Consequences of Discontinuing Long-Term Drugs in HFrEF

Khawaja M. Talha, MBBS; Javed Butler, MD, MPH, MBA; Milton Packer, MD

J Am Coll Cardiol. 2025;84(22)

Abstract and Introduction

Abstract

There is uncertainty regarding the clinical effects of discontinuation of drugs for heart failure after long-term use. The withdrawal of long-term treatment can follow 1 of 4 distinct patterns: 1) loss of on-treatment effect with no observed changes following discontinuation (eg, prazosin); 2) attenuation or loss of on-treatment effect with rebound clinical worsening following discontinuation (eg, nitroprusside); 3) persistence of deleterious on-treatment effect followed by clinical worsening after discontinuation (eg, milrinone and flosequinan); and 4) persistence of favorable on-treatment effect followed by clinical worsening after discontinuation (eg, digoxin and sodium-glucose cotransporter 2 inhibitors). Persuasive evidence for persistence of efficacy has been demonstrated for the use of digoxin, diuretic agents, sodium-glucose cotransporter 2 inhibitors, and (to a limited extent) for angiotensin-converting enzyme inhibitors. Available evidence for worsening of clinical status following the withdrawal of neurohormonal antagonists largely consists of observational studies. However, their findings are difficult to interpret because of considerable confounding related to the fact that drugs were withdrawn for clinical reasons, which represented a more important contributor to the poor outcome of these patients than the withdrawal of an effective drug. Nevertheless, the totality of available evidence points to a meaningful clinical deterioration within a few weeks following the withdrawal for most drugs that have been evaluated for the treatment of heart failure. These findings suggests that that our current emphasis on the implementation of foundational drugs needs to include an equally important emphasis to avoid even short-term gaps in treatment.

Introduction

Current recommendations support the use of angiotensin receptor neprilysin inhibitors, beta-blockers, mineralocorticoid receptor antagonists, and sodium-glucose cotransporter 2 (SGLT2) inhibitors in all patients with heart failure and a reduced ejection fraction, with these treatments to be continued indefinitely for the lifetime of each patient.^[1,2] However, these trials that established the efficacy of these drugs were not designed to provide evidence that the benefits of these drugs persist for the duration of treatment.

The conventional approach to the evaluation of a drug for the treatment of heart failure is to define a patient population of interest and to randomly assign (double-blind) one-half of the patients to placebo and the other half to active treatment, which is added to all appropriate other therapies for heart failure and continued for the entire duration of the trial, typically lasting for years. Patients are analyzed according to their original randomized treatment, whether or not therapy with placebo or the new treatment is sustained for the duration of follow-up, and whether or not background therapy is altered in a differential manner in the 2 treatment groups during the course of the trial. Such an intention-to-treat approach minimizes bias, but if adherence to the study medication changes dramatically and background therapy is differentially intensified in the placebo group, the estimated treatment effect is driven toward the null, thus distorting the ability of a trial to assess the true magnitude of benefit that might be produced by a drug’s pharmacological action.^[3]

Many physicians believe that visual inspection of the pattern of separation of Kaplan-Meier curves displaying event rates in the 2 treatment arms provides reliable information about maintenance of a drug effect. However, visual interpretation of the onset and maintenance of separation of event curves is highly subjective and is prone to considerable confounding. Because patients are enrolled in the trial over a long period of recruitment, those represented on the right side of a Kaplan-Meier plot (providing the longest duration of follow-up) are those who were recruited early in the trial. These may or may not show a similar response to treatment than those recruited later in the study. Yet, the small number of events in these early-recruited patients can cause the Kaplan-Meier plots after years of follow-up to swing wildly, often causing the plots to separate or converge in ways that are inconsistent with the pattern of response of the treatment effect seen in most patients who provided the most followup data. Furthermore, if a disorder is lethal and the follow-up period is long, the survival rates in both treatment arms of a trial converge to 0%, even when drugs produce dramatic treatment effects.^[4]

Interpretation of survival plots is further complicated by the fact that perceptions of a long-term treatment effect may be driven by a dramatic shortterm benefit experienced by a small fraction of patients. Imagine that a trial enrolled 10,000 patients hospitalized with myocardial infarction, of whom the 1,000 patients with the highest troponins would be at extreme risk of sudden death but only during the first month following coronary occlusion. Assume that the risk of sudden death during this vulnerable period in the at-risk population could be reduced from 90% to 10% with the use of a beta-blocker, but that the betablocker would have no benefit beyond the vulnerable period even in the highest-risk patients and would have no benefit at any time in patients with lower troponins. However, the investigators are not aware of these patterns of response when the trial is designed, and thus, the trial randomizes 10,000 patients to either placebo or a beta-blocker (in a 1:1 ratio) with treatment to be maintained for a median of 3 years. In the group with the highest troponins, at the end of 1 month, we would observe 450 deaths in the high-risk placebo group and 50 patients in the highrisk beta-blocker group. This 400-death betweengroup difference would cause the Kaplan-Meier curves to remain separated for the entire 3 years of double-blind therapy—even though, for most of the duration of follow-up, treatment with the betablocker was unnecessary. If the beta-blocker is withdrawn at the end of the trial, patients would not experience adverse consequences, even though the investigators might propose that continued separation of the event curves for the duration of follow-up implied a persistent treatment effect. For example, in CONSENSUS (Cooperative North Scandinavian Enalapril Survival Study), which enrolled patients with class IV heart failure, after 1 year of randomized therapy, most patients assigned to both the placebo and enalapril groups were treated with enalapril. The event curves maintained their separation for many years, even though the 2 treatment arms were being treated in a similar manner.^[4] The outcome in the placebo and enalapril group converged only when nearly all patients had died.

Therefore, to evaluate whether a drug has a persistent effect, a trial must be designed to discern the presence or absence of an effect in an unbiased manner, typically by conducting a formal randomized evaluation of the withdrawal of therapy in patients with chronic heart failure who have taken the drug for long periods of time, ideally for several years. Such trials may be carried out in 1 of 2 ways. First, patients taking active therapy for many years and who are clinically stable can be randomly assigned (doubleblind) to be maintained on the drug (drug continuation group) or to receive placebo (drug discontinuation group) with the pharmacological and clinical responses being compared during and at the end of specified period of follow-up.^[5] Second, patients who have participated in a long-term double-blind, placebo-controlled trial and are clinically stable can be required by protocol to stop their study medication at the trial’s end, leading to one-half of the patients being withdrawn from active therapy and the other half being withdrawn from placebo in a blinded manner.^[6] Assuming that the trials are large enough and the duration of follow-up is long enough, either trial design can lead to interpretable results regarding identification of persistence of a drug effect. Withdrawal trials are a powerful approach to determining the persistence of a drug effect and informing guidance as to whether effective drugs need to be maintained for the lifetime of patients.^[7]

1 2 3 4 5 6 7 8 9

Next Section

Table 1. Studies Demonstrating Pharmacological Tolerance During Repetitive or Prolonged Treatment
First Author (Year)	Study Design	Drug Tested	N	Patient Characteristics	Duration of Withdrawal Period	Background Therapy	Responses Following Drug Discontinuation
Packer et al⁹ (1979)	Single-arm prospective trial	Nitroprusside	20	Severe HFrEF refractory to digoxin and diuretic agents.	3 h	Digoxin and diuretic agents	Rebound decrease in cardiac index and increase in LV filling pressure from baseline. Pulmonary edema in 3 (15%) patients.
Packer et al⁵ (1986)	Single-arm prospective trial	Prazosin	27	Severe HFrEF refractory to digoxin and diuretic agents, patients already on oral prazosin therapy.	48 h	Digoxin and diuretic agents	No changes in mean cardiac index with long-term therapy and subsequent withdrawal. No significant changes in hemodynamic parameters in 14 (58%) patients with longterm therapy; no changes following withdrawal in 13 patients.
Packer et al⁶ (1982)	Single-arm prospective trial	Hydralazine	11	HFrEF patients who experienced a loss of symptomatic efficacy of hydralazine during longterm treatment.	48 h	Digoxin and diuretic agents	No changes in mean cardiac index, LV filling pressure, heart rate, systemic vascular resistance, LVEF, or LV enddiastolic dimensions following withdrawal, and no response to dose escalation or intravenous drug administration.
Packer et al⁸ (1986)	Single-arm prospective trial	Transdermal nitroglycerin	22	Severe HFrEF refractory to digoxin and diuretic agents.	2 h	Digoxin and diuretic agents	Rebound decreases in cardiac index and rebound increases in mean arterial pressure and systemic vascular resistance after nitroglycerin discontinuation in first 30 min but hemodynamic parameters returned to pretreatment baseline after 2 h.
Wieshammer et al²⁵ (2001)	Randomized placebocontrolled withdrawal trial	Isosorbide dinitrate	29	HFrEF patients with history of myocardial infarction >3 mo before study.	6 wks	Captopril and diuretic agents	Reduction in LVEF, increase in LVEDV and LVESV with exercise in the nitrate withdrawal group, compared with placebo.

ACEI = angiotensin-converting enzyme inhibitor; AF = atrial fibrillation; ARB = angiotensin receptor blocker; BUN = blood urea nitrogen; GDMT = guideline-directed medical therapy; HFrEF = heart failure with reduced ejection fraction; LAV = left atrial volume; LVEDV = left ventricular end-diastolic volume; LVEF = left ventricular ejection fraction; LVESV = left ventricular end-systolic volume; MRA = mineralocorticoid receptor antagonist; NT-proBNP = N-terminal pro-brain natriuretic peptide; SBP = systolic blood pressure.

Table 2. Studies Reporting the Responses to Withdrawal of Cardiotoxic Positive Inotropic Drugs
First Author (Year)	Study Design	Drug Tested	N	Patient Characteristics	Duration of Withdrawal Period	Background Therapy	Responses Following Drug Discontinuation
DiBianco et al⁵¹ (1984)	Randomized, placebocontrolled, withdrawal trial	Amrinone	52	HFrEF patients with limited exercise capacity.	12 wks	Digoxin, loop diuretic agents, and vasodilators	Worsening in exercise times in amrinone withdrawal groups, but amrinone continuation group was worse than pretreatment baseline.
Packer et al¹¹ (1984)	Prospective singlearm study	Amrinone	31	Severe HFrEF patients refractory to digoxin and diuretic agents.	48 h	Digoxin and loop diuretic agents	Initial short-term improvement in hemodynamic and clinical parameters followed by worsening with long-term amrinone use and further deterioration following withdrawal.
Maskin et al⁴⁸ (1982)	Prospective singlearm study	Amrinone	6	Severe HFrEF patients on amrinone for 41 wks.	16-64 h	Digoxin, loop diuretic agents, and vasodilators	Reduction in cardiac index, increase in pulmonary capillary wedge pressure and severe symptomatic and hemodynamic deterioration in all patients.
Evans et al⁵² (1984)	Randomized, placebocontrolled, withdrawal trial	Amrinone	18	Severe HFrEF patients on amrinone for 4 mo.	16 wks	Digoxin, loop diuretic agents, and vasodilators	No change in clinical status, LVEF, exercise capacity, and functional class between amrinone continuation and discontinuation groups.
Ribeiro et al⁵⁰ (1986)	Prospective singlearm study	Milrinone	9	HFrEF patients with NYHA functional class III and IV symptoms.	24 h	Digoxin, loop diuretic agents, and vasodilators	Postwithdrawal decreases in maximal oxygen uptake and anaerobic threshold compared with pretreatment measurements.
Monrad et al⁴⁹ (1986)	Prospective singlearm study	Milrinone	13	HFrEF patients on milrinone for at least 3 mo.	36 h	ACEI, digoxin, and loop diuretic agents	No changes in cardiac index, left-, and right-sided filling pressures with milrinone discontinuation.
Rettig et al⁵³ (1986)	Randomized, placebocontrolled, crossover, withdrawal trial	Amrinone	14	HFrEF patients on amrinone for 8-15 mo.	12 wks	Digoxin and loop diuretic agents	No significant change in exercise capacity, LVEF, and systolic time intervals between amrinone continuation and discontinuation groups.
Packer et al¹⁰ (2017) (PROFILE)	Randomized controlled trial	Flosequinan	1,560	HFrEF patients with NYHA functional class III or IV symptoms.	28 d	ACEI, digoxin, and loop diuretic agents	Higher rates of worsening HF, urgent visits or hospitalization for HF, and need for intravenous diuretic agents, vasodilators, and inotropes in patients in whom flosequinan was withdrawn compared with patients in whom placebo was withdrawn.

PROFILE-HF = Prospective Randomized Flosequinan Longevity Evaluation; other abbreviations as in Table 1.

Table 3. Studies Reporting the Responses to Withdrawal of Digoxin
First Author (Year)	Study Design	Drug Tested	N	Patient Characteristics	Duration of Withdrawal Period	Background Therapy	Responses Following Drug Discontinuation
DiBianco et al⁴⁷ (1989) (The Multicenter Milrinone trial)	Randomized controlled withdrawal trial	Digoxin	230	HFrEF patients on chronic digoxin therapy. Patients randomized to the following: 1) digoxin continuation; 2) digoxin discontinuation; 3) addition of milrinone; and 4) replacement of digoxin with milrinone.	12 wks	Loop diuretic agents	Worsening of HF requiring withdrawal from the trial in patients in digoxin withdrawal group compared with digoxin continuation group, whether or not they received milrinone. Worsening exercise tolerance and LVEF in patients withdrawn from digoxin compared with those that continued digoxin.
Captopril-Digoxin Multicenter Research Group⁵⁸ (1988)	Randomized controlled withdrawal trial	Digoxin	300	HFrEF patients on chronic digoxin and diuretic therapy. Patients randomized to the following: 1) digoxin continuation; 2) placebo (digoxin discontinuation in 60%-70%); and 3) captopril without digoxin (digoxin discontinuation).	6 mo	Loop diuretic agents	Worsening HF requiring withdrawal from the study occurred more commonly in the placebo (digoxin discontinuation) vs digoxin continuation group. LVEF worsened significantly in the placebo (digoxin discontinuation) vs digoxin continuation group, not prevented by treatment with captopril.
German and Austrian Xamoterol trial⁵⁹ (1989)	Randomized controlled withdrawal trial	Digoxin	433	HFrEF patients on chronic digoxin therapy. Patients randomized to the following: 1) digoxin continuation; 2) placebo (digoxin discontinuation in 40%-50%); and 3) xamoterol without digoxin (digoxin discontinuation).	3 mo	Loop diuretic agents and nitrates	Symptoms, signs, and radiographic evidence of HF more common in the digoxin withdrawal as compared with the digoxin continuation group.
Uretsky et al⁵⁴ (1993) (PROVED)	Randomized controlled withdrawal trial	Digoxin	88	HFrEF patients on digoxin for at least 1 mo (ACEi excluded).	3 mo	Loop diuretic agents	Decrease in exercise time, higher rates of treatment failure, decrease in LVEF, increase in body weight and heart rate, and worsening kidney function in the digoxin discontinuation compared with digoxin continuation group.
Packer et al⁵⁵ (1993) (RADIANCE)	Randomized controlled withdrawal trial	Digoxin	178	HFrEF patients on digoxin for at least 3 mo (ACEi included).	3 mo	Loop diuretic agents and ACEI	Higher rates of worsening HF events necessitating withdrawal, worsening functional capacity, worsening maximal exercise tolerance, worsening submaximal exercise endurance, and decrease in LVEF in the digoxin discontinuation compared with digoxin continuation group.
Shammas et al⁶⁹ (2001)	Case series	Digoxin	8	HFrEF patients in sinus rhythm whose LVEF had normalized after treatment with betablockers.	7 mo	Beta-blockers	Significant reduction in LVEF after digoxin withdrawal.
Khand et al⁷⁰ (2003)	Randomized controlled trial	Digoxin	47	HFrEF patients with concomitant atrial fibrillation.	6 mo	Beta-blockers	Increase in heart rate and decline in LVEF in the digoxin discontinuation, despite use of background carvedilol treatment.
Ahmed et al⁵⁷ (2007)	Randomized controlled trial	Digoxin	3,365	Patients with HFrEF receiving diuretic agents and digoxin before randomization.	40 mo	Loop diuretic agents and ACEI	In patients with low serum digoxin concentration, digoxin continuation was accompanied by decreased HF and total hospitalizations and all-cause mortality compared with digoxin discontinuation.
Hopper et al⁶⁸ (2015)	Randomized controlled crossover trial	Digoxin	16	HFrEF patients on stable medical therapy, including digoxin.	3 mo	ACEI/ARB, betablockers, and MRA	Significant increase in NT-proBNP and decrease in 6-min walk distance in the digoxin withdrawal group.

DIG = Digitalis Investigation Group; PROVED = Randomized Study Assessing the Effect of Digoxin Withdrawal in Patients with Mild to Moderate Chronic Congestive Heart Failure; RADIANCE = Randomized Assessment of the effect of Digoxin on Inhibitors of the Angiotensin-Converting Enzyme; other abbreviations as in Table 1.

Table 4. Studies Reporting the Responses to Withdrawal of Diuretic Agents
First Author (Year)	Study Design	Drug Tested	Number of Patients	Patient Characteristics	Duration of Withdrawal Period	Background Therapy	Responses Following Drug Discontinuation
Richardson et al⁷² (1987)	Randomized crossover trial	Diuretic agents	14	HFrEF patients on diuretic agents for 8 wks; randomized to captopril or placebo groups.	8 wks	None	4 (29%) patients withdrew during diuretic discontinuation phase caused by worsening HF, not prevented by captopril.
Grinstead et al⁷³ (1994)	Randomized controlled trial	Diuretic agents	41	HFrEF patients on diuretic agents for at least 3 mo; randomized to captopril only and placebo groups.	12 wks	Loop diuretic agents	29 (72%) patients required reinitiation of diuretic agents.
Walma et al⁷⁵ (1997)	Randomized controlled withdrawal trial	Diuretic agents	84	HFrEF patients receiving diuretic agents for at least 6 mo.	6 mo	Loop diuretic agents	30 of 46 patients (65%) required reinitiation of diuretic agents in withdrawal group vs 4 of 38 patients (11%) required escalation of diuretic agents in continuation group.
Braunschweig et al⁷¹ (2002)	Prospective singlearm study	Diuretic agents	4	HFrEF patients on 3 mo of stable medical therapy withdrawn from diuretic agents.	2 wks	Beta-blocker, ACEi/ ARB, digoxin, and loop diuretic agents	Increase in RV systolic and diastolic pressures. Increase in pulmonary artery pressure. Decrease in exercise tolerance. Increase in natriuretic peptides. Worsening HF symptoms.
Galve et al⁷⁶ (2005)	Prospective singlearm study	Diuretic agents	26	HFrEF patients with stable NYHA functional class II symptoms.	3 mo	ACEi, digoxin, and diuretic agents	Reinitiation of diuretic agents in 9 patients (35%).
de Silva et al⁷⁷ (2007)	Prospective singlearm study	Diuretic agents	67	HFrEF patients with renal dysfunction on stable medical therapy for 8 wks.	4 wks	ACEi, beta-blocker, and loop diuretic agents	Worsening HF in 25 (37%) patients.
Rohde et al⁷⁸ (2019) (ReBIC-1)	Randomized controlled withdrawal trial	Diuretic agents	188	HFrEF patients receiving diuretic agents for at least 6 mo with no recent worsening HF events and optimized on ACEi/ARB/ARNI and beta-blockers.	90 d	ACEi/ARB/ARNI, beta-blocker, and MRA	No significant difference in dyspnea score or rate of diuretic reinitiation between the diuretic continuation and discontinuation groups, but patients on furosemide 80 mg (<20% of total) more frequently required reinstitution of therapy.

ReBIC = Rede Brasileira de Estudos em Insuficiência Cardíaca; other abbreviations as in Table 1.

Table 5. Studies Reporting the Responses to Withdrawal of Foundational Drugs for HFrEF
First Author (Year)	Study Design	Drug Tested	Number of Patients	Patient Characteristics	Duration of Withdrawal Period	Background Therapy	Responses following Drug discontinuation
Nicholls et al⁸⁵ (1982)	Case series	Captopril	5	Stable HFrEF patients on captopril therapy for mean 4.5 mo.	4 d	Digoxin and loop diuretic agents	Pulmonary edema in 2 of 5 patients. Increase in heart rate, SBP, angiotensin II, and aldosterone, but no changes in pulmonary artery pressure or cardiac output.
Swedberg et al⁸⁶ (1980)	Case series	Beta-blocker	15	HFrEF patients on betablocker for 6 mo.	72 d	Digoxin and loop diuretic agents	Clinical deterioration in 6 patients. Decrease in LVEF in 11 of 13 patients in whom it was measured.
Waagstein et al⁸⁷ (1989)	Case series	Beta-blocker	24	HFrEF patients on betablockers for 6 mo.	7.7 mo	Digoxin, loop diuretic agents, spironolactone, captopril	Clinical deterioration in 16 patients. Death in 4 patients. Decrease in LVEF. Increase in left atrial dimension.
Konstam et al⁹⁵ (1992) (SOLVD)	Placebocontrolled withdrawal trial	Enalapril	18	HFrEF patients on enalapril for mean 33 mo.	15 d	Digoxin, loop diuretic agents and non- ACEI vasodilators	4 vs 0 major cardiovascular events in the enalapril discontinuation and placebo withdrawal group, respectively. No between-group differences in LVESV and LVEDV between the enalapril and placebo discontinuation groups.
Pflugfelder et al⁹⁶ (1993)	Placebocontrolled withdrawal trial	Quinapril	224	HFrEF patients with NYHA II and III symptoms who were maintained on quinapril therapy for 10 wks.	16 wks	Digoxin and diuretic agents	Reduction in exercise tolerance and mild worsening of quality-of-life scores in quinapril discontinuation vs quinapril continuation groups. Higher proportion of patients in the quinapril discontinuation group (18 of 110 patients vs 5 of 114 patients; P < 0.001) withdrew because of worsening HF.
Morimoto et al⁸⁸ (1999)	Case series	Beta-blocker	13	HFrEF patients on betablockers for ≥30 mo.	3.2 mo	Digoxin and loop diuretic agents	Worsening HF in 7 patients (including 4 deaths). Decrease in LVEF. Increase in LV diastolic dimension.
Dovancescu et al⁹⁷ (2017)	Randomized crossover trial	Abrupt discontinuation of multiple medications	20	Ambulatory HFrEF patients on and loop diuretic agents.	48 ho	ACEI/ARB, MRA, beta-blockers, digoxin, and loop diuretic agents	Increase in NT-proBNP, SBP, and left atrial volume in withdrawal group with worsening functional capacity in 6 patients in withdrawal group.
Halliday et al⁹⁹ (2019) (TRED-HF)	Crossover, randomized controlled trial	Phased discontinuation of multiple medications	50	HFrEF patients (LVEF <40%) with normalized LVEF >50% and LVEDV with NTproBNP <250 ng/L.	6 mo	ACEI/ARB, MRA, beta-blockers, and loop diuretic agents	Overall, 40% patients showed decrease in LVEF and increase in LVEDV, LAV, heart rate, SBP, and NT-proBNP followed withdrawal compared with control group.
Nijst et al⁹⁸ (2020) (STOP-CRT)	Randomized controlled withdrawal trial	ACEi/ARB, betablockers, and MRA	80	HFrEF patients with recovered LV function after CRT randomized to the following: 1) continuation of neurohumoral blockade; 2) withdrawal of RAAS inhibitors; 3) withdrawal of betablockers; and 4) withdrawal of both RAAS inhibitors and beta-blockers.	24 mo	ACEI/ARB, betablockers, and MRA	No difference in LVESV index among the 4 treatment groups. No difference in composite of allcause deaths and HF hospitalizations among the 4 groups.
Packer et al¹⁰⁰ (2023)	Randomized controlled withdrawal trial	SGLT2 inhibitor (empagliflozin)	3,981	HFrEF and HFpEF patients on empagliflozin for 1-3 y.	30 d	ACEI/ARB/ARNI, beta-blockers, and MRA	Increase in cardiovascular death and hospitalization in the empagliflozin withdrawal group compared with no change in placebo withdrawal group. Worsening KCCQ scores following empagliflozin withdrawal compared with placebo withdrawal.

ARNI = angiotensin receptor neprilysin inhibitor; CRT = cardiac resynchronization therapy; eGFR = estimated glomerular filtration rate; HFpEF: heart failure with preserved ejection fraction; KCCQ = Kansas City Cardiomyopathy Score; SGLT-2 = sodium-glucose cotransporter-2; SOLVD = Studies of Left Ventricular Dysfunction; STOP-CRT: Systematic Withdrawal of Neurohumoral Blocker Therapy in Optimally Responding CRT Patients; TRED-HF = Withdrawal of pharmacological treatment for heart failure in patients with recovered dilated cardiomyopathy; other abbreviations as in Table 1.