Kizer JR, Kimmel SE. Epidemiologic Review of the Calcium Channel Blocker DrugsAn Up-to-date Perspective on the Proposed Hazards. Arch Intern Med. 2001;161(9):1145-1158. doi:10.1001/archinte.161.9.1145
In the setting of soaring popularity, postmarketing studies of calcium channel blockers came to suggest an increase in a variety of major adverse end points. The evidence, however, was largely observational, and large-scale trials capable of addressing the concerns were wanting. Clinical trials now support the safety and efficacy of the long-acting dihydropyridines for patients with both uncomplicated and diabetic hypertension, although conventional therapies and, in the latter case, angiotensin-converting enzyme inhibitors have superior proof of benefit. By contrast, short-acting dihydropyridines should be avoided. In the acute coronary syndromes, β-blockers remain the treatment of choice; the evidence for nondihydropyridines remains inconclusive. Stable angina calls for β-blockers as first-line therapy and nondihydropyridines as second-line therapy, whereas in ventricular dysfunction, safety data for nondihydropyridines are lacking. Initial reports of cancer, bleeding, and suicide have been contradicted by subsequent data, making the associations uncertain or unlikely. Remaining questions await completion of ongoing trials to better define the indications for these agents.
During the past several years, few issues in medical therapeutics have achieved more prominence than the debate over the safety and efficacy of calcium channel blockers (CCBs) in cardiovascular disease.1- 9 This relates in no small part to the status of CCBs, in the early 1990s, as the most widely used class of agents for the treatment of hypertension in the United States.10 It is also a consequence of the wide variety of conditions to which these agents were applied after their introduction. The emergence of unfavorable observational data more than a decade after their first release forced the medical community to reassess the drugs' potential benefits and harms—only to find great uncertainty stemming from a paucity of reliable evidence. Purported deleterious effects of CCBs have ranged from increased cardiac events and mortality11- 14 to cancer,15- 17 major hemorrhage,18- 20 and suicide.21
Now, 5 years and numerous studies later, we have gained important insights, although uncertainty persists. Interpretation is complicated by the heterogeneous properties of different CCB classes and by the development of physiologically distinct long-acting preparations. While the subject has been reviewed lately,22,23 to our knowledge no recent work has approached the data from a detailed epidemiologic perspective. In this article, we review the literature reporting on the major outcomes above—selected from a systematic search of MEDLINE articles and their reference sections—paying particular attention to study methodology. Although not a formal meta-analysis, this review should guide us to epidemiology-based, up-to-date conclusions on the merits, shortcomings, and continuing unknowns of this important group of agents.
The approval of CCBs in the 1980s rested on their proved efficacy on the surrogate outcomes of blood pressure lowering and angina relief.24 Few data on major cardiovascular outcomes or long-term safety were available. In this context, the appearance of 3 studies detailing adverse coronary events in 1995 produced grave concerns among the medical and lay communities alike.25 Psaty et al11 presented the results of a case-control study of hypertensive patients, demonstrating a 60% increase in the risk of myocardial infarction for CCBs compared with thiazide diurectics (TDs) or β-blockers (BBs).11 Soon after, Furberg and colleagues12,13 published a meta-analysis of coronary heart disease (CHD) trials documenting a significant association between high doses of nifedipine and increased overall mortality. To these reports, Pahor et al14 added their analyses of hypertensive subjects in the Established Populations for Epidemiologic Studies of the Elderly (EPESE) cohort, documenting higher mortality and coronary events for CCBs than for BBs. These reports generated an enduring controversy centered on the interpretation of observational studies to define suitable roles for these medications pending information from large-scale trials. This section will examine the totality of the evidence bearing on the association between CCBs and fatal and nonfatal coronary events.
Any assessment of Psaty and coworkers' and Pahor and colleagues' data must take into account the shortcomings intrinsic to observational studies evaluating medication effects. These studies present the special methodologic problem of "confounding by indication."26 In the observational design, treatment allocation is not under the control of the investigator, but at the discretion of the patient's physician. Since the physician's choice of the drug of interest depends on a host of distinguishing patient characteristics, exposed and unexposed subjects will be fundamentally different. When such differences in turn affect the outcome under evaluation, the requisite conditions for confounding will be satisfied. Investigators may apply standard techniques to adjust for these dissimilarities. These adjustments, however, are generally incomplete: the totality of clinical characteristics driving treatment selection is almost never fully measured and may, in fact, never be fully measurable.26,27 In consequence, attribution of an observed effect to the drug in question rather than to the clinical characteristics that prompted its use, especially when the risk estimate is modest, becomes problematic.26- 28
The odds ratios of myocardial infarction reported by Psaty et al were approximately 1.6, whereas Pahor and colleagues' nifedipine-associated relative risk of mortality was 1.7 (Table 1). In these observational settings, the modest magnitude of the estimates makes the reported associations suspect. Physicians responsible for the care of patients analyzed in these studies were at liberty to prescribe the antihypertensive agent of their choice. With regard to overall mortality in Pahor and coworkers' study, differential prescription of CCBs to patients with higher comorbidities would lead to confounding. Moreover, approved by the Food and Drug Administration for the treatment of stable angina pectoris, CCBs would tend to be used preferentially in this subset of patients at higher risk of myocardial infarction. This predilection for CCB use among patients with CHD was indeed observed by investigators in these studies. Although they adjusted for all measured confounders, such adjustments may not have entirely accounted for clinical differences tied to the selective use of CCBs that confer an increased risk of events.35 Last, while Pahor and associates' CHD risk estimates are high, their broad confidence intervals attest to their uncertainty.
Several observational studies failed to reproduce the findings in these original reports (Table 1), but these had few outcomes and often could not exclude large increases in risk (ie, 70% or greater).29- 31,33 Recognition of potential uncontrolled confounding led one subsequent study that did document an association to characterize its results as inconclusive.34 Thus, the true implication of the association remained unsettled. In addition, the Psaty et al and Pahor et al studies raised the question of whether their results, drawn from short-acting preparations, were applicable as well to the increasingly dominant long-acting CCBs. Unlike their short-acting counterparts, long-acting preparations do not cause the wide fluctuations in blood pressure or the same degree of neurohormonal activation—if any—believed to mediate the adverse effects.36 The case-control study by Alderman et al32 seemed to answer this question in the negative by documenting a 4-fold hazard for short-acting, but none for long-acting, preparations (Table 1). Yet, aside from the imprecision of some of these estimates, the study is prone to the same confounding-by-indication concerns that plague its peers. The prevalence of prior CHD was substantially higher in the short-acting than in the long-acting CCB group, which was in turn higher than in the alternative-drug group. These limitations make it impossible to draw firm conclusions from the findings.
Early clinical trials of CCBs in hypertension were small and not specifically designed to assess major cardiovascular risks.37,38 In a meta-analysis of 98 published randomized trials of long-acting nifedipine, some 70% had enrolled 100 patients or less, and approximately two thirds had a follow-up of less than 13 weeks.39 The overview found no significant difference between monotherapy and active controls, but documented a significant reduction in cardiovascular events in the combination therapy comparison (Table 2). Nevertheless, lack of uniformity in the comparison arms limits interpretation of the results.
More evidence on dihydropyridines (DHPs) came in the form of larger trials from China,40- 42 but failure to blind investigators to drug assignment and alternate treatment allocation leaves these trials open to the influence of investigator bias in assigning active therapy to patients with better baseline health profiles. Nonetheless, these trials consistently showed reductions in stroke risk and, driven primarily by this end point, cardiovascular risk by DHPs (mostly long-acting) compared with placebo. None demonstrated differences in myocardial infarction events, but their confidence intervals were wide.
Data from the Systolic Hypertension in Europe (Syst-Eur) study,43 the largest double-blind, placebo-controlled randomized trial of CCBs in hypertension, provide confirmation of these findings. Investigators evaluated the DHP nitrendipine, with addition of enalapril maleate or hydrochlorothiazide as needed to achieve target blood pressure, in elderly patients with isolated systolic hypertension. The trial was stopped prematurely when a statistically significant 42% reduction in the primary end point of stroke was reached. Early termination, however, limited the power to detect differences in secondary end points, so that the trial was able to suggest, but not demonstrate, a reduction in myocardial infarction (Table 2). Subsequently, the Swedish Trial in Old Patients With Hypertension–2 (STOP-2)44 compared CCBs, angiotensin-converting enzyme inhibitors (ACEIs), and conventional therapy (BBs and/or TDs), supplemented by alternative treatment as necessary, and found no significant differences in the primary end point of cardiovascular mortality among these groups (Table 2). Nevertheless, the study did document higher CCB risks of myocardial infarction and congestive heart failure compared with the ACEI group.
Most recently, 2 additional randomized trials comparing CCBs with conventional therapies have reported their findings. The Nordic Diltiazem (NORDIL) Study45 compared short-acting diltiazem hydrochloride with TD/BB-based therapy in patients with mild hypertension, while the Intervention as a Goal in Hypertension Treatment study (INSIGHT)46 evaluated long-acting nifedipine vs TD-based therapy in hypertensive patients with 1 additional cardiovascular risk factor (Table 2). Neither trial demonstrated a difference in its primary end point of combined cardiovascular events. The NORDIL Study found a marginally significant diltiazem-associated reduction in stroke, but INSIGHT did not reproduce this finding for nifedipine. By contrast, the latter trial showed increases in fatal myocardial infarction and nonfatal heart failure not observed in the former.
Analysis of the results from these 4 large-scale trials requires consideration of several key points. First, a substantial proportion of patients in these trials withdrew from their assigned treatment (30%-40% in STOP-2 and INSIGHT). This is of significance for the active control trials, because it would tend to bias the drug comparisons in these negative studies toward the null hypothesis (ie, to dampen any true differences). Second, important numbers of patients were receiving more than 1 drug. In Syst-Eur and STOP-2, this was the case for more than 40%. Interpretation of Syst-Eur, then, must recognize that its benefits reflect a DHP-based regimen and not the exclusive effects of nitrendipine. For the active control trials, this would similarly blunt any differences between drug classes. The STOP-2, NORDIL, and INSIGHT figures likely overestimate the similarity between CCBs and the other drug groups. Third, the analysis of multiple outcomes, for instance, 48 separate end points in STOP-2, increases the likelihood that any given difference would arise merely by chance. Thus, the observed differences in secondary end points, especially when accompanied by broad confidence intervals, need to be viewed cautiously.
Insight into the validity of the STOP-2 differences, however, is provided by several trials evaluating CCBs and ACEIs in patients with diabetes mellitus. In addition to several post hoc subgroup analyses unfavorable to CCBs,47,52,53 2 clinical trials have shown increased cardiovascular events for CCBs compared with ACEIs in hypertensive diabetic populations. The Fosinopril vs Amlodipine Cardiovascular Events Randomized Trial (FACET)48 found the ACEI-associated risk to be half that of the DHP arm for the combined end point of myocardial infarction, stroke, and hospitalized angina. The Appropriate Blood Pressure Control in Diabetes (ABCD) Trial,49 whose primary objective was to compare intensive to moderate blood pressure reduction strategies, was forced to terminate prematurely its arm of hypertensive subjects randomized to nisoldipine or enalapril on the secondary finding of a 5-fold–plus risk of myocardial infarction in the CCB group.
The results of ABCD and the FACET, however, cannot alone settle whether the differences in cardiac events stem from neutrality of CCBs and benefit of ACEIs, harm of CCBs and neutrality of ACEIs, or some combination thereof. Neither can the findings of STOP-2 in a population with a 10% prevalence of diabetes. But a number of additional trials shed light on this issue (Table 2). Hypertension Optimal Treatment (HOT)50 demonstrated a 67% decrease (95% confidence interval, 22%-86%) in major cardiovascular events in the diabetic subgroup randomized to the most intensive blood pressure reduction. To be sure, the absence of a felodipine-free control arm and the high concurrent use of ACEIs (41%) or BBs (28%) render impossible a conclusive claim of felodipine safety. Nevertheless, the analysis does attest to the superiority of intensive vs moderate felodipine-based regimens in diabetic patients, making important CCB hazards improbable. Moreover, post hoc analysis of the Syst-Eur diabetic subset documented significant reductions in all cardiovascular end points, and these reductions were substantially higher for diabetic patients than for nondiabetic patients.51
Further support for superior ACEI benefit comes from the recently reported Captopril Prevention Project (CAPPP)54 and Heart Outcomes Prevention Evaluation (HOPE) Study.55 While unable to demonstrate a difference in its primary cardiovascular end point among hypertensive patients randomized to captopril vs TD-BB, CAPPP documented significant ACEI-associated reductions in cardiac end points and mortality in its diabetic subgroup analyses. HOPE randomized patients with vascular disease or diabetes plus 1 additional cardiovascular risk factor to ramipril or placebo, demonstrating significant decreases in every major cardiovascular end point. In subgroup analyses, these findings held up irrespective of the presence of cardiovascular disease or diabetes mellitus. Thus, different event rates between diabetic patients receiving CCBs and ACEIs likely reflect ACEIs' superior cardioprotective properties and not CCB-mediated adverse effects.
In summary, based on clinical trials comparing long-acting DHPs with placebo, existing evidence supports the safety and efficacy of these agents in patients with mild to moderate hypertension. In populations at modest risk of CHD, such evidence is in the form of a reduction in cardiovascular events, driven primarily by a decrease in stroke. The data are insufficient to demonstrate a benefit for coronary events, as contrasted with existing data on TD-BB regimens. Large-scale trials evaluating long-acting DHPs or a short-acting non-DHP vs TDs, BBs, and/or ACEIs exclude moderate differences in combined cardiovascular events. Nevertheless, substantial treatment withdrawal and use of combination therapy across comparison groups may have biased the findings toward no difference. Moreover, significant differences in cause-specific outcomes, such as myocardial infarction, stroke, and heart failure, remain plausible. Data from the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT)56 and from a projected meta-analysis of antihypertensive trials57 should help refine treatment selection. At present, continued adherence to the latest recommendation from the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC-VI) of TDs and BBs as preferred, and long-acting DHPs and ACEIs as second-line, therapy for patients with uncomplicated hypertension (at modest risk of CHD) represents a sensible strategy58 (although the use of long-acting DHPs and diltiazem has received an important boost). Recent evidence, however, supports a different approach among diabetic hypertensive patients. In this population, regardless of the presence of proteinuria, strong consideration should be given to ACEIs as first-line therapy; TDs and long-acting DHPs are reasonable alternatives. Finally, short-acting DHPs have never demonstrated a reduction in cardiovascular outcomes in any group of patients with hypertension. Given their potential hazards, these agents should be avoided in the routine treatment of hypertension.
In contrast to hypertension, data on CCBs in acute coronary syndromes have come primarily from clinical trials and overviews. In the time since the controversy, however, 2 long-term registries of patients surviving 28 days after myocardial infarction have supplied additional data.59,60 Both documented increases in coronary events and mortality when short-acting CCBs were compared with BBs, but when the comparison groups excluded patients receiving BBs, the differences in outcomes largely disappeared. A third large cohort study similarly showed no significant mortality difference in comparisons of patients with CHD on and off a regimen of short-acting CCBs.61 Taken together, these studies strengthen the notion of superior BB efficacy in secondary prevention but cannot settle CCB safety or efficacy questions in the absence of BB administration.
Numerous randomized trials evaluated CCBs in acute coronary syndromes62- 78 well before the onset of the controversy. These assessed the effects of short-acting DHPs, primarily nifedipine, and non-DHPs, namely diltiazem and verapamil hydrochloride, on major CHD events compared with either placebo or BBs. In myocardial infarction, the drugs were assessed when instituted in the acute phase of the event, as well as with initiation thereafter and continuation long term. Yet, notwithstanding the observation of both favorable and unfavorable trends depending on the agent studied, no single trial could demonstrate a statistically significant effect on mortality or myocardial infarction outcomes—whether against placebo or BBs. This occurred despite the fact that, although most studies were insufficiently powered to detect moderate (20%) benefit (or harm), several were large enough to detect moderately large (30%-40%) effects in either direction.79 The lack of proof was summed up by an overview in 1989 by Held et al79 involving 19 100 patients that, moreover, found no evidence of heterogeneity among the different CCBs (Table 3).
Nonetheless, post hoc data review79 identified the possibility of benefit among non-DHPs, in contrast to a nonsignificant harmful effect for nifedipine. Furberg and coworkers' subsequent nifedipine meta-analysis12,13 included a single additional trial in patients with angiography-proven coronary disease.87 It documented a marginally significant overall association between nifedipine use and all-cause mortality, together with a significant data-derived dose-response relationship (Table 3). The pooled trials, however, were variously inhomogeneous, calling into question the validity of the findings. Furthermore, the overview left unaddressed the safety of the long-acting DHP preparations. The latter have since been evaluated by a single postinfarction trial that detected a nonsignificant effect involving very few outcomes.80
With respect to the non-DHPs, although individual placebo-controlled trials of diltiazem68,73 and verapamil71,77,88 in myocardial infarction yielded little evidence of cardiac benefit, exploratory analyses showed differential effects based on the presence or absence of congestive heart failure71,73 (Table 3). In these analyses, both diltiazem and verapamil reduced coronary events in patients without pulmonary congestion, and diltiazem appeared to increase their occurrence in patients with radiographic pulmonary edema. Subsequent overviews pooling the overall results of non-DHP trials documented nominally significant reductions in reinfarction (P<.05), but not mortality81,82 (Table 3). These findings, however, must be viewed with caution. Post hoc analyses increase detection of false-positive results,89 whereas cumulative meta-analyses are susceptible to the effects of statistical multiplicity81 and inclusion bias.90
More recent studies,83- 86,91,92 while providing additional evidence of non-DHPs' anti-ischemic properties, have failed to demonstrate benefits on myocardial infarction or mortality outcomes. Two small placebo-controlled trials did show benefit in composite end points that included ischemic symptoms—one, nonrandomized, of patients after infarction with stabilized heart failure receiving ACEIs,84 the other of thrombolytic-treated subjects with preserved left ventricular function.85 However, a moderate-sized randomized trial of diltiazem vs placebo after thrombolysis found no significant advantage to active therapy even when the combined primary end point included refractory ischemia.86 Thus, the reported differences by CCB class (non-DHP vs DHP) are without prospective confirmatory evidence of non-DHP benefits for major cardiovascular outcomes.
In conclusion, given the overwhelming evidence of BB benefit on coronary mortality when initiated early or late after myocardial infarction,93 and in accordance with the latest US guidelines,94 these agents remain the treatment of choice for this condition. Similarly, in line with evidence-based national recommendations,95 β-blockade constitutes first-line therapy in unstable angina. In both settings, the short-acting DHPs should be avoided. Scant data can be brought to bear on the long-acting DHPs. For the non-DHPs, there is no convincing proof of efficacy in the acute coronary syndromes for infarction or mortality outcomes, although the agents appear to afford benefits in symptom-driven end points. Some support does exist for the use of non-DHPs in patients with preserved left ventricular function when β-blockade is clearly contraindicated. The merits of this practice, however, require validation in the form of large-scale trials.
Long before the controversy, 3 small placebo-controlled DHP trials96- 98 in stable angina pectoris that found favorable effects on surrogate ischemic end points all reported more coronary events in the treatment arms, albeit too few to assess significance. Subsequently, an overview of 24 placebo-controlled trials accompanying new drug applications to the Food and Drug Administration found an increased rate of cardiovascular-related withdrawals in the CCB arms99 (Table 4). Lately, the Prospective Randomized Evaluation of the Vascular Effects of Norvasc Trial (PREVENT) documented an amlodipine-associated reduction in its composite end point, but this primarily reflected fewer procedures and recurrent anginal episodes (Table 4).100
Multiple trials have compared the efficacy of CCBs with BBs in symptom relief, but these also have had short follow-up and small numbers of cardiac events. The only 2 trials to have evaluated the long-term effects of CCBs vs BBs in stable angina, the Angina Prognosis Study in Stockholm (APSIS)101 and the Total Ischaemic Burden European Trial (TIBET),102 failed to show differences in major cardiovascular end points (Table 4). Because of their size, however, they cannot exclude large differences (47% in APSIS) in cardiac risk. The 2 studies account for most (100 of 116) of the cardiac events documented in a recently published meta-analysis.103 It is therefore not surprising that this overview did not detect differences in myocardial infarction and cardiac death (Table 4). Interestingly, it did find that discontinuation because of adverse events was lower for BBs. Regardless, because the stable angina data still allow moderate differences in risk, and in view of unproved differential symptom relief, β-blockade should be first-line (and long-acting calcium antagonism, second-line) therapy for this condition.
In patients with left ventricular dysfunction, CCBs' reflex sympathetic activation and negative inotropic effects have from early on raised concerns about their safety.104 Yet, while an exploratory analysis of CCB therapy in the Studies of Left Ventricular Dysfunction (SOLVD) showed an increase in myocardial infarction (Table 5),105 3 placebo-controlled trials of long-acting DHPs in patients receiving standard congestive heart failure therapy failed to demonstrate worse overall outcomes (Table 5).106- 108 Nonetheless, amlodipine-associated event reductions in patients with nonischemic cardiomyopathy, highlighted by post hoc analyses in the first Prospective Randomized Amlodipine Survival Evaluation (PRAISE),106 were not reproduced prospectively by the second Prospective Randomized Amlodipine Survival Evaluation (PRAISE-2).108 The PRAISE-2 results underscore the pitfalls of exploratory analyses, confirming long-acting DHPs' largely neutral effects among patients with treated heart failure, irrespective of cause.
For the non-DHPs, negative inotropic effects have been cited behind the post hoc differences in outcomes observed for subsets with impaired left ventricular function in myocardial infarction trials.73,77,110 Among patients receiving ACEI-based heart failure therapy, the nonrandomized trandolapril-verapamil study84 (Table 3) and the Diltiazem in Dilated Cardiomyopathy (DiDi) Trial109 provide some evidence of benefit. In the latter, randomization to diltiazem conferred significant improvements in left ventricular function and exercise capacity, although broad confidence bounds preclude conclusions on mortality (Table 5).109 In the aggregate, however, the lack of prospective or adequately powered data on CHD events for the non-DHPs, together with reports of increased heart failure,77,110 leaves their safety in left ventricular dysfunction very much in question.
Early in the controversy's development, studies suggesting an increased risk of cancer15- 17 were also published (Table 6). The biological plausibility of the association was provided by reports that CCBs could interfere with calcium-mediated127- 129 Pahor and associates' findings,15,16 however, need to be viewed in the context of important limitations. In their analyses of the EPESE cohort, patients administered CCBs appeared to be sicker than their counterparts. Although the authors conducted adjustments for factors such as smoking, alcohol intake, and body mass index, these probably do not account for the full scope of clinical differences underlying differential cancer risk. Also, proved carcinogens generally exert selective tissue effects and are associated with latent periods beyond 4 years, the follow-up period in question.35 The absence of a latency period, together with the observation that the increased risk was not confined to a particular type of cancer, supports the notion that the elevated risk may have been the effect of residual confounding. Moreover, because ill patients are more prone to undergo workups that would uncover malignant neoplasms, the existence of potential detection bias in the analyses cannot be dismissed.
Despite 2 additional observational studies showing increased risks of colon17 and breast115 cancer, the preponderance of the ensuing clinical evidence has failed to substantiate an elevation in cancer risk, overall or site specific (Table 6).111- 114,116- 126 While a nested case-control study documented a nonsignificant increase in cancer risk among CCB vs BB users, it found no relationship between cancer and increasing treatment duration.111 Three-year follow-up of nearly 18 000 patients prescribed CCBs found no difference in observed (412) vs predicted (414) cancers based on national incidence rates.112 Retrospective analysis of a registry of hypertensive patients did not show a difference in incidence among CCB users, nonusers, and controls from 2 other populations.116 A case-control study that included 9513 cases and 6492 controls did not identify higher overall or site-specific risk for CCBs, with the exception of renal cell carcinoma, which was also found for BBs and ACEIs.117 Similarly, long-term follow-up in a large cohort study118 and in 2 randomized trials120,121 has failed to demonstrate an increased cancer incidence.
The observed association with renal cell carcinoma117 (Table 6) could reflect confounding by indication. This neoplasm has been found to be associated with hypertension, as well as with several other antihypertensive agents,117,130 suggesting that hypertension (and not the drugs used to treat it) may be the responsible factor. Alternatively, the association may represent reversal of cause and effect. Furthermore, a comprehensive review of the cellular and animal data regarding CCB-specific effects on apoptosis showed highly variable results, with the observation that effects in either direction required doses in the suprapharmacologic range.131 Thus, both biologically and epidemiologically, the evidence of a link to cancer is so far unpersuasive; studies with prospective collection of cancer incidence are needed before the issue can be settled conclusively.
The possibility of a CCB-related hemorrhagic risk gained notice when a randomized trial in valvular surgery found an excess of major bleeding in the nimodipine arm compared with placebo (Table 7).18,19 Earlier, a post hoc analysis of the Thrombolysis in Myocardial Infarction (TIMI)–II trial had shown more intracerebral hemorrhage with CCBs,132 and postmarketing surveillance had found greater hemoglobin decreases for nifedipine relative to lisinopril.146 Concerns were heightened when Pahor and coworkers' EPESE analysis showed a greater risk of gastrointestinal hemorrhage for CCBs compared with BBs.20 The adverse effect was postulated to result from CCB-mediated inhibition of platelet aggregation and of protective vasoconstriction in response to hemorrhage.20,147 Since, additional reports have appeared demonstrating higher rates of in-hospital141 and post–hip surgery136 hemoglobin loss for patients taking CCBs. Moreover, in line with Pahor and associates' results, 2 subsequent observational studies have documented increased CCB risks of gastrointestinal hemorrhage.139,142
Nonetheless, much of the data on CCBs and bleeding is at odds with these findings. For gastrointestinal hemorrhage, 4 additional observational studies134,137,138,140 have failed to reproduce the Pahor et al results (Table 7). They also point up the shortcomings of the positive reports. β-Blockers, the reference category in both Pahor and coworkers' and Kaplan and associates'142 studies, have been suggested to protect against gastrointestinal bleeding,137 which could account for the higher CCB risk ratios observed therein. Misclassification of drug exposure, reflected in the Pahor et al study's implausibly low bleeding risk associated with nonsteroidal anti-inflammatory drugs,134,138,140 makes residual confounding by these agents in the sicker population prescribed CCBs difficult to discount. Similarly, in the Rodriguez et al study,139 failure of bleeding risk to decrease with time since discontinuation of CCBs may reflect uncontrolled confounding or bias. Further, it is significant that the Pahor et al study was prone to substantial misclassification of hospitalizations as resulting from gastrointestinal hemorrhage.138 This would have differentially affected the CCB-treated patients (higher comorbidities), artificially inflating the risk ratio for this outcome.140
Additional studies have not supported the CCB-bleeding association. In the case of perioperative bleeding in patients undergoing cardiac surgery, 2 studies133,135 failed to confirm the observations of the nimodipine trial for patients receiving various CCBs, one in a cohort of 5157 patients.135 Among patients with ischemic stroke or subarachnoid hemorrhage, another population at high risk of bleeding, randomized trials143 and meta-analyses144,145 have demonstrated CCB-related improvements in neurologic outcomes and no excess of bleeding or rebleeding. In coronary disease, a review of several major randomized trials (14 000 patient-years of follow-up)35 did not document any CCB-associated increase in bleeding, although data on this end point may not have been recorded systematically. In sum, the data bearing on a possible link between CCBs and bleeding are contradictory, although much of them speak against a materially increased risk. Definitive proof, however, can only come with prospective, systematic collection of data on this outcome in large-scale trials.56
Two years after Hallas' prescription sequence symmetry analysis148 linked use of CCBs and ACEIs—but not BBs—to prescription of antidepressant medications, Lindberg and coworkers' Swedish cohort and cross-sectional ecologic studies21 found an association between CCBs and suicide not observed for other cardiovascular medications. Their work has significant limitations, however, including a small number of outcomes and limited adjustment for potential confounders. The latter is particularly important because BBs may be prescribed less frequently to patients with depression.
Subsequently, Bergman and colleagues (as reported by Kizer and Kimmel24) examined national Swedish pharmacy and forensic toxicology data to calculate medication-related rates of suicide in their country. They found figures of 0.01, 0.18, and 0.11 suicides per 1000 person-years for verapamil, diltiazem, and propranolol hydrochloride, respectively, which differ markedly from that reported by Lindberg et al (1.1 suicides per 1000 person-years) and are lower than the general suicide rate in Sweden (0.2/1000 person-years). Dunn and coworkers' recent cohort study149 using prescription-event monitoring similarly could not document an increased risk of depression for diltiazem or nicardipine relative to ACEIs. Thus, the proposed association between CCBs and suicide remains unconfirmed.
The foregoing is not a formal meta-analysis, nor did we undertake to grade systematically the studies in question. Instead, the article represents our views after comprehensive review of the published literature. We attempted to present an unbiased summary of the available data based on study methodology. Nevertheless, the incompleteness of the data precludes definitive conclusions, and the controversial nature of the subject makes disagreement inevitable. Ultimate resolution of differences in opinion will have to await emergence of more robust evidence.
The characteristically limited data furnished by premarketing studies of CCBs, coupled with the absence of large-scale postmarketing trials, set the stage for alarm when observational studies reported a variety of serious hazards for what had become the most widely prescribed cardiovascular drugs worldwide. Nevertheless, some of the reported deleterious associations for this class of agents have not been borne out. Firmer evidence from large-scale trials has since appeared, principally for the less hemodynamically disruptive long-acting preparations. Together with improved knowledge of the clinical applications of other classes of antihypertensive–anti-ischemic agents, this allows for more informed therapy selection tailored to specific conditions.
In patients with uncomplicated hypertension at modest risk of CHD, long-acting DHPs and probably diltiazem seem safe and effective and remain reasonable alternatives after attempts with the first-line TDs and BBs (JNC-VI). Recent data, while buttressing safety contentions for CCBs in diabetic subgroups, support the superiority of ACEIs and compel extension of this therapy to patients with known cardiovascular disease. In accordance with national guidelines,58,94,95 BBs should continue to be first-line therapy for both acute coronary syndromes and stable angina, with ACEIs the mainstay of congestive heart failure management.
Many unknowns remain, such as the role of non-DHPs in patients with CHD intolerant of BBs or the relative merits (especially for cause-specific outcomes) of the various agents for hypertensive patients at modest CHD risk. In the latter, more data documenting coronary event reduction exist for TD-BBs than for long-acting DHPs, diltiazem, or ACEIs. In both cases, however, data from ongoing randomized trials and future meta-analyses should help to refine our therapeutic choices.56,150 Regardless, clinicians will have to continue to combine a critical assessment of the literature with sound clinical judgment to guide their decisions for the optimal welfare of these patients.
Accepted for publication November 7, 2000.
We thank Brian Strom, MD, MPH, for his helpful comments during the preparation of the manuscript.
Corresponding author: Stephen E. Kimmel, MD, MSc, University of Pennsylvania School of Medicine, 717 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104-6021 (e-mail: firstname.lastname@example.org).