Diuretics, β-Blockers, and the Risk for Sudden Cardiac Death in Hypertensive Patients

Methods

Case-Patients and Controls

Eligible case-patients were all residents of Rotterdam, the Netherlands, with a sudden cardiac death between 21 November 1988 and 21 November 1990 who had been receiving drug therapy for hypertension on the date of death. Sudden death was defined as death occurring within 1 hour of the onset of symptoms or as unwitnessed death [11, 12]. The death was considered to be of cardiac origin unless, according to the attending physician, evidence from hospital or autopsy records suggested a noncardiac cause [13]. The age- and sex-matched controls were also receiving drug treatment for hypertension on the day the corresponding case-patient died but were alive on that date. The study was approved by the ethics committee of the Academic Hospital Dijkzigt/Erasmus University, Rotterdam.

Selection of Case-Patients and Controls

For 2 years, municipal authorities prospectively provided information on all persons aged 20 years or older who died in Rotterdam. The physicians who signed the death certificate were mailed a questionnaire comprising three brief questions on the period between the onset of symptoms and the occurrence of death, the possibility of a noncardiac cause, and the name and address of the patient's general practitioner. If the patient met the criteria for sudden cardiac death, the general practitioner was asked by mailed questionnaire whether the patient was receiving antihypertensive medication on the date of death and, if so, whether hypertension was the indication for the prescription and not another condition such as congestive heart failure or arrhythmias. We sent a questionnaire to physicians of 10 649 patients who were reported to have died of natural causes during the study period. A total of 8314 questionnaires was returned (a response rate of 78%), and 7834 (74%) contained adequate information on the circumstances of death. Responding physicians reported 1847 deaths as sudden cardiac deaths; 272 (15%) of these patients were receiving drugs for hypertension on the date of death. Fifteen patients (5%) were excluded because general practitioners refused to cooperate or because evidence of antihypertensive drug treatment on the date of death was missing from the files. Thus, we identified 257 cases of sudden cardiac death.

When a case-patient was identified, a general practitioner practicing in Rotterdam was randomly selected and visited at his or her office. The first patient of the same sex as and born within 2.5 years of the birth of the corresponding case-patient was selected from the alphabetically ordered patient registry. The general practitioner was subsequently asked the same questions asked of the case-patient's general practitioner on the use of antihypertensive medication and whether hypertension was the indication for treatment. The first randomly selected patient who met the criteria was selected as the control.

Ascertainment of the Use of Antihypertensive Medications

All information on case-patients and controls, available at the general practitioner's office, was thoroughly examined by one of the investigators (AWH). This information was obtained from patient charts, correspondence from medical specialists, laboratory results, and electrocardiograms. Only printed or handwritten (that is, reproducible) information was used. From these sources, the specific antihypertensive medication use on the index date (the day the case-patient died), the duration of use, and the dosage were assessed.

We first categorized the prescribed antihypertensive medication into three groups: non-potassium-sparing diuretics (predominantly hydrochlorothiazide, chlorthalidone, and furosemide) regardless of concomitant use of β-blockers; β-blockers (predominantly atenolol and metoprolol) regardless of concomitant use of non-potassium-sparing diuretics; and antihypertensive medication without either non-potassium-sparing diuretics or β-blockers. The latter category served as the reference group. We chose this categorization to assess the risk for sudden death among persons receiving non-potassium-sparing diuretics or β-blockers, regardless of the concomitant use of the other antihypertensive drugs.

We also decided in advance to categorize the antihypertensive medication into four mutually exclusive medication groups: 1) non-potassium-sparing diuretic therapy without β-blocking agents; 2) β-blocking agents without non-potassium-sparing diuretic therapy; 3) non-potassium-sparing diuretics and β-blockers used concomitantly; and 4) antihypertensive medication without either non-potassium-sparing diuretics or β-blockers.

Accordingly, category 4 included antihypertensive drugs that have not been recommended for first-choice therapy in most published national guidelines on hypertension therapy [14, 15]. In all analyses, this category was the reference group, and it primarily included potassium-sparing diuretics (used by 65% of the patients in this group), calcium antagonists, and angiotensin-converting enzyme inhibitors. The concomitant use of a non-potassium-sparing diuretic and either a potassium-sparing diuretic or an angiotensin-converting enzyme inhibitor was considered to be a potassium-sparing combination and was thus placed in category 4.

Assessment of Comparability of Patient Groups

Given the nonexperimental approach taken in this study, it was important to ascertain that patients who died suddenly had not received non-potassium-sparing diuretics or β-blockers for medical reasons that eventually led to their death. We had to determine whether clinical variables were present, such as indications or contraindications, that were related both to the use of non-potassium-sparing diuretics or β-blockers and to the patient's propensity to sudden cardiac death. We obtained data on the following potential confounders from the general practitioners' records:

1. History of cardiovascular disease, including previous myocardial infarction, heart failure, angina pectoris, stroke, intermittent claudication, atrial fibrillation, and other cardiac arrhythmias (tachycardia, bradycardia, and atrioventricular blocks). These diseases were considered to be present when the diagnosis was explicitly mentioned in the patient file.

2. Cardiovascular risk indicators, including smoking habits, last available plasma cholesterol concentration, body weight, and left ventricular hypertrophy (measured by electrocardiography).

3. Use of digitalis.

4. Comorbid conditions, such as diabetes mellitus (that is, a diagnosis of diabetes recorded in the patient file), kidney dysfunction (the last measured serum creatinine level more than 100 µmol/L), and chronic obstructive pulmonary disease (categorized into two groups: a diagnosis of asthma, chronic bronchitis, or emphysema without current use of pulmonary medication and current use of β-agonists, oral corticosteroids, or theophylline).

We also recorded information on the severity of hypertension, including blood pressure at the initiation of therapy, the mean pressure during the last 5 years, the last available blood pressure, and the number of antihypertensive drugs currently used.

Validation of Data Collection

Because investigators could not be blinded to the vital status of the patient when collecting the data, efforts were made to check the validity of the data collection. First, the use of antihypertensive and concomitant medications on the index date as collected from the general practitioner's files was compared with data obtained from computerized pharmacy databases. For logistic reasons, medication histories from pharmacies were available for 28% of the case-patients and for 46% of the controls. These different proportions primarily resulted from the fact that persons enrolled in the mandatory collective insurance (most of the Rotterdam population) are usually deleted from the computerized pharmacy databases within 1 to 2 months of their death. We found only a few discrepancies between the findings recorded in the general practitioner's office and the pharmacy data. To further check the validity of the data collected from the patient files, a physician who was unaware of the specific research question and was blinded to the vital status of the patient reexamined the files of a random sample of 10% of both the case-patients and controls. No differences in the categorization of antihypertensive medication by the two investigators were found, and only minor discrepancies between the reported prevalence of potential confounders were detected (unpublished data).

Statistical Analysis

To estimate the association between the use of antihypertensive drugs and the occurrence of sudden cardiac death, we used conditional logistic regression [16] to calculate matched odds ratios (as an approximation of relative risks) among persons receiving non-potassium-sparing diuretics regardless of the concomitant use of β-blockers and among persons receiving β-blockers regardless of the use of non-potassium-sparing diuretics relative to the use of antihypertensive medication without either of these two types of drugs. We also computed the risks for sudden death associated with the use of non-potassium-sparing diuretics and no β-blockers, the use of β-blockers and no non-potassium-sparing diuretics, and the concomitant use of both drugs relative to the same reference group.

To further quantify the possibility of the interaction between non-potassium-sparing diuretics and β-blockers with regard to the risk for sudden death, we did an analysis that included a multiplicative interaction term. To evaluate the stability of the odds ratios and 95% CIs, we added potential confounders to a multivariate model one by one. When appropriate, we included continuous variables in the model as dichotomous variables or dummy variables representing three or more categories. In case of missing data for a categorized continuous variable (for example, blood pressure or plasma cholesterol level), the risk indicator was considered to be absent. To assess the consequences of this approach, dummy variables indicating missing values were included in the model.

We did subgroup analyses for men and women and for patients aged 75 years or younger and those older than age 75. In these analyses, we applied the same multivariate model as that used for the entire study population. To study the possibility of a dose-response relation between non-potassium-sparing diuretics and sudden cardiac death, we categorized the prescribed dosage according to the number of defined daily dosages received per day.

Results

The mean age of all patients included in the study was 74 years (range, 42 to 93 years); 46% were male (Table 1). The mean blood pressure during the last 5 years was somewhat higher in the controls than in the case-patients. More case-patients than controls were prescribed combination therapy for hypertension. As expected, a history of cardiovascular diseases was more prevalent in the case-patients than in the controls. Similarly, more case-patients had smoked cigarettes.

Twenty-two percent of the case-patients had received non-potassium-sparing diuretics; 9% had received concomitant β-blocker therapy, and 13% had not (Table 2). Eighteen percent of the controls had received non-potassium-sparing diuretics; 9% had received concomitant β-blocker therapy, and 9% had not. In addition, more case-patients than controls were taking β-blockers. After adjustment for potential confounders, patients receiving non-potassium-sparing diuretics (regardless of additional use of β-blockers) had an increased risk for sudden cardiac death compared with hypertensive patients receiving antihypertensive medication without non-potassium-sparing diuretics and β-blockers (relative risk, 1.8 [CI, 1.0 to 3.1]). The corresponding relative risk of patients receiving β-blockers (regardless of additional use of non-potassium-sparing diuretics) was 1.7 (CI, 1.1 to 2.6). The use of non-potassium-sparing diuretics without β-blockers was associated with a higher risk for sudden death (relative risk, 2.2 [CI, 1.1 to 4.6]) than concomitant use of non-potassium-sparing diuretics and β-blockers (relative risk, 1.4 [CI, 0.6 to 3.0]). The relative risk of patients receiving β-blockers without non-potassium-sparing diuretics was 1.8 (CI, 1.1 to 2.9). In the multivariate analysis that included a multiplicative interaction term of the use of non-potassium-sparing diuretics and β-blockers, the odds ratio for the interaction term was 0.4 (CI, 0.1 to 1.0).

Analyses further showed that previous myocardial infarction and a history of heart failure and angina pectoris were the strongest confounders of the relation between the use of diuretics or β-blockers and the risk for sudden death. Additional inclusion of the other potential confounders in the model did not appreciably change the estimated relative risks. Further, all relative risks for sudden death associated with the potential confounding variables lay between 0.7 and 1.9, expect for myocardial infarction (relative risk, 3.0 [CI, 1.5 to 5.7]) and heart failure (relative risk, 2.7 [CI, 1.4 to 5.3]).

Relative risks for sudden death associated with the use of non-potassium-sparing diuretics and β-blockers were similar for men and women. However, the association between non-potassium-sparing diuretic therapy and sudden cardiac death was more pronounced in hypertensive patients aged 75 years or younger than in those older than age 75 years (Figure 1). Further subgroup analyses did not show a clear dose-response relation between the dose of non-potassium-sparing diuretic therapy prescribed per day and the risk for sudden cardiac death. However, almost all patients were prescribed one defined daily dose or less, suggesting that the range in dosage may have been too small to detect a dose-response association. Furthermore, the duration of therapy with non-potassium-sparing diuretics seemed to influence the risk for sudden cardiac death. The adjusted relative risk for sudden death among persons receiving short-term therapy (< 1 year) with non-potassium-sparing diuretics and without β-blockers was 3.1 (CI, 1.2 to 8.2), whereas the corresponding relative risk for persons receiving long-term therapy (> 1 year) was 1.3 (CI, 0.4 to 3.6). Hypertensive patients who received non-potassium-sparing diuretics in combination with digitalis had an increased risk for sudden cardiac death compared with those who received non-potassium-sparing diuretics without digitalis (crude relative risks, 1.9 and 1.6, respectively). However, this increased risk disappeared after adjustment for potential confounders.

Discussion

In this population-based study of hypertensive patients, those who received non-potassium-sparing diuretics had a twofold increased risk for sudden cardiac death compared with a reference group that primarily received potassium-sparing diuretics. Patients receiving β-blockers for hypertension also had a higher incidence of sudden cardiac death.

Our finding of an increased risk for sudden death among patients receiving non-potassium-sparing diuretics for hypertension is remarkably similar to that of a recent case–control study by Siscovick and colleagues [10]. In their study of 114 patients with a sudden cardiac death between 1977 and 1990, the relative risk for sudden cardiac death among hypertensive patients receiving potassium-sparing diuretics compared with those receiving thiazides was 0.3 (CI, 0.1 to 0.7). Several trials in patients with mild to moderate hypertension reported that the incidence of sudden cardiac death was higher in hypertensive patients randomly assigned to interventions that included thiazide therapy compared with a reference group assigned to receive placebo, no treatment, or “usual care” [3, 17-21]. Other trials, including the Systolic Hypertension in the Elderly Program trial, could not show an increased risk for sudden death in patients receiving non-potassium-sparing diuretic treatment [22, 23].

Diuretic-induced potassium or magnesium depletion leading to cardiac arrhythmias has been suggested as the mechanism underlying an increased risk for sudden death [24]. Although our study was not designed to directly address this mechanism, the increased risk for sudden cardiac death in patients treated with non-potassium-sparing diuretics compared with those receiving potassium-sparing diuretic therapy (included in the reference category) supports the possibility that potassium or magnesium depletion may be involved. The relatively higher risk in persons who recently received non-potassium-sparing diuretics compared with persons receiving them for more than 1 year further supports the view that drug-induced electrolyte depletion may be implicated.

The increased risk for sudden death among recipients of non-potassium-sparing diuretics was less pronounced in those who also received concomitant β-blocker therapy (relative risk, 1.4) than in those who did not (relative risk, 2.2). The analysis that included a multiplicative interaction term further indicated that an interaction between the two drugs exists. The antiarrhythmic properties of β-blockers possibly reduce the occurrence of severe arrhythmias in some patients with diuretic-induced electrolyte depletion. Furthermore, concomitant β-blocker use may prevent an additional adrenaline-mediated intracellular loss of potassium during cardiac ischemia.

Our finding of an increased risk for sudden death among hypertensive patients prescribed β-blockers (relative risk of 1.7 [CI, 1.1 to 2.6] compared with the reference group) was unexpected in view of the evidence suggesting a cardioprotective effect of β-blockers in such patients [7, 8]. In their recent case–control study, however, Siscovick and coworkers [10] found a similar increased risk for sudden death among hypertensive patients receiving β-blockers; however, the investigators did not emphasize this finding.

Several explanations for our finding should be considered. First, differences in clinical characteristics (such as symptoms of angina and cardiac arrhythmias) between patients prescribed β-blockers and those receiving other antihypertensive medications could lead to an artificially increased risk for sudden death with β-blocker use. However, the availability of extensive clinical information in the patients' files allowed us to adjust for these characteristics in the statistical analysis. Nevertheless, small differences between patients cannot be completely excluded. Further, the excess risk for sudden death during β-blocker therapy in relatively old patients with hypertension (mean age, 74 years) could be partly explained by the fact that these drugs may be less efficacious in preventing coronary heart disease in elderly persons than in middle-aged persons [25].

Future studies are needed to establish the clinical importance and mechanistic nature of our finding on the risk for sudden death associated with β-blockers and to further address the possibility that concomitant β-blocker use improves the risk for sudden death that is related to the use of non-potassium-sparing diuretics. The difference between a moderate cardioprotective modification by β-blockers on the effect of non-potassium-sparing diuretics in persons receiving these drugs and the possibly increased risk for sudden death associated with the use of β-blockers in persons not receiving non-potassium-sparing diuretics requires further attention.

A limitation of our study was the absence of randomization. Nevertheless, it is possible to compare patients receiving different types of antihypertensive drugs. Although specific classes of antihypertensive drugs have several important indications and contraindications, a large gray area exists in which the choice of drug therapy depends more on the beliefs of the physician than on the actual condition of the hypertensive patient. Thus, many patients who are similar in most respects are treated differently. Still, our greatest concern about the nonexperimental study design we used is the possibility that the choice of antihypertensive therapy is influenced by conditions of the patients that are related to the risk for sudden cardiac death [26]. We used several methods to limit this problem of confounding.

First, we restricted our case-patients to patients who died within 1 hour of the onset of symptoms and those whose deaths were unwitnessed. Most of these patients died instantaneously. Thus, conditions that existed before their death are unlikely to have led to a change in antihypertensive therapy. Second, we assured maximum comparability of patients who received different classes of antihypertensive drugs by adjusting for differences in the risk for sudden cardiac death in a multivariate statistical analysis. The most important factors thought to be related to the risk for sudden death include concomitant heart failure, previous myocardial infarction, angina pectoris, cardiac arrhythmias, and severity of hypertension. Because of the central position of the general practitioner in the health care system in the Netherlands, data on most of these potential confounders were available in the general practitioner's patient files.

Nevertheless, slight differences in therapeutic indications among the patients cannot be ruled out, even after the multivariate analysis. The presence of mild heart failure leading to prescription of non-potassium-sparing diuretics could result in an artificially increased risk for sudden cardiac death among persons receiving these drugs. However, the choice of the reference group in our study (65% of whom received potassium-sparing diuretics) probably resulted in a similar prevalence of mild symptoms of heart failure in patients in the non-potassium-sparing diuretic group and in the reference group. This is supported by the similar prevalence of a diagnosis of heart failure and digitalis use in these two medication groups. Further, the general practitioner was explicitly asked whether the primary indication for the antihypertensive therapy was hypertension rather than heart failure. Mild symptoms of angina leading to preferential prescription of β-blockers could have resulted in an artificially increased risk for sudden death among persons receiving β-blockers; however, adjustment for current use of nitrate therapy as an indicator of angina did not alter the findings. Thus, we believe that differences in therapeutic indications for non-potassium-sparing diuretics and β-blockers cannot explain the increased risk for sudden cardiac death among persons receiving these drugs in our study.

Because our study could not be blinded, bias may have been induced if the misclassification of the use of specific antihypertensive drugs differed among case-patients and controls. Comparison with the findings from the pharmacy databases, however, did not show differential misclassification, nor did the comparison with the recorded findings from the second physician who was unaware of the research question and the underlying hypothesis.

A major advantage of our study compared with previous studies (particularly the randomized trials) is its large number of case-patients (n = 257). This is almost three times the number reported in the largest hypertension trial to date [27]. Furthermore, the case–control approach permits assessment of the influence of many aspects of antihypertensive medication use (including specific drugs, dosages, and the duration of therapy) on the risk for sudden death.

An increased risk for sudden cardiac death among hypertensive patients receiving non-potassium-sparing diuretics or β-blockers may reduce the potential efficacy of these drugs. Assuming an incidence of sudden cardiac death among users of potassium-sparing diuretics of 2/1000 patient-years, the use of non-potassium-sparing diuretics for hypertension would increase the risk for sudden death with another 2/1000 patient-years, given our estimated twofold risk for sudden cardiac death. This would imply that in view of the potential reduction in mortality from antihypertensive therapy among relatively old hypertensive patients of 4/1000 patient-years [9], as much as half of the potential effect of non-potassium-sparing diuretics on survival could be lost as a result of an increased risk for sudden death.

In conclusion, our findings among treated hypertensive patients indicate that the use of non-potassium-sparing diuretics and β-blockers is associated with an increased risk for sudden cardiac death compared with other antihypertensive medications. This may offset part of the mortality benefit of these drugs in the treatment of hypertension.

Drs. van der Does and Hoes: Department of General Practice, Erasmus University Medical School, PO Box 1738, 3000 DR Rotterdam, the Netherlands.

Dr. Lubsen: Societe pour la Recherche Cardiologique (SOCAR) SA, Domaine de Leydefeur, CH 1261 Givrins, Switzerland.

Dr. Man in 't Veld: Department of Internal Medicine I, Academic Hospital Dijkzigt, Erasmus University Medical School, PO Box 1738, 3000 DR Rotterdam, the Netherlands.