Natriuresis-Guided Diuretic Therapy in Acute Heart Failure (PUSH-AHF)
Authors, Journal, Affiliations, Type, DOI
- Authors: Jozine M. ter Maaten (first and corresponding author), Iris E. Beldhuis, Peter van der Meer, Jan A. Krikken, Douwe Postmus, Jenifer E. Coster, Wybe Nieuwland, Dirk J. van Veldhuisen, Adriaan A. Voors, and Kevin Damman (senior author)
- Affiliations: Department of Cardiology and Department of Epidemiology, University Medical Center Groningen, University of Groningen, the Netherlands
- Journal: Nature Medicine, volume 29, October 2023, pages 2625–2632 (published online 28 August 2023)
- Type: Prospective, single-centre, pragmatic, open-label, randomized controlled trial
- DOI: 10.1038/s41591-023-02532-z
- Registration: NCT04606927 (PUSH-AHF — Pragmatic Urinary Sodium-based algoritHm in Acute Heart Failure)
- Funding: Personal Dekker grant from the Dutch Heart Foundation (2020T012) to the first author; the funder had no role in design, analysis or writing
Overview
PUSH-AHF asked whether titrating intravenous loop diuretics to measured urinary sodium improves decongestion in acute heart failure. In 310 patients at one Dutch centre, a protocol that measured spot urinary sodium at 2, 6, 12, 18, 24 and 36 hours and escalated treatment when it fell below 70 mmol/L raised 24-hour natriuresis from 345 to 409 mmol (P=0.0061) — meeting the first of its dual primary endpoints — and roughly doubled the cumulative in-hospital loop diuretic dose without any excess of renal events. It did not reduce the second primary endpoint of death or heart-failure rehospitalisation at 180 days (31% vs 31%; HR 0.92), and the trial was powered only to detect a hazard ratio of 0.49, with a lower event rate than anticipated. The result is therefore a proof of concept for urine sodium as a treatment target, on a surrogate, in a population in which the clinical endpoint was never realistically testable at this size.
Keywords
Acute heart failure, decompensated heart failure, loop diuretics, bumetanide, natriuresis, spot urinary sodium, diuretic resistance, decongestion, sequential nephron blockade, hydrochlorothiazide, acetazolamide, pragmatic trial
Key Takeaways
Background and Rationale
- Decongestion is the main treatment goal in acute heart failure, but monitoring of the diuretic response is inconsistent and usually relies on weight loss, which is insensitive, inaccurate and slow to change
- Insufficient diuretic response is common early after starting loop diuretics and is associated with residual congestion, mortality and heart-failure rehospitalisation; observational data link poor natriuresis to the same outcomes
- Urinary sodium is attractive because it is inexpensive, reliable, readily available on routine laboratory assessment, and is a direct measure of what a loop diuretic does
- The 2021 ESC heart-failure guidelines already recommended early and repeated assessment of spot urinary sodium to guide diuretic treatment — PUSH-AHF was designed to supply the randomized evidence behind that recommendation
Trial Design
- Single centre (University Medical Center Groningen), pragmatic, open-label; randomization built into the electronic health record (EPIC) with a hard-coded random number, allocated 1:1 at the point of starting intravenous loop diuretics
- Deferred consent was permitted by the ethics committee to allow immediate randomization at the emergency department; written consent was obtained within the first four days in all patients
- Inclusion deliberately broad: adults with acute heart failure requiring intravenous loop diuretics, no ejection-fraction or natriuretic-peptide entry criteria, 44% with de novo heart failure
- Physicians and investigators were blinded to all urinary sodium measurements (timed collections and spot values) in the standard-of-care arm, to prevent contamination
- The protocol: baseline loop dose set from eGFR and the outpatient dose, maximum bolus 5 mg bumetanide, given twice daily every 12 hours. In the guided arm, spot urinary sodium was measured at 2, 6, 12, 18, 24 and 36 hours; an insufficient response — urinary sodium <70 mmol/L and/or diuresis <150 mL/h — triggered escalation in a still-congested patient
- The escalation ladder: (1) an additional bolus at double the previous dose, to a maximum of 5 mg bumetanide, with the maintenance dose raised to match; (2) after two consecutive maximum boluses, add hydrochlorothiazide 25 mg; (3) then acetazolamide 500 mg daily or an SGLT2 inhibitor; (4) ultrafiltration as bail-out
- After 48 hours, diuretic adjustment reverted to physician discretion; the algorithm ran for 36 hours only
- 24-hour urine collections were obtained on the first three days in both arms, with standard-of-care values blinded until study end
- Dual primary endpoints: total 24-hour natriuresis at day 1, and time to all-cause mortality or adjudicated heart-failure rehospitalisation at 180 days
- Sample size: 80% power for 24-hour natriuresis at a two-sided alpha of 0.025 (Bonferroni for the dual endpoint); 81% power to detect a hazard ratio of 0.49 for the clinical endpoint assuming a 38% event rate
- No data safety monitoring board was used, the trial being considered low-risk; primary analyses were run by an independent statistician because the trial was open-label
Population
- 310 randomized (150 natriuresis-guided, 160 standard of care) at one centre between 11 February 2021 and 17 November 2022; last follow-up 9 May 2023
- Median age 74 years (IQR 65–82); 45% women; 96–98% white
- Median NT-proBNP 4,710 ng/L (IQR 2,553–8,750); median LVEF 35% in the guided arm and 38% in standard care; 26% vs 18% had HFpEF
- Median eGFR 54 vs 53 ml/min/1.73 m²; median creatinine 106 µmol/L in both arms
- 71–75% were NYHA class IV; rales in 71–73%, oedema in 67–68%
- Background therapy was modest by contemporary standards: SGLT2 inhibitor in 5% vs 8%, ACE inhibitor/ARB/ARNI 55% vs 54%, beta-blocker 67% vs 72%, MRA 35% vs 34%
- 56–58% were on a home loop diuretic (median 2 mg bumetanide equivalents); 44% vs 43% had new-onset heart failure
Decongestive Treatment Delivered
- Starting daily intravenous dose was identical (median 4 mg bumetanide in both arms)
- 128 of 150 (85%) patients in the guided arm had treatment intensified per protocol at some point in the first 36 hours
- Insufficient spot urinary sodium was the trigger in 100% of intensifications at 2 h, 80% at 6 h, 63% at 12 h, 72% at 18 h and 90% at 36 h; response was most often insufficient at 12 hours, when 58% of the guided arm had an inadequate response
- 228 additional boluses were given in 123 of 150 guided patients within 24 hours; a second diuretic was added in 32 (21%) — hydrochlorothiazide in 31, acetazolamide in 3, two of whom received triple nephron blockade; one SGLT2 inhibitor was added per protocol; no patient required ultrafiltration in the first 24 hours
- Cumulative dose was far higher in the guided arm: 12 mg vs 6 mg bumetanide over 24 h and 26 mg vs 15 mg in hospital (both P<0.0001)
Primary End Points
| End point | Natriuresis-guided | Standard of care | Effect (95% CI) | P |
|---|---|---|---|---|
| 24-hour natriuresis — mmol | 409±178 | 345±202 | difference 63 (18–109) | 0.0061 |
| Death or HF rehospitalisation at 180 days | 46/150 (31%) | 50/159 (31%) | HR 0.92 (0.62–1.38) | 0.6980 |
- The first primary endpoint was met; the second was neutral, and the authors state the trial was "probably underpowered to detect a smaller than hypothesized difference," with a lower event rate than the 38% assumed
- A major protocol deviation occurred in 40 of 310 (14.2%) — 28 (18.7%) in the guided arm and 16 (10.0%) in standard care; the per-protocol analysis gave similar findings
Secondary End Points
| End point | Natriuresis-guided | Standard of care | Effect (95% CI) | P |
|---|---|---|---|---|
| 48-hour natriuresis — mmol | 653±249 | 575±290 | 78 (10–145) | 0.0241 |
| 24-hour diuresis — ml | 3,900 (3,200–4,945) | 3,330 (2,510–4,500) | 534 (160–908) | 0.0053 |
| 48-hour diuresis — ml | 6,655 (5,401–7,824) | 5,915 (4,600–7,400) | 672 (137–1,206) | 0.0140 |
| Length of hospital stay — days | 6 (5–9) | 7 (5–10) | — | 0.1436 |
| HF rehospitalisation — % | 17 | 17 | HR 0.96 (0.56–1.67) | 0.8904 |
| 180-day all-cause mortality — % | 19 | 21 | HR 0.89 (0.54–1.46) | 0.6369 |
| NT-proBNP change at 48 h — % | −22 | −18 | — | 0.4351 |
| NT-proBNP change at 72 h — % | −33 | −33 | — | 0.7881 |
- The effect did not persist to 72 hours: 72-hour natriuresis was 832±323 vs 746±350 mmol (P=0.0706), and 72-hour diuresis 8,720 vs 8,255 ml (P=0.1104). The authors attribute this to the algorithm stopping at 36 hours
- Net fluid balance favoured the guided arm at 24 hours (−2,833±1,673 vs −2,380±1,573 ml; P=0.0218) and 48 hours (−4,728±2,318 vs −4,110±2,137; P=0.0297) but not at 72 hours (−6,216±3,000 vs −5,728±2,825; P=0.2400)
- Weight change did not differ at any time point (24 h median −1.2 vs −0.4 kg, P=0.1103) — the surrogate the trial was designed to replace did not detect the difference the urinary measurements did
Safety
- No signal of renal harm despite roughly double the diuretic dose: doubling of serum creatinine at 24 hours in 0% vs 1%, at 48 hours 1% vs 1%; true worsening renal function 1% vs 1%; worsening heart failure 6% vs 9% (P=0.3689)
- Serious adverse events 40% vs 44% (P=0.5799); adverse events 57% vs 60% (P=0.7180)
- Hypokalaemia (<3.5 mmol/L) was numerically more frequent at 23% vs 15% (P=0.0849); hyperkalaemia was less frequent (2% vs 6%, P=0.1136)
Interpretation
- The trial is described as "the first randomized clinical trial to show that natriuresis-guided diuretic therapy improves natriuresis and diuresis" in acute heart failure, and as "a first step towards a personalized treatment approach"
- The safety argument is the substantive contribution: DOSE showed that giving high-dose loop diuretic to everyone produced more diuresis but transient worsening of renal function, whereas individualising the escalation to those measured as under-responding produced more natriuresis without renal deterioration
- The authors argue that venous congestion, not the diuretic, drives worsening renal function in this setting, so decongesting the congested patient may be renoprotective
- Achieved natriuresis was higher than in ENACT and than in ADVOR's acetazolamide arm, which the authors attribute to a population including 44% diuretic-naïve de novo heart failure and to higher loop doses
- The authors note that implementing the protocol may itself be less pragmatic than the trial design — the frequency of sampling, including out of hours — and suggest embedding the algorithm in the electronic health record and nurse-led execution as the route to practice
Limitations of the document
- Single centre, in the Netherlands, with more than 90% white participants — generalisability to other health systems and populations is untested
- Open-label design. Blinding of the treating physician to assignment was impossible by construction, because the intervention is a response to a measured value; the standard-of-care urinary values were blinded to limit contamination, but the physicians caring for guided patients knew they were escalating
- The clinical endpoint was underpowered by design. The sample size assumed a hazard ratio of 0.49 and a 38% event rate; the observed event rate was 31%, and the authors state the trial was probably underpowered for a smaller difference
- The intervention lasted 36 hours, then stopped. Any effect on 180-day outcomes could only arise through the decongestion achieved in that window, and the natriuresis advantage had disappeared by 72 hours
- Pragmatic design brought missing data and protocol deviations — 14.2% major deviations, 32 patients without assessable 24-hour natriuresis (2 deaths, 30 missing collections) — and there was no systematic assessment of congestion status
- Hydrochlorothiazide was chosen as first-line second diuretic on the evidence available at design, before ADVOR was published; whether acetazolamide first would have produced a different result is untested
- No data safety monitoring board
- The dual primary endpoint used a Bonferroni-corrected alpha of 0.025 for each component, which is conservative for the natriuresis endpoint and modest for the clinical one
- Absolute natriuresis differences are large but the outcome separation is absent — the paper does not resolve whether the surrogate is a valid treatment target or merely a valid marker
Key Concepts Mentioned
- concepts/Diuretic-Resistance — the mechanism the protocol is designed to overcome, and the page where the escalation ladder now sits
- concepts/Decompensated-HF-Management — PUSH-AHF is the trial behind the ESC 2026 Class IIb/B1 recommendation for urinary-sodium-guided therapy
- concepts/Cardiorenal-Syndrome — the congestion-versus-perfusion trade-off that the renal safety data speak to
- concepts/Right-Ventricular-Failure — venous congestion as the driver of apparent diuretic failure
- concepts/Cardiac-Rehabilitation-HF — not addressed by this trial, but the post-discharge window is where a decongestion benefit would have to act
Key Entities Mentioned
- entities/Heart-Failure — the syndrome and its phenotypes
- entities/HFpEF — 18–26% of the enrolled population
- Bumetanide, hydrochlorothiazide and acetazolamide — the drugs in the escalation ladder; each is covered within concepts/Diuretic-Resistance rather than as a separate molecule page
Wiki Pages Updated
wiki/sources/push-ahf-natmed-2023.md— created (this page)wiki/concepts/Diuretic-Resistance.md— new natriuresis-guided section with the protocol, endpoints and safety data; two contradictions addedwiki/concepts/Decompensated-HF-Management.md— the Class IIb/B1 recommendation now carries its trial data; one contradiction revisedwiki/sourceindex.md— index rowwiki/wikiindex.md— Diuretic-Resistance and Decompensated-HF-Management rows revisedlog.md— ingest entry