was trimmed to 1,212, had the duration and sample to detect a mortality separation if one existed. One did not. And bundling a null mortality signal with a potentially real hospitalization signal in a single composite is how you generate a negative trial from a drug that may still have a defined role in a subset of patients.

What Sponsors Must Take From This

The digoxin result is not a story about a dead drug. It is a masterclass in endpoint selection under uncertainty — and a cautionary signal for any sponsor running a cardiovascular Phase 3 program where the mechanism of action is more likely to affect hospitalization and quality of life than survival curves.

The FDA’s heart failure endpoints guidance explicitly supports hospitalization rate as a standalone approvable endpoint. Sponsors who are designing trials for agents with inotropic, neurohormonal, or symptom-modifying mechanisms should be having pre-IND and Type B meetings specifically to interrogate whether a mortality-inclusive composite serves their drug’s actual biology — or whether it introduces a null-signal contaminant that will swamp the genuine effect in the final analysis.

The digoxin trial’s investigators did not make an error in judgment. They made a scientifically defensible bet that a mortality signal would emerge at low concentrations if the toxicity confound from the 1997 DIG data was removed. With 6,800 patients, the DIG trial could detect small mortality differences. With 1,212 patients and a contemporary background therapy that already reduces mortality by 30–40% through SGLT2 inhibitors and sacubitril/valsartan, detecting an incremental mortality signal from a fourth or fifth agent requires either an enormous sample or a population deliberately under-treated on background therapy — an increasingly difficult and ethically fraught design choice.

That structural problem will not resolve itself as heart failure pharmacotherapy matures further. If anything, the arrival of additional guideline-directed agents means that future trials testing adjunctive therapies face mounting competition for mortality signal in an already-treated population. The solution is not to abandon composite endpoints — it is to architect those composites around the events the drug is mechanistically positioned to affect, and to defend that architecture aggressively in regulatory meetings before a single patient is enrolled.

The data monitoring committee that reviewed those final 2023 results was looking at a trial that had done nearly everything right. The primary endpoint failed. The hospitalization signal, arguably the one the drug’s mechanism predicted, sits in the secondary data — real, visible, and unable to carry the regulatory weight of an approval without a pre-specified primary designation. For every cardiovascular sponsor reading this trial report and thinking “that won’t happen to us” — it will, unless the protocol team asks the harder question before lock: are we building a trial around what this drug actually does, or around what we wish it would do?

References

1. Nature Medicine — “Low-dose digoxin in patients with heart failure with reduced or mildly reduced ejection fraction: a randomized controlled trial”
2. PubMed / New England Journal of Medicine — “The DIG Trial: Effect of Digoxin on Mortality and Morbidity in Patients with Heart Failure” (1997)
3. FDA — “Treatment for Heart Failure: Endpoints for Drug Development Guidance for Industry”
4. Cardiology Trials — “Review of the DAPA-HF Trial” (dapagliflozin, primary composite outcome data)
5. Cleveland Clinic Journal of Medicine — DIGIT-HF trial enrollment and patient population characteristics
6. DrOracle — “Guidelines for Initiating and Managing Digoxin: Therapeutic Drug Monitoring and Dosing Safety”