Minimization or withdrawal of oral pharmacotherapy in chronic heart failure patients with improved myocardial function: A systematic review

Abstract

Aims

The necessity of lifelong treatment and polypharmacy in chronic heart failure (HF) patients with improved myocardial function remains debated. This systematic review aims to synthesize current literature regarding this issue.

Methods and results

A systematic literature search was performed in MEDLINE, Embase, and Cochrane Central Register of Controlled Trials from the inception to 18 October 2024. Seven studies (n = 552) reporting minimization or withdrawal of pharmacotherapy in chronic HF patients with improved ejection fraction or stable New York Heart Association status were included. Findings were heterogeneous due to variations in study design and protocols. Loop diuretic withdrawal was favoured by one non‐randomized study (n = 26) and one randomized controlled trial (RCT) (n = 188). Minimization of angiotensin receptor–neprilysin inhibitors (n = 77) or withdrawal of mineralocorticoid receptor antagonists (MRA) (n = 70) was not favourable. Carvedilol monotherapy was favoured by one small‐sample RCT (n = 60). One RCT (n = 51) reported a high overall relapse rate (65%) following multiple drug withdrawal in recovered patients with dilated cardiomyopathy. Another RCT (n = 80) found a low occurrence of cardiac dimensional deterioration (7.5%) following multiple drug withdrawal in post‐cardiac resynchronization therapy patients with normalized ejection fraction. However, 28% required drug re‐initiation due to cardiac comorbidities.

Conclusion

The existing evidence on minimizing or withdrawing oral pharmacotherapy in chronic HF patients with improved myocardial function remains very limited and heterogeneous, supporting only loop diuretic withdrawal and possibly carvedilol monotherapy, but not the minimization or withdrawal of renin–angiotensin system inhibitors, MRA, or the combination of HF medications. Large RCTs are needed to determine the appropriate treatment strategy.

Introduction

Heart failure (HF), a clinical syndrome involving structural and functional abnormality of the heart, is affecting more than 64 million people worldwide. ¹ With the advancement of modern biomedical science, remarkable progress has been made in the management of chronic HF patients. ² Pharmacotherapy is regarded as the cornerstone of HF treatment, which has been shown to reduce mortality, prevent recurrent hospitalizations, and improve clinical status, functional capacity, and quality of life. ² , ³ , ⁴ , ⁵ For a significant number of HF patients who receive guideline‐directed medical therapy, it is possible not only to stabilize symptoms, but often to observe significant improvements in myocardial function. ⁶ These patients are typically identified by an increase in left ventricular ejection fraction (LVEF) and are referred to as HF with improved ejection fraction (HFimpEF). According to the universal definition of HF, HFimpEF is characterized by a baseline LVEF ≤40%, a ≥10% absolute increase from this baseline, and a second measurement >40%. ⁷

Heart failure patients with improved myocardial function show a better prognosis compared to those with other classes of HF. ⁸ Upon the resolution of symptoms and the restoration of cardiac function, the necessity of continuation in lifelong treatment is under discussion. Patients frequently express reluctance to adhere to prolonged polypharmacy due to concerns regarding potential side effects. ⁹ Reducing unnecessary medications may enhance long‐term adherence and lower healthcare costs for both patients and society. ¹⁰ Various studies have investigated the optimization of HF pharmacotherapy in this population; however, findings have been inconsistent due to the heterogeneous nature of the studies and their diverse emphases. ¹¹ , ¹² Hence, a systematic review of the existing literature on this topic is warranted. This systematic review assessed the feasibility, safety, and clinical outcomes of minimizing or withdrawing oral pharmacotherapy in HF patients with improved myocardial function. It aims to improve pharmacological management, review the scope of existing research, and suggest future research directions based on identified gaps in the field.

Methods

This systematic review was registered at PROSPERO (https://www.crd.york.ac.uk/PROSPERO/; registration ID: CRD42024611161), and was reported following the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) statement. ¹³ Databases including MEDLINE, Embase, and Cochrane Central Register of Controlled Trials were searched from inception to 18 October 2024. The search terms and strategies (online supplementary Table  S1 ) were formulated and provided by the biomedical information specialist from our institute.

Study selection

Search results were screened with Covidence (Melbourne, Australia) after removal of duplicate records. Titles and abstracts of identified records were screened by two independent reviewers (Y.L. and W.X.) to select potentially eligible studies, followed by a full‐text review to confirm inclusion. Additionally, reference lists of the retrieved studies and potentially relevant congress presentations were screened manually to identify any further relevant literature. Disagreements were resolved through consensus with the supervisor (K.C.). Studies reporting minimized or withdrawn pharmacotherapy outcomes of chronic HF patients with improved LVEF or stable New York Heart Association (NYHA) status, regardless of the aetiology, and involving participants aged over 18, were included. The definition of LVEF improvement as applied in each individual study was acknowledged. Stable NYHA status was defined as consistently remaining in NYHA class I or II for at least 3 months. Detailed definitions of myocardial function improvement and primary outcomes of the included studies are presented in online supplementary Table  S2 . Studies were excluded if the minimization or withdrawal of pharmacotherapy was caused by adverse events. Case reports with fewer than 10 cases, review articles, editorials, commentaries, and guidelines were also excluded. Considering time and efficiency constraints, articles written in languages other than English were not included.

Information collection

The metadata of the included studies were collected, including the name of the first author, year of publication, total number of participants, study intervention, publication source, study site, study design, and study registration. Baseline characteristics of the recruited patients were collected, including demographic data (age, gender, aetiology, and comorbidities) and cardiovascular parameters (heart rate, blood pressure, N‐terminal pro‐B‐type natriuretic peptide [NT‐proBNP], baseline LVEF, and baseline left ventricular end‐diastolic diameter [LVEDD]), documented for both the intervention group and the control group. Protocols and main findings of the included studies were recorded and listed in a table containing information about the aetiology of HF, key inclusion criteria, detailed intervention protocol, study duration, primary outcomes, main findings, and conclusion. Non‐randomized studies and randomized controlled trials (RCTs) were documented separately in each table. Dichotomous variables are presented in numbers and percentages. Continuous variables are presented in means ± standard deviations or medians with interquartile ranges.

Quality assessment

The quality of the included studies in this systematic review was assessed by two independent reviewers (Y.L. and W.X.), with the supervisor (K.C.) solving the disagreements. For non‐randomized studies, the Newcastle–Ottawa scale was employed to assess the methodological quality and risk of bias. ¹⁴ Each study was assigned a score reflecting its overall quality with a maximum of 9 points, and classified as low (0–3 points), moderate (4–6 points), or high quality (7–9 points). Higher scores indicate a lower risk of bias. For RCTs, the Cochrane risk of bias 2 tool was employed to assess the risk of bias. ¹⁵ The overall risk of bias for each study was classified as low, high, or unclear.

Results

A total of 5324 records from databases were screened after removing the duplicates. After excluding studies through record screening and adding others via reference screening and additional sourcing, 56 studies were assessed for eligibility through full‐text review. Forty‐nine studies were further excluded, and seven studies were included in the systematic review, consisting of three non‐randomized studies ¹⁶ , ¹⁷ , ¹⁸ and four RCTs, ¹¹ , ¹² , ¹⁹ , ²⁰ with a total of 552 patients. A PRISMA flow diagram of the literature search is presented in Figure  1 .

Figure 1.

Flow diagram of the literature search. Exclusion criteria: case reports with fewer than 10 cases, review articles, editorials, commentaries, guidelines, and non‐English articles.

Study metadata

Of the seven included studies, six were published within the last 5 years and one was published in 2005. The largest study ¹⁹ involved a total of 188 patients, while the smallest ¹⁶ had only 26 enrolled patients. The included studies were conducted in various locations, including two in Asia, ¹⁷ , ¹⁸ three in Europe, ¹¹ , ¹² , ¹⁶ and two in South America (Brazil), ¹⁹ , ²⁰ with four in outpatient settings and three conducted via in‐hospital settings. Five are single‐centre studies and two are multicentre studies. The vast majority are prospective studies. Table  1 presents the detailed metadata information of each included study.

Table 1.

Metadata of the included studies

Author	Year of publication	Total, n	Study intervention	Study site	Study design	Study registration
Non‐randomized studies
Galve et al. 16	2005	26	Loop diuretic withdrawal	Spain, single‐centre, outpatient	Prospective, non‐randomized	NA
Chang et al. 17	2020	77 a	ARNi minimization	China, single‐centre, outpatient	Retrospective, non‐randomized, observational	NA
Chen et al. 18	2021	70	MRA withdrawal	China, single‐centre, outpatient	Prospective, non‐randomized, observational, open‐label	NCT02998788 (REM‐HF trial)
Randomized controlled trials
Rohde et al. 19	2019	188	Loop diuretic withdrawal	Brazil, multicentre, outpatient	Prospective, randomized, placebo‐controlled, double‐blind	NCT02689180 (ReBIC‐1 trial)
Halliday et al. 11	2019	51	Multiple drug minimization	United Kingdom, single‐centre, hospital	Prospective, randomized, pilot, open‐label	NCT02859311 (TRED‐HF trial)
Nijst et al. 12	2020	80	Multiple drug minimization	Belgium, multicentre, hospital	Prospective, randomized, pilot, open‐label	NCT02200822 (STOP‐CRT trial)
Belfort et al. 20	2024	60	Multiple drug minimization	Brazil, single‐centre, hospital	Prospective, randomized, pilot, open‐label	CATHEDRAL‐HF trial

ARNi, angiotensin receptor–neprilysin inhibitor; MRA, mineralocorticoid receptor antagonist; NA, not available.

^a Only patients with recovered left ventricular ejection fraction were calculated.

Baseline characteristics

Loop diuretic withdrawal was assessed by one non‐randomized study (n = 26) ¹⁶ and one RCT (n = 188), ¹⁹ respectively. Minimization of angiotensin receptor–neprilysin inhibitors (ARNi) (n = 77) ¹⁷ and withdrawal of mineralocorticoid receptor antagonists (MRA) (n = 70) ¹⁸ were each evaluated by one non‐randomized study. The remaining three RCTs investigated the outcomes of multiple drug minimization or withdrawal. ¹¹ , ¹² , ²⁰ A total of five classes of HF medications were minimized or withdrawn across the included studies; however, no study specifically assessed the single drug minimization of angiotensin‐converting enzyme inhibitors (ACEi)/angiotensin receptor blockers (ARB) or beta‐blockers. The classes of medications evaluated in each study are listed in online supplementary Table  S3 .

Table  2 presents the baseline demographics of the patients in the included studies. Among the recruited patients, the age distribution predominantly ranged from 54 to 70 years, with a higher predominance of males. According to the studies that provide aetiology information, the proportion of ischaemic causes is relatively low. The RCT conducted by Belfort et al. ²⁰ included a notable portion of patients with hypertension, whereas comorbidities were less prevalent in the remaining studies. Table  3 presents the baseline cardiovascular parameters of the patients in the included studies. In studies where heart rate was recorded, the recruited patients typically exhibited values ranging from 60 to 80 bpm. According to the studies in which blood pressure was recorded, values were predominantly at or slightly above the upper limit of the reference range. NT‐proBNP levels exhibited substantial between‐group variability, ranging from under 100 to over 700 pg/ml. In studies where loop diuretics were withdrawn, the mean or median baseline LVEF was lower, ranging from 30% to 40%; in contrast, other studies reported a mean or median baseline LVEF >40% across all groups, with a maximum of 62%. Measurements of baseline LVEDD appeared to be comparable between studies.

Table 2.

Baseline characteristics of the patients in the included studies

Author	Year of publication	Total, n	Study intervention	Patients, n (%)		Age (years)		Men, n (%)		Ischaemic, n (%)		Diabetes, n (%)		Atrial fibrillation, n (%)		Hypertension, n (%)
				Intervention	Control	Intervention	Control	Intervention	Control	Intervention	Control	Intervention	Control	Intervention	Control	Intervention	Control
Non‐randomized studies
Galve et al. 16	2005	26	Loop diuretic withdrawal	17 (65)	9 (35)	61 ± 14	61 ± 16	12 (71)	8 (89)	8 (47)	3 (33)	NA	NA	2 (12)	2 (22)	NA	NA
Chang et al. 17	2020	77 a	ARNi minimization	26 (34)	51 (66)	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
Chen et al. 18	2021	70	MRA withdrawal	40 (57)	30 (43)	57 ± 16	56 ± 13	34 (85)	25 (83)	0 (0)	0 (0)	8 (20)	6 (20)	2 (5)	0 (0)	NA	NA
Randomized controlled trials
Rohde et al. 19	2019	188	Loop diuretic withdrawal	95 (50)	93 (50)	61 (52–67)	61 (54–68)	70 (74)	70 (75)	34 (36)	29 (31)	29 (31)	23 (25)	15 (16)	9 (10)	NA	NA
Halliday et al. 11	2019	51	Multiple drug minimization	25 (49)	26 (51)	54 (46–64)	56 (45–64)	16 (64)	18 (69)	0 (0)	0 (0)	0 (0)	1 (4)	8 (32)	4 (15)	3 (12)	1 (4)
Nijst et al. 12	2020	80	Multiple drug minimization	20 (50) b	20 (50)	68 (61–75)	70 (67–77)	7 (35)	13 (65)	NA	NA	1 (5)	3 (15)	0 (0)	7 (35)	5 (25)	5 (25)
Belfort et al. 20	2024	60	Multiple drug minimization	30 (50)	30 (50)	54 (48–59)	56 (47–69)	17 (57)	15 (50)	18 (60)	17 (57)	3 (10)	2 (7)	NA	NA	14 (47)	19 (63)

Dichotomous variables are presented as n (%) and continuous variables are presented as mean ± standard deviations or median (interquartile range).

ARNi, angiotensin receptor–neprilysin inhibitor; MRA, mineralocorticoid receptor antagonist; NA, not available.

^a Only patients with recovered left ventricular ejection fraction were calculated.

^b The withdrawal of renin–angiotensin–aldosterone system inhibitors and beta‐blockers group in this study was regarded as the intervention group.

Table 3.

Baseline cardiovascular parameters of the patients in the included studies

Author	Year of publication	Total, n	Study intervention	Heart rate (bpm)		Systolic blood pressure (mmHg)		Diastolic blood pressure (mmHg)		NT‐proBNP (pg/ml)		LVEF (%)		LVEDD (mm)
				Intervention	Control	Intervention	Control	Intervention	Control	Intervention	Control	Intervention	Control	Intervention	Control
Non‐randomized studies
Galve et al. 16	2005	26	Loop diuretic withdrawal	78 ± 14	72 ± 9	135 ± 22	122 ± 13	80 ± 11	77 ± 5	NA	NA	36 ± 13	30 ± 4	60 ± 11	66 ± 8
Chang et al. 17	2020	77 a	ARNi minimization	NA	NA	NA	NA	NA	NA	NA	NA	54 ± 7	55 ± 5	50 ± 6	49 ± 5
Chen et al. 18	2021	70	MRA withdrawal	NA	NA	131 ± 21	130 ± 24	NA	NA	721 (371–1793)	598 (298–1609)	44 ± 5	47 ± 7	60 ± 7	61 ± 8
Randomized controlled trials
Rohde et al. 19	2019	188	Loop diuretic withdrawal	NA	NA	120 (110–140)	120 (110–135)	NA	NA	644 (287–1527)	696 (282–1776)	32 (26–40)	32 (25–38)	NA	NA
Halliday et al. 11	2019	51	Multiple drug minimization	62 (58–74)	70 (60–75)	123 (117–133)	127 (117–134)	72 (68–80)	76 (70–80)	72 (44–147)	75 (37–133)	62 (55–66)	60 (55–61)	NA	NA
Nijst et al. 12	2020	80	Multiple drug minimization	64 (60–70)	63 (60–69)	132 (110–140)	130 (111–133)	72 (63–79)	71 (63–83)	169 (124–399)	335 (105–766)	56 (53–59)	55 (53–59)	NA	NA
Belfort et al. 20	2024	60	Multiple drug minimization	76 (68–82)	69 (64–74)	129 (117–135)	128 (117–134)	79 (74–87)	82 (70–84)	77 (33–137)	95 (44–143)	57 (54–62)	59 (54–64)	50 (47–54)	51 (44–56)

Dichotomous variables are presented as n (%) and continuous variables are presented as mean ± standard deviations or median (interquartile range).

ARNi, angiotensin receptor–neprilysin inhibitor; LVEDD, left ventricular end‐diastolic diameter; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; NA, not available; NT‐proBNP, N‐terminal pro‐B‐type natriuretic peptide.

^a Only patients with recovered LVEF were calculated.

Risk of bias

The quality assessment using the Newcastle–Ottawa scale suggested that two of the included non‐randomized observational studies ¹⁷ , ¹⁸ were of high quality, while the other one ¹⁶ was classified as having moderate quality (online supplementary Table  S4A ). The quality assessment using the Cochrane risk of bias 2 tool indicated an overall low risk of bias for all included RCTs ¹¹ , ¹² , ¹⁹ , ²⁰ (online supplementary Table  S4B ), reflecting their sound design and rigorous methodology.

Study findings

The study protocols and main findings of the included studies are presented in Table  4 . The aetiology of HF is heterogeneous in each study. Additionally, the relapse and recovery were reported with varying definitions. Four studies involved patients with mixed aetiologies, while two studies specifically targeted patients with dilated cardiomyopathy (DCM). One study focused exclusively on HF patients who underwent cardiac resynchronization therapy (CRT). The inclusion criteria varied across studies. Apart from two studies evaluating loop diuretic withdrawal, the others employed LVEF to define myocardial function improvement. In the loop diuretic withdrawal studies, the key inclusion criterion was stable status assessed by NYHA class (see online supplementary Table  S2 for details). For interventions, the majority of studies were to withdraw the medications being investigated, whereas in the study conducted by Chang et al. ¹⁷ the use of ARNi was tapered or switched to ACEi/ARB. The RCT conducted by Belfort et al. ²⁰ was to withdraw other HF medications but maintain carvedilol and evaluate its role as single‐agent maintenance therapy in chronic HF management. The study duration ranged from 3 to 60 months. ¹⁹ , ²¹

Table 4.

Protocols and main findings of the included studies

Author	Year of publication	Total, n	Intervention protocol	Aetiology of HF	Key inclusion criteria	Study duration	Primary outcomes	Main findings	Conclusion
Non‐randomized studies
Galve et al. 16	2005	26	Withdrawal of loop diuretics (furosemide)	Mixed	(1) NYHA class II (2) LVEF <45%	12 months	(1) Clinical change (2) Modification in neurohumoral parameters	Improvement in renal function, glucose metabolism, and neurohumoral parameters following withdrawal	Favourable for loop diuretic withdrawal
Chang et al. 17	2020	77 a	Tapering the dose of ARNi (sacubitril/valsartan) or switching from ARNi to ACEi/ARB	Mixed	Baseline LVEF <40%, follow‐up LVEF ≥50%	24 months	(1) Follow‐up LVEF (2) Cardiovascular death (3) HF‐related hospitalization	LVEF deterioration, more frequent cardiovascular death or HF‐related hospitalization following minimization	Unfavourable for ARNi minimization
Chen et al. 18	2021	70	Withdrawal of MRA (spironolactone)	DCM	Baseline LVEF <40%, ΔLVEF ≥10%, follow‐up LVEF >40%	12 months	Relapse of DCM	(1) Increased risks of DCM relapses and symptom aggravations following withdrawal (2) Cessation of cardiac structure improvement following withdrawal	Unfavourable for MRA withdrawal
Randomized controlled trials
Rohde et al. 19	2019	188	Withdrawal of loop diuretics (furosemide)	Mixed	(1) NYHA class I or II (2) LVEF ≤45%	3 months	(1) Dyspnoea (2) Intolerance to furosemide withdrawal	No significant difference in dyspnoea, intolerance to withdrawal, hospitalization, emergency visit, or death between groups	Favourable for loop diuretic withdrawal
Halliday et al. 11	2019	51	Sequential withdrawal of loop diuretics, MRA, beta‐blockers, and ACEi/ARB	DCM	Baseline LVEF ≤40%, follow‐up LVEF ≥50%	60 months b	Relapse of DCM	(1) 6‐month relapse rate following withdrawal: 44% (2) Overall relapse rate following withdrawal: 65% b	Unfavourable for multiple drug minimization
Nijst et al. 12	2020	80	Sequential withdrawal of loop diuretics, MRA, ACEi/ARB, and beta‐blockers	Post‐CRT HF	(1) NYHA class I or II (2) LVEF ≥50%	24 months	Recurrence of negative remodelling	(1) Low occurrence of cardiac dimensional deterioration (7.5%) (2) Drug re‐initiation required in 28% of patients due to cardiac comorbidities	Partially favourable for multiple drug minimization in the setting of CRT
Belfort et al. 20	2024	60	Sequential withdrawal of loop diuretics, MRA, and ACEi/ARB, while maintaining carvedilol	Mixed	Baseline LVEF ≤40%, follow‐up LVEF ≥50%	13 months	Recurrence of ventricular dysfunction	No significant difference in recurrence of ventricular dysfunction, hospitalization, death, or sustained arrhythmia between groups	Favourable for carvedilol monotherapy

ACEi, angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; ARNi, angiotensin receptor–neprilysin inhibitor; CRT, cardiac resynchronization therapy; DCM, dilated cardiomyopathy; HF, heart failure; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; NYHA, New York Heart Association.

^a Only patients with recovered LVEF were calculated.

^b Long‐term follow‐up data were accessed from the publication by Cheng et al. ²¹

Four studies designated ‘relapse’ or ‘recurrence’ as the primary outcomes. The remaining studies primarily focused on clinical outcomes such as LVEF change, symptoms, mortality, and hospitalization (see online supplementary Table  S2 for details). Because the included studies focused on different outcomes, the descriptions of the findings were highly varied. However, each study typically reached a conclusion that either supports or opposes the intervention under investigation. The feasibility of loop diuretic withdrawal was supported by one non‐randomized study and one RCT. ¹⁶ , ¹⁹ Single minimization or withdrawal of ARNi and MRA was not favourable by non‐randomized studies. ¹⁷ , ¹⁸ A small‐sample RCT by Belfort et al. ²⁰ was favourable for carvedilol monotherapy. The RCT by Halliday et al. ¹¹ reported a high overall relapse rate (65%) following multiple drug withdrawal, suggesting the need to maintain pharmacotherapy in recovered DCM patients. Nijst et al. ¹² found a low occurrence of cardiac dimensional deterioration (7.5%) following multiple drug withdrawal in post‐CRT patients with normalized ejection fraction. This study observed only partially favourable outcomes for multiple drug withdrawal in the setting of CRT, as 28% of patients required drug re‐initiation due to cardiac comorbidities.

Discussion

To the best of our knowledge, this is the first systematic review summarizing the existing literature on minimizing or withdrawing oral pharmacotherapy in chronic HF patients with improved myocardial function. We identified seven studies evaluating the feasibility and clinical outcomes of HF drug minimization or withdrawal, but the heterogeneity and limited number of studies are insufficient for drawing definitive conclusions. Based on the findings of this review, it seems that large‐scale minimization or withdrawal of pharmacotherapy in chronic HF patients with improved myocardial function should not be attempted before firm evidence is generated by large RCTs (Graphical Abstract).

Previous reviews focused on drug withdrawal in older HF patients or stable chronic HF patients, who were conceptually different from chronic HF patients with improved myocardial function. ²² , ²³ , ²⁴ , ²⁵ The improvement of myocardial function refers to the normalization of molecular, cellular, and left ventricular geometric changes, along with the sign of cardiac reverse remodelling and the increase in LVEF. ²⁶ Compared with previous study populations, such as older HF patients or those with reduced ejection fraction, patients with improved myocardial function tend to be younger and have a lower prevalence of comorbidity, ischaemic aetiology, or myocardial fibrosis, which contributes to an overall better prognosis. ⁶ , ²⁷ , ²⁸ , ²⁹ Thus, their request for minimizing oral pharmacotherapy should have been met with a more positive response.

Loop diuretics

Loop diuretics were widely prescribed for HF patients to relieve congestion, improve symptoms, and prevent worsening conditions. ³⁰ However, the need for diuretic therapy in stable HF patients is debated due to its negative effect on neurohormonal activation. ²⁴ , ³¹ The two studies discussing diuretic withdrawal in this review included chronic HF patients with stable NYHA status, rather than those with improved LVEF. They both concluded that loop diuretics can be safely discontinued, consistent with findings from other unincluded studies. ³² , ³³ , ³⁴ , ³⁵ Notably, most of these studies focused on loop diuretic withdrawal in older, stable HF patients. While it is reasonable to apply these findings to chronic HF patients with improved myocardial function due to their better cardiac performance compared to older patients, further investigation is still needed to assess the feasibility of loop diuretic withdrawal in this specific study population.

Angiotensin‐converting enzyme inhibitors/angiotensin receptor blockers

Before the recent introduction of ARNi, ACEi and ARB were core medications for inhibiting the renin–angiotensin system. ³⁶ There is no study that evaluated the single drug withdrawal of ACEi/ARB in this review. However, a Korean registry study, ³⁷ which we excluded due to the discrepancy driven by acute aetiology, evaluated the impact of discontinuing renin–angiotensin system inhibitors (RASi) in HF patients with restored ejection fraction after acute myocardial infarction. It concluded that RASi withdrawal was associated with a significantly increased risk of all‐cause death, myocardial infarction, and HF‐related hospitalization. Even though the myocardial dysfunction of these patients was often rescued by revascularization, RASi still show protective effects on this population, underlining the necessity of consistent neurohormonal regulation driven by inhibiting the renin–angiotensin system.

Angiotensin receptor–neprilysin inhibitors

The PARADIGM‐HF trial demonstrated that ARNi is superior to ACEi in improving the prognosis of HF patients, leading the current European Society of Cardiology and American Heart Association HF guidelines to recommend ARNi as the preferred choice among RASi medications. ² , ³⁰ , ³⁸ In this review, only Chang et al. ¹⁷ explored the effect of tapering the dose of ARNi or switching from ARNi to ACEi/ARB in chronic HF patients with improved myocardial function. In their study, even partial adjustments to the use of ARNi, such as dose reduction or change of medication rather than complete withdrawal, have shown worsened clinical outcomes. This highlights the critical role of ARNi in chronic HF treatment, even in cases where myocardial function is improved. Based on current evidence, it is not recommended to completely stop RASi in chronic HF patients with improved myocardial function. For patients unsuitable for ARNi, introducing ACEi/ARB may still be required. ² , ³⁰

Beta‐blockers

Beta‐blockers were recommended to reduce mortality and hospitalizations for chronic HF patients. ³⁹ , ⁴⁰ , ⁴¹ Among all seven studies included in this review, none of them evaluated the effect of single beta‐blocker withdrawal. Only in the PRESERVE‐HF trial involving HF patients with preserved ejection fraction was beta‐blocker withdrawal found to be positive for short‐term functional capacity improvement. ⁴² A study conducted by Enzan et al. ⁴³ analysed the clinical impact of beta‐blocker use in patients with recovered DCM, concluding that beta‐blockers can prevent left ventricular systolic function from deteriorating. Meanwhile, Halliday et al. ¹¹ demonstrated that the rise in heart rate following beta‐blocker withdrawal was most pronounced in HF patients who relapsed through the TRED‐HF trial, emphasizing the critical role of beta‐blockers in stabilizing heart rate and potentially mitigating relapse risk. Another included RCT, the CATHEDRAL‐HF trial conducted by Belfort et al., ²⁰ examined the non‐inferiority of carvedilol monotherapy compared to conventional polypharmacy in patients with HFimpEF. These findings strongly underscore the importance of beta‐blocker therapy in optimizing oral pharmacotherapy for chronic HF patients and highlight its potential as a viable option for female patients with pregnancy wishes, given that RASi is teratogenic. ⁴⁴

Mineralocorticoid receptor antagonists

Mineralocorticoid receptor antagonists were proven to significantly reduce mortality and hospitalizations in HF patients with reduced ejection fraction. ⁴⁵ , ⁴⁶ However, it was unknown whether HF patients with improved myocardial function would benefit from MRA withdrawal. Chen et al. ¹⁸ discovered that most DCM patients with improved ejection fraction would relapse after spironolactone withdrawal, suggesting the necessity of MRA maintenance. Nevertheless, the long‐term use of MRA requires careful monitoring due to potential risk of hyperkalaemia and renal insufficiency. The minimization of MRA in specific patient populations is still worth exploring. ⁴⁷ , ⁴⁸

Other heart failure medications

Current research is still mainly focused on the optimization of core medications. Discussions on the minimization or withdrawal of other HF medications following improved myocardial function remain limited. Sodium–glucose co‐transporter 2 inhibitors (SGLT2i) such as dapagliflozin or empagliflozin were listed as a novel vital class of HF medications for reducing the risk of rehospitalization and mortality. ⁴⁹ , ⁵⁰ Limited studies have revealed that SGLT2i withdrawal did not demonstrate benefits for HF patients. However, these patients did not exhibit formal myocardial improvement. ⁵¹ , ⁵² Digoxin withdrawal is worth considering, as it has not been shown to reduce mortality. ⁵³ , ⁵⁴ With the introduction of guideline‐directed medical therapies, the contemporary benefits of digoxin for patients with improved myocardial function should be questioned. ³⁰ There is also no study evaluating the safety or feasibility of discontinuing I _f‐channel inhibitor (ivabradine) or soluble guanylate cyclase stimulator (vericiguat) in chronic HF patients with improved myocardial function. In many instances, these medications have not been prescribed, making it unavailable to conduct clinical trials and further investigate the clinical impact following their withdrawal.

Future perspectives

The heterogeneous findings given by the included studies in this review reflected that, without a clear identification of the target population or a solid understanding of pharmacological mechanisms on how to sustain myocardial functional homeostasis, conclusions drawn after drug minimization—whether positive or negative—are inherently biased and overly simplistic.

Future exploration on minimizing oral pharmacotherapy in chronic HF patients with improved myocardial function should focus on addressing the following aspects. The primary task is to clarify the definition of myocardial function improvement. Currently, each study generally established its own definition and used varying terminologies to describe myocardial improvement, complicating the identification of the target population. ⁷ , ³⁰ , ⁵⁵ Research has demonstrated that neurohormonal activation and abnormal biomarker profile persist even when LVEF normalizes. ⁵⁶ This indicates that effective reverse remodelling and LVEF recovery generally signify a ‘remission’ of myocardial dysfunction rather than a complete cure of the underlying heart disease. ⁵⁷ Therefore, relying solely on LVEF measurement seems not sufficient. Personalized pharmacological management should be introduced through the adoption of more advanced evaluation methods, such as global longitudinal strain measurement, ⁵⁸ , ⁵⁹ cardiovascular magnetic resonance imaging, ⁶⁰ , ⁶¹ and novel biomarkers. ⁶² , ⁶³ , ⁶⁴ , ⁶⁵ The development of artificial intelligence may aid in identifying patient subgroups and managing heart failure. ⁶⁶ , ⁶⁷ Only with a deeper understanding and improved management of HF can pharmacological minimization or withdrawal be scientifically guided.

Another aspect is the renewal of the concept updated by future clinical trials. The TRED‐HF trial has proven that too aggressive a withdrawal regimen is undesirable and will lead to a worsening of the condition. ¹¹ Further clinical trials and large‐scale database analyses are warranted to address the critical questions regarding the optimal timing for drug withdrawal, the specific classes of HF medications to be discontinued, and whether the pharmacotherapy should be completely withdrawn or whether dosage down‐titration would be more appropriate.

Limitations

The limitations of this systematic review primarily stemmed from the small number of studies included, which restricts the generalizability of the findings and their applicability across diverse clinical settings. To broaden the scope and identify more studies for synthesis, this review adopted a relatively inclusive definition of myocardial function improvement. We accepted the criteria for LVEF improvement as reported in each individual study and included studies involving populations with stable NYHA class I or II. While this approach facilitated a more comprehensive exploration of strategies to minimize HF pharmacotherapy, it may have, to some extent, compromised the specificity and precision of the review. The absence of pooled estimates or statistical analyses further constrained the ability to quantify effect sizes due to the substantial heterogeneity of the included studies. The variability among studies regarding the aetiologies of HF, inclusion criteria, intervention protocols, and recorded clinical outcomes significantly complicates the interpretation of the findings. Although the non‐randomized observational studies included in this review are of high quality, their findings should be interpreted with extreme caution, as their drug withdrawal may have occurred for a myriad of reasons. Most critically, the types of HF medications evaluated vary in each study, undermining the ability to draw unified conclusions.

Conclusion

The literature on minimizing or withdrawing oral pharmacotherapy in chronic HF patients with improved myocardial function is currently limited and heterogeneous. Available evidence supports only the withdrawal of loop diuretics and possibly carvedilol monotherapy, but not the minimization or withdrawal of RASi, MRA, or the combination of HF medications. Large RCTs are needed to determine the appropriate treatment strategy and identify the specific target populations in this area.

Funding

This work was supported by the Scientific and Technological Research Council of Türkiye (TUBITAK) Research grant [1059B192302470].

Conflict of interest: none declared.

Supporting information

Table S1. Search terms and strategies.

Table S2. Definitions of myocardial function improvement and primary outcomes of the included studies.

Table S3. The classes of medications evaluated in each included study.

Table S4. Quality assessment of the included studies.