Meta-analysis of Trials Comparing β-Blockers, Calcium Antagonists, and Nitrates for Stable Angina FREE

Stable angina affects more than 7 million people in the United States,¹ with an estimated 350,000 new cases annually. Long-term pharmacotherapy to prevent anginal symptoms has consisted of β-blockers, calcium antagonists, long-acting nitrates, or their combinations.² The choice of a first-line agent has been controversial. Because long-term use of β-blockers has been shown to reduce mortality following myocardial infarction,³ recent clinical guidelines have suggested that β-blockers be used as first-line agents in patients with stable angina.^4- 5 However, calcium antagonists are often used as initial therapy because they are assumed to have equal efficacy in relieving angina, and fewer adverse effects.^6- 7 Others have recommended that therapy begin with long-acting nitrates,^6,8 but concern about the need for a nitrate-free period has led many physicians to use long-acting nitrates as a second-line agent.² No single class of therapy has been proven superior.^9- 10

Recent case-control studies and meta-analyses of randomized trials have suggested that short-acting calcium antagonists increase the risk of cardiac events among patients with hypertension¹¹ and that nifedipine increases mortality following acute ischemic syndromes.^12- 13 These observations have led to considerable discussion and controversy.¹⁴ Even if these concerns prove to be valid in the specific populations studied, it is uncertain whether the findings in these patient groups can be applied to patients who have stable angina.

To determine the relative efficacy and safety of antianginal therapies, we performed a meta-analysis of trials that directly compared β-blockers, calcium antagonists, and long-acting nitrates in patients who had stable angina. We also assessed the efficacy and safety of nifedipine and short-acting calcium antagonists by examining trials that compared these agents with β-blockers.

We searched MEDLINE (1966-1997) and EMBASE (1974-1997), and reviewed cited references of retrieved articles to identify relevant studies. Our search criteria were (1) the Medical Subject Headings angina pectoris or the text word angina and (2) publication type randomized controlled trial or text word containing a form of the word random. For this criteria set, we searched MEDLINE, then EMBASE. We also performed searches replacing the second criterion with (3) publication type controlled clinical trial or (4) text word double-blind. Finally, we reviewed the bibliographies of identified trials to locate other relevant studies.

We restricted this review to controlled trials that directly compared drugs from 2 or 3 major antianginal drug classes, including long-acting nitrates, β-blockers, and calcium antagonists. Trials were included if the enrolled patients had a history of stable angina. Studies were included if they were either randomized parallel design or crossover studies. Studies of the following design were excluded: 1 drug vs placebo, comparison of different dosage of the same drug, and 1 drug vs a 2-drug combination. Studies with a duration of less than 1 week were excluded. Trials that used a study medication during a run-in period prior to randomization were also excluded.

For studies that compared more than 1 drug from a given class, we limited our analysis to the drug most commonly used for stable angina. To determine the most commonly used agents, we obtained pharmaceutical data from IMS Health, Plymouth Meeting, Pa, that described the frequency of medications prescribed for stable angina (in the absence of hypertension) in 1998. For studies comparing more than 1 dosage, we chose the dosage closest to the midpoint of the recommended treatment dosage range from the Physicians' Desk Reference.

We examined measures of antianginal efficacy and tolerability, conducting separate meta-analyses for each outcome. Measures of antianginal efficacy included number of angina episodes per week (patient recorded), nitroglycerin use per week (patient recorded), total exercise time, and time to 1-mm ST depression on a treadmill or upright bicycle. For crossover studies, mean between-person values were used to allow comparison with parallel studies. The tolerability measure was the rate of subject withdrawal due to adverse events, defined as death or cardiac or noncardiac symptoms. Withdrawal due to protocol violation or loss to follow-up were not counted as adverse events. We also examined the combined end point of cardiac death and nonfatal myocardial infarction for studies in which either was reported.

Study selection was performed initially by title review (P.A.H. and M.A.H.). Candidate abstracts were then reviewed and appropriate studies were selected for data retrieval. Two independent reviewers extracted data from each article, compared their results, and settled any differences by consensus. One of the 2 reviewers had training in cardiology and the other had training in epidemiology or health services research. All outcome data were abstracted a third time by 1 of the investigators (P.A.H., B.F., or B.K.L.).

We examined several subgroups of trials to determine their contribution to the summary differences between β-blockers and calcium antagonists. The 2 prespecified comparisons of interest were nifedipine vs nonnifedipine calcium antagonists and short- vs long-acting calcium antagonists. Short-acting preparations were defined as those with usual dosing of 3 or more times per day according to the Physicians' Desk Reference. Additional comparisons were made based on the funding source (industry vs other) and the location of the trial (Europe vs United States). Because several trials examined drugs that are not in common use for treatment of stable angina, we repeated our analysis after limiting the trials to those agents that had at least 2% market share in the United States during the first quarter of 1998 (data from IMS Health). Based on these data, common therapies were limited to 4 calcium antagonists (amlodipine, diltiazem, nifedipine, and verapamil) and 5 β-blockers (atenolol, carvedilol, propranolol, metoprolol, and nadolol). We compared trials that were double-blind with those that were not to determine the effect of this quality measure. We did not evaluate subgroups of long-acting nitrates because the number of trials of this class of agents was small.

We combined estimates of the mean difference for number of angina episodes and number of nitroglycerin tablets used per week, the standardized mean difference for exercise time, and the odds ratio (OR) for adverse events, cardiac death, and myocardial infarction.¹⁵ The mean difference_i between the drug treatment arms is equal to the mean_ai−mean_bi, where a and b are the 2 arms of the ith study. The estimated variance is v_i=σ²_i (1/n_ai+1/n_bi), where n is the within-study sample size for the 2 study arms and σ² is estimated as the pooled within-group variance.

We used the standardized mean difference to compare exercise times because exercise protocols varied widely across studies. This measure expresses the difference in mean exercise time as a fraction of the SD. The measure is calculated as d_i=(mean_ai – mean_bi)/pooled SD. The variance for this value is estimated as v_i=[(n_ai+n_bi)/ n_ai n_bi] + [d_i²/2(n_ai + n_bi)]. For this comparison, a positive value indicates increased exercise time in the treatment group. We pooled both mean differences and standardized mean differences by weighting the individual differences by the inverse of the variance. Random effects were assumed.

We used the Peto (fixed-effects) method to estimate summary ORs for the outcomes of death or myocardial infarction and withdrawal due to adverse events.¹⁶ Although we found no statistical evidence of heterogeneity (P>.20 for all) among studies using the Q statistic,¹⁶ we chose also to display random-effects results using the DerSimonian-Laird method¹⁷ where stated for comparison. For studies that reported no events, we substituted 0.1 in place of 0 for the random-effects calculation. A summary rate difference was calculated for trials with 1 or more adverse events.^16- 17 This measure complements the OR by providing an absolute difference in the adverse event rate. We examined differences between study subgroups of trials using analysis of variance.¹⁸

For the outcomes of cardiac death, myocardial infarction, and adverse events, the study sample size was defined as the number of patients randomized. For the outcomes of angina episodes, nitroglycerin use, and exercise time, the sample size was defined as the number of patients who completed the study. P values are 2-tailed.

A total of 143 articles were identified by the MEDLINE and EMBASE searches and citation review. We excluded 25 studies that may have fulfilled the inclusion criteria but were not written in English; 17 that were duplicates or substudies of an included study; 6 that used a study drug during a run-in period prior to randomization; and 5 that did not report any of the primary outcomes. The remaining 90 articles (63%) (Table 1) were abstracted and included in the analysis.^{19- 4950- 8990- 108}

The majority of studies (n=72) compared β-blockers with calcium antagonists (Table 1, Table 2, and Table 3). The included studies had a mean length of 8 weeks and only 2 trials^67,71 were longer than 6 months. Most studies (n=64) that compared β-blockers with calcium antagonists required evidence of coronary artery disease or ischemia, in addition to a history of stable angina. A positive treadmill test result was required in 57 trials. Exclusion criteria were mentioned in 56 of the reports and included congestive heart failure (n=46), recent myocardial infarction (n=45), bradyarrhythmia or heart block (n=31), significant lung disease (n=26), and diabetes mellitus (n=13). The trials were predominately crossover in design (n=49), and were most often conducted in Europe (n=50) or North America (n=16). Source of study funding was stated in 31 studies; of those, 14 received support from the pharmaceutical industry. Significant differences in the abstracted outcome measures between β-blockers and calcium antagonists were reported in only 17 of 72 trials.

Cardiac Death or Nonfatal Myocardial Infarction. Only the TIBET⁷¹ and APSIS⁶⁷ trials were long-term (>6 months) comparisons of β-blockers and calcium antagonists; these 2 studies accounted for 103 of the 116 cardiac events recorded in all trials (Figure 1). The pooled OR for these 2 trials for risk with β-blockers was 0.98 (95% confidence interval [CI], 0.66-1.47). For the 59 short-term trials, there was no obvious difference in cardiac death or myocardial infarction (OR, 0.90; 95% CI, 0.40-2.00). When both long- and short-term trials were combined (Table 3), there were 57 events in the β-blocker and 59 in the calcium antagonist groups. The OR for risk with β-blockers was 0.97 (95% CI, 0.67-1.38). A random-effects analysis (DerSimonian-Laird method) produced similar results (OR, 0.98; 95% CI, 0.67-1.43). No significant differences were found when the trials were grouped by type of calcium antagonist (nifedipine vs nonnifedipine) or duration of action (Figure 2, top).

Angina Episodes. The patients enrolled in the trials that compared β-blockers with calcium antagonists had moderately frequent angina (mean [SD], 4.0 [5.9] episodes per week during treatment). Compared with calcium antagonists, β-blockers were associated with 0.31 (95% CI, 0.00-0.62) fewer episodes of angina per week (P=.05) (Table 3). When the comparison was restricted to trials comparing β-blockers with nifedipine, there were 0.63 (95% CI, 0.23-1.00) fewer angina events per week with β-blockers (Figure 3, top). We did not observe a significant effect of the duration of action of calcium antagonists when short- and long-acting calcium antagonists were compared separately with β-blockers (Figure 3, top). Short-acting calcium antagonists were associated with an increase of 0.44 angina episodes per week (95% CI, −0.87 to −0.02), while long-acting calcium antagonists decreased angina frequency by 0.08 episodes per week (95% CI, −0.75 to 0.91).

Nitroglycerin Use. Despite the differences in the frequency of angina, nitroglycerin tablet use per week was not significantly different between patients treated with β-blockers and with calcium antagonists (Table 3). There was a nonsignificant reduction of 0.14 (95% CI, −0.19 to 0.46) fewer nitroglycerin tablets per week with β-blockers in studies that compared β-blockers with nifedipine. We found no differences when short- and long-acting calcium antagonist preparations were examined separately.

Exercise Time. There were no significant differences in time to ischemia (1-mm ST depression) between calcium antagonists and β-blockers (Table 3). Similarly, no difference in exercise time to ischemia was observed when only trials comparing β-blockers with nifedipine were evaluated (Figure 3, bottom). No differences were noted when short- and long-acting preparations of calcium antagonists were evaluated separately.

Withdrawal Due to Adverse Events. Adverse events that led to study withdrawal occurred in 8% of patients. Patients were less likely to discontinue β-blockers than calcium antagonists in the 51 studies that compared these agents (OR, 0.72; 95% CI, 0.60-0.86) (Figure 4 and Table 3). The absolute difference in the adverse event rate was 2.0 (95% CI, 0.5-3.4) fewer events with β-blockers compared with calcium antagonists for every 100 patients treated.

Trials comparing nifedipine with β-blockers had significantly fewer adverse events with β-blockers (OR, 0.60; 95% CI, 0.47-0.77) (Figure 2, bottom). In contrast, there was no significant difference in the rate of adverse events between nonnifedipine calcium antagonists and β-blockers. Adverse events were not significantly different between long- and short-acting calcium antagonists when these agents were compared separately with β-blockers (Figure 2, bottom).

Secondary Analyses. When we limited the trials to those that studied calcium antagonists and β-blockers that had at least 2% market share in the United States in 1998, we noted significantly fewer adverse events (OR, 0.73; 95% CI, 0.59-0.92), and angina episodes per week (absolute difference, 0.40; 95% CI, 0.05-0.75) for β-blockers compared with calcium antagonists. No significant differences in any of the outcome measures were found when the trials were grouped by funding source or study location.

We compared β-blockers with dihydropyridine calcium antagonists separately (nifedipine, n=26; nicardipine, n=5; amlodipine, n=1; felodipine, n=1) and found fewer adverse events (OR, 0.63; 95% CI, 0.50-0.80), and angina episodes per week (OR, 0.61; 95% CI, 0.23-0.99) with β-blocker therapy. However, nifedipine accounted for 79% of all dihydropyridines, and an effect of dihydropyridine therapy independent of nifedipine could not be demonstrated. Random- and fixed-effects methods of pooling gave similar estimates for all comparisons.

There were 12 studies that compared calcium antagonists with long-acting nitrates (Table 1). Patient characteristics were similar to those in the trials previously described that compared β-blockers with calcium antagonists. There were no significant differences in outcome between these classes of medications (Table 4); however, a trend (P=.10) was noted for greater number of angina episodes per week in patients who were taking long-acting nitrates.

Only 6 studies compared β-blockers with long-acting nitrates (Table 1). Patient characteristics were similar to those in the trials that compared β-blockers with calcium antagonists. There were no significant differences between β-blockers and long-acting nitrates in any of the outcomes measured ( Table 5). A trend (P=.08) toward increased nitroglycerin use per week was noted for patients who were taking long-acting nitrates.

To determine the effect of excluding articles written in languages other than English, we reviewed the abstracts (in English) of 12 (48%) of 25 studies whose full text was not written in English. None of the studies reported a significant difference in death or myocardial infarction rates, exercise time, angina episodes, or nitroglycerin use for comparisons between β-blockers and calcium antagonists (n=6), β-blockers and nitrates (n=2), or calcium antagonists and nitrates (n=4). Two studies that compared calcium antagonists with nitrates noted fewer adverse effects with calcium antagonists.

This quantitative overview shows that calcium antagonists were associated with a greater number of adverse events compared with β-blockers in trials of patients who have stable angina. Angina relief, as measured by number of angina episodes, nitroglycerin use, and exercise time, was not significantly different between β-blockers and calcium antagonists. Trial duration was too short to determine differences in long-term rates of survival or myocardial infarction between calcium antagonists and β-blockers. Similarly, too few studies compared long-acting nitrates with calcium antagonists or β-blockers to determine differences in survival, symptoms, or adverse events.

The safety of calcium antagonists for patients who have coronary artery disease has been questioned. Prior meta-analyses and case-control studies have demonstrated an increased risk of myocardial infarction in hypertensive patients who take short-acting calcium antagonists.¹¹ In patients who have acute coronary syndromes, mortality increased with increasingly large dosages of short-acting nifedipine.^12- 13 In a recent case-control study, short-acting calcium antagonists were associated with a 4-fold increase in the risk of cardiovascular deaths and hospital admission compared with β-blockers.¹⁰⁹ We were unable to determine whether there were important differences in mortality or myocardial infarction rates between antianginals because follow-up was less than 6 months in all but 2 studies. This important question could be addressed by larger long-term comparisons of these agents.

Our finding that β-blockers were as well tolerated as calcium antagonists was not expected. Past reviews have noted frequent adverse effects with β-blockers,⁸ whereas calcium antagonists have been recommended and used widely because of their low adverse-effect profile.^6- 7 Our findings must be tempered by noting that patients who were thought unlikely to tolerate β-blockers (such as those with congestive heart failure, heart block, or significant pulmonary disease) were generally excluded from the randomized studies. Nonetheless, our findings suggest that patients without these contraindications should have success in finding angina relief and freedom from adverse events when they use a β-blocker instead of a calcium antagonist.

Our study suggests that short- and long-acting nifedipine may not be as well tolerated as other calcium antagonists, consistent with past investigations of long-acting nifedipine. In 6 small randomized studies of at least 1 week in duration that compared long-acting nifedipine with other calcium antagonists, nifedipine was associated with more symptoms and adverse effects in 2^110- 111 and was equivalent to other calcium antagonists in 4.^112- 115

Our findings have several clinical implications. They suggest that β-blockers should be considered first-line agents because they are as effective as and better tolerated than calcium antagonists and have been shown to improve prognosis in other populations with coronary disease. For patients who have a history of myocardial infarction, β-blockers are already recommended as first-line agents because they are associated with decreased rates of cardiac death and myocardial infarction.³ These recommendations do not apply to patients with contraindications to β-blockers (eg, severe reactive airway disease).

Current treatment of angina in the United States frequently does not include β-blockers. A study using the National Ambulatory Medical Survey found that treatment with β-blockers was reported in only 21% of office visits by patients with coronary artery disease and no strong contraindication to β-blockade.¹¹⁶ Pharmaceutical data from the first quarter of 1998 (IMS Health) suggest that physicians prefer long-acting nitrates to calcium antagonists or β-blockers. European data also suggest that calcium antagonists are more often used than β-blockers for first-line therapy for patients with stable angina.⁷

The choice of a first-line agent for use in patients who have stable angina has been addressed by recent clinical guidelines. The European Society of Cardiology recommends β-blockers for patients who have clear-cut effort-related angina or a prior myocardial infarction.⁵ The North of England Stable Angina Guideline Development Group recommends β-blockers for all patients who have no contraindication,⁴ citing evidence that β-blockers lower mortality after myocardial infarction, reduce mortality for patients who have a subsequent myocardial infarction while taking medication, and reduce adverse vascular events in patients who have hypertension. Our study data support these recommendations and add evidence that β-blockers are equally well or better tolerated than and provide angina relief equivalent to that of calcium antagonists and long-acting nitrates.

Like those of any meta-analysis, our findings may have been affected by publication bias.¹¹⁷ It is possible that studies that failed to demonstrate significant differences were not submitted or accepted for publication. However, few (25%) of the studies included demonstrated significant differences between drugs, suggesting that any such bias may be small. Patients who have congestive heart failure, chronic obstructive lung disease, and nonexertional chest pain were rarely included in the studies; thus, our findings are not applicable to those types of patients. Only a few small studies compared long-acting nitrates with calcium antagonists or β-blockers; thus, our analysis was unlikely to detect any clinically important differences between these agents. We did not include studies reported in foreign-language publications in our meta-analysis. A recent study has found that studies published in English were more likely to report P <.05.¹¹⁸ However, review of the English abstracts that were available (50% of foreign-language studies) did not demonstrate a difference in significant findings. Our study did not examine measures of quality of life. Although adverse events severe enough to lead to withdrawal from the study occurred less frequently in patients treated with β-blockers than in those treated with calcium antagonists, less severe adverse effects were not consistently reported, so we could not measure differences in their rates.

In summary, our quantitative review suggests that β-blockers provide an equivalent reduction in angina and lead to similar or reduced rates of adverse events compared with either calcium antagonists or long-acting nitrates. The differences in adverse event rates and frequency of angina between β-blockers and calcium antagonists are accounted for in large part by nifedipine, which was associated with a higher rate of adverse events and angina symptoms. More long-term comparative trials are required to determine whether the different therapies confer a significant mortality difference in patients who have stable angina.