Calcium channel blockers (CCBs) are widely used to control hypertension. Previous work suggested that their use could increase heart failure (HF), which is 1 of the consequences of uncontrolled hypertension. Information about the effect of CCBs on incident HF in patients with hypertension is scarce. A systematic review was conducted to evaluate patients with hypertension treated with CCBs and incident HF. An electronic search of publications was conducted using 8 major databases. Studies were eligible if they (1) were randomized clinical trials, (2) performed comparisons of CCBs versus active control, (3) randomized >200 patients, (4) had follow-up periods >6 months, and (5) provided data regarding incident HF. Trials of renal transplantation patients, placebo-controlled trials, and HF trials were excluded. A total of 156,766 patients were randomized to CCBs or control, with a total of 5,049 events. The analysis indicated a significant increase in the diagnosis of HF in patients allocated to CCBs (odds ratio 1.18, 95% confidence interval 1.07 to 1.31). The effect observed was independent of incident myocardial infarction. Subgroup analyses indicated that patients with diabetes were at higher risk for developing HF (odds ratio 1.71, 95% confidence interval 1.21 to 2.41). In conclusion, the results suggest that patients with hypertension treated with CCBs have increased incident HF.
The aim of this systematic review was to evaluate clinical trials in patients with hypertension that involved calcium channel blockers (CCBs) and incident heart failure (HF).
Methods
Two librarians (MD and TC) searched Medline (1950 to 2009), Embase (1988 to 2009), the Cochrane Library, PubMed, International Pharmaceutical Abstracts (1970 to July 2009), Web of Science, BIOSIS Previews, and Scopus using relevant subject headings and keywords. The search terms used included, but were not limited to, “calcium channel blockers,” “calcium antagonists,” “calcium inhibitors,” “heart failure,” “heart disease,” and “cardiovascular disease.” Explosion of subject headings and truncation of terms were used as appropriate to make the search strategies as comprehensive as possible. Validated search filters were used when available to limit the results to randomized controlled trials and systematic reviews. When such filters were unavailable, text words related to desired study designs were used. For complete search strategies for individual databases, please contact the investigators.
Studies were eligible if they (1) were randomized clinical trials, (2) performed comparisons of CCBs versus active control, (3) randomized >200 patients, (4) had follow-up periods >6 months, and (5) provided data regarding incident HF. Trials of renal transplantation patients and placebo-controlled trials were excluded, as well as trials in which the main eligibility criteria was HF (HF trials).
Two reviewers (MCS and HL) independently assessed the eligibility of each citation. Discrepancies were resolved by consensus. Citations were selected according to eligibility criteria, and the full reports were retrieved for data acquisition. Cohen’s κ statistic was used to evaluate the concordance between abstractors. The quality of the retrieved reports was graded according to the Jadad score.
RevMan versions 4.2.9 and 5 (Cochrane Collaboration, Copenhagen, Denmark) were used for data analysis. DerSimonian-Laird random-effects or Mantel-Haenszel fixed-effects models were used as appropriate. Higgins’s statistical heterogeneity test (I 2 ) was applied to detect and quantify heterogeneity.
Continuous data are expressed as mean ± SD or as medians as appropriate. Categorical data are expressed as percentages. Odds ratios (ORs) with 95% confidence intervals (CIs) express treatment effects. The criterion for statistical significance was set at α = 0.05.
Optimal information size has been proposed as an improvement in the quality of meta-analysis. A priori calculation of the optimal sample size to allow the detection of clinically important differences in treatments (as in any well-designed clinical trial) will give the reader important information to evaluate and interpret the results properly. Optimal information size was calculated using data from Pahor’s meta-analysis ( Table 1 ).
| Alpha Error (%) | Power (%) | Sample Size (per Arm) |
|---|---|---|
| 5 | 70 | 14,200 |
| 5 | 80 | 17,955 |
| 5 | 90 | 24,041 |
| 1 | 90 | 34,123 |
Publication bias was minimized by carrying out an extended search in the 8 major databases mentioned previously. The reference lists of all reports retrieved were also checked. A funnel plot was used to formally assess publication bias.
Results
After searching 8 major databases, 5,326 citations were identified, from which 44 full reports were retrieved. Nineteen trials were included in the final analysis. Figure 1 shows the search flow of the trials and reasons for the exclusions. Cohen’s κ coefficients for inter reviewer agreement were 0.99 for the selection of the trials and 0.84 for the quality evaluation using the Jadad score.
Particulars of the 19 trials are listed in Table 2 . Seven trials recruited patients with hypertension, and 12 trials recruited patients with hypertension and ≥1 other risk factor for heart disease. Mean age varied from 54 to 76 years, with a mean follow-up period ranging from 2 to 5 years. Mean systolic and diastolic blood pressure at study entry ranged from 146 to 194 and 82 to 102 mm Hg, respectively. Most of the trials evaluated the dihydropyridine types of CCBs (15 trials), 3 trials evaluated benzothiazepines (verapamil), and 1 evaluated phenylalkylamine (diltiazem). Several drugs were used for the control group: angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARB), diuretics, and β blockers.
| Trial | n | Setting | Mean Age (years) | Follow-Up (years) | Men (%) | Mean BP, Baseline (mm Hg) | CCB (mg/day) | Control (mg/day) | End Point(s) |
|---|---|---|---|---|---|---|---|---|---|
| IDNT | 1,715 | HTN + DM + nephropathy | 58.3 | 2.6 | 66.4 | 160 ± 20/87 ± 11 | Amlodipine (10) | Irbesartan (300) | MI, HF, stroke, coronary revascularization |
| MIDAS | 883 | HTN | 58.2 | 3 | 78.3 | 150 ± 16/96 ± 5 | Isradipine (2.5–5) | HCTZ (12.5–25) | Carotid intima-media thickness |
| VHAS | 1,414 | HTN | 54.5 | 2 | 48.9 | 169 ± 10/102 ± 5 | Verapamil SR (240) | Chlorthalidone (25) | BP, carotid lesion |
| NICSEHS | 429 | HTN | 69 | 4 | 33.1 | 171 ± 13/94 ± 10 | Nicardipine SR (20) | Trichlormethiazide (2) | Stroke, MI, angina, HF |
| STOP-2 9 | 6,614 | HTN | 76 | 4 | 36.2 | 194/98 | Felodipine/isradipine † | BB, diuretic, ACE inhibitor † | CV mortality |
| INSIGHT | 6,321 | HTN + 1 RF | 65 | 3 | 46.3 | 167/96 | Nifedipine GITS (30) | Co-amlodipine (25/2.5) | CV mortality, MI, stroke, HF |
| NORDIL | 10,881 | HTN | 60 | 4 | 48.6 | 173 ± 17/105 ± 8 | Diltiazem (180–360) | BB/diuretic ‡ | Stroke, MI, CV death |
| J-MIND | 464 | HTN + DM | 60.2 | 2 | 50.5 | 162 ± 17/90 ± 12 | Nifedipine R (20–60) | Enalapril (5–20) | Macroalbuminuria >300 mg/day |
| ALLHAT | 33,357 | HTN + 1 RF | 66.9 | 5 | 53.1 | 146 ± 16/84 ± 10 | Amlodipine (2.5–10) | Lisinopril/chlorthalidone § | CAD or MI |
| CONVINCE | 16,602 | HTN + 1 RF | 65.6 | 3 ⁎ | 44.0 | 150 ± 15/86 ± 10 | Verapamil (180) | Atenol/HCTZ ∥ | Stroke, MI, CV death |
| INVEST | 22,576 | HTN + CAD | 66 | 3 | 47.9 | 149 ± 19/86 ± 12 | Verapamil ¶ SR (240) | Atenol # (50) | Death, MI, stroke |
| SHELL | 1,882 | HTN | 72 | 3 ⁎ | 38.7 | 178 ± 10/87 ± 6 | Lacidipine (4) | Chlorthalidone (12.5) | Stroke, sudden death, MI, HF, revascularization, CE |
| JMIC-B 17 | 1,650 | HTN + CAD | 65 | 3 | 68.8 | 147 ± 19/82 ± 11 | Nifedipine SR (10–20) | ACE inhibitor ⁎⁎ | Cardiac events |
| VALUE | 15,245 | HTN + RFs | 67 | 4 | 57.6 | 155 ± 19/87 ± 10 | Amlodipine (5–10) | Valsartan (80–160) | Cardiac mortality/morbidity |
| AASK | 1,094 | HTN + nephropathy | 54.4 | 3 ⁎ | 64.5 | 150 ± 25/96 ± 14 | Amlodipine (5–10) | Ramipril (5–10) | Renal function |
| ASCOT-BPLA | 19,257 | HTN + 3 RFs | 63 | 5 | 76.5 | 164 ± 18/94 ± 10 | Amlodipine †† (5–10) | Atenolol ‡‡ (50–100) | MI, CAD |
| MOSES | 1,405 | HTN + stroke | 68.1 | 2 | 52.2 | 152 ± 18/87 ± 10 | Nitrendipine (10) | Eprosartan (600) | Total mortality/CV events |
| CASE-J 22 | 4,728 | HTN + any RF | 63.9 | 3.2 | 55.2 | 163.2 ± 14.1/91.8 ± 11.4 | Amlodipine (2.5–10) | Candesartan (4–12) | CV events/renal disease |
| ACCOMPLISH | 11,506 | HTN + RF | 68.4 | 60.5 | 145.3 ± 18.4/80.1 ± 10.8 | Benazepril + amlodipine §§ | Benazepril + HCTZ ∥∥ | CV event/CV death |
† Felodipine 2.5 mg, isradipine 2 to 5 mg, atenolol 50 mg, metoprolol 100 mg, pindolol 5 mg.
‡ Thiazide diuretic and or β blocker at the discretion of the treating physician.
§ Lisinopril 10 to 40 mg and chlorthalidone 12.5 to 25 mg.
∥ Atenolol 50 mg ± hydrochlorothiazide 12.5 mg.
¶ Trandolapril could be added.
# Hydrochlorothiazide could be added.
⁎⁎ Enalapril 5 to 10 mg, imidapril 5 to 10 mg, lisinopril 10 to 20 mg.
†† Perindopril could be added.
§§ Benazepril 20 to 40 mg, amlodipine 5 to 10 mg.
∥∥ Benazepril 20 to 40 mg, hydrochlorothiazide 12.5 to 25 mg.
We collected information from individual trials with regard to the criteria used to define and adjudicate HF. We were able to obtain the definitions of HF in 11 of the 19 trials (58%). Most criteria were based on clinical features, laboratory tests, chest imaging, and response to fluid removal by either diuretics or interventions (such as dialysis or ultrafiltration). However, clear predefined criteria were obtained from 4 of the 19 trials included in the final analysis (21%) (the Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT], the Japan Multicenter Investigation for Cardiovascular Diseases–B [JMIC-B], the Irbesartan in Diabetic Nephropathy Trial [IDNT], and the International Verapamil-Trandolapril Study [INVEST]). Most of the definitions were centered on systolic HF, and no description of diastolic HF was found.
In terms of outcome assessment, 15 of 19 trials (79%) clearly reported that the evaluation of incident HF was reviewed by a blinded committee. Four trials provided either vague or no information about adjudication.
There was no significant difference between systolic blood pressure reduction in the CCB and control arms. Patients allocated to CCBs had slightly lower diastolic pressures, with a difference of 0.66 mm Hg. This difference was statistically significant, although clinically irrelevant, and would not explain nor alter the interpretation of the results ( Figure 2 ).
A total of 156,766 patients were randomized to CCBs or control, with a total of 5,049 events. Analysis indicated a significant increase in the diagnosis of HF in patients allocated to CCBs (OR 1.18, 95% CI 1.07 to 1.31), while results of the test for heterogeneity were moderate (I 2 = 38%). The increase in HF was consistent among trials in that all but 4 had point estimates showing a higher risk for HF in the CCB group ( Figure 3 ). Incident HF was analyzed by type of active control (ACE inhibitors or ARBs, diuretics, and β blockers and/or diuretics; Figure 4 ). There were 47,688 patients in trials comparing CCBs to ACE inhibitors or ARBs, with a total of 2,722 events. There was a significant increase in incident HF in patients allocated to CCBs (OR 1.21, 95% CI 1.10 to 1.32). The magnitude of the effect was even more pronounced when CCBs were compared with diuretics. There was a total of 46,723 patients randomized, with 1,859 events. The OR was 1.32 (95% CI 1.04 to 1.66). A total of 52,714 patients were randomized to CCBs or β blockers and/or diuretics, with 771 events. The OR was 1.01 (95% CI 0.82 to 1.25). Publication bias was unlikely, as assessed by the funnel-plot method ( Figure 5 ).
