Calcium Channel Blockers and Mortality in Elderly Patients with Myocardial Infarction

ORIGINAL INVESTIGATION Calcium Channel Blockers and Mortality in Elderly Patients With Myocardial Infarction

James G. Jollis, MD; Ross J. Simpson, Jr, MD, PhD; Mridul K. Chowdhury, PhD; Wayne E. Cascio, MD; John R. Crouse III, MD; Mark W. Massing, MD, MPH; Sidney C. Smith, Jr, MD

Background: Although calcium channel blockers are Results: Calcium channel blockers were widely pre- a useful therapy in relieving angina, lowering blood pres- scribed at hospital discharge to elderly patients with sure, and slowing conduction of atrial fibrillation, grow- myocardial infarction between 1994 and 1995 ing evidence has cast doubt on their safety in patients with (n = 51 921), the most commonly prescribed being dil- coronary disease. tiazem (n = 21 175), nifedipine (n = 12 670), amlodipine (n = 11 683), and verapamil (n = 3639). After Objective: To examine the association between cal- adjusting for illness severity and concomitant medica- cium channel blocker therapy at hospital discharge and tion use, patients who were prescribed calcium chan- mortality in a population-based sample of elderly pa- nel blockers at hospital discharge did not have tients hospitalized with acute myocardial infarction. increased risk for 30-day or 1-year mortality, with the exception of the few (n = 116) treated with bepridil. Design: Retrospective cohort study using data from medi- Bepridil differs from other calcium channel blockers cal charts and administrative files. because of its tendency to prolong repolarization, and its association with proarrhythmic effects in elderly Setting: All acute care hospitals in 46 states. patients.

Patients: All Medicare patients with a principal diag- Conclusion: We did not identify a mortality risk in a nosis of acute myocardial infarction consecutively dis- large consecutive sample of elderly patients with myo- charged from the hospital alive during 8-month periods cardial infarction, which supports the need for addi- between 1994 and 1995 (N = 141 041). tional prospective trials examining calcium channel blocker therapy for ischemic heart disease. Main Outcome Measure: Mortality at 30 days and 1 year. Arch Intern Med. 1999;159:2341-2348

ALCIUM CHANNEL block- nel blocker therapy, and further studies in- ers represent a poten- volving large numbers of patients are tially useful therapy for needed to examine the safety of their use. elderly patients with coro- Given current apprehension among clini- nary artery disease, given cians regarding calcium channel blocker theirC ability to relieve angina, lower blood therapy, such studies are unlikely to be From the Duke Clinical pressure, and slow conduction of atrial fi- performed, particularly involving formu- Research Institute, Duke brillation. However, growing evidence has lations for which patents have expired. University, Durham, NC cast doubt on their safety. In post hoc Decisions regarding the use of calcium (Dr Jollis); Medical Review of analyses, the Multicenter Diltiazem Postin- channel blockers in the elderly are even North Carolina Inc, Cary farction Trial (MDPIT)1 found higher rates more difficult, given the relatively few (Drs Simpson and Chowdhury); of nonfatal reinfarction or cardiac death elderly patients enrolled in previous tri- Division of Cardiology associated with diltiazem treatment in pa- als. In 1995, calcium channel blockers— (Drs Simpson, Cascio, and tients with myocardial infarction (MI) and including nifedipine, diltiazem, and ver- Smith) and Department of pulmonary congestion or reduced ejec- apamil—were widely used among Epidemiology (Dr Massing), tion fraction. More recently, a meta- Medicare patients. We examined the re- University of North Carolina, 2 Chapel Hill; and the analysis of nifedipine therapy after acute lationship between calcium channel Department of Medicine, Wake MI identified higher mortality associated blocker use after acute MI and survival for Forest University School of with nifedipine therapy on a dose-re- 141 041 elderly patients, as part of the Medicine, Winston-Salem, NC sponse basis. These findings have led to Health Care Financing Administration’s (Dr Crouse). much uncertainty regarding calcium chan- Cooperative Cardiovascular Project (CCP).

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2341

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Table 1. Thirty-Day Mortality Logistic Regression Model*

PATIENTS AND METHODS Variable Odds Variable Estimate ␹2 P Ͼ Ratio PATIENT POPULATION Intercept −3.57 727.79 .001 The CCP abstracted hospital charts of Medicare patients Age from 65 y 0.04 51.39 .001 1.04 Age from 65 y squared −0.0003 1.87 .17 1.00 withaprincipaldiagnosisofacuteMI(InternationalClas- Female −0.15 29.48 .001 0.86 sification of Diseases codes 410.x0 and 410.x1) consecu- Black −0.12 4.29 .04 0.89 tively discharged from the hospital during 8-month pe- Anterior infarction 0.18 38.56 .001 1.20 3-5 riods between 1994 and 1995, in 46 states. Informa- Inferior infarction −0.01 0.25 .62 0.99 tion collected for each Medicare patient included patient Non–Q wave infarction −0.19 45.80 .001 0.82 identifiers, hospitalization dates, demographics, chest Systolic blood pressure −0.007 263.17 .001 0.99 pain history, physical examination findings, medica- Pulse 0.003 31.62 .001 1.00 tions used, presence or absence of contraindications to Respiratory rate 0.01 23.06 .001 1.01 therapy,electrocardiograms,cardiacenzymelevels,treat- Rales or pulmonary congestion 0.15 26.23 .001 1.16 ment, complications, and survival status. Accuracy of Ejection fraction 40%-59% −0.53 206.58 .001 0.59 hospital chart abstraction was evaluated on a monthly Ejection fraction Ͼ60% −0.97 177.00 .001 0.38 basis by masked reabstraction, with agreement rates by Ejection fraction missing −0.14 16.32 .001 0.87 data element in the 85% to 95% range. Calcium chan- Previous myocardial infarction 0.03 1.30 .25 1.03 nel blockers were identified at 2 points in the CCP chart Previous congestive heart failure 0.08 6.82 .009 1.08 abstraction, according to discharge medications entered Previous bypass surgery 0.06 2.19 .14 1.06 by free text and by a specific variable for “calcium chan- Previous angioplasty −0.37 33.45 .001 0.69 Electrocardiogram—atrial 0.15 25.50 .001 1.16 nel blocker at discharge.” These 2 variables agreed for fibrillation 51 843 of 51 921 patients we identified as being treated Electrocardiogram—heart block 0.10 2.59 .11 1.11 with calcium channel blockers, and for 89 099 of 89 120 Electrocardiogram—myocardial 0.11 10.25 .001 1.12 patientsnotbeingtreated.Allpatientsolderthan65years infarction who were discharged from the hospital alive were in- Diabetes 0.13 20.32 .001 1.14 cludedinthepresentstudy.Datesofdeathwereobtained Hypertension −0.16 29.31 .001 0.86 from the Medicare Enrollment Database and the Social Previous stroke 0.06 2.96 .08 1.06 SecurityAdministration’sMasterBeneficiaryRecordFile. Peripheral vascular disease 0.10 7.62 .006 1.11 Obstructive pulmonary disease 0.08 5.57 .018 1.08 DATA ANALYSIS Current cigarette smoker 0.18 18.58 .001 1.20 Dementia 0.39 98.80 .001 1.48 The primary analysis of the present study compared Serum urea nitrogen level, Ͻ5 −0.13 7.25 .007 0.88 the mortality of patients who were prescribed cal- mmol/L Serum urea nitrogen level, 7-8 0.19 19.05 .001 1.21 cium channel blockers at hospital discharge after acute mmol/L MI with that of patients discharged but not taking Serum urea nitrogen level, 9-10 0.24 24.63 .001 1.27 these drugs. Calcium channel blockers were classi- mmol/L fied as nifedipine, amlodipine, other dihydropteri- Serum urea nitrogen level, Ͻ10 0.60 234.58 .001 1.82 dines, diltiazem, verapamil, and bepridil hydrochlo- mmol/L ride. Baseline characteristics and outcomes were Serum urea nitrogen level 0.17 5.15 .023 1.19 compared between treatment categories using ␹2 tests missing for categorical variables and analysis of variance for Walks with assistance 0.30 90.61 .001 1.35 continuous variables. Unable to walk 1.22 715.83 .001 3.39 Thirty-day and 1-year mortality after hospital dis- Unable to determine mobility 0.88 92.84 .001 2.41 charge were examined in logistic regression models Congestive heart failure 0.56 304.32 .001 1.74 after adjusting for illness severity, medications taken, Recurrent angina 0.23 64.10 .001 1.26 and propensity for treatment with calcium channel Shock 0.17 6.65 .01 1.19 blockers (propensity score). Logistic model vari- Reinfarction 0.24 11.49 .001 1.28 Stroke 0.55 93.28 .001 1.73 ables were selected on the basis of their association Hemorrhage −0.008 0.05 .82 0.99 with mortality according to previous work, strength Cardiac arrest 0.13 4.66 .03 1.14 of association, and clinical intuition (Table 1 and Discharged to nonacute hospital 0.74 448.40 .001 2.10 Table 2). Specific model components included age, Discharged to home health care 0.27 50.16 .001 1.30 sex, race, descriptors of MI and coronary disease se- Propensity decile 1 −0.73 68.36 .001 0.48 verity, comorbid illnesses, mobility at discharge, dis- Propensity decile 2 −0.48 35.28 .001 0.62 charge destination, and propensity score. Propen- Propensity decile 3 −0.27 12.42 .001 0.76 sity score was derived according to the methods of Propensity decile 4 −0.23 9.44 .002 0.79 Rubin6 using an additional logistic model to exam- Propensity decile 5 −0.19 6.87 .009 0.82 ine characteristics associated with being discharged Propensity decile 6 −0.22 9.29 .002 0.80 while taking a calcium channel blocker.7 Compo- Propensity decile 7 −0.14 3.97 .05 0.87 nent variables for the propensity model were se- Propensity decile 8 −0.10 1.89 .17 0.91 lected in a similar manner as those for the mortality Propensity decile 9 −0.05 0.56 .46 0.95 models, using association with calcium channel Discharged while taking calcium −0.04 2.20 .14 0.96 blocker therapy as the gauge of importance (Table 3). channel blockers

*C-index = 0.80. To convert serum urea nitrogen from millimoles per liter to milligrams per deciliter, divide by 0.357.

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2342

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Table 2. One-Year Mortality Logistic Regression Model* Table 3. Propensity Score Logistic Regression Model*

Variable Odds Variable Odds Variable Estimate ␹2 P Ͼ Ratio Variable Estimate ␹2 P Ͼ Ratio Intercept −2.72 1377.64 .001 Intercept −1.94 1574.41 .001 Age from 65 y 0.04 124.38 .001 1.04 Age from 65 y 0.01 25.68 .001 1.01 Age from 65 y squared Ͻ0.01 1.69 .19 1.00 Age from 65 y squared −0.001 34.06 .001 0.10 Female −0.17 111.94 .001 0.84 Female 0.13 106.19 .001 1.14 Black 0.07 5.65 .02 1.08 Black 0.03 1.04 .31 1.03 Anterior infarction 0.12 50.69 .001 1.13 Anterior infarction −0.09 46.00 .001 0.91 Inferior infarction −0.06 13.76 .0002 0.94 Inferior infarction −0.08 30.38 .001 0.92 Non–Q wave infarction −0.05 9.44 .002 0.95 Non–Q wave infarction 0.39 940.14 .001 1.48 Systolic blood pressure −0.005 382.19 .001 0.10 Systolic blood pressure 0.007 1187.73 .001 1.01 Pulse 0.006 341.88 .001 1.01 Pulse 0.0002 0.73 .39 1.00 Respiratory rate 0.02 167.72 .001 1.02 Respiratory rate −0.009 66.18 .001 0.99 Rales or pulmonary congestion 0.21 158.32 .001 1.24 Rales or pulmonary congestion −0.05 11.05 .0009 0.95 Ejection fraction 40%-59% −0.55 686.48 .001 0.58 Ejection fraction 40%-59% 0.39 477.79 .001 1.48 Ejection fraction Ͼ60% −0.85 545.33 .001 0.43 Ejection fraction Ͼ60% 0.58 554.55 .001 1.78 Ejection fraction missing −0.14 46.64 .001 0.87 Ejection fraction missing 0.30 270.41 .001 1.35 Previous myocardial infarction 0.15 79.10 .001 1.16 Previous myocardial infarction 0.28 412.31 .001 1.32 Previous congestive heart failure 0.35 379.72 .001 1.41 Previous congestive heart failure 0.02 1.90 .17 1.02 Previous bypass surgery 0.14 39.00 .001 1.16 Previous bypass surgery 0.29 253.27 .001 1.34 Previous angioplasty −0.28 74.43 .001 0.76 Previous angioplasty 0.38 275.42 .001 1.47 Electrocardiogram—atrial 0.16 75.84 .001 1.17 Electrocardiogram—atrial 0.09 29.44 .001 1.09 fibrillation fibrillation Electrocardiogram—heart block 0.05 1.62 .20 1.05 Electrocardiogram—heart block −0.28 62.60 .001 0.76 Electrocardiogram—myocardial −0.001 0.00 .96 0.10 Electrocardiogram—myocardial −0.15 97.57 .001 0.86 infarction infarction Diabetes 0.21 157.71 .001 1.23 Diabetes 0.12 77.68 .001 1.13 Hypertension −0.12 51.30 .001 0.89 Hypertension 0.49 1337.30 .001 1.62 Previous stroke 0.21 108.67 .001 1.24 Previous stroke 0.09 27.14 .001 1.10 Peripheral vascular disease 0.20 77.45 .001 1.22 Peripheral vascular disease 0.27 191.52 .001 1.31 Obstructive pulmonary disease 0.22 134.27 .001 1.24 Current cigarette smoker 0.17 56.83 .001 1.19 Obstructive pulmonary disease 0.19 158.49 .001 1.22 Dementia 0.44 253.20 .001 1.56 Current cigarette smoker −0.09 24.11 .001 0.92 Serum urea nitrogen level, Ͻ5 −0.12 23.81 .001 0.88 Dementia −0.23 67.60 .001 0.80 mmol/L Serum urea nitrogen level, Ͻ5 0.01 0.68 .41 1.02 Serum urea nitrogen level, 7-8 0.15 42.73 .001 1.17 mmol/L mmol/L Serum urea nitrogen level, 7-8 0.03 3.62 .06 1.04 Serum urea nitrogen level, 9-10 0.29 119.92 .001 1.34 mmol/L mmol/L Serum urea nitrogen level, 9-10 0.05 5.97 .01 1.06 Serum urea nitrogen level, Ͻ10 0.75 1137.40 .001 2.12 mmol/L mmol/L Serum urea nitrogen level, Ͻ10 0.13 44.28 .001 1.14 Serum urea nitrogen level 0.10 5.42 .02 1.10 mmol/L missing Serum urea nitrogen level 0.11 12.71 .0004 1.11 Walks with assistance 0.34 366.70 .001 1.41 missing Unable to walk 1.06 932.74 .001 2.89 Walks with assistance −0.04 5.01 .02 0.96 Unable to determine mobility 0.58 74.17 .001 1.78 Unable to walk −0.28 62.05 .001 0.75 Congestive heart failure 0.49 740.56 .001 1.62 Unable to determine mobility −0.26 15.78 .001 0.77 Recurrent angina 0.11 43.40 .001 1.12 Bypass surgery during −1.18 2540.73 .001 0.31 Shock 0.06 1.86 .17 1.06 hospitalization Reinfarction 0.14 9.11 .002 1.15 Angioplasty during 0.43 643.61 .001 1.53 Stroke 0.30 50.74 .001 1.35 hospitalization Hemorrhage −0.04 4.55 .03 0.96 Discharged while taking −0.87 4107.61 .001 0.42 Cardiac arrest 0.14 13.93 .0002 1.15 ␤-adrenergic blocking agents Discharged to nonacute hospital 0.56 609.38 .001 1.75 Discharged while taking −0.66 2176.83 .001 0.52 Discharged to home health care 0.22 112.97 .001 1.25 angiotensin-converting Propensity decile 1 −0.62 169.43 .001 0.54 enzyme inhibitors Propensity decile 2 −0.38 74.40 .001 0.69 Discharged while taking aspirin 0.31 503.25 .001 1.36 Propensity decile 3 −0.26 38.28 .001 0.77 Congestive heart failure −0.26 299.54 .001 0.77 Propensity decile 4 −0.13 10.35 .001 0.88 Recurrent angina 0.29 463.93 .001 1.34 Propensity decile 5 −0.15 14.23 .0002 0.86 Shock −0.45 91.25 .001 0.64 Propensity decile 6 −0.14 12.83 .0003 0.87 Reinfarction −0.07 3.43 .06 0.93 Propensity decile 7 −0.07 3.98 .05 0.93 Stroke −0.14 11.70 .0006 0.87 Propensity decile 8 −0.04 1.07 .30 0.96 Hemorrhage −0.08 23.49 .001 0.92 Propensity decile 9 −0.05 2.08 .15 0.95 Cardiac arrest −0.20 32.99 .001 0.82 Discharged while taking calcium Ͻ0.01 Ͻ0.01 .99 1.00 Discharged to nonacute hospital −0.29 173.40 .001 0.75 channel blockers Discharged to home health care −0.02 1.83 .180.98

*C-index = 0.80. To convert serum urea nitrogen from millimoles per liter to *C-index = 0.73. To convert serum urea nitrogen from millimoles per liter to milligrams per deciliter, divide by 0.357. milligrams per deciliter, divide by 0.357.

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2343

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Table 4. Demographic Characteristics and Illness Severity by Calcium Channel Blocker Therapy (N = 141 041)*

Any Calcium No Calcium Other Channel Blocker Channel Blocker Nifedipine Amlodipine Dihydropteridines (n = 51 921) (n = 89 120) (n = 12 670) (n = 11 683) (n = 1705) Age, mean (percentiles), y 76.1 (70, 81) 76.3 (70, 82) 76.1 (70, 81) 76.4 (71, 82) 77.0 (71, 82) Female, % 50.5 47.3 50.1 50.7 53.7 Race, % White 88.2 88.3 86.1 88.6 88.9 Black 6.6 5.8 8.4 6.6 5.7 Other 5.2 5.9 5.5 4.8 5.4 Coronary disease Infarct location, % Anterior 38.0 46.3 37.0 40.5 36.4 Inferior 41.4 46.4 42.8 41.9 40.8 Non–Q wave 51.1 38.0 49.6 50.5 49.3 Systolic blood pressure, mean (percentiles), mm Hg 149 (128, 169) 141 (122, 160) 153 (130, 172) 148 (126, 168) 150 (130, 170) Pulse, mean (percentiles), beats/min 86 (70, 100) 87 (70, 100) 84 (68, 96) 86 (70, 100) 87 (71, 100) Respiratory rate, mean (percentiles), /min 22 (18, 24) 22 (18, 24) 21 (18, 24) 22 (18, 24) 22 (18, 24) Rales, % 32.1 35.1 30.6 37.6 34.2 Left ventricular ejection fraction, % Ͼ60 12.3 8.6 12.6 9.9 11.0 40-59 40.1 37.1 40.2 39.0 37.4 Ͻ39 15.6 24.5 13.4 22.6 16.6 Missing 32.1 29.9 33.8 28.5 35.0 Chest pain within 2 d, % 79.3 74.1 79.1 79.4 81.5 Infarction confirmed clinically, % 84.6 86.0 83.7 85.4 87.6 Peak creatine kinase level, mean (percentiles), 662 (204, 759) 988 (257, 1206) 638 (204, 750) 698 (204, 773) 671 (213 744) U/L Previous cardiac history Myocardial infarction 35.1 29.1 35.0 38.7 39.7 Congestive heart failure 22.2 21.6 21.0 28.0 27.9 Bypass surgery 16.2 10.9 17.1 18.7 20.4 Angioplasty 10.2 5.9 10.7 10.5 12.8 Electrocardiogram, % Atrial fibrillation 9.5 9.2 7.9 7.8 10.2 Heart block 2.8 3.8 3.0 3.5 3.6 Comorbid illness Diabetes, % 32.9 29.3 34.4 35.6 34.7 Hypertension, % 69.2 58.7 76.5 71.1 73.0 Previous stroke, % 14.4 13.0 15.6 15.0 15.3 Peripheral vascular disease, % 13.1 9.7 14.0 15.1 13.4 Obstructive pulmonary disease, % 23.5 19.6 18.4 21.8 19.5 Current cigarette smoker, % 14.8 15.4 14.3 13.4 12.9 Dementia, % 4.8 6.6 4.8 4.4 4.8 Serum urea nitrogen level, mean (percentiles), mmol/L 8.2 (5, 10) 8.2 (5, 9) 8.6 (5, 10) 8.9 (6, 10) 8.6 (6, 10) Mobility, % Independent 73.0 69.5 72.9 72.5 71.1 Walks with assistance 23.3 24.9 23.1 24.0 26.0 Unable to walk 3.0 4.7 3.3 2.8 1.8 Unable to determine 0.8 1.0 0.7 0.8 1.2

*PϽ.001 for comparison between specific calcium channel blocker categories (nifedipine, amlodipine, other dihydropteridines, diltiazem, verapamil, and bepridil) using the likelihood ratio ␹2 test for categorical variables, with data given as percentage of patients, or using analysis of variance for continuous variables, with data given as means (25th, 75th percentiles).

The propensity model included all variables in the mor- blood pressure (Ͻ80, 80-100, and Ͼ100 mm Hg), and tality model, plus revascularization procedures after ad- pulse (Ͻ90, 90-110, and Ͼ110 beats/min). mission (bypass surgery and angioplasty) and hospital discharge medications (aspirin, ␤-adrenergic blocking RESULTS agents, and angiotensin-converting enzyme inhibitors). Mortality for patients who were prescribed bepri- PATIENT CHARACTERISTICS dil at hospital discharge was also compared with that for control patients matched for age (65-69, 70-74, 75-79, A total of 141 041 patients with acute MI who survived and Ն80 years), race (black), serum urea nitrogen level to hospital discharge were included in this study (Table 4 (Ͼ10.7 mmol/L [Ͼ30 mg/dL]), mobility (unable to walk), and Table 5). Of these patients, 51 921 were pre- congestive heart failure or pulmonary edema, systolic scribed calcium channel blockers at hospital discharge,

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2344

Downloaded From: https://jamanetwork.com/ on 09/28/2021 ergic blocking agents or angiotensin-converting enzyme inhibitors at hospital discharge, or to have undergone bypass surgery. Bepridil Among patients who were prescribed calcium Diltiazem Verapamil Hydrochloride (n = 21 175) (n = 3639) (n = 116) channel blockers at hospital discharge, those treated with amlodipine were most likely to have rales, con- 75.9 (70, 81) 76.0 (70, 81) 78.6 (73, 84) gestive heart failure, or a low ejection fraction, whereas 49.9 52.8 64.7 those treated with diltiazem were least likely to have 89.0 89.4 93.1 these conditions. Patients prescribed nifedipine at hos- 5.6 6.2 3.5 pital discharge were most likely to have hypertension 5.4 4.4 3.4 and to be prescribed ␤-adrenergic blocking agents. According to the propensity model, blood pressure, left ventricular ejection fraction greater than 40%, and 37.2 37.7 47.4 40.9 39.0 39.7 non–Q wave MI had the strongest associations with 51.9 52.4 56.9 calcium channel blocker treatment at hospital dis- 147 (127, 166) 148 (127, 168) 144 (123, 166) charge (Table 3).

87 (70, 100) 93 (72, 107) 88 (78, 98) MORTALITY 22 (18, 24) 22 (18, 24) 23 (20, 26) 29.4 32.1 51.7 Crude 30-day and 1-year mortality rates were lower for 13.4 13.4 4.3 patients who were prescribed calcium channel blockers 41.1 40.0 28.5 at hospital discharge compared with those discharged but 13.2 13.0 24.1 not taking such drugs (Table 5). The lowest unadjusted 32.3 33.6 43.1 30-day and 1-year mortality rates were seen in nifedi- 80.0 73.8 92.2 pine- and diltiazem-treated patients. By 1 year, 61 (52.6%) 84.7 82.4 94.0 of 116 patients discharged while taking bepridil had died 667 (205, 769) 623 (202, 738) 549 (217, 706) compared with 29 361 of 141 041 (a 20.8% mortality rate) 33.5 27.6 63.8 for the entire cohort. 19.1 21.4 50.9 After adjusting for illness severity, medications used, 14.5 11.1 31.9 and treatment propensity, the likelihood of death at 30 days 9.7 7.8 12.9 or 1 year for patients who were prescribed calcium chan- 9.7 18.9 9.6 nel blockers at hospital discharge was similar to that for 2.2 3.2 2.6 those not treated (Figure 1 and Figure 2). Stratifying the adjusted analyses according to the most commonly pre- 30.5 28.4 46.6 scribed calcium channel blockers (diltiazem, nifedipine, 62.5 72.1 72.4 amlodipine, and verapamil), all 95% confidence intervals 13.0 15.0 10.3 overlapped 1, consistent with no statistically significant 11.7 10.8 19.0 26.2 32.2 19.0 differences in mortality. However, trends for mortality 16.0 15.4 7.8 for amlodipine-treated patients were somewhat higher at 4.7 5.7 6.9 30 days and, for verapamil-treated patients, were some- 7.9 (5, 9) 7.9 (5, 9) 10 (7, 11) what lower at 1 year. Bepridil therapy represented an exception to the 74.4 70.0 66.4 lack of association between calcium channel blocker 22.1 25.6 31.0 therapy and mortality. Mortality for 116 patients who 2.8 3.2 2.6 0.8 1.2 0 were prescribed bepridil at hospital discharge was sub- stantially higher than for 116 control patients matched for age and illness severity (30-day mortality, 13.8% vs 4.3%; PϽ.01, and 1-year mortality, 52.6% vs 27.6%; PϽ.001).

COMMENT the most commonly diltiazem (21 175 patients), nifedipine (12 670 patients), amlodipine (11 683 patients), Calcium channel blockers were widely prescribed at and verapamil (3639 patients). Compared with patients hospital discharge to elderly patients with MI between who were not prescribed calcium channel blockers at 1994 and 1995. With the exception of the few patients hospital discharge, these patients were similar in age; treated with bepridil, this treatment was not associated more likely to have had a non–Q wave MI, hyperten- with increased mortality. After adjusting for illness sion, and previous coronary disease; and less likely to severity and concomitant medication use, the one third have a low, left ventricular ejection fraction or to have of patients with MI who were prescribed calcium chan- developed shock or congestive heart failure during hos- nel blockers at hospital discharge had similar 1-year pitalization. Patients treated with calcium channel mortality rates as those discharged but not taking these blockers were also less likely to be prescribed ␤-adren- drugs. This failure to identify a mortality risk in a large,

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2345

Downloaded From: https://jamanetwork.com/ on 09/28/2021 Table 5. Treatment, Complications, and Mortality by Calcium Channel Blocker Therapy*

Any Calcium No Calcium Channel Channel Other Bepridil Blocker Blocker Nifedipine Amlodipine Dihydropteridines Diltiazem Verapamil Hydrochloride Treatment, % Catheterization 44.1 40.5 45.1 45.7 40.6 44.0 40.7 20.7 Bypass surgery 6.0 14.8 6.8 3.2 3.2 5.7 15.2 0.9 Angioplasty 23.0 16.0 24.1 23.9 21.9 23.8 15.5 6.0 Discharge medications ␤-Adrenergic blocking agent 30.7 42.2 43.3 38.7 38.0 20.3 18.7 49.1 Angiotensin-converting enzyme 27.0 38.8 25.6 33.3 28.6 24.1 26.2 37.9 inhibitor Aspirin 70.6 65.6 70.4 72.8 69.3 71.4 61.7 62.1 Complications, % Congestive heart failure 35.6 44.1 33.0 43.1 38.6 31.9 37.4 69.0 Recurrent angina 30.8 25.7 30.1 33.8 33.3 29.5 26.5 59.5 Shock 1.4 3.3 1.2 1.9 1.1 1.3 1.8 3.5 Reinfarction 2.2 2.8 2.1 2.2 2.3 2.3 2.0 4.3 Stroke 2.0 2.8 2.4 1.8 1.8 1.7 3.0 0 Hemorrhage 15.2 18.0 14.4 16.2 14.1 15.1 16.0 11.2 Cardiac arrest 2.9 4.6 2.8 3.1 3.5 2.7 3.2 3.5 Discharge destination, % Home 77.5 71.4 78.3 76.7 76.4 78.7 73.0 61.2 Nonacute hospital 9.7 14.8 9.9 9.2 8.7 9.3 12.2 14.7 Home health care 12.8 13.8 11.8 14.1 15.0 12.0 14.9 24.1 Mortality, % 30 d 4.3 5.7 3.8 5.1 5.1 3.8 4.3 13.8 1 y 19.6 21.5 18.3 22.0 21.9 18.3 19.2 52.6

All Calcium Blockers N = 51 921 All Calcium Blockers N = 51 921

Diltiazem n = 21 175 Diltiazem n = 21 175

Nifedipine n = 12 670 Nifedipine n = 12 670

Amlodipine n = 11 683 Amlodipine n = 11 683

Verapamil n = 3639 Verapamil n = 3639

0.5 0.60.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0.5 0.60.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 Odds Ratios (and 95% Confidence Intervals) Odds Ratios (and 95% Confidence Intervals) Less Likely More Likely Less Likely More Likely

Figure 1. Relative likelihood of 30-day mortality for patients who were Figure 2. Relative likelihood of 1-year mortality for patients who were prescribed calcium channel blockers at hospital discharge vs those not prescribed calcium channel blockers at hospital discharge vs those not prescribed. prescribed.

consecutive sample of elderly patients with MI suggests POTENTIAL EXPLANATIONS FOR FINDINGS that calcium channel blockers—particularly diltiazem, nifedipine, amlodipine, and verapamil—can be used to There are several potential explanations for why we did treat angina, hypertension, and atrial fibrillation in not identify a mortality risk with calcium channel blocker elderly patients with MI without adversely affecting therapy. The first explanation may be that, despite con- their 1-year mortality. The negative findings further cerning findings in some previous studies, calcium chan- support the need for additional prospective trials exam- nel blocker therapy does not increase mortality after MI. ining calcium channel blocker therapy in ischemic Primary analyses of 2 randomized controlled trials, the heart disease. MDPIT1 and the Danish Verapamil Infarction Trial,8 did

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2346

Downloaded From: https://jamanetwork.com/ on 09/28/2021 not identify excess mortality associated with long-term dil- cium channel blockers at hospital discharge, suggesting tiazem or verapamil therapy, respectively. Meta-analyses that the MDPIT affected treatment selection. However, by Held et al9 and Pepine et al10 and a cohort study by Braun there still were many patients with pulmonary conges- et al11 also did not detect a mortality association for cal- tion or congestive heart failure who were prescribed calcium channel blocker therapy. Our findings might differ cium channel blockers at hospital discharge, making treat- from the meta-analysis by Furberg et al2 that found dose- ment selection an unlikely explanation for the lack of related mortality for short-acting nifedipine because of mortality difference. changes in drug formulation. In 1989, longer-acting ni- Our population of elderly patients is a final explana- fedipine formulations became available; several investi- tory factor for our findings. Hypertension and strokes are gators12-15 have since suggested that longer-acting prepa- more common in the elderly.18 In this elderly cohort, cal- rations might have sufficiently different pharmacodynamic cium channel blocker therapy might have played a greater properties so as to result in lower mortality rates. We had role in lowering blood pressure and preventing fatal limited information regarding the specific formulation of strokes compared with their effect in younger patients the nifedipine that our patients took. For 33% of nifedipine- typically included in randomized trials (mean age, 58 treated patients, long-acting formulations were indicated years; fewer than 15% were Ͼ69 years in the MDPIT1). by text such as “Procardia XL” or “nifedipine GITS,” but In this scenario, such a benefit might have offset poten- we cannot determine whether the remaining patients were tial deleterious effects of calcium channel blocker therapy treated with long-acting formulations—for which docu- observed in younger cohorts, resulting in the absence of mentation was not specific—or short-acting prepara- a net mortality risk in the elderly. tions. Based on our data, we can only assert that increased mortality was not seen for patients of whom a BEPRIDIL substantial portion were discharged while taking long- acting nifedipine formulations. Although few patients were treated with bepridil, the ex- Our study design might also have impacted our find- traordinarily high 1-year mortality rate associated with ings. As in all observational cohort studies, it is possible this therapy is concerning. Bepridil, like other calcium that these findings might have been caused by confound- channel blockers, inhibits voltage-dependent L-type cal- ing or insufficient consideration of a factor related to both cium channels, but, unlike other calcium channel block- the selection of calcium channel blockers and mortality. ers, it also has a propensity to inhibit myocardial repo- Our approach was designed to limit the possibility of con- larization and to lengthen the QT interval.19 The drug has founding. The CCP identified an extensive amount of per- also been associated with torsade de pointes, particu- tinent illness severity and medication data, and the simi- larly in patients older than 70 years.20 The increased mor- lar prevalence of these characteristics compared with other tality with bepridil therapy in this elderly cohort is con- MI cohorts suggests that the measures were reliable.16,17 sistent with QT prolongation and proarrhythmia. With the CCP data, we specifically considered factors that Underlying disease severity may also explain the high mor- led to the selection of calcium channel blockers using the tality observed with bepridil therapy. Patients selected propensity model. After this initial adjustment to com- for bepridil therapy were older and more likely to have pare patients with a similar likelihood of treatment, we diabetes, previous MI, and previous congestive heart fail- used a second regression model to further adjust for fac- ure than were all other patients. Although adjustment for tors related to mortality. The ability of our approach to differences in patient characteristics did not account for identify mortality risk factors was demonstrated by mod- the excess mortality, it is possible that additional factors els that indicated higher mortality for all other factors related to the selection of bepridil therapy and mortality known to be associated with greater risk. that were not included in our data led to the worse out- Because we had a 90% chance of identifying a 1% come. Given the small number of patients who were pre- difference in 1-year mortality at the 95% confidence level, scribed bepridil at hospital discharge, it is also possible if adequate data were available to balance comparisons, that the high mortality rate is a spurious finding. With we were unlikely to lack adequate sample size to iden- previous studies identifying an association with torsade tify a clinically significant mortality effect for calcium de pointes, as well as the extremely high mortality of be- channel blockers. The study lacked information about pridil-treated patients in our study compared with compliance during the year, and without such informa- matched controls, the safety of this drug after MI should tion, we cannot be sure that the lack of difference in 1-year be further investigated. mortality was not related to discontinuation of calcium channel blocker therapy in the treated cohort or initia- CONCLUSIONS tion of such therapy in the untreated cohort. Our findings might also have been affected by cli- We examined more than 50 000 elderly patients with MI nicians avoiding calcium channel blocker therapy in pa- who were prescribed calcium channel blockers at hos- tients whom they believed were likely to have adverse pital discharge and found no evidence of increased mor- outcomes with such therapy. Findings of the MDPIT1 re- tality, except for the few patients treated with bepridil, a garding increased mortality with low ejection fraction or drug that is also associated with proarrhythmia. Our pulmonary congestion were published in 1988, 6 years findings suggest that calcium channel blockers— before the start of our study. The propensity models con- particularly diltiazem, nifedipine, amlodipine, and ver- firmed that patients with pulmonary congestion or con- apamil—can be used to treat angina, hypertension, and gestive heart failure were less likely to be prescribed cal- atrial fibrillation among elderly patients with MI with-

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2347

Downloaded From: https://jamanetwork.com/ on 09/28/2021 out adversely affecting mortality. These negative find- 3. Ellerbeck EF, Jencks SF, Radford MJ, et al. Quality of care for Medicare patients ings also support the need for additional prospective tri- with acute myocardial infarction: a four-state pilot study from the Cooperative Cardiovascular Project. JAMA. 1995;273:1509-1514. als examining the efficacy of commonly prescribed 4. Krumholz HM, Radford MJ, Ellerbeck EF, et al. Aspirin in the treatment of acute calcium channel blockers in ischemic heart disease. myocardial infarction in elderly Medicare beneficiaries. Circulation. 1995;92: 2841-2847. 5. International Classification of Diseases, Ninth Revision, Clinical Modification. Wash- Accepted for publication February 19, 1999. ington, DC: US Dept of Health and Human Services, Centers for Disease Control Supported by contract 500-94-0613 from the Health and Prevention, Health Care Financing Administration; 1997. Care Financing Administration, Department of Health and 6. Rubin DB. Estimating causal effects from large data sets using propensity scores. Human Services, Baltimore, Md. Ann Intern Med. 1997;127:757-763. The conclusions and opinions expressed and the meth- 7. Connors AF, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA. 1996;276:889-897. ods used herein are those of the authors. They do not nec- 8. Anonymous. Effect of verapamil on mortality and major events after acute myo- essarily reflect Health Care Financing Administration policy, cardial infarction (the Danish Verapamil Infarction Trial II—DAVIT II). Am J Car- and the authors assume full responsibility for the accuracy diol. 1990;66:779-785. and completeness of the ideas presented. This article is a di- 9. Held PH, Yusuf S, Furberg CD. Calcium channel blockers in acute myocardial infarction and unstable angina: an overview. BMJ. 1989;299:1187-1192. rect result of the Health Care Quality Improvement Pro- 10. Pepine CJ, Faich G, Makuch R. Verapamil use in patients with cardiovascular dis- gram initiated by the Health Care Financing Administra- ease: an overview of randomized trials. Clin Cardiol. 1998;21:633-641. tion, which has encouraged identification of quality 11. Braun S, Boyko V, Behar S, et al. Calcium antagonists and mortality in patients improvement projects derived from analysis of patterns of with coronary artery disease: a cohort study of 11,575 patients. J Am Coll Car- care, and therefore required no special funding on the part diol. 1996;28:7-11. 12. Opie LH, Messerli FH. Nifedipine and mortality: grave defects in the dossier. Cir- of this contractor. We welcome ideas and contributions to culation. 1995;92:1068-1073. the corresponding author concerning experiences in engag- 13. Messerli FH. Nifedipine trials. Circulation. 1997;95:1671-1672. ing with the issues presented. 14. Kloner RA. Nifedipine in ischemic heart disease. Circulation. 1995;92:1074- We thank Tracey A. Simons, MA, for editorial sup- 1078. 15. Yusuf S. Calcium antagonists in coronary artery disease and hypertension: time port. for reevaluation? Circulation. 1995;92:1079-1082. Reprints: James G. Jollis, MD, Duke University Medi- 16. Williams DO, Braunwald E, Knatterud G, et al. One-year results of the Throm- cal Center, DUMC Box 3254, Durham, NC 27710 (e-mail: bolysis in Myocardial Infarction Investigation (TIMI) Phase II Trial. Circulation. [email protected]). 1992;85:533-542. 17. Lee KL, Woodlief LH, Topol EJ, et al. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction: results from an international trial of REFERENCES 41,021 patients. Circulation. 1995;91:1659-1668. 18. Mulrow CD, Cornell JA, Herrera CR, Kadri A, Farnett L, Aguilar C. Hypertension in the elderly: implications and generalizability of randomized trials. JAMA. 1994; 1. The Multicenter Diltiazem Postinfarction Trial Research Group. The effect of dil- 272:1932-1938. tiazem on mortality and reinfarction after myocardial infarction. N Engl J Med. 19. Gill A, Flaim SF, Damiano BP, Sit SP, Brannan MD. Pharmacology of bepridil. 1988;319:385-392. Am J Cardiol. 1992;69:11D-16D. 2. Furberg CD, Psaty BM, Meyer JV. Nifedipine: dose-related increase in mortality 20. Singh BN. Bepridil therapy: guidelines for patient selection and monitoring of in patients with coronary heart disease. Circulation. 1995;92:1326-1331. therapy. Am J Cardiol. 1992;69:79D-85D.

ARCH INTERN MED/ VOL 159, OCT 25, 1999 WWW.ARCHINTERNMED.COM 2348

Downloaded From: https://jamanetwork.com/ on 09/28/2021