Results of the search

The results of the search are shown in the PRISMA diagram (Figure 1), We identified 4,700 records from database searches. 4,649 records remained after removal of duplicates. After screening titles and abstracts, we obtained 65 full‐text articles. Of these articles, we excluded 42 studies based on them not meeting our inclusion criteria.

Figure 1

---

---

Study flow diagram.

Included studies

See Characteristics of included studies for details.

We included 23 RCTs (AASK; ABCD; ALLHAT; ASCOT‐BPLA; CASE‐J; CONVINCE; ELSA; FACET; HOMED‐BP; IDNT; INSIGHT; INVEST; J‐MIC(B); MIDAS; NAGOYA; NICS‐EH; NORDIL; SHELL; STOP‐Hypertension‐2; TOMHS; VALUE; VART; VHAS) with a total of 153,849 participants. Five of the 23 trials were new in this update (CASE‐J; HOMED‐BP; J‐MIC(B); NAGOYA; VART).

All the included RCTs supplied explicit inclusion and exclusion criteria. Twenty trials included only hypertensive participants, but these were defined differently, as follows: 140/90 mmHg or more (FACET; INVEST; NAGOYA; VART); 150/90 mmHg or more (J‐MIC(B)); more than 160/95 mmHg (VHAS); more than 135/85 mmHg for participants with diabetes mellitus (IDNT); 140 to 179 mmHg systolic and/or 90 to 109 mmHg diastolic (ALLHAT); 150 to 210 mmHg systolic and 95 to 115 mmHg diastolic (ELSA); systolic BP ≥ 180 mmHg and/or diastolic BP ≥ 105 mmHg (STOP‐Hypertension‐2); diastolic BP of 100 mmHg or more, NORDIL, or of 90 to 99 mmHg (TOMHS); treated hypertension with an upper limit of 175/100 mmHg or untreated hypertension of 140 to 190 mmHg systolic or 90 to 110 mmHg diastolic (CONVINCE); BP ≥ 160/100 mmHg for participants with untreated hypertension or BP ≥ 140/90 mmHg for participants on antihypertensive treatment (ASCOT‐BPLA); systolic BP ≥ 150 mmHg and diastolic BP ≥ 95 mmHg, or only systolic BP ≥ 160 mmHg (INSIGHT); only diastolic BP ≥ 95 mmHg, AASK, or 90 to 115 mmHg (MIDAS); 160 to 210/220 mmHg systolic and less than 115 mmHg diastolic (NICS‐EH; VALUE); ≥ 160 mmHg systolic and ≤ 95 mmHg diastolic (SHELL); systolic BP ≥ 140 mmHg or diastolic BP ≥ 90 mmHg in participants < 70 years old, or systolic BP ≥ 160 mmHg or diastolic BP ≥ 90 mmHg in participants ≥ 70 years old (CASE‐J); mild‐to‐moderate hypertension (HOMED‐BP). Only one trial did not limit participants to elevated BP (diastolic BP ≥ 80 mmHg) (ABCD), but it reported outcomes on participants with elevated BP (diastolic BP ≥ 90 mmHg) separately, so data for hypertensive participants could be extracted.

Additional inclusion criteria varied for each study, as follows: with other risk factor(s) for coronary heart disease or cardiovascular disease (ALLHAT; ASCOT‐BPLA; CASE‐J; CONVINCE; INSIGHT); with coronary heart disease (INVEST; J‐MIC(B)); with cardiovascular risk factors or cardiovascular disease (VALUE); with type 2 diabetes mellitus (non‐insulin‐dependent diabetes mellitus), ABCD; FACET, or type 2 diabetes mellitus and nephropathy (IDNT); with glucose intolerance (type 2 diabetes or impaired glucose tolerance) (NAGOYA); African‐Americans with hypertensive kidney disease (AASK).
 

All 23 included RCTs recruited participants of both sexes, but age requirements differed amongst the trials, as follows: ≥ 30 years (VART); > 40 years (HOMED‐BP; MIDAS); > 50 years (INVEST; VALUE); > 55 years (ALLHAT; CONVINCE); > 60 years (NICS‐EH; SHELL); 18 to 70 years (AASK); 30 to 70 years (IDNT); 30 to 75 years (NAGOYA); 40 to 65 years (VHAS); 40 to 74 years (ABCD); 40 to 79 years (ASCOT‐BPLA); 45 to 69 years (TOMHS); 45 to 75 years (ELSA); 50 to 74 years (NORDIL); 55 to 80 years (INSIGHT); under 75 years (J‐MIC(B)); 70 to 84 years (STOP‐Hypertension‐2). In the CASE‐J trial, differing BP levels were required for participants aged < 70 years and ≥ 70 years.

Most trials followed a goal BP in their protocols, mostly less than 140/90 mmHg (ALLHAT; ASCOT‐BPLA; CONVINCE; FACET; INSIGHT; INVEST; VALUE; VART); or less than 150/90 mmHg (J‐MIC(B)); less than 130/80 mmHg for hypertensive participants with glucose intolerance (NAGOYA); less than 130/85 mmHg for participants with diabetes or renal impairment (ASCOT‐BPLA; INVEST); ≤ 135/85 mmHg or a decrease ≥ 10 mmHg systolic for diabetic participants (IDNT); ≤ 160/95 mmHg (STOP‐Hypertension‐2); less than 90 mmHg diastolic, NORDIL, or 95 mmHg (TOMHS); less than 95 mmHg with a fall of at least 5 mmHg (ELSA); less than 90 mmHg with a fall of at least 10 mmHg (MIDAS); reduction more than 20 mmHg or systolic BP ≤ 160 mmHg (SHELL); ≤ 90 mmHg or ≤ 95 mmHg with a reduction of at least 10% from baseline value (VHAS); 75 mmHg or less diastolic in the intensive‐treatment group and 80 to 89 mmHg diastolic in the moderate‐treatment group (ABCD); 102 to 107 mmHg of mean arterial pressure in the usual‐goal group and 92 mmHg or less in the lower‐goal group (AASK); a decrease ≥ 20 mmHg of BP if systolic BP was more than 160 mmHg or diastolic BP was more than 110 mmHg (FACET); 60 years old, systolic BP/diastolic BP 130/85 mmHg; 60 to 69 years old, systolic BP/diastolic BP 140/90 mmHg; 70 to 79 years old, systolic BP/diastolic BP 150/90 mmHg; 80 years old, systolic BP/diastolic BP 160/90 mmHg (CASE‐J); usual control 125 to 134/80 to 84 mmHg and tight control < 125/< 80 mmHg (HOMED‐BP).

Of CCBs for hypertension, dihydropyridines (DHPs) were the most commonly studied, especially amlodipine (AASK; ALLHAT; ASCOT‐BPLA; CASE‐J; FACET; IDNT; NAGOYA; TOMHS; VALUE; VART). Other DHPs studied included nifedipine (INSIGHT; J‐MIC(B)), felodipine (STOP‐Hypertension‐2), nisoldipine (ABCD), nicardipine (NICS‐EH), lacidipine (ELSA; SHELL), and isradipine (MIDAS). Other trials evaluated non‐DHPs such as an aralkylamine derivative verapamil, CONVINCE; INVEST; VHAS, and a benzothiazepine derivative diltiazem (NORDIL). One study did not describe the specific CCBs used (HOMED‐BP). The included RCTs compared one of the above CCBs to other classes of antihypertensive drugs, including: a diuretic (ALLHAT; INSIGHT; MIDAS; NICS‐EH; SHELL; TOMHS; VHAS); a beta‐blocker (AASK; ASCOT‐BPLA; ELSA; INVEST; TOMHS); a diuretic or beta‐blocker, or both, data of which could not be separated for each drug (CONVINCE; NORDIL; STOP‐Hypertension‐2); an alpha‐1‐antagonist (TOMHS); an ACE inhibitor (AASK; ABCD; ALLHAT; FACET; HOMED‐BP; J‐MIC(B); STOP‐Hypertension‐2; TOMHS); or an ARB (CASE‐J; HOMED‐BP; IDNT; NAGOYA; VALUE; VART).

Supplemental antihypertensive agents other than the study drugs were permitted in most of the included trials, often administered sequentially to achieve BP goals (AASK; ABCD; ALLHAT; ASCOT‐BPLA; CASE‐J; CONVINCE; ELSA; HOMED‐BP; IDNT; INSIGHT; INVEST; J‐MIC(B); MIDAS; NAGOYA; NORDIL; SHELL; STOP‐Hypertension‐2; VALUE; VART; VHAS). The FACET trial added the study drug of the other group to participants whose BP was not controlled well. The TOMHS trial studied five classes of first‐line antihypertensive drugs, and added chlortalidone or enalapril, both of which were study drugs, to participants to control BP. NICS‐EH prohibited the use of any other antihypertensive drugs.

Outcomes differed amongst studies, but results for our planned outcomes of cardiovascular events and BP changes were reported in most trials. However, fatal MI, stroke, and heart failure were sometimes contained in death events, and in some trials components of cardiovascular events were not reported separately. As a result, not every trial supplied data to each meta‐analysis for outcomes of this review. Only two trials explicitly presented the mean BP changes with standard deviations (SDs), INVEST, or standard errors, TOMHS, which could be directly inputted into Review Manager 5 for analysis. In some other trials, mean BP change could be calculated by subtracting the baseline value at randomisation from the value reported at the end of the trial, but SDs for changes were not reported. We calculated change‐from‐baseline SDs when baseline and final values were known (Higgins 2011b) (AASK; ALLHAT; FACET; NICS‐EH; NORDIL; VALUE). However, when trials did not supply SDs for the baseline and final BPs, the BP results were not entered into the meta‐analysis (ABCD; ASCOT‐BPLA; CASE‐J; CONVINCE; ELSA; HOMED‐BP; IDNT; INSIGHT; J‐MIC(B); MIDAS; NAGOYA; SHELL; STOP‐Hypertension‐2; VART; VHAS).

Mean duration of follow‐up ranged from 2 to 5.3 years. One trial stated that no participant was lost to follow‐up and no participant refused to continue in the study (STOP‐Hypertension‐2), whilst loss to follow‐up and withdrawal were reported in the other 22 trials. All trials with the exception of NICS‐EH stated that an intention‐to‐treat analysis was performed.

Excluded studies

See Characteristics of excluded studies for details.

The reasons for exclusion included: non‐randomised design (Abascal 1998; Bhad 2011; DHCCP; Pahor 1995; Psaty 1995); the control group used placebo instead of other classes of antihypertensive drugs (Chen 2013; STONE; Syst‐China; Syst‐Eur); the comparison was performed between different kinds of CCBs, without any other classes of antihypertensive drugs (Abe 2013; Kes 2003); the follow‐up was shorter than two years (Espinel 1992; GLANT; Gottdiener 1997; Kereiakes 2012; Leon 1993; Papademetriou 1997; PRESERVE; Schneider 1991; Van Leeuwen 1995; Weir 1990; Zhang 2012); small sample of participants (fewer than 100 were randomised) (Bakris 1996; Bakris 1997; FACTS; Kim 2011; Maharaj 1992; Mesci 2011; Radevski 1999); cannot make a separate comparison of CCB with other classes antihypertensive drugs because of combination drug use (ACCOMPLISH; BEAHIT; Calhoun 2013; Cicero 2012; COLM; DEMAND; FEVER; Kojima 2013; Lauria 2012; OSCAR; Wen 2011); study groups differed in target BPs instead of drug classes (HOT); to avoid repeated inclusion of the research population in extended trial (CASE‐J Ex).

Allocation

All of the included studies were stated as randomised controlled trials. A computer‐generated code for randomisation was often used, but eight trials did not report the methods of allocation (ABCD; HOMED‐BP; INSIGHT; NICS‐EH; NORDIL; STOP‐Hypertension‐2; VART; VHAS). Allocation concealment was seldom described; only four trials stated that their randomisation codes were concealed at the clinical trials centre (ALLHAT; ASCOT‐BPLA; IDNT; INVEST), whilst in the CONVINCE trial, an interactive voice response system for randomising, assigning, and tracking blinded medication was used. Information was insufficient to assess this 'Risk of bias' domain for the remaining trials.

Blinding

All the included trials compared two first‐line antihypertensive drug classes to each other. With exception of the FACET trial, which was open‐label, the included studies were stated as blinded. In some trials active drugs were described as of indistinguishable appearance, but it was still impossible to know the extent of blinding (Higgins 2011a). Nine trials used a Prospective, Randomised, Open‐label, Blinded Endpoint (PROBE) design (ASCOT‐BPLA; CASE‐J; HOMED‐BP; INVEST; J‐MIC(B); NAGOYA; NORDIL; STOP‐Hypertension‐2; VART), which differs from the classical double‐blind method. In a PROBE study, outcomes are evaluated by a blinded endpoint committee to avoid detection bias; in this way treatment allocation might be open to risk of performance bias from participants and doctors (Hansson 1992).

Incomplete outcome data

Missing data caused by loss to follow‐up or withdrawals were on the whole equal amongst the treatment groups, and an intention‐to‐treat analysis, which meant data were analysed according to randomised treatment assignments regardless of the subsequent medications (Fergusson 2002), was performed in most of the included trials, with the exception of the STOP‐Hypertension‐2 trial (with negligible loss) and the NICS‐EH trial. Some sites and their participants were excluded after randomisation because of poor documentation of informed consent, data integrity concerns, or misconduct (ALLHAT; ASCOT‐BPLA; CONVINCE; INSIGHT), which could have led to attrition bias.

Selective reporting

In this review, we judged all included studies to have a low risk of reporting bias.

Other potential sources of bias

In FACET trial, when BP was not controlled well on monotherapy, the other study drug was added. In NORDIL trial, a diuretic or blocker was added in step 3, and any other antihypertensive compound could be added as step 4 in the diltiazem group. This could have affected the evaluation of effect of each study drug. 

All‐cause mortality

The effect of CCBs on death from any cause was not significantly different from that of any other evaluated agents: diuretics (5 trials with 35,057 participants: risk ratio (RR) 0.98, 95% confidence interval (CI) 0.92 to 1.04, I² = 0%; moderate‐certainty evidence); beta‐blockers (4 trials with 44,825 participants: RR 0.94, 95% CI 0.88 to 1.00, P = 0.54, I² = 0%; moderate‐certainty evidence); diuretics and beta‐blockers (3 trials with 31,892 participants: RR 1.03, 95% CI 0.94 to 1.12, I² = 0%; moderate‐certainty evidence); ACE inhibitors (7 trials with 27,999 participants: RR 0.97, 95% CI 0.91 to 1.03, I² = 0%; low‐certainty evidence); and ARBs (6 trials with 25,611 participants: RR 1.00, 95% CI 0.92 to 1.08, I² = 0%; moderate‐certainty evidence) (Analysis 1.1; Figure 4).

Figure 4

---

---

Forest plot of comparison: 1 All‐cause mortality, outcome: 1.1 CCBs versus other classes of antihypertensive agents.

MI (non‐fatal and fatal MI plus sudden or rapid death)

The effect of CCBs on MI was not significantly different from that of diuretics (5 trials with 34,072 participants: RR 1.00, 95% CI 0.92 to 1.08, I² = 0%; moderate‐certainty evidence); beta‐blockers (3 trials with 22,249 participants: RR 0.90, 95% CI 0.79 to 1.02, I² = 0%; moderate‐certainty evidence); diuretics and beta‐blockers (3 trials with 31,892 participants: RR 1.05, 95% CI 0.93 to 1.19, I² = 72%; moderate‐certainty evidence); and ACE inhibitors (7 trials with 27,999 participants: RR 1.05, 95% CI 0.97 to 1.14], I² = 66%; low‐certainty evidence). The incidence of MI was statistically significantly lower (P = 0.004) for CCBs compared to ARBs (6 trials with 25,611 participants: RR 0.82, 95% CI 0.72 to 0.94, I² = 0%; moderate‐certainty evidence) (Analysis 2.1).

We found significant statistical heterogeneity between trials comparing CCBs to diuretics and beta‐blockers (I² = 72%, P = 0.03) and CCBs to ACE inhibitors (I² = 66%, P = 0.007). A possible explanation for the heterogeneity may be that the type of CCB studied was different in each trial. The three trials involving diuretics and beta‐blockers respectively studied an aralkylamine derivative (verapamil, CONVINCE), a benzothiazepine derivative (diltiazem, NORDIL), and a DHP (felodipine, STOP‐Hypertension‐2). Another possible explanation is difference in participants. In the CONVINCE trial, participants diagnosed as having hypertension and who had one or more additional risk factors for cardiovascular disease were enrolled, but participants enrolled in the other two trials had no additional risk factors for cardiovascular disease. Six of seven trials involving ACE inhibitors studied DHPs, but three of them gave participants amlodipine (AASK; ALLHAT; FACET), and two administered felodipine (STOP‐Hypertension‐2), or nisoldipine (ABCD) and one gave nifedipine (J‐MIC(B)). One study did not describe the the specific ACE inhibitors and CCBs that were used (HOMED‐BP). The pooled RR for the trials comparing amlodipine and ACE inhibitors was 1.00 (95% CI 0.91 to 1.10, I² = 0%; low‐certainty evidence) (Analysis 2.2).

Stroke (non‐fatal and fatal stroke)

The incidence of stroke was not significantly different between CCB and diuretic groups (5 trials with 34,072 participants: RR 0.94, 95% CI 0.84 to 1.05, I² = 0%; moderate‐certainty evidence) or between CCB and diuretic and beta‐blocker groups (3 trials with 31,892 participants: RR 0.92, 95% CI 0.81 to 1.03, I² = 55%; moderate‐certainty evidence). Hypertensive participants treated with CCBs had a significantly lower risk of developing a stroke than those treated with a beta‐blocker (3 trials with 22,249 participants: RR 0.77, 95% CI 0.67 to 0.88, I² = 0%; moderate‐certainty evidence) or an ACE inhibitor (7 trials with 27,999 participants: RR 0.90, 95% CI 0.81 to 0.99, I² = 28%; low‐certainty evidence). There was no difference in risk of stroke between on CCB and ARB groups (6 trials with 25611 participants: RR 0.87, 95% CI 0.76 to 1.00, p = 0.05, I² = 15%; moderate‐certainty evidence) (Analysis 3.1), but the incidence of stroke was lower for amlodipine compared to ARBs (5 trials with 23265 participants:RR 0.85, 95% CI 0.74 to 0.98, I² = 0%)(Analysis 3.2; Figure 5).

Figure 5

---

---

Forest plot of comparison: 3 Stroke, outcome: 3.1 CCBs versus other classes of antihypertensive agents.

The reason for significant statistical heterogeneity between trials comparing CCBs to diuretics and beta‐blockers (I² = 55%, P = 0.11) might be related to the type of CCBs, similar to the description above in the MI results. Explanation for the heterogeneity may be that the type of CCB studied and inclusion criteria of participants were different in each trial. Regarding trials comparing CCBs to ARBs, one trial did not describe the specific CCBs used (HOMED‐BP), whilst five of six trials gave participants amlodipine (CASE‐J; IDNT; NAGOYA; VALUE; VART). The pooled RR for the trials comparing amlodipine to ARBs was 0.85 (95% CI 0.74 to 0.98, I² = 0%) (Analysis 3.2).

Congestive heart failure

There was no significant difference in development of congestive heart failure between CCB and beta‐blocker groups (2 trials with 19,915 participants: RR 0.83, 95% CI 0.67 to 1.04, I² = 0%; low‐certainty evidence) and between CCB and diuretic and beta‐blocker groups (3 trials with 31,892 participants: RR 1.15, 95% CI 0.99 to 1.33, I² = 0%; low‐certainty evidence). However, the risk of developing congestive heart failure was markedly higher in participants given CCBs than those given diuretics (5 trials with 34,072 participants: RR 1.37, 95% CI 1.25 to 1.51, I² = 17%; moderate‐certainty evidence); ACE inhibitors (5 trials with 25,276 participants: RR 1.16, 95% CI 1.06 to 1.28, I² = 0%; low‐certainty evidence); and ARBs (5 trials with 23,265 participants: RR 1.20, 95% CI 1.06 to 1.36, I² = 66%; low‐certainty evidence) (Analysis 4.1).

The lack of homogeneity between the five trials comparing a CCB to an ARB may be due to the different inclusion criteria of participants: the IDNT trial included hypertensive individuals with type 2 diabetic nephropathy, and the NAGOYA trial included hypertensive individuals with glucose intolerance, whilst the VALUE, CASE‐J, and VART trials only required participants to have hypertension and cardiovascular risk factors. There was a significant increase in congestive heart failure events among the diabetic nephropathic participants in IDNT (RR 1.58, 95% CI 1.17 to 2.14) and glucose intolerance participants in NAGOYA (RR 5.00, 95% CI 1.46 to 17.18]) treated with a CCB compared to those treated with an ARB.

Cardiovascular mortality

We added death caused by cardiovascular disease as a supplemental outcome, which differed from the published protocol, as we judged it to be important and it was reported in most of the included trials.

We found only a marginally lower cardiovascular mortality in the CCBs group compared to the beta‐blocker group (4 trials with 44,825 participants: RR 0.90, 95% CI 0.81 to 0.99, I² = 62%; low‐certainty evidence). The effect of CCBs on cardiovascular mortality was not significantly different from that of diuretics (4 trials with 32721 participants: RR 1.02, 95% CI 0.93 to 1.12, I² = 0%; moderate‐certainty evidence); diuretics and beta‐blockers (3 trials with 31892 participants: RR 1.04, 95% CI 0.92 to 1.18, I² = 0%); ACE inhibitors (6 trials with 27619 participants: RR 0.98, 95% CI 0.89 to 1.07, I² = 0%; moderate‐certainty evidence) or ARBs (3 trials with 4642 participants: RR 0.79, 95% CI 0.54 to 1.15, I² = 0%; moderate‐certainty evidence) (Analysis 5.1)

The heterogeneity amongst trials involving beta‐blockers (I² = 62%, P = 0.05) might be explained by the different types of CCBs that were evaluated: a non‐DHP in the INVEST trial (verapamil) and a DHP in the other three trials (amlodipine in AASK and ASCOT‐BPLA, and lacidipine in ELSA). After deselecting the INVEST trial, the pooled RR was 0.77 (95% CI 0.66 to 0.90, I² = 0%), still showing a significant decrease in cardiovascular mortality in the CCB group (Analysis 5.2).

Major cardiovascular events (MI, congestive heart failure, stroke, and cardiovascular mortality)

Compared to beta‐blockers, CCBs significantly reduced major cardiovascular events (3 trials with 22,249 participants: RR 0.84, 95% CI 0.77 to 0.92, I² = 0%). In contrast, when compared to diuretics, CCBs probably increased major cardiovascular events (4 trials with 33,643 participants: RR 1.05, 95% CI 1.00 to 1.09, I² = 0%, P = 0.03). There was no significant difference in total major cardiovascular events when CCBs were compared to diuretics or beta‐blockers (2 trials with 21,011 participants: RR 1.02, 95% CI 0.95 to 1.10, I² = 0%); to ACE inhibitors (5 trials with 25,186 participants: RR 0.98, 95% CI 0.94 to 1.02, I² = 45%); or ARBs (3 trials with 6874 participants: RR 0.97, 95% CI 0.78 to 1.22, I² = 32%) (Analysis 6.1; Figure 6).

Figure 6

---

---

Forest plot of comparison: 6 Major cardiovascular events, outcome: 6.1 CCBs versus other classes of antihypertensive agents.

The poor methodological quality of the FACET trial might be a source of heterogeneity amongst the five trials comparing CCBs with ACE inhibitors. We undertook a sensitivity analysis on this effect by deselecting the FACET trial; the results were unchanged (4 trials with 24,806 participants: RR 0.98, 95% CI 0.94 to 1.02, I² = 0%) (Analysis 6.2).

Systolic and diastolic BP reduction

Using the weighted mean difference method and the fixed‐effect model, we found that the mean systolic BP reduction of the CCB group was 0.81 mmHg (95% CI 0.56 to 1.06) less than that of the diuretic‐based regimen group, and 3.00 mmHg (95% CI 2.59 to 3.41) less than the diuretic‐and‐beta‐blocker‐based regimen group. Systolic BP reduction was ‐1.11 mmHg (95% CI −1.40 to −0.82) more with CCBs than with ACE inhibitors, and ‐2.10 mmHg (95% CI −2.46 to −1.74]) more than with ARBs. There was no significant difference between the CCB group and beta‐blocker group (P = 0.38), or between the CCB group and alpha‐1‐antagonist group (P = 0.27) (Analysis 7.1).

For diastolic BP, the mean reduction of the CCB group was −0.68 mmHg (95% CI −0.84 to −0.52) more than the diuretic group; −0.63 mmHg (95% CI −0.81 to −0.44) more than the ACE inhibitor group; −1.70 mmHg (95% CI −1.91 to −1.49) more than the ARB group; and −1.20 mmHg (95% CI −2.39 to −0.01) more than the alpha‐1‐antagonist group. Mean diastolic changes between the CCB and beta‐blocker groups were not significantly different (Analysis 7.2).

There was heterogeneity (I² of 85%) for the four trials comparing the effect of CCBs versus ACE inhibitors on systolic BP reduction, however there was no heterogeneity for the same comparison evaluating diastolic BP reduction. The heterogeneity was most likely due to the poor methodological quality of the FACET trial. Sensitivity analyses conducted without the FACET trial resulted in homogeneous significant mean differences for both systolic and diastolic BP: mean difference −1.00 (95% CI −1.29 to −0.70) and −0.62 (95% CI −0.81 to −0.44), respectively (Analysis 7.3).