Angiotensin-converting enzyme inhibitors and angiotensin receptor blockers are drugs commonly used for the treatment of hypertension. However, studies on their comparative efficacy have not been extensively investigated. The current systematic review and network meta-analysis studied the comparative efficacy of the two antihypertensive treatment categories in reducing blood pressure, mortality, and morbidity in essential hypertension patients. A literature search was carried out in Medline and Cochrane Central Register of Controlled Trials for placebo- and active-controlled, double-blind randomized clinical trials, which had reported blood pressure effects, mortality, and/or morbidity. Blood pressure results were found in 30 studies with 7370 participants and 8 studies with 25,158 participants with mortality/morbidity results included in the analysis. The two drug classes had similar effectiveness in lowering systolic (weighted mean difference (WMD): 0.59, 95% CI: −0.21 to 1.38) and diastolic blood pressure (WMD: 0.62, 95% CI: −0.06 to 1.30), all-cause mortality (risk ratio (RR)): 0.96, 95% CI 0.80 to 1.14), cardiovascular mortality (RR: 0.87, 95% CI 0.67 to 1.14), fatal and non-fatal myocardial infarction (RR: 1.02, 95% CI 0.75 to 1.37) and stroke (RR: 1.13, 95% CI 0.87 to 1.46). Angiotensin-converting enzyme inhibitors were more helpful in the prevention and/or the hospitalization for heart failure than angiotensin receptor blockers (RR: 0.71, 95% CI 0.54 to 0.93). Angiotensin-converting enzyme inhibitors and angiotensin receptor blockers were similarly effective in decreasing blood pressure, mortality, and morbidity in essential hypertension. Angiotensin-converting enzyme inhibitors were more protective in the advancement and/or hospitalization of the hypertensive patient for heart failure than angiotensin receptor blockers.
Hypertension (HT) is an independent risk factor for cardiovascular (CV) disease, related dramatically with myocardial infarction (MI), heart failure (HF), stroke, and renal disease. According to the World Health Organization, elevated blood pressure (BP) “kills” prematurely about 9.4 million people daily. In particular, HT is causative for at least 45% of deaths attributable to CV disease and for 51% of deaths from the cerebral vascular disease .
There are many pharmacological treatment options for HT. Two widely used drug classes are angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin receptor blockers (ARBs). The first inhibit ACE activity reducing the conversion of angiotensin I to angiotensin II and the vasoconstricting activity of angiotensin II, while the second one by blocking the type 1 receptor of the angiotensin II . Over the years, the extensive research investigated these two drug classes for the treatment of HT and other chronic diseases. Systematic reviews and meta-analyses explored the effect of one of the two pharmacological agents on BP reduction compared to placebo [3, 4] or their effect on mortality and morbidity . Systematic reviews and meta-analyses have also compared the BP reduction effect of these two drug classes [6, 7] and one studied telmisartan effectiveness in BP reduction compared to four different ACEIs . Nevertheless, currently available data are not solely based on double-blind randomized control trials (RCTs). One network meta-analysis (NMA) studied the comparative efficacy of ARBs in BP reduction and CV event rates in patients with HT, but again it did not only include double-blind RCTs . Finally, another NMA studied the comparative efficacy of ACEIs and ARBs in mortality and morbidity of patients with high CV risk, but it did not only include patients with HT . Thus the comparative efficacy of ACEIs and ARBs on BP reduction, mortality, and morbidity in essential HT patients through the use of double-blind RCTs have not been clearly demonstrated.
The current study aimed to provide an updated systematic review and meta-analysis on the comparative efficiency of ACEIs and ARBs to reduce BP as well as in preventing CV events in hypertensive patients.
The Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Network Meta-analyses extension statement methods were used for the present systematic review . The methods used for this review were in advance specified and registered in the PROSPERO database: CRD42017078157.
Eligibility and exclusion criteria
General eligibility criteria were double-blinded RCTs, published in English language, with a publication date after 1990 and including patients aged >18 years. The included studies should have at least 4 weeks of treatment duration and compare directly ACEIs with ARBs on office/clinic BP measurements obtained in a seated position in patients with essential HT. In case of using add-on treatments in the studies, these treatments had to include exactly the same drug, in the same dose, in both treatment groups. For the mortality/morbidity arm, studies should compare ACEIs directly with ARBs or the two categories with placebo on mortality and/or morbidity rate occurrence in hypertensive patients. For this arm, the inclusion criteria may include essential HT or not as the small number of studies included patients with essential HT. Mortality and morbidity were defined as all-cause mortality, CV mortality and morbidity, fatal and non-fatal MI, fatal and non-fatal stroke, sudden cardiac death, fatal congestive HF, hospitalization, and development of HF.
Studies that had more than one add-on treatment and that did not denote statistically important differences between the intervention groups or not were excluded from the analysis. Moreover, studies that used fimasartan or allisartan as HT treatment, as well as studies with no published results, were excluded. At the time the study was initiated, fimasartan was approved for use only in South Korea and allisartan in China. Therefore, it was considered appropriate to exclude studies with fimasartan and allisartan since they are not used widely and worldwide.
Literature search strategy and study selection
Eligible studies by searching Medline and the Cochrane Library were identified using pertinent keywords and a filter of maximum sensitivity to restrict our results to RCTs (Supplementary material 1). Two separate searches were performed, one for the BP reduction arm and one for the mortality/morbidity arm. Finally, we conducted a manual search by scanning the reference lists of pertinent articles and clinicaltrials.gov.
Records retrieved from the literature search were imported in a reference management software. After removing the duplicate records, two reviewers (CD and ID) screened titles and abstracts independently, and full texts were investigated for eligible studies. Cases with disagreement were resolved by consensus together with a third reviewer (CA).
From each included study, study characteristics (first author’s name, title, year and journal of publication, duration, drugs and doses in the treatment groups, any add-on treatment), patients’ characteristics (number of participants in each treatment group, age, gender, race, smoking and alcohol habits, previous antihypertensive treatment, co-morbidities), pre-specified outcomes of BP levels, and number of mortality/morbidity events were extracted. In case of studies that reported multiple BP measurements in different periods during follow-up, the BP measurements that were finally extracted and used were those at the end of the intervention. If SD was not reported in a study, but SEM or confidence interval (CI) was reported as a variability measure, SD was calculated using the RevMan program implementation. Also, if the BP reduction from its baseline value or its SD was not reported in the text of a study, but it was shown in a graph, then it was extracted from the graph. If it was not possible to derive the BP reduction from the baseline or the corresponding SD from the available data, the final systolic and diastolic BP (SBP and DBP, respectively) and the corresponding SD were used .
Quality assessment in individual studies
The quality of the selected studies was evaluated using the Cochrane Collaboration’s tool to assess the risk of bias in randomized trials. This tool helps to find six types of bias in the studies through assessing the domains of random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and anything else that could result in a systematic bias. This was attained through the judgment of the reviewers, who responded to the above questions with “yes,” “no,” and “unclear” and consequently indicated low, high, and unknown risk of bias, respectively .
For continuous variables, such as SBP and DBP, results were reported as the weighted mean difference (WMD) with the corresponding 95% CI, and for dichotomous variables of mortality and morbidity, results were reported as risk ratio (RR) with the corresponding 95% CI. Analysis of the studies’ results was based primarily on intention-to-treat analysis. RevMan version 5.3 (Nordic Cochrane Center, Copenhagen, Denmark) was used for data synthesis and analysis of traditional meta-analysis. Presence of heterogeneity was tested by chi2 test, considering that a p value <0.10 showed significant heterogeneity, and by I2 statistic, considering that value of I2 > 50% indicated moderate heterogeneity. Funnel plots assessed visually the risk of bias across studies and statistically, the risk was estimated using Egger’s test. Funnel plot asymmetry tests were used only when at least 10 studies included in the meta-analysis. For the two quantitative variables of SBP and DBP, inverse variance and random-effects model were used, as there was an expectation of heterogeneity between the studies. Subsequently, in order to investigate this heterogeneity, subgroup analyses were performed based on gender, mean age, presence of patients with diabetes mellitus (DM) in the studies, smoking and alcohol consumption habits, previous antihypertensive treatment, baseline BP measurements, doses of the drugs used in the studies, whether there was add-on treatment or not, and whether there was a wash-out period or not. In order to explore potential causes of heterogeneity, studies with high risk of bias and studies with sponsoring were excluded with sensitivity analysis. Finally, meta-regression analyzed possible effect of mean age, gender, DM, funding, duration, drugs’ dose and random sequence generation, allocation concealment, and blinding of outcome assessment on the meta-analysis results . Meta-regression analysis was carried out with R version 3.3.3., using the package “metaphor” and mixed effects model.
Random-effects NMA was performed using a graph theoretical approach within a frequentist framework. In the absence of direct comparisons in RCTs, interventions of interest were analyzed with indirect comparisons using a common comparator. NMA was carried out with R version 3.3.3., using package “netmeta.” Ranking interventions were determined by using p scores. For each outcome, p scores were used to rank the treatments on a continuous 0–1 scale and so it was determined which treatment was the best, the second best, etc . Homogeneity and consistency were assessed by decomposing the Q statistic into variation of the effect estimates within and between designs (heterogeneity and inconsistency, respectively). Direct and indirect evidence inconsistency was investigated by “node-splitting” . p Values were two-tailed and were considered significant at a 5% significance level unless otherwise stated.
Quality of evidence
Quality of evidence evaluated by means of GRADE (Grading of Recommendations Assessment, Development, and Evaluation). In the traditional meta-analysis, the assessment of quality is performed by considering that evidence from RCTs starts at high quality. Quality may be downgraded based on the risk of bias in the design, consistency, directness of the evidence, and precision of estimates. Publication bias to lower levels of quality reflects the confidence we have in the estimated effect size . In the NMA, quality from the direct evidence is assessed as mentioned above, and quality assessment from the indirect evidence is based on first-order loops, which promote information to the indirect estimate. It starts at the lowest rating of the two direct comparisons that contribute to the indirect estimate and it then can be downrated if there is imprecision or intransitivity. If both direct and indirect evidence for a specific comparison is available, the higher of the two ratings is considered to be the quality rating for the NMA .
Search strategy revealed 30 eligible trials that were inserted in the meta-analysis for the BP reduction outcome [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47], while for mortality/morbidity outcome, the search strategy bright out 8 eligible trials that were included in the NMA [48,49,50,51,52,53,54,55]. Figure 1a, b show the study selection process for the BP reduction and the mortality/morbidity endpoints, respectively.
Characteristics of included studies
Tables 1 and 2 summarize the characteristics of the studies analyzed for the BP reduction and the mortality/morbidity outcomes, respectively. For the BP reduction arm, the analysis involved 7370 participants, with 3543 of them receiving ACEI and 3827 of them receiving ARB. The intervention duration ranged from 6 to 156 weeks, and the sample size extended from 20 to 1185 patients. Mean age ranged from 47.9 to 72.5 years and male percentage ranged from 0 to 65.0%. Regarding comorbidities and other characteristics, 8 studies reported DM ranging from 10.5 to 100.0% [19, 26, 32, 33, 36, 38, 39, 42], smoking was reported in 7 studies ranging from 2.8 to 29.0% [19, 22, 30, 32, 38, 39, 42], alcohol consumption in 5 studies ranging from 22.0 to 49.6% [19, 22, 30, 32, 42], and previous antihypertensive treatment ranging from 32.9 to 100.0% [19, 21, 23, 25, 26, 29, 31,32,33,34, 36,37,38, 40,41,42, 45, 46]. For the mortality/morbidity outcome, the analysis involved 25,158 participants, with 7566 of them receiving ACEI, 9855 of them receiving ARB, and 7737 of them receiving placebo. The intervention duration ranged from 26 to 60 months with a sample size from 250 to 17,607 patients. Mean age ranged between 60.6 and 76.4 years and male percentage ranged between 35.5 and 76.6%. Regarding co-morbidities and other characteristics, 6 studies reported DM, smoking, and previous antihypertensive treatment with a range of 12.1–100.0%, 8.7–50.9%, and 52.7–100.0%, respectively [48,49,50,51,52, 54]; previous MI was reported in 4 with a range of 4.6–43.1% [50,51,52, 54]; and previous stroke in 3 with a range of 3.9–100.0% [48, 50, 54].
Assessment of risk of bias
For the BP reduction outcome, the risk of bias assessment was done for 30 studies. All of them were of double-blind design. Risk of bias was unclear in the majority of them (n = 25), mainly because of inadequate random sequence generation and allocation concealment [20,21,22,23,24,25,26,27,28,29,30,31, 33, 35, 36, 38,39,40,41,42,43,44, 46, 47]. Five of them were not carefully weighed for risk of bias, because there was baseline imbalance between the different treatment groups, or because participants were receiving prior antihypertensive treatment without an adequate wash-out period, or because sponsor’s involvement was considered to have affected the results [19, 32, 34, 37, 45]. For the mortality/morbidity outcome, the risk of bias categorization was performed for eight studies. All of them were of double-blind design and all of them were characterized as studies with unclear risk of bias. All studies had pharmaceutical funding and all of them include participants, which had previously received antihypertensive treatment or were receiving additional antihypertensive therapy during the trial or received the study drug as a tolerance test. Supplementary material 2 shows the authors’ evaluation for risk of bias for each included citation for the BP reduction and the mortality/morbidity outcomes, respectively.
ACEIs and ARBs were compared in SBP reduction in 28 studies [19,20,21, 23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44, 46, 47] and in DBP reduction in 29 studies [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37, 39,40,41,42,43,44,45,46,47]. The effect of the two drug classes did not differ in the SBP (WMD: 0.59, 95% CI: −0.21 to 1.38) or DBP reduction (WMD: 0.62, 95% CI: −0.06 to 1.30) as Fig. 2a, b show, respectively.
The effect of ACEIs and ARBs was compared in SBP reduction for patients aged >60 years in three studies [33, 36, 38] and in DBP reduction in two [33, 36]. ARBs were superior in reducing SBP (WMD: 2.02, 95% CI: 0.72 to 3.32) and DBP (WMD: 1.47, 95% CI: 0.69 to 2.26) in those patients as it is shown in supplementary material 3.
Subgroup and sensitivity analyses
The main analysis showed insignificant heterogeneity among the studies included in the SBP (chi2: 34.48, p: 0.15, I2: 22%), while there was high heterogeneity (chi2: 77.51, p < 0.00001, I2: 64%) among the studies included in the DBP reduction analysis. Further investigation for the heterogeneity that emerged in the DBP reduction analysis was done with subgroup analyses. The results showed that, from the characteristics examined, a possible explanation for the heterogeneity could be the inclusion of male patients in a proportion >50% in a study (chi2: 5.04, p: 0.02, I2: 80.1%), as supplementary material 4 shows.
For DBP reduction analysis, sensitivity analyses were carried out including the 24 main analysis’ studies not evaluated in any field as high risk (analysis 1), and the 13 main analysis’ studies that did not report pharmaceutical funding (analysis 2). Supplementary material 5 show that sensitivity analyses, in accordance with the main analysis, showed no difference between ACEIs and ARBs in DBP reduction (analysis 1: WMD: 0.65, 95% CI: −0.12 to 1.42, analysis 2: WMD: −0.07, 95% CI: −1.32 to 1.17) but could not explain the main analysis’ heterogeneity (analysis 1: chi2: 61.01, p < 0.00001, Ι2: 62%, analysis 2: chi2: 35.75, p: 0.0004, Ι2: 66%).
Meta-regression analysis was conducted to investigate the effect of various factors on the results of the DBP reduction analysis due to significant heterogeneity resulting both from the main analysis as well as from the sensitivity analyses. Statistically significant was considered a result if p value <0.10. Table 3 shows the factors studied in the univariate meta-regression analysis. Of these, the drugs’ dose used in the studies, male gender, and funding was found to affect the main DBP analysis results. Afterward, a multivariate meta-regression analysis model with these three factors was performed, which showed that using the maximum drug dose in one of the two intervention groups compared to using maximum drug dose in neither group was the only statistically significant factor (p value: 0.0843). Nevertheless, this model explains 28.13% of the main analysis’ heterogeneity between the studies.
Mortality and morbidity
The results of the NMA for each comparison and for each outcome are presented in Table 4. The two pharmacological categories did not show differences in all-cause mortality (RR: 0.96, 95% CI 0.80 to 1.14), CV mortality (RR: 0.87, 95% CI 0.67 to 1.14), fatal and non-fatal MIs (RR: 1.02, 95% CI 0.75 to 1.37), and fatal and non-stroke (RR: 1.13, 95% CI 0.87 to 1.46). Development and/or hospitalization for HF was less in patients treated with ACEIs than ARBs (RR: 0.71, 95% CI 0.54 to 0.93) As far as the ranking of the treatments for all-cause mortality, CV mortality, and HF (development and/or hospitalization for), ACEIs was found superior over the ARBs and placebo (p score: 0.84, 0.91, and 1.00 respectively). The best treatment for fatal and non-fatal MIs and stroke was ARBs (p score: 0.68 and 0.91, respectively). The detailed results of the NMA analyses (comparisons, ranking, and homogeneity/consistency results) are presented in Supplementary material 6.
Quality of the information
No serious risk of bias, imprecision, or indirectness were found as regards the two direct comparisons of the traditional meta-analysis. Inconsistency was found in the DBP outcome due to high heterogeneity between the studies. Publication bias was detected only for SBP because the funnel plot (Supplementary material 7) seemed to have some kind of asymmetry with the small studies being absent in the lower right of the graph. However, Egger’s test recovered insignificant publication bias (p value: 0.7169). Serious risk of bias, indirectness, or publication bias for any of the direct comparisons of the NMA were not found. Inconsistency was found serious in some comparisons due to heterogeneity. Imprecision was considered serious in some comparisons because the 95% credible interval intersected unity (underlining the possibility of sizeable benefit and serious harm in terms of the risk of all-cause death, CV death, MI, and stroke). Applying the GRADE criteria for direct and indirect comparisons of the network, serious risk of bias and indirectness for studied outcomes were not revealed. Studies with high heterogeneity in the all-cause mortality outcome led to serious inconsistency between them and in CV mortality due to inconsistency in the NMA. In summary, there was moderate confidence in SBP and DBP reduction estimates, as well as in fatal and non-fatal MI and stroke estimates, supporting that there was no difference between ACEIs and ARBs. There was high confidence in development and/or hospitalization for HF estimates providing evidence for the use of ACEIs over ARBs and low confidence in mortality from any cause and CV cause estimates supporting that there was no difference between the two categories. Supplementary material 8 and material 9 summarize the GRADE assessment of outcome quality of evidence for direct and indirect comparison in the NMA.
This systematic review and meta-analysis analyzed the available evidence on the efficacy of ACEIs and ARBs in BP reduction and CV events in essential HT patients. Traditional meta-analysis for BP reduction included 30 studies and NMA for mortality and morbidity included 8 studies. The main conclusions of the above analyses are that ACEIs and ARBs do not differ in SBP and DBP reduction, fatal events from any cause and CV causes, fatal and non-fatal MI, and stroke, but ACEIs were shown to be more protective than ARBs in development and/or hospitalization for HF.
Although HT has a high worldwide prevalence and it is proven to be an independent risk factor for CV and renal disease, the two most commonly used pharmacological agents for HT, ACEIs, and ARBs have been investigated head to head for their efficacy on BP reduction, mortality, and morbidity in essential HT patients in a relatively limited number of double-blind RCTs. Results from the present analysis are in agreement with those from a recent systematic review regarding the total and CV mortality in hypertensive patients but not separately the development and/or hospitalization for HF . In this systematic review, Li et al. attribute the dilute evidence for the efficacy of ARBs, regarding mortality and morbidity outcomes, to the few numbers of placebo-controlled trials for HT compared with ACEIs. Results from those placebo-controlled trials providing the significance of ACEIs for the development and/or hospitalization for HF have been established years ago but only in HF or/and post-MI patients and not exclusively in a hypertensive population [56,57,58]. Similarly, our results confirm the higher evidence of ACEIs in an effort to isolate the hypertensive population possibly with other co-morbidities, which could be the reason for these results as the above trials indicate.
Moreover, an NMA that investigated the comparative efficacy of ARBs in BP reduction and CV event rates in patients with HT conclude the same results. This NMA does not include exclusively double-blind RCTs and the outcome of development and/or hospitalization for HF was not studied .
The outcome of the development of HF (but not hospitalization) was studied in an NMA in which ARBs and ACEIs do not have a statistically significant difference. The difference between this NMA and our effort was that Ricci et al. do not include only patients with HT but also with other health problems .
Thomopoulos et al. , published the results of a meta-analysis using RCTs in hypertensive patients or in cohorts with at least 40% of hypertensive patients investigating the CV effects of the five major classes of antihypertensive drugs. Similarly to our results, when ACEIs compared to ARBs they did not find differences in CV outcomes, except HF approaching a statistical significance in favor of ACEIs.
The strengths of this meta-analysis are related to the variety of the assessed outcomes and the investigation of plausible causes of heterogeneity in the DBP reduction analysis by subgroup, sensitivity, and meta-regression analyses. In addition, for the data synthesis, an intention-to-treat basis was used. When the studies had multiple BP measurements at different times during the intervention, the final measurements at the end of each intervention were extracted and used, avoiding a unit-of-analysis error.
However, several limitations should be appreciated for the present analysis. The use of these drugs can also be influenced by factors in the clinical setting such as tolerability and side effects. ACEIs are supposed to carry a higher risk for some side effects (a dry cough) that can impact on the use and prescription pattern of these drug classes. Even if this is not the focus of our systematic review, tolerability and side effects also have to be considered in the wider clinical perspective when drug classes are evaluated. In addition, drug dosages may have a great variety in different trials, so when drugs are compared for clinical efficacy, a crucial question is pharmacological equipotency. However, this comparison seems to be impossible most of the times due to the studies’ design, as in many protocols there was a prediction to increase the dosage according to the achievement of BP target at a predetermined period of follow-up. Another possible argument could be that at least the BP-lowering properties were equal between these two classes of drugs. The NMA included studies that reported subgroup analyses including only the hypertensive patients of studies with different population inclusion criteria. This could partly explain the heterogeneity and inconsistency found in the all-cause and CV mortality analyses. Finally, solid evidence for mortality and morbidity outcomes were limited because studies while eligible for inclusion could not be analyzed as the data for the hypertensive subgroup were not extractable.
This systematic review did not exclude studies in which patients with essential HT had other co-morbidities to explore the efficacy of these drug classes in these patients in the presence of different co-morbidities. Unfortunately, only eight studies were found eligible and included. Patients’ co-morbidities were not reported in many studies or a small number of patients had co-morbidities or in most cases of studies including patients with co-morbidities, the effect of ACEIs and ARBs in BP reduction for these patients group were not reported separately. Therefore, ACEI and ARB impact on essential HT patients and co-morbidities for BP reduction and long-term effect on mortality and morbidity needs to be further studied.
In conclusion, the present analysis found similar efficacy of ACEIs and ARBs in BP reduction and mortality and morbidity rates in treated essential HT patients. Nevertheless, these results should be interpreted with caution due to the high heterogeneity in DBP reduction, all-cause mortality analyses, and the inconsistency in the CV mortality analysis. Further research is required to investigate the short- and long-term effects of ACEIs and ARBs in patients with essential HT and co-morbidities.
What is known about topic?
ACEIs and ARBs have been investigated head to head for their efficacy on BP reduction, mortality, and morbidity in essential HT patients in a relatively limited number of double-blind RCTs.
Over the years, the extensive research investigated these two drug classes for the treatment of HT and other chronic diseases.
What this study adds?
ACEIs and ARBs do not differ in terms of SBP and DBP reduction, fatal events from any cause and CV causes, fatal and non-fatal MI and stroke.
ACEIs were shown to be more protective than ARBs in development and/or hospitalization for HF.
Further research is required to investigate the effects of ACEIs and ARBs in essential HT patients and co-morbidities in BP reduction and long-term mortality and morbidity risk.
An interim analysis of the results has been presented during the European Society of Hypertension 2018 Congress in Barcelona, Spain.