TRED-HF - Withdrawal of HF meds in patients with recovered (non-ischemic) dilated cardiomyopathy

Halliday BP, et al. Withdrawal of pharmacological treatment for heart failure in patients with recovered dilated cardiomyopathy (TRED-HF): an open-label, pilot, randomised trial. Lancet 2019 Jan 5;393(10166):61-73.

Bottom Line: In patients with recovered dilated cardiomyopathy (DCM), even careful withdrawal of HF medications will result in relapse of DCM (based on clinical signs, imaging or biomarkers) in approximately 4 out of 10 patients within 6 months, compared to no deterioration in this timeframe if these medications are continued.

These medications should be considered “lifelong” medications until we have tools that can reliably predict which patients can stop them without deteriorating.


  • In patients who initially have HF with reduced ejection fraction (HFrEF), recovery of ejection fraction >50% portends a more favorable prognosis

    • e.g. In one study, vs patients who had initial HFrEF followed by LVEF recovery to >50%, patients with non-recovered HFrEF had an increased risk of death, transplant or VAD placement (HR 3.4) & CV hospitalization (HR 1.8)

  • The 2017 Canadian Cardiovascular Society (CCS) heart failure (HF) guidelines recommend consideration of monitored, sequential discontinuation of HF meds in certain subsets of patients with recovered non-ischemic cardiomyopathy

    • Including: chemotherapy-related, ETOH overuse-related, peripartum, tachycardia-related, or valvular cardiomyopathy

    • If: Asymptomatic (NYHA 1), LVEF and LV volumes normalized, trigger eliminated (e.g. ETOH abstinence, HR controlled, valve repaired/replaced)

  • There is limited evidence for pharmacological treatment withdrawal in patients with HFrEF who get EF recovery

    • e.g. in an early observational study of 13 participants with DCM taking metoprolol for >2.5 years who weaned off metoprolol, 54% (7/13) experienced clinical deterioration (4 deaths & 3 patients who worsened by 1 NYHA functional class).

Design: Open-label RCT (pilot trial designed to plan larger trial)

Patients (n=51)

  • Included if:

    • 16+ y/o

    • Previous dx of dilated cardiomyopathy (DCM) with LVEF 40% or lower

    • Currently:

      • NYHA functional class 1 (no current HF symptoms)

      • LVEF 50% or higher & left ventricular end diastolic volume indexed (LVEDVi) WNL (based on cardiac MRI, or 3D echo if MRI contraindicated)

      • NT-proBNP <250 ng/L

      • Treatment with 1+ of the following HF meds: Loop diuretic, ACEI, ARB, mineralocorticoid-receptor antagonist (MRA; spironolactone or eplerenone)

  • Key exclusion criteria

    • Uncontrolled HTN (>160/100 mmHg in clinic)

    • Mod-severe valvular disease

    • Angina

    • Beta-blocker required for AF/flutter, VT, or SVT

    • GFR <30

    • Pregnant.

  • Baseline characteristics (average of both groups unless specified)

    • Median age 55 y/o (IQR 45-64), male (67%)

    • Time since dx (4.9 y), median LVEF at dx 25%

    • Cause: Idiopathic (69%), familial (14%), trigger (excess ETOH, pregnancy, anthracycline, hyperthyroidism or myocarditis; 18%), pathogenic TTN truncation (22%)

    • Time since LVEF >50% (2 y)

    • CV symptom burden (0=none, 185=severe): 10-11

    • Quality of life using Kansas City Cardiomyopathy Questionnaire (KCCQ; 0=worst, 100=best)): 94-97

    • LVEF 60%, LVEDVi 83 mL/m^2, NT-proBNP 72 ng/L

      • Global longitudinal strain median 14% (values <16% considered abnormal)

    • Meds: ACEI/ARB (100%), beta-blocker (88%), MRA (47%), loop diuretic (12%)

Intervention & Comparator

  • Intervention: Sequential discontinuation of HF meds over max 4 months, total 6 months follow-up

    • Order of drug dose reduction/discontinuation:

      • (1) Loop diuretic (reduced by 50% q2 weeks until furosemide 40 mg/d-equivalent, then D/Ced)

      • (2) MRA (reduced by 50% until equivalent to spiro 50 mg/d, then D/Ced)

      • (3) Beta-blocker (reduced by 50% until 25% target dose or lower, then D/Ced)

      • (4) ACEI/ARB (reduced by 50% until 25% target dose or lower, then D/Ced)

    • Follow-up schedule:

      • Baseline: Clinic visit, symptom & QoL questionnaire, exercise stress test, cardiac MRI, NT-proBNP

      • q4 weeks: Clinic visit & NT-proBNP

      • @ week 16: Repeat cardiac MRI

      • @ month 6: Same as baseline

  • Comparator:

    • Phase 1 (randomized phase) x6 months: Continued all HF meds per baseline

      • @ baseline & month 6: Same as intervention group

      • @ weeks 8 & 16: Clinic visit, NT-proBNP

    • Phase 2: Then, non-randomized crossover to sequential discontinuation of HF meds as per intervention protocol


Primary outcome: DCM relapse in 6-month randomized phase

  • Defined as meeting 1+ of:

    • Clinical HF based on signs & symptoms

    • LVEF reduced by >10%, to <50%

    • LVEDVi increased by >10%, to above normal range

    • NT-proBNP doubled, to >400 ng/L

  • Discontinuation group 44%, control group 0% (p=0.0001) - “number needed to harm” = 3 (rounded up from 2.3)

Figure 3 from TRED-HF. Kaplan-Meier curve of time to relapsed DCM comparing discontinuation versus continuation of HF meds

Secondary outcomes:

  • Composite safety outcome (CV death, major adverse CV events, unplanned CV hospitalization): 0 in both groups

  • (Select) differences in means between groups from baseline to month 6:

    • KCCQ: -5.1 (95% CI -9.9 to -0.4; lower with discontinuation vs continuation of HF meds)

    • LVEF -9.5% (lower with discontinuation vs continuation)

    • LVEDVi +4.7 mL/m^2 (95% CI -1.5 to +11.0, p=0.14)

    • Vitals: HR +15 bpm, BP +7/+7 mmHg

    • Inconclusive: CV symptom burden, exercise time, peak VO2, log-transformed NT-proBNP

Secondary analyses including withdrawals from phase 1 + phase 2

  • DCM relapse in control group phase 2: 36%

  • Overall DCM relapse rate after HF med discontinuation: 40% (26% relapse <2 months of discontinuation)

Figure 4 from TRED-HF. Venn diagram breakdown of component of primary outcome met (includes all withdrawals from randomized phase + single-arm crossover phase)

Figure 4 from TRED-HF. Venn diagram breakdown of component of primary outcome met (includes all withdrawals from randomized phase + single-arm crossover phase)

Internal validity

  • Allocation bias: Low risk

    • Computer-generated random sequence, 1:1 allocation in permuted blocks, stratified by baseline NT-proBNP

    • Centralized allocation via online system

  • Performance bias: Low/unclear risk

    • Patients & their clinicians aware of treatment allocation; however, the study employed a standardized protocol to wean & D/C HF meds, as well as standardized monitoring

  • Detection bias

    • Low risk of bias for objective outcomes (core lab MDs reading imaging unaware of study group allocation)

    • High risk of bias for QoL outcomes (patients completed the questionnaires aware of treatment allocation)

  • Attrition bias: Low risk

    • Loss to follow-up 2% (1 participant in withdrawal group left trial after 7 days)

    • Analyzed intention-to-treat population

Other Considerations

  • We can’t yet predict which stable, recovered DCM patients will deteriorate with D/C of HF meds

    • In this trial, predictors of DCM relapse after withdrawal of therapy included: greater age, use of >2 meds, use of MRA, higher NT-proBNP, lower global radial strain on cardiac MRI, & possibly lower peak VO2

      • However, based on univariable analysis only (no adjusting for other variables) & small n of events

    • DCM etiology did not clearly predict risk of deterioration with therapy withdrawal. Some patients with a seemingly reversible cause of DCM (e.g. ETOH use, pregnancy) did have DCM relapse upon D/Cing HF meds. Therefore, presence of a trigger does not indicate that D/Cing HF meds after HF remission will be safe.

Mineralocorticoid receptor antagonists (MRAs) in HFrEF or post-MI LV dysfunction (EPHESUS, EMPHASIS, RALES)

EPHESUS: Pitt B, et al. Eplerenone, a selective aldosterone blocker, in patients with left ventricular dysfunction after myocardial infarction. N Engl J Med 2003;348:1309-21.

EMPHASIS-HF: Zannad F, et al. Eplerenone in patients with systolic heart failure and mild symptoms. N Engl J Med 2011;364:11-21.

RALES: Pitt B, et al. The effect of spironolactone on morbidity and mortality in patients with severe heart failure. N Engl J Med 1999;341:709-17.


  • In patients with HFrEF with NYHA functional class 2-4, MRAs reduced the risk of death (NNT 18-60 per year) and hospitalization (NNT 24-30 per year), but increased the risk of hyperkalemia.

  • Proper monitoring of renal function & serum potassium (e.g. 1 week after start/dose change, monthly x3 months, then q3-6 months) is critical to ensure that harms do not outweigh benefits of this important therapy.

Patients, Interventions, Controls & Duration of Follow-Up



MRA O.png

Additional Results/Analyses

Generalizability & internal validity

  • RALES set the indication for mineralocorticoid antagonists in HFrEF NYHA class III-IV; EPHESUS expanded it to HFrEF or LV dysfunction+diabetes post-MI; EMPHASIS-HF further expanded the indication to HFrEF NYHA II with high risk for HF hospitalization.

  • Assessed together, these 3 trials evaluated MRAs in all stages in a varity of etiologies for HFrEF, as early as 3 days post-MI, with background therapy ranging from the full gamut to only ACEI + diuretic therapy.

  • Trials employed routine monitoring for renal & potassium abnormalities:

    • EPHESUS: Serum potassium 48h after treatment start, then at weeks 1, 4, 5, 12 of treatment, then q3 months

    • RALES: Serum potassium q4 weeks x3 months, then q3 months x1 year, then q6 months

  • Internal validity: Low risk of bias in all 3 trials

TOPCAT - Spironolactone in HFpEF

Pitt B, et al. Spironolactone for heart failure with preserved ejection fraction. NEJM 2014;370:1383-92.

Bottom line: In patients with HFpEF, spironolactone did not reduce the risk of death or hospitalization over 3.3 years. Spironolactone does not noticeably improve quality of life in HFpEF.

Consistent with other "negative" studies in HFpEF, the results of TOPCAT supports the notion that patients with "HFpEF" should receive symptomatic treatment for HF, & interventions to reduce morbidity & mortality according to the underlying etiology. In this context, spironolactone remains useful for patients with HFpEF & resistant hypertension. The label of "HFpEF" does not yet offer an actionable management plan.



  • Heart failure with preserved ejection fraction (HFpEF):

    • Defined as a clinical syndrome of signs & symptoms of HF with normal systolic function (LVEF >50% & LV end-diastolic volume index <97 mL/m^2) & evidence of diastolic LV dysfunction (relaxation, filling, diastolic distensibility or diastolic stiffness)

  • Interventions that reduce morbidity & mortality in HFrEF have not translated to benefits in HFpEF, including ACEIs & ARBs, beta-blockers & digoxin

  • Previous studies have demonstrated inconsistent results for aldosterone antagonists in HFpEF:

    • Aldo-DHF trial: In patients with HFpEF with a good prognosis, spironolactone improved E/e' (echo-derived surrogate for filling pressure), but not VO2 (stress test-derived measure of aerobic exercise capacity)

    • Cohort study: In patients with HFpEF, use of an aldosterone antagonist was not associated with reduced mortality or HF hospitalization

Patients (n=3445)

  • Inclusion

    • Age 50+

    • Clinical HF

    • LVEF >45% (measured within 6 months prior to randomization & after any prior ACS)

    • Controlled SBP (<140, or <160 if receiving 3+ antihypertensives)

    • Either,

      • 1+ HF-related hospitalization in prior 12 months, or

      • BNP >100 pg/mL (or N-terminal pro-BNP >360 pg/mL) within 30 days, not explained by another disease entity

  • Exclusion

    • Pericardial constriction

    • Hemodynamically significant uncorrected primary valvular heart disease

    • Infiltrative or hypertrophic obstructive cardiomyopathy (HoCM)

    • Stroke, MI or CABG in past 90 days

    • AF with resting HR >90 bpm

    • Heart transplant recipient or currently implanted LVAD

    • Chronic pulmonary disease: Requiring home O2 or PO steroids, hospitalization for exacerbation within 12 months, or significant in the opinion of the investigator

  • "Average" patient

    • Age 69 y (age 75+ 27%)

    • Female 52%

    • White 89%

    • HFpEF characteristics

      • Hospitalization in last 12 months 71%

      • NYHA functional class I (3%), II (64%), III (33%), IV (<1%)

      • BNP 236

      • LVEF 56%

    • BP 130/80 mm Hg

    • Meds

      • Diuretic 80%

      • ACEI/ARB 85%

      • Beta-blocker 75%

      • ASA 65%

      • Statin 55%


  • I: Spironolactone

    • Initial dose: 15 mg PO daily x4 weeks

    • If initial dose tolerated, increased to 30 mg daily

    • If HF still symptomatic at month 4, increased to 45 mg daily

  • C: Matching placebo

  • Monitoring: Measurement of SCr & serum K required <1 week after start or change of study drug dose

Results @ mean 3.3 years

  • Efficacy

    • Death: Spironolactone 14.6% versus placebo 15.9%, hazard ratio (HR) 0.91 (95% confidence interval 0.77-1.08)

    • Hospitalization for any cause: 44.5% vs 46%, HR 0.94 (0.85-1.04)

    • Primary outcome (CV death, aborted cardiac arrest, HF hospitalization): 18.6% vs 20.4%, HR 0.89 (0.77-1.04)

    • Quality of life

      • Kansas City Cardiomyopathy Questionnaire (KCCQ; 100-point scale, minimal clinically important difference -5): -1.86 vs placebo at 36 months (p=0.02)

      • No statistically or clinically significant differences in EQ5D-VAS or McMaster Overall Treatment Evaluation score

  • Safety

    • Serious adverse event: 48.5% vs 49.6%

    • Discontinuation of study drug: 34.3% vs 31.4%

    • Doubling of SCr: 10.2% vs 7% (NNH 32)

    • Hyperkalemia (serum K >5.5 mmol/L): 18.7% vs 9.1% (NNH 11)

Issues with internal validity?

  • Low risk of bias characteristics: Randomized, allocation-concealed, double-blind trial analyzed using intention-to-treat population

  • Unclear risk of bias: Loss-to-follow-up on vital status ~4%

  • Subgroup analysis: Primary outcome varied based on region (Americas vs Russia/Georgia, p<0.001 for interaction)

    • Americas: Spironolactone 27.3% vs placebo 31.8%, HR 0.82 (0.69-0.98) (NNT=23)

    • Russia/Georgia: 9.3% vs 8.4%, HR 1.10 (0.79-1.51)

    • Credibility of this subgroup effect has been increased by a substudy demonstrating that participants from Russia were far more likely than those from North America to have no detectable serum concentrations of spironolactone metabolites. This indicates that significantly more participants from Russia did not receive the study drug, and raises the potential of misconduct at these study sites. Therefore, the results of the 'Americas' subgroup may be the most accurate reflection of the effect of spironolactone in HFpEF.

    • Secondary outcomes from the Americas subgroup:

      • Death: 20.1% vs 23.5%, HR 0.83 (0.68-1.02)

      • All-cause hospitalization: 59.4% vs 60.7%, HR 0.94 (0.83-1.05)

        • HF hospitalization: 20.8% vs 24.5%, HR 0.82 (0.67-0.99) (NNT=28 over 3.3 years)

        • Hyperkalemia: 25.2% vs 8.9% (NNH=7)

    • Exclusion of the Russia/Georgia participants renders the primary outcome statistically significant, but does not materially affect the effect estimate for the primary or secondary outcomes. Additionally, TOPCAT still fails to demonstrate a significant reduction in death or all-cause hospitalizations.