Network meta-analysis comparing angiotensin receptor-neprilysin inhibitors, angiotensin-receptor blockers, and angiotensin-converting enzyme inhibitors in heart failure with reduced ejection fraction

Abstract

The superiority of ARNI over ACE-I and ARB have not been reassessed following the publication of recent trials that did not find clinical benefits. Therefore, we performed an updated network meta-analysis comparing the efficacy and safety of ARNI, ACE-I, ARB, and placebo in HFrEF. We included randomized clinical trials that compared ARNI, ARB, ACE-I, and placebo in HFrEF. We extracted pre-specified efficacy endpoints and produced network estimates, P-scores, and SUCRA scores using frequentist and Bayesian network meta-analysis approaches. A total of 28 RCTs including 47,407 patients were included. ARNI was associated with lower risk of all-cause mortality (RR 0.81, 95% CI 0.68–0.96), cardiac death (RR 0.79, 95% CI 0.64–0.99), and MACE (RR 0.83, 95% CI 0.72–0.97), but higher risk of hypotension (RR 1.46, 95% CI 1.02–2.10) compared with ARB. ARNI was associated with lower risk of MACE (RR 0.85, 95% CI 0.74–0.97), but higher risk of hypotension (RR 1.69, 95% CI 1.27–2.24) compared with ACE-I. P-scores and SUCRA scores demonstrated superiority of ARNI over ARB and ACE-I in all-cause mortality, cardiac death, MACE, and hospitalization for heart failure. In conclusion, ARNI was associated with improved clinical outcomes, except for higher risk of hypotension, compared with ARB and ACE-I.

Graphical Abstract. Comparison of endpoints among ARNI, ARB, and ACE-I

The figure illustrates the difference in endpoints among ARNI, ARB, and ACE-I that were statistically significant from the network meta-analysis

Abbreviations: ACE-I = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; ARNI = angiotensin receptor-neprilysin inhibitor; MACE = major adverse cardiac events

Introduction

In heart failure with reduced ejection fraction (HFrEF), renin-angiotensin-aldosterone system (RAAS) inhibition using either an angiotensin-converting enzyme inhibitor (ACE-I) or angiotensin-receptor blocker (ARB) has consistently shown mortality benefit². However, after the landmark Prospective Comparison of ARNI with ACEI to Determine Impact on Global Mortality and Morbidity in Heart Failure (PARADIGM-HF) trial, the first angiotensin receptor-neprilysin inhibitor (ARNI) was approved in 2015, and currently, both North American and European guidelines recommend the replacement of ACE-I or ARB by ARNI to reduce mortality and morbidity with class I recommendation^3–7. Following the PARADIGM-HF trial, additional trials on ARNIs have been performed, namely Randomized Study Using Accelerometry to Compare Sacubitril/valsartan and Enalapril in Patients With Heart Failure (OUTSTEP-HF) trial, Study of Efficacy and Safety of LCZ696 in Japanese Patients With Chronic Heart Failure and Reduced Ejection Fraction (PARALLEL-HF) trial, and LCZ696 in Hospitalized Advanced Heart Failure (LIFE) trial, which did not find a reduction in clinical composite of number of days alive, out of hospital, and free from heart failure events among patients with advanced heart failure^8–10. Given the conflicting data gleaned from more recent trials, we performed a network meta-analysis to draw direct and indirect comparisons between ARNI, ACE-I and ARBs in patients with HFrEF using both frequentist and Bayesian approaches.

Methods

All supporting data are available within the article. Our systematic review was conducted according to a published protocol available on Open Science Framework (10.17605/OSF.IO/NT6XC). Each step of our study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines¹¹. Our study was exempt from institutional review board as only data from previously published papers were used.

Two authors (D.P. and S.An) systematically searched the literature comprising of PubMed, Medline, Embase, and Cochrane Library from inception to February 4, 2022. The medical subject heading terms used to find the trials are listed in Supplementary Table 1. The inclusion criteria were as follows: (1) RAAS inhibitors limited to ARNI, ARB, and ACE-I as these agents are not used simultaneously; (2) randomized controlled trials (RCTs) with one class of RAAS inhibitor in case group and either another class of RAAS inhibitor or placebo in the control group; (3) HFrEF defined as left ventricular ejection fraction (LVEF) <50%; and (4) pre-specified efficacy endpoints which included all-cause mortality, cardiac death, major adverse cardiac events (MACE), myocardial infarction (MI), stroke, and hospitalization for heart failure, or pre-specified safety endpoints which included hypotension, hyperkalemia, renal failure, and angioedema. Full papers written in languages other than English were excluded. If an RCT had multiple publications, the first publication containing data on primary and secondary endpoints was selected.

Two authors (D.P. and S.An) reviewed full text articles to winnow eligible articles from the literature. Selected articles were also reviewed to identify other eligible studies. Conflicts were resolved after discussions with a third author (A.V.). For each selected RCT, author, publication year, follow up period, class of medication, and the generic name of the medication were arranged into tables (Table 1). Demographics, percentage of ischemic cardiomyopathy, laboratory values, symptomatology, comorbidities, past medical history, and other concomitant treatments were organized into tables (Supplementary Table 2). Definitions of endpoints in the trials were dictated into tables (Supplementary Table 3). For quality assessment, the Cochrane Collaboration’s tool was used to assess the risk of biases (Supplementary Table 4)¹².

Table 1.

General characteristics of the selected trials

Triala	Author	Year	Follow Up (Days)	LVEF (%)	Case			Control
					N	Class	Generic	N	Class	Generic
LIFE	Mann et al.	2021	168	≤35	167	ARNI	Sacubitril/valsartan	168	ARB	Valsartan
OUTSTEP-HF	Piepoli et al.	2021	84	≤40	309	ARNI	Sacubitril/valsartan	310	ACE-I	Enalapril
PARALLEL-HF	Tsutsui et al.	2021	183	≤35	111	ARNI	Sacubitril/valsartan	112	ACE-I	Enalapril
EVALUATE-HF	Desai et al.	2019	84	≤40	231	ARNI	Sacubitril/valsartan	233	ACE-I	Enalapril
PRIME	Kang et al.	2019	365	25–50	60	ARNI	Sacubitril/valsartan	58	ARB	Valsartan
PIONEER-HF	Velazquez et al.	2018	56	≤40	440	ARNI	Sacubitril/valsartan	441	ACE-I	Enalapril
PARADIGM-HF	McMurray et al.	2014	810	≤35	4187	ARNI	Sacubitril/valsartan	4212	ACE-I	Enalapril
ARCH-J	Matsumori et al.	2003	168	≤45	148	ARB	Candesartan	144	Placebo	Placebo
CHARM-Alternative	Granger et al.	2003	1011	≤40	1013	ARB	Candesartan	1015	Placebo	Placebo
VALIANT	Pfeffer et al.	2003	741	≤45	4909	ARB	Valsartan	4909	ACE-I	Captopril
HEAVEN	Willenheimer et al.	2002	84	≤45	70	ARB	Valsartan	71	ACE-I	Enalapril
OPTIMAAL	Dickstein et al.	2002	986	≤35	2744	ARB	Losartan	2733	ACE-I	Captopril
REPLACE	Dunselman et al.	2001	84	≤40	301	ARB	Telmisartan	77	ACE-I	Enalapril
Val-HeFT	Cohn et al.	2001	690	≤40	2511	ARB	Valsartan	2499	Placebo	Placebo
ELITE II	Pitt et al.	2000	555	≤40	1578	ARB	Losartan	1574	ACE-I	Captopril
SPICE	Granger et al.	2000	84	≤35	179	ARB	Candesartan	91	Placebo	Placebo
RESOLVD	McKelvie et al.	1999	301	≤40	327	ARB	Candesartan	109	ACE-I	Enalapril
STRETCH	Riegger et al.	1999	84	35–45	633	ARB	Candesartan	211	Placebo	Placebo
ELITE	Pitt et al.	1997	336	≤40	352	ARB	Losartan	370	ACE-I	Captopril
LPES	Lang et al.	1997	84	≤45	78	ARB	Losartan	38	ACE-I	Enalapril
Dickstein et al. (1995)	Dickstein et al.	1995	56	≤35	108	ARB	Losartan	58	ACE-I	Enalapril
FEST	Erhardt et al.	1995	84	≤35	155	ACE-I	Fosinopril	153	Placebo	Placebo
FHFSG	Brown et al.	1995	168	≤35	116	ACE-I	Fosinopril	125	Placebo	Placebo
TRACE	Kober et al.	1995	720–1500	≤35	876	ACE-I	Trandolapril	873	Placebo	Placebo
CASSIS	Widimsky et al.	1995	84	≤40	200	ACE-I	Spirapril, Enalapril	48	Placebo	Placebo
Colfer et al. (1992)	Colfer et al.	1992	84	≤35	114	ACE-I	Benazepril	58	Placebo	Placebo
SAVE	Pfeffer et al.	1992	1260	≤40	1115	ACE-I	Captopril	1116	Placebo	Placebo
SOLVD	SOLVD Investigators	1991	1242	≤35	1285	ACE-I	Enalapril	1284	Placebo	Placebo

^aReferences of all the trials are shown in Supplementary Table 2.

Abbreviations: ACE-I = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; ARNI = angiotensin receptor-neprilysin inhibitor; LVEF = left ventricular ejection fraction

For consistency, risk ratios were manually calculated from all selected trials. Zero-cells were corrected using the modified Haldane-Ascombe method¹³. A frequentist network meta-analysis was initially conducted to calculate pooled estimates. Node-splitting analysis was performed to evaluate for inconsistencies between direct and indirect estimates (Supplementary Table 5). Tau-squared and I-squared values were used to assess for heterogeneity in the network models (Supplementary Table 6). P-scores of each class of medication were calculated for each endpoint but were interpreted only for endpoints in which the network meta-analysis showed significant differences among the different classes of medications. P-scores represent the mean extent of certainty that one class of RAAS inhibitor is better than its competitors averaged over all competitors with equal weights¹⁴. Theoretically, P-score of 1 and 0 signifies best and worst, respectively. In addition, we conducted a Bayesian network meta-analysis in which estimates were calculated by a generalized linear model fitted under a hierarchic Bayesian random-effects framework. Models were generated by Markov-chain Mote Carlo simulations incorporating 4 chains, 5,000 adaptations, and 100,000 iterations. Time-series and density plots were visualized to confirm convergence. Surface Under the Cumulative Ranking (SUCRA) scores were calculated from the Bayesian framework to validate the P-scores from the frequentist counterpart. Rankograms were produced to display relative hierarchy from first to fourth. Finally, meta-regression was modeled using mixed-effects logistic regression to assess for the association between selected endpoints in ARNI trials and LVEF as well as NT-proBNP levels. Beta-coefficients with P values, tau-square, I-square, H-squared, and R-squared indexes were generated from all meta-regression models (Supplementary Table 7). Frequentist network meta-analysis was performed using meta and netmeta packages, Bayesian network meta-analysis using gemtc and rjags packages, and meta-regression using meta package in R version 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria).

Results

After screening 44,983 articles, we identified 28 RCTs for inclusion in the final analysis (Figure 1). A total of 5,505 (11.6%), 15,177 (32.0%), 19,108 (40.3%), and 7,617 (16.1%) patients were randomized to receive ARNI, ARB, ACE-I, and placebo, respectively. Seven older studies from the years 1991 to 1995 compared ACE-I with placebo, and seven newer studies from the years 2014 to 2021 compared ARNI with ARB (2 studies) or ACE-I (5 studies). Studies in between compared ARB with ACE-I (9 studies) or placebo (5 studies). The network plot is illustrated in Figure 2. The generic drugs used in each class of RAAS inhibitors were variable, but all ARNIs were sacubitril/valsartan. Follow up period ranged from 56 to 1500 days, with the average at 363 days. Definitions of HFrEF varied with most trials including heart failure patients with LVEF ≤40%, but five trials^15–19 included those with LVEF ≤45% and one trial²⁰ those with LVEF <50%.

Figure 1. PRISMA flow diagram of the search for relevant trials.

The PRISMA flow diagram shows the process whereby the databases were screened and assessed for relevant trials that met the inclusion criteria.

Figure 2. Network plot of selected trials.

The network plot demonstrates the number of trials and patients included among trials that compared ARNI, ARB, ACE-I, and placebo. The size of the blue circles and blue lines are proportional to the total sample size and number of relevant trials, respectively.

Baseline patient characteristics varied widely across the trials (Supplementary Table 2). The mean age of the patients in most studies was in their sixties. The etiology of HFrEF was ischemic cardiomyopathy in most trials, and most of the patients had New York Heart Association class II or III symptoms. Only the LIFE trial was an outlier in that more than a third of its patients had class IV symptoms. Comorbidities, past medical histories, and concomitant treatments were variable (Supplementary Table 2).

Treatment with an ARNI was associated with significantly lower risk of all-cause mortality (RR 0.81; 95% CI 0.68–0.96), cardiac death (RR 0.79; 95% CI 0.64–0.99), and MACE (RR 0.83; 95% CI 0.72–0.97) compared with ARB. Treatment with an ARNI was also associated with significantly lower risk of MACE (RR 0.85; 95% CI 0.74–0.97) compared with ACE-I. There were no differences in the risk of all-cause mortality, cardiac death, and MACE between ARB and ACE-I (Table 2). No differences were observed among ARNI, ARB, and ACE-I in the endpoints of MI, stroke, and hospitalizations for heart failure. Pooled estimates from Bayesian network meta-analysis showed similar findings (Supplementary Table 8). P-scores from the frequentist network model revealed the superiority of ARNI for the outcomes of all-cause mortality, cardiac death, and MACE (Figure 3 and Supplementary Table 9). Similar results were found with SUCRA scores (Supplementary 10 and Supplementary Figure 1). Rankograms from the Bayesian network model redemonstrated these findings (Figure 4).

Table 2.

Pooled estimates of frequentist network meta-analysis for each outcome^a

All-cause mortality (27 studies)	ARNI	1.24 (1.04–1.47)	1.16 (0.99–1.34)	1.34 (1.12–1.59)
	0.81 (0.68–0.96)	ARB	0.94 (0.86–1.02)	1.08 (0.99–1.19)
	0.87 (0.74–1.01)	1.07 (0.98–1.16)	ACE-I	1.16 (1.06–1.26)
	0.75 (0.63–0.89)	0.92 (0.84–1.01)	0.86 (0.79–0.94)	Placebo
Cardiac death (16 studies)	ARNI	1.26 (1.01–1.57)	1.17 (0.96–1.42)	1.41 (1.13–1.75)
	0.79 (0.64–0.99)	ARB	0.93 (0.83–1.03)	1.12 (0.99–1.26)
	0.86 (0.71–1.04)	1.08 (0.97–1.20)	ACE-I	1.20 (1.08–1.34)
	0.71 (0.57–0.89)	0.90 (0.79–1.01)	0.83 (0.74–0.93)	Placebo
Major adverse cardiac events (16 studies)	ARNI	1.20 (1.03–1.40)	1.18 (1.03–1.35)	1.41 (1.19–1.67)
	0.83 (0.72–0.97)	ARB	0.98 (0.91–1.07)	1.18 (1.07–1.30)
	0.85 (0.74–0.97)	1.02 (0.94–1.10)	ACE-I	1.20 (1.07–1.34)
	0.71 (0.60–0.84)	0.85 (0.77–0.94)	0.84 (0.75–0.93)	Placebo
Myocardial infarction (11 studies)	ARNI	0.14 (0.01–2.77)	0.14 (0.01–2.66)	0.14 (0.01–2.85)
	7.04 (0.36–137.32)	ARB	0.95 (0.78–1.16)	1.01 (0.76–1.35)
	7.39 (0.38–145.10)	1.05 (0.86–1.28)	ACE-I	1.07 (0.82–1.38)
	6.94 (0.35–137.17)	0.99 (0.74–1.31)	0.94 (0.72–1.22)	Placebo
Stroke (10 studies)	ARNI	0.93 (0.23–3.72)	0.89 (0.22–3.57)	0.98 (0.24–4.02)
	1.07 (0.27–4.27)	ARB	0.96 (0.75–1.22)	1.05 (0.76–1.46)
	1.12 (0.28–4.48)	1.05 (0.82–1.34)	ACE-I	1.10 (0.81–1.50)
	1.02 (0.25–4.18)	0.95 (0.68–1.32)	0.91 (0.67–1.24)	Placebo
Hospitalization for heart failure (18 studies)	ARNI	1.18 (0.97–1.44)	1.16 (0.97–1.38)	1.67 (1.33–2.10)
	0.85 (0.69–1.03)	ARB	0.98 (0.87–1.11)	1.42 (1.23–1.63)
	0.86 (0.72–1.03)	1.02 (0.90–1.15)	ACE-I	1.44 (1.23–1.69)
	0.60 (0.48–0.75)	0.71 (0.61–0.81)	0.69 (0.59–0.81)	Placebo
Hypotension (19 studies)	ARNI	0.68 (0.48–0.98)	0.59 (0.45–0.79)	0.34 (0.21–0.52)
	1.46 (1.02–2.10)	ARB	0.87 (0.66–1.14)	0.49 (0.34–0.70)
	1.69 (1.27–2.24)	1.15 (0.88–1.52)	ACE-I	0.57 (0.39–0.82)
	2.98 (1.91–4.67)	2.04 (1.43–2.92)	1.77 (1.23–2.55)	Placebo
Hyperkalemia (14 studies)	ARNI	0.84 (0.51–1.38)	0.78 (0.57–1.07)	0.39 (0.20–0.76)
	1.20 (0.73–1.97)	ARB	0.93 (0.58–1.49)	0.47 (0.25–0.89)
	1.29 (0.94–1.76)	1.07 (0.67–1.71)	ACE-I	0.51 (0.28–0.92)
	2.55 (1.32–4.90)	2.13 (2.12–4.03)	1.98 (1.09–3.60)	Placebo
Renal failure (16 studies)	ARNI	1.50 (1.04–2.15)	1.07 (0.82–1.39)	0.81 (0.54–1.21)
	0.67 (0.47–0.96)	ARB	0.72 (0.55–0.94)	0.54 (0.40–0.74)
	0.93 (0.72–1.21)	1.40 (1.07–1.83)	ACE-I	0.76 (0.55–1.04)
	1.23 (0.82–1.84)	1.84 (1.35–2.51)	1.32 (0.96–1.80)	Placebo
Angioedema (12 studies)	ARNI	0.56 (0.22–1.42)	0.95 (0.44–2.06)	0.48 (0.11–2.16)
	1.78 (0.71–4.49)	ARB	1.70 (0.96–2.99)	0.85 (0.26–2.81)
	1.05 (0.48–2.27)	0.59 (0.33–1.04)	ACE-I	0.50 (0.13–1.88)
	2.10 (0.46–9.50)	1.18 (0.36–3.89)	2.00 (0.53–7.51)	Placebo

^aAll pooled estimates are given in risk ratio (95% confidence interval) with each class of renin-angiotensin-aldosterone system inhibitor or placebo as the reference horizontally.

Abbreviations: ACE-I = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; ARNI = angiotensin receptor-neprilysin inhibitor

Figure 3. Bar graph showing P-scores of each type of medication for every outcome.

The bar graphs show the P-scores of ARNI (blue), ARB (red), ACE-I (violet), and placebo (green) from the frequentist network meta-analysis for each outcome. P-scores measure the extent of certainty that the class of medication is better than competing classes. Outcomes with significant differences in the pooled estimates of network meta-analysis (Table 2) are marked with an asterisk (*).

Abbreviations: ACE-I = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; ARNI = angiotensin receptor-neprilysin inhibitor

Figure 4. Rankograms for each outcome.

Rankograms illustrate the probability of ARNI (blue), ARB (red), ACE-I (violet), and placebo (green) being the best, second, third, and fourth most efficacious for each of the outcomes. The ordinate and abscissa show the probability and rank, respectively.

Abbreviations: ACE-I = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; ARNI = angiotensin receptor-neprilysin inhibitor

Treatment with an ARNI was associated with significantly higher risk of hypotension compared with ARB (RR 1.46; 95% CI 1.02–2.10) and ACE-I (RR 1.69; 95% 1.27–2.24). Use of an ARB was associated with significantly higher risk of renal failure compared with ARNI (RR 1.50, 95% CI 1.04–2.15) and ACE-I (RR 1.40, 95% CI 1.07–1.83). No differences were found among ARNI, ARB, and ACE-I in the endpoints of hyperkalemia and angioedema. These findings were also graphically demonstrated in P-scores, SUCRA scores, and rankograms (Figure 3–4).

Mixed-effects logistic meta-regression of trials that compared ARNI with either ARB or ACE-I revealed weak increasing trends with a potential inverse relationship between the risk of all-cause mortality (β= −0.035, p=0.396), cardiac death (β= −0.080, p=0.122), MACE (β= −0.027, p=0.590), and hospitalization for heart failure (β= −0.025, p=0.581) with decreasing LVEF (Figure 5). In other words, ARNI, as compared with ARB or ACE-I, was weakly associated with increasing adverse outcomes in relation to decreasing LVEF. Additional meta-regression analysis exploring association of identical efficacy outcomes with NT-proBNP levels suggested weak increasing trends of all-cause mortality (β= 0.0002, p=0.464) and cardiac death (β= 0.0002, p=0.595) with increasing NT-proBNP levels, but not in MACE (β= 0.0000, p=0.977) and hospitalization for heart failure (β= −0.0001, p=0.719) (Supplementary Figure 2). However, none of these trends reached statistical significance. Residual and accounted heterogeneity of the models are shown in Supplementary Table 7.

Figure 5. Meta-regression of ARNI versus ACE-I or ARB to left ventricular ejection fraction.

The four graphs above show the association of the risk ratio of all-cause mortality, cardiac death, major adverse cardiac events, and hospitalization for heart failure with mean left ventricular ejection fraction of the trials on a logarithmic scale. Positive beta-coefficient (β) signifies positive correlation and vice versa. P-value of the trend is shown at the center of each graph.

Abbreviations: ACE-I = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; ARNI = angiotensin receptor-neprilysin inhibitor

Discussion

In line with the most recent guideline recommendations, our contemporary network meta-analysis including the more recent ARNI trials confirms the superiority of ARNI when compared with ARB and ACE-I for the endpoints of all-cause mortality, cardiac death, MACE, and hospitalization for heart failure⁷. Treatment with an ARNI was associated with a higher risk of hypotension compared with ARB and ACE-I. ARB and ACE-I performed similarly across the endpoints of interest, with the exception of a higher risk of renal failure with ARB (Graphical Abstract). Meta-regression of trials comparing ARNI with ARB or ACE-I showed modest non-significant trends toward an increased risk of all-cause mortality, cardiac death, MACE, and hospitalization for heart failure, respectively, as LVEF decreased.

Our findings are consistent with those of previous studies, including a network meta-analysis by Burnett et al., who found the greatest mortality reduction in the combination of ARNI, mineralocorticoid receptor antagonist (MRA), and beta-blocker among combinations of ARNI, ARB, ACE-I, MRA, and beta-blocker²¹. We chose not to stratify by MRA and beta-blocker since doing so would have introduced even greater heterogeneity via the different proportions of beta-blockers and MRAs used in the trials (Supplementary Table 2). Nielsen et al. performed a conventional pairwise meta-analysis, including 39 trials written in Chinese language, which also found a beneficial effect of ARNI over ARB or ACE-I when assessing all-cause mortality and serious adverse events²². Similar findings were also observed by Zhang et al. and Tai et al^23,24. Our study was unique in that we used both frequentist and Bayesian approaches, employed network meta-analysis to combine direct and indirect evidence, included the most recently published trials, pooled outcomes from the greatest number of trials, and performed meta-regression analysis to assess the association between treatment with ARNI and endpoints of interest as LVEF decreased and as NT-proBNP levels increased.

ARNI treatment, however, was associated with increased risk of hypotension compared with both ARB and ACE-I potentially because of the vasodilatory, natriuretic, and diuretic effects of natriuretic peptides²⁵. Despite this safety concern, PARADIGM-HF did not find increased discontinuation of ARNI due to hypotension-related adverse effects⁵. This increased risk of hypotension did not lead to increased risk of pre-renal or intrinsic acute kidney injury leading to renal failure, as demonstrated by ARNI being associated with lower risk of renal failure compared with ARB and showing no difference compared to ACE-I. This is in line with the findings of Prospective ARNI vs. ACE inhibitor trial to DetermIne Superiority in reducing heart failure Events after Myocardial Infarction (PARADISE-MI) trial, which showed no difference in renal impairment between patients with acute AMI taking sacubitril/valsartan and those taking ramipril²⁶. However, the caveat is that the definitions of renal failure used in the trials were variable (Supplementary Table 3).

Meta-regression analysis suggested a weak non-significant trend towards poorer efficacy of ARNI relative to ARB or ACE-I as LVEF decreased and as NT-proBNP levels increased. The trends were not significant, but the inverse relationship was consistent in all the four outcomes with decreasing LVEF. Neither establishment of causal relationships nor definitive conclusions can be made, but these findings are hypothesis-generating in that there may be certain subpopulations of patients with HFrEF who benefit more from ARNI. This analysis was motivated by the LIFE trial, which did not observe differences in NT-proBNP levels or clinical outcomes in advanced heart failure, defined by NYHA class IV symptoms, LVEF ≤35%, NT-proBNP ≥800pg/mL, and an objective finding as defined by the trial protocol despite ≥3 months of goal-directed medical therapy⁸. Previous studies have noted blunting of the natriuretic peptide system with the progression of heart failure as demonstrated by reduced renal, vascular, and endocrine responsiveness^27,28. Other implicated mechanisms include increased clearance of natriuretic peptides, reduced or desensitized receptors, inhibition of downstream cascade, and hyperactivation of the RAAS and sympathetic nervous system in chronic heart failure²⁹. Although patients with advanced heart failure have poorer tolerance to ARNI, as evidenced by the high discontinuation rate and two-thirds of the patients not reaching the target dose in the LIFE trial, efficacy of neprilysin inhibitors in advanced heart failure requires further investigation. Among Medicare beneficiaries, the incremental annual cost of ARNI instead of ACE-I is estimated to be $1,400, and data hitherto has not definitively demonstrated a clear clinical benefit among those with the most advanced disease³⁰. Therefore, guidelines should be implemented more carefully among those with advanced heart failure.

Our study was limited by the absence of patient-level data. Therefore, the findings of our study should be interpreted in the study-level to prevent ecological bias. We assumed that generic drugs within the same pharmacological class have the same effect, which may not be accurate. Trials were heterogenous in baseline demographics, comorbidities, concurrent treatment, and duration of follow-up. Trials spanned 30 years from 1991 to 2021, with earlier trials focusing on ACE-I and recent trials on ARNI. Trials also differed in the definition of certain endpoints (Supplementary Table 3). Risks of biases in the trials were low overall, although a few trials exhibited high risk of bias. Endpoints of MI and stroke showed significant inconsistencies between direct and indirect evidence, but the differences in these endpoints among ARNI, ARB, ACE-I, and placebo were non-significant in the network meta-analysis. We were also limited by the relatively small number of trials on ARNI to include in our meta-regression analyses, impeding the production of more robust results. The trends suggested by meta-regression analyses should be interpreted with caution as none of the trends were significant and should only be used for hypothesis-generating purposes.

Our network meta-analysis offers a comprehensive comparison of ARNI, ARB, and ACE-I in patients with HFrEF. ARNI was superior to ARB and ACE-I in terms of all-cause mortality, cardiac death, and hospitalization for heart failure. This comes at the small cost of a higher risk of hypotension associated with ARNI treatment. Interestingly, the superiority of an ARNI treatment strategy may be somewhat attenuated in advanced heart failure patients with very low LVEF, though these findings were not statistically significant. Further research is needed to determine the appropriate use of ARNI in those with advanced heart failure with very low LVEF.

Highlights.

• ARNI has class I recommendation in heart failure with reduced ejection fraction.

• Updated network meta-analysis supported the superiority of ARNI over ACE-I and ARB.

• ARNI was associated with better efficacy endpoints and comparable safety endpoints.

• Meta-regression suggested attenuated efficacy of ARNI in advanced heart failure.

• ARNI remains first line, but its use in advanced heart failure needs more studies.

Abbreviations

ACE-I

Angiotensin-converting enzyme inhibitor

ARB

Angiotensin-receptor blocker

ARNI

Angiotensin receptor-neprilysin inhibitor

HFrEF

Heart failure with reduced ejection fraction

LVEF

Left ventricular ejection fraction

MACE

Major adverse cardiac events

MI

Myocardial infarction

MRA

Mineralocorticoid receptor antagonist

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-analysis

RAAS

Renin-angiotensin-aldosterone system

RCT

Randomized clinical trial

SUCRA

Surface under the cumulative ranking