The oldest heart failure drug missed its goal in a 1,001-patient trial
Added to modern treatment in the Netherlands, low-dose digoxin came with 19% fewer heart-failure emergencies and cardiovascular deaths, a gap that could be chance. Deaths weren't clearly lower. The "25% fewer hospitalizations" in the headlines comes from a separate pooled analysis in which a trial from the 1990s supplied three-quarters of the patients.
8 October 2026 · How we rate evidence
- Claim
- Finding (Early): Adding low-dose digoxin to modern treatment led to 15.7 heart-failure emergencies (hospital stays or urgent visits) and cardiovascular deaths per 100 patient-years, against 19.3 on placebo: 19% fewer in relative terms, a difference not clear of chance (rate ratio 0.81, range 0.61 to 1.07). This was the trial's main, pre-stated outcome.
Heart-failure emergencies (Early): Counted alone, 10.2 against 13.5 per 100 patient-years (rate ratio 0.76, range 0.54 to 1.05). A secondary outcome, and not clear of chance.
Safety (Early): No clear increase in serious adverse events with blood levels kept at 0.5 to 0.9 ng/ml: 19.9 against 18.3 per 100 patient-years (1.09, range 0.86 to 1.37). Serious events that led to stopping the drug or were flagged as drug-related: 40 patients against 25 (rate ratio 1.53, range 0.91 to 2.55). The trial wasn't built to detect harms, and these ranges still allow meaningful ones.
Popular claim (Overstated): "A low dose of digoxin may help people with heart failure avoid hospitalization and reduce the risk of death" (UMCG press release, reprinted by ScienceDaily on 16 Aug 2026 under "This 10-cent heart drug cuts hospitalizations by 25%"). Medical Xpress's reprint in May: it "ensures that people with heart failure are hospitalized and die less frequently." Deaths weren't clearly lower, and the 25% isn't this trial's result - Studied in
- People: 1,001 adults with chronic heart failure and an ejection fraction of 50% or less, at 43 Dutch hospitals, enrolled 2020 to 2023. Mean age 72; 28% women; 29% had atrial fibrillation; 88% had mild symptoms (NYHA class II); 41% took an SGLT2 inhibitor at the start. Followed for a median of 36.5 months
- Design
- Double-blind randomized trial against placebo. The dose was adjusted centrally to a target blood level, with dummy adjustments for placebo patients. Main outcome stated in advance and registered (NCT03783429); events judged by a committee that didn't know who took what. Of patients alive at the end, more than a fifth had stopped the study drug (24% digoxin, 21% placebo). The per-100 figures are the paper's measured rates
- Status
- Peer-reviewed, Nature Medicine 32(7):2647, online 10 May 2026
- Replicated
- No repeat of this trial. On heart-failure hospital stays, two other large trials point the same way: DIG (1997, digoxin, before today's drugs), clearly; DIGIT-HF (2025, digitoxin, a related drug), in the same direction on hospital stays (not clear of chance alone) and clearly on its combined outcome of death or hospital stay. None of the three found fewer deaths
- Funding
- Dutch Heart Foundation; digoxin makers Disphar, Tiofarma and TEVA Nederland supplied the drug and funding; UMCG. Investigator-led: the Groningen team designed the trial and managed the data. Several authors report fees from heart-drug companies, mostly paid to their hospital
- We read
- Full paper (free via PubMed Central, CC BY-NC-ND) with tables, figures, extended data, supplement, protocol and analysis plan; not the image-only reporting summary. For DIG, DIGIT-HF, the pooled analysis and the withdrawal study, abstracts only. The UMCG release as reprinted by ScienceDaily, and Medical Xpress's headline and summary line
- Would change our mind
- Up: a second trial on modern treatment, or pooled patient data from DECISION and DIGIT-HF alone, showing fewer heart-failure emergencies clear of chance. Down: more serious side effects or deaths in pooled data at low doses, or in women
Digoxin, the paper's authors write, is "the oldest drug in cardiovascular medicine," and their university's press release says it costs less than ten cents a day. In a trial that followed patients for about three years, 500 Dutch patients with heart failure were given a low dose of it, and 501 a dummy pill. At the end, the digoxin group had fewer trips to hospital for heart failure. But the gap was too small, given the numbers involved, to count as more than possible chance.
The claim going around. "This 10-cent heart drug cuts hospitalizations by 25%," ScienceDaily headlined on 16 August, reprinting a release from the University Medical Center Groningen (UMCG). Its opening line says the drug "may help people with heart failure avoid hospitalization and reduce the risk of death." The same release, reprinted by Medical Xpress in May, opened more bluntly: digoxin "ensures that people with heart failure are hospitalized and die less frequently." The paper itself reports the miss plainly, and the release does say that this trial's main result "did not reach statistical significance."
What they did. The DECISION trial, led by Dirk Jan van Veldhuisen and Peter van der Meer at UMCG, enrolled adults with heart failure whose heart pumped out half or less of its blood per beat (the ejection fraction). Most were already on the standard modern drugs. Digoxin is dangerous at high blood levels, so a central lab checked each patient's level and adjusted the dose to a low target of 0.5 to 0.9 ng/ml. To keep the secret, placebo patients got fake adjustments. The main outcome, set in advance, counted every heart-failure hospital stay or urgent visit plus cardiovascular deaths. With blinding down to the dose checks and events judged by a committee that didn't know who took what, it is a clean test.
What they found. The digoxin group had 238 main-outcome events and the placebo group 291. That is 15.7 against 19.3 per 100 patient-years, 19% lower in relative terms, about 3.6 fewer events per 100 patients each year. The range runs from 39% fewer to 7% more, so the trial can't rule out no effect. It was designed to detect a 22% reduction, and collected 529 events, more than the 472 it planned for.
| Over a median of 3 years | Digoxin | Placebo | Ratio (range) |
|---|---|---|---|
| Main outcome, per 100 patient-years | 15.7 | 19.3 | 0.81 (0.61–1.07) |
| Heart-failure emergencies, per 100 patient-years | 10.2 | 13.5 | 0.76 (0.54–1.05) |
| Cardiovascular deaths | 83 (17%) | 88 (18%) | 0.93 (0.69–1.26) |
| Deaths from any cause | 118 (24%) | 126 (25%) | 0.93 (0.72–1.19) |
| Serious adverse events, per 100 patient-years | 19.9 | 18.3 | 1.09 (0.86–1.37) |
Hospital stays for any reason were no different (rate ratio 1.00). Side effects patients reported, such as blurred or yellow vision (6.8% against 5.4%) and nausea (5.8% against 5.6%), weren't clearly more common. Dose control mostly worked: after up to two checks, 84% of tested patients were in range.
Where the 25% comes from. Not from this trial. The same team, with the leaders of a German trial, pooled three large placebo-controlled trials in JAMA. According to its abstract, 26% of patients on digitalis drugs had a first worsening heart-failure event (mostly hospital stays), against 33% on placebo, a 25% lower rate. Deaths were no different. By our count, 6,800 of the 9,013 patients came from DIG, a trial run in the early 1990s at higher doses, before most of today's heart-failure drugs were in use. The release's body does say the 25% comes from this pooled analysis; the headline doesn't. It also drops "heart failure": this trial found no fall in hospital stays overall.
The release's line on a third study, in which 288 digoxin patients were switched to placebo at the end, says 14 "were hospitalized or died." That study's abstract counts 14 events, not patients: 12 hospital stays and 2 urgent visits, with no deaths in that group.
What it doesn't show. It doesn't show digoxin works on top of modern treatment, and it doesn't show it's useless. Its range is wide enough to hold a worthwhile benefit and none at all. More than a fifth of patients stopped the study drug, more often during the pandemic, which blurs any difference. The authors report a stronger effect when counting only time on the drug (a 34% reduction), and label that analysis post hoc. It drops patients when they stop, and the ones who stop may be sicker, so it can't stand in for the randomized comparison.
The paper's description of low-dose digoxin as "well tolerated and safe" also goes further than the numbers can show. The trial wasn't sized to detect harms, and its ranges still allow up to 37% more serious adverse events, or 2.5 times the rate of serious events that led to stopping the drug or were flagged as drug-related. The patients were mostly men with mild symptoms, so it says little about women or sicker patients.
What it means beyond the study. For people with heart failure, the question is whether a cheap old pill can prevent some hospital trips on top of today's four standard drugs. This trial alone can't answer yes. Two trials in the modern era point that way: this one, and DIGIT-HF, which tested digitoxin, a related drug, in sicker patients and found fewer deaths or hospital stays combined. Low-dose digoxin doesn't look dangerous when blood levels are watched, which matters for the many people already taking it.
Why it's still interesting. Older studies suggested low blood levels of digoxin were where any benefit lived, and high levels where the harm was. DECISION shows a simple dosing scheme can hold most patients low for years, in older people with other conditions. Every heart-failure measure leaned the digoxin way. And according to the withdrawal study's abstract, a striking signal came when the drug was taken away: in the six weeks after, there were 14 heart-failure events among those switched off digoxin, against 2 events (one a death) among those switched off placebo. That rests on few events, and the release itself says it doesn't directly prove digoxin works. For a drug costing cents, that's reason enough for a decisive trial.
Open questions.
Does it help on top of all four modern drugs? The pooled analysis found no clear difference by background treatment, but it leaned on the 1990s trial. Pooling patient-level data from DECISION and DIGIT-HF alone, or a larger new trial, would answer it. If yes, one of the cheapest drugs available would join the standard regimen.
Digoxin or digitoxin? DIGIT-HF's digitoxin met its main target; DECISION's digoxin didn't, in a less sick group. A direct comparison would show whether the drug or the patients made the difference, and which one doctors should reach for.
Is the benefit only for those who stay on it? Analyses that keep randomization intact, rather than dropping people who stop, could estimate the effect of taking the drug. If stopping carries its own risk, that changes advice on when to start it, and on taking people off it.
A note on timing. The paper came out on 10 May 2026, with the release that week. ScienceDaily's repost came three months later, on 16 August; the results hadn't changed.