Diuretics, Hypokalemia, and Cardiac Arrhythmia: A 20‐Year Controversy

Diuretics were first introduced in the management of hypertension in 19571 when early clinical trial data demonstrated that chlorothiazide was effective in reducing blood pressure (BP) either as monotherapy or in combination with other antihypertensive agents. Later hydrochlorothiazide (HCTZ), a longer-acting diuretic, was tested and found to be effective and well tolerated. In the early 1960s, when Edward D. Freis and the Veterans Administration (VA) Cooperative Study Group designed the landmark VA trials that demonstrated the benefits of hypertension control, HCTZ-based regimens were utilized. These landmark studies2,3 and most of the subsequently published placebo-controlled trials demonstrated consistently that antihypertensive therapy was effective in preventing strokes and most cardiovascular (CV) complications of hypertension, but appeared to have a lesser effect on myocardial infarctions (MIs) and cardiac death. Later trials in the elderly that were also diureticbased, however, reported that MIs were significantly reduced by treatment.4 Most of the trials summarized in review articles and meta-analyses5–8 used diuretic-based regimens to control BP. The question debated intensely in the 1980s and 1990s was whether metabolic side effects of diuretics blunted the expected beneficial effects from hypertension control on cardiac mortality. In fact, the issue became so heated that many “thought leaders” were suggesting withdrawal of diuretics from the market, and cartoons appeared in scientific journals depicting tombstones of the lifespan of diuretics, e.g., “1956–1987.” I recall advertisements published in reputable journals depicting the heart leaking potassium, wrapped in bundles of dynamite ready to explode. Television advertisements showed ambulances rushing to patients' homes to resuscitate them from sudden death because of diuretic therapy. Yet all this noise was based on circumstantial evidence and opinions, not hard data.9 Drug therapy in hypertension, using diuretic-based regimens, had proved beneficial in repeated studies. Data from well designed trials constantly affirmed the efficacy and safety of diuretics. A few lone investigators recommended calm and reason. Edward D. Freis, Marvin Moser, and few others (this author included) continued publishing editorials and original articles in defense of diuretics.10–16 In this article, I review the controversy that surrounded diuretic therapy in the 1980s, some of the evidence, and the changes in perception over the past two decades—as well as the status of diuretic-based treatment today. In 1986, in an editorial in The American Heart journal,13 I noted that “Major changes in the established treatment of disorders such as hypertension, which have a high prevalence in the population, should be approached with caution, because they may have considerable financial and public health implications.” I noted the fact that effective management of hypertension with diuretic-based regimens was a major contributor to the decrease in CV morbidity and mortality and that diuretics were routinely recommended as initial therapy in the stepped-care approach.17,18 Diuretics have a unique mode of action and are the only agents to reduce BP, at least in part, by volume contraction; this enhances the effect of other antihypertensive agents. I pointed out the controversies and concerns about their safety. Although other potential metabolic side effects of diuretics, such as at least a transient increase in serum cholesterol, hyperuricemia, and some increase in plasma glucose were of concern, the major focus was on diuretic-induced hypokalemia and sudden cardiac death.19 Fears that mild-to-moderate hypokalemia due to diuretic therapy might cause malignant ventricular arrhythmia initially resulted in a dramatic increase in prescriptions for potassium supplements and potassium-sparing diuretics and initiated a trend away from diuretic use in favor of newer, more expensive regimens.20 Here are some of the questions posed at that time: • Is diuretic therapy, per se, a risk in patients with uncomplicated systemic hypertension? • Is diuretic-induced hypokalemia a cause of arrhythmias in these patients? • Is hypokalemia causally related to malignant arrhythmias in patients suffering from acute myocardial infarction? The notion that diuretics may increase the risk of sudden death among patients with uncomplicated hypertension originated from the Multiple Risk Factor Intervention Trial (MRFIT).21 MRFIT was a primary prevention trial designed to investigate the effect of multiple risk factor reduction on coronary artery disease events. Participants in the study had one or more of three major risk factors: high cholesterol, high BP, or a history of cigarette smoking. Patients were randomized to a special intervention (SI) or usual care (UC) group. The hypothesis was that more aggressive treatment would result in fewer fatal and nonfatal coronary events. (Both groups were treated with diuretics.) The relationship between coronary deaths and diuretic therapy was not an initial focus of the study. When the study failed to demonstrate the expected benefit in the intensively treated group, a retrospective analysis of a number of subgroups was carried out. An unfavorable trend was identified among hypertensive men with baseline ECG abnormalities in the SI group. Although the authors of the study advised caution in interpreting their findings, and further analyses22 softened the initial impressions, their findings were interpreted as establishing an association between thiazide therapy and increased risk of sudden death. Here are some of the comments about this interpretation of the MRFIT data: • The difference in mortality rates between the SI and UC groups (29.2 vs. 17.7/1000) was not statistically significant. It was only when the estimated differences in the relative risks (SI/UC) for hypertensive men with or without ECG abnormalities were compared that statistical significance was achieved.22 • There was no relationship between the dose of diuretic used or the presence or absence of hypokalemia and increased mortality rates. Nevertheless, it was assumed that some toxic effect of diuretics (perhaps hypokalemia) caused an increased mortality rate. The most likely explanation for the difference in mortality rates, however, was the unexpected low mortality rate observed in hypertensive men with resting ECG abnormalities in the UC group (Table I) (17.7/1000 in patients with ECG abnormalities and 20.7/1000 with normal ECG). The observed difference between hypertensive men in the SI group with or without ECG abnormalities was as expected (29.2 vs. 15.8/1000). These data have been extensively reviewed.23 Contrary to expectations, UC participants with normal resting ECGs had higher mortality rates than UC participants with abnormal resting ECGs—a most unusual finding. Thus the difference in mortality rates was not so much a result of a high mortality rate in the SI group as it was the result of an atypically low rate in the UC group. • Most ECG abnormalities were consistent with left ventricular hypertrophy, which is a well known risk factor for sudden death. • This concept was supported by a subsequent analysis of the Hypertension Detection and Follow-up Program (HDFP) (Table II).24 Following the original MRFIT publication, the HDFP Cooperative Research Group performed extensive reanalysis of their data using criteria similar to those of MRFIT. Their findings24 did not confirm the MRFIT findings. As shown in Table II, mortality rates were similar in the SI (MRFIT) and stepped-care groups (HDFP) for hypertensive participants with resting ECG abnormalities (20.3 vs. 21.4/1000). The UC mortality rate in MRFIT, however, was 40% lower than the referred-care mortality rate in HDFP (10.7 vs. 17.5/1000). The HDFP Cooperative Research Group concluded that their findings “offer no support for the hypothesis raised in MRFIT that intensive diuretic therapy might have increased the mortality rate of hypertensive patients with resting ECG abnormalities.” If diuretics had a deleterious CV effect, it should have manifested itself in patients with abnormal exercise tests and a high prevalence of coronary artery disease. This was not the case, however, in MRFIT.22 Hypertensive SI participants with positive exercise ECG responses had considerably lower mortality rates as compared with similar patients in the UC group. On the basis of these findings, it was concluded that the assumption that diuretics produce deleterious effects was unwarranted, but the original message had been repeated at medical meetings and in journals—retractions or corrections are often buried or ignored.25 Diuretic-induced hypokalemia is dose dependent. With moderate/high doses of diuretics, the reported incidence of hypokalemia ranges between 20% and 40%.22,23 Serum potassium changes are usually quite small and serum potassium <3.0 mEq/L is only occasionally encountered. The incidence of hypokalemia with low doses of diuretics is 5%–7%.26–28 Although there is no evidence that mild-to-moderate hypokalemia is harmful, it has been speculated that low serum potassium may trigger cardiac arrhythmias and sudden death. The initial observations associating hypokalemia with cardiac arrhythmias were made in digitalized patients.29 These observations were later confirmed,30 and there is consensus that hypokalemia may aggravate digitalis-induced arrhythmias. These observations were generalized, however, to include patients with uncomplicated hypertension, and ventricular ectopy has been used as a surrogate for sudden death. In 1982, Harrington et al.20 reviewed the published literature and concluded that “the published experience discloses no evidence that fatal ventricular ectopy results from diuretic-induced hypokalemia.” In 1986, we reviewed reports published after 1981 and reached similar conclusions.13 Since ventricular ectopic activity became the debated surrogate end point for sudden death, it is of interest to discuss the published reports. The most frequently quoted study providing evidence of an association between diuretic-induced hypokalemia and cardiac arrhythmia is the study by Holland et al.31 These investigators performed 24-hour ECG monitoring before and after diuretic therapy in 21 patients with uncomplicated hypertension who were known to become hypokalemic on thiazides. Patients with more than minimal ectopy at baseline were excluded. Following diuretic therapy and hypokalemia, seven of these 21 patients had increased ventricular ectopy, and correction of hypokalemia improved arrhythmia. Serum potassium was not corrected, and repeat monitoring was not done in the remaining 14 patients. The study was criticized for the following reasons: • Patients were separated based on the presence or absence of cardiac arrhythmias, not based on the potassium level. • Patients with more than minimal ectopy at baseline were excluded, thus biasing results favoring more arrhythmias during the second recording. • Plasma potassium was corrected only in the seven patients with arrhythmias, predisposing the data to regression toward the mean. • The average decrease in plasma potassium with thiazide therapy was the same in both groups of patients—those with and without arrhythmias. • Day-to-day variability32 and regression toward the mean could explain most of the findings. Several other studies33–40 failed to confirm the results of Holland et al. In one study33 from our laboratory, 44 patients with uncomplicated hypertension underwent 48-hour ambulatory ECG recording before and after diuretic therapy (100 mg of HCTZ q.d. for 4–8 weeks). There was no increase in ventricular arrhythmias despite the fact that in 27 patients serum potassium was reduced below 3.4 mEq/L. In seven patients, serum potassium was reduced to <3.0 mEq/L, but no increase in ectopy was found. In a second study,10 hypertensive patients with proven left ventricular hypertrophy by echocardiography were monitored before and after high-dose diuretic therapy. Although serum potassium was reduced by 0.57 mEq/L, there was no increase in the frequency of premature ventricular beats, couplets, or runs of ventricular tachycardia. Thus it appeared that short-term diuretic therapy/hypokalemia was not associated with increased ventricular ectopy. Madias et al.34 found no increase in cardiac arrhythmia in diuretic-treated patients with mild or moderate hypokalemia. The Medical Research Council (MRC) of Great Britain reported findings in two groups of patients which are rather conflicting.35 In one group of patients receiving long-term treatment with thiazide diuretics, increased ventricular ectopy was noted compared with a placebo-treated group. The increased ectopy could not, however, be attributed to hypokalemia; there was no correlation with plasma potassium in this particular subgroup. Furthermore, these patients had no baseline ECG monitoring for comparison. In another group of patients, 24-hour ECG monitoring was performed before and after 9–10 weeks of diuretic therapy. No difference in cardiac arrhythmias was found. In a study by Siegel et al.,36 212 hypertensive men were randomized to placebo or one of five diuretic treatment regimens. No significant association was found between any of the treatment regimens and ventricular ectopy. The effect of chlorthalidone on ventricular arrhythmias was examined in a subgroup of 238 patients in the Systolic Hypertension in the Elderly Program (SHEP).37 No significant differences were noted in ventricular arrhythmias between placebo or diuretic-treated patients. In a similar study, Narayan et al.38 examined the effect of HCTZ therapy on ventricular arrhythmias in 45 hypertensive African-American men with moderate-to-severe left ventricular hypertrophy. Systolic BP was reduced from 168/103 mm Hg at baseline to 146/89 mm Hg after HCTZ. Serum potassium decreased significantly, from 4.2 mmol/L to 3.7 mmol/L. The average hourly incidence of ventricular premature contractions was 22 with placebo and 25 with HCTZ. There were three and one couplets, and 0.2 and 0.2 runs of ventricular tachycardia per patient per hour, respectively. Variables of signal-averaged ECGs did not differ between the two treatments.38 The effects of potassium supplements or potassium-sparing diuretics on cardiac arrhythmias was also examined in a limited number of patients. In Holland's study,31 correction of mild hypokalemia in seven diuretic-treated patients improved arrhythmias. In contrast, in one of our studies8 in 16 patients with severe hypokalemia (average serum potassium of 2.8 mEq/L), correction of hypokalemia with high-dose potassium supplements, triamterene, or both, had no effect on ventricular ectopy. Exercise-induced arrhythmias deserve special consideration. The combination of low serum potassium with high catecholamine levels produced during exercise can potentially create an arrhythmogenic environment. Catecholamine infusion has been shown to shift potassium intracellularly and exaggerate diuretic-induced hypokalemia.39 These relationships were studied in our laboratory in a group of 20 hypertensive patients39 who underwent a symptom-limited exercise stress test before and after high-dose diuretic therapy (100 mg HCTZ q.d. for 4 weeks). Catecholamine levels, potassium levels, and ventricular ectopy were assessed before, during, and after exercise testing. Although diuretic therapy resulted in a significant reduction of serum potassium, and exercise resulted in a substantial rise in serum norepinephrine levels, there was no increase in ventricular ectopy.40 These studies strongly suggest that diuretic therapy is not associated with increased arrhythmogenicity during exercise in patients with uncomplicated systemic hypertension. Mclnnes et al.41 summarized the data on the cardiotoxicity of diuretics and concluded that there was little substance and much speculation regarding this topic. The effect of diuretic-induced hypokalemia on health outcomes has been examined in a limited number of studies. In a secondary analysis of the SHEP trial,42 approximately 7.2% of the participants randomized to chlorthalidone developed serum potassium <3.5 mmol/L at the first annual visit. A total of 151 patients on diuretics and 21 on placebo developed hypokalemia. During the follow-up period, patients with hypokalemia were found to have more strokes and nonfatal Mis, but a lower total mortality (although the differences were not statistically significant). Overall, the CV risk in the treated patients with hypokalemia was less than in placebo subjects. Possible confounders are the relatively small number of patients who developed hypokalemia (151 of the 1102 participants in the diuretic group were hypokalemic at 1 year) and the fact that hypokalemia might have been a marker of more severe hypertension necessitating higher doses of diuretics. The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) also examined the effect of low serum potassium on CV outcomes. Results have been presented in an abstract form.43 Of the 15,255 patients participating in the chlorthalidone group, 1185 developed hypokalemia. The authors concluded that “Although chlorthalidone-treated participants were significantly more likely to develop hypokalemia, no evidence of an adverse CV effect was observed.” Full publication is in progress. But the proof of the pudding is in the eating. If diuretic-induced hypokalemia or other possible metabolic effects of diuretics had detrimental consequences on health outcomes, differences would have become manifest in the many comparative trials that have tested the effect of newer agents compared with older ones, specifically diuretics compared with angiotensin-converting enzyme inhibitors (ACEIs), calcium channel blockers (CCBs), or angiotensin receptor blockers (ARBs). Careful review of the literature indicates that the effects of diuretic-based treatments remain unsurpassed, particularly on cardiac outcomes.44 Below are some of the supporting clinical trial data. ALLHAT,28 the largest trial ever carried out in hypertension, demonstrated that the effect of chlorthalidone-based therapy on coronary heart disease mortality and nonfatal MI was equivalent to lisinopril- or amlodipine-based regimens. In fact, there were more strokes in the lisinopril group and more heart failure events in both the amlodipine- and lisinopril-treated patients. Although some of the differences in these secondary end points can be explained by the small differences in BP, the fact is that the protective effect of diuretic treatment on health outcomes remained unsurpassed by the newer therapies. The Swedish Trial in Old Patients with Hypertension (STOP-2)45 compared the effects of conventional (diuretic/β blocker) and newer antihypertensive drugs (ACEIs, CCBs) on CV morbidity and mortality in elderly patients. The primary combined end point of fatal stroke, fatal MI, and other fatal CV events occurred in 221 of 2213 patients in the conventional drug group (19.8 events/1000 patient years) and in 438 of 4401 in the newer drug groups (19.8 events/1000 patient years). There was no difference in cardiac mortality among the different groups.45 The International Nifedipine Gastrointestinal Therapeutic System Study: Intervention as a Goal for Hypertension compared the effects of with the diuretic combination on CV mortality and morbidity in patients with hypertension. The primary was CV death, MI, heart or outcomes occurred in patients in the group and in in the group. the protective effect of diuretic therapy was unsurpassed by newer agents. The compared the effects of with those of a or both, on CV morbidity and mortality in hypertensive patients. The combined primary end point was fatal and nonfatal stroke, MI, and other CV death. The primary end point occurred in patients in the group and in in the diuretic and group. and nonfatal occurred in patients in the group and in in the group, and fatal and nonfatal MI occurred in and patients, respectively. There was no significant difference between The of compared to an or HCTZ in hypertensive patients. There were primary CV events in the group compared with in the or HCTZ The compared the effects of to conventional therapy on CV morbidity and mortality in patients with hypertension. end points occurred in patients in the group patient years) and in the group patient years). CV mortality was lower with than with conventional treatment vs. the rate of fatal and nonfatal MI was similar vs. significant difference between but fatal and nonfatal were significantly more with vs. other studies published In a et summarized the clinical trial data regarding the safety and efficacy of antihypertensive used as initial treatment and their effects on major CV events and mortality. Data were combined from clinical trials that patients randomized to seven major treatment all outcomes, diuretics were to placebo in preventing coronary heart disease heart stroke, CV CV and total mortality. of the other antihypertensive agents and were significantly than diuretics for any primary or secondary health diuretics are recommended by all major The of the on and Treatment of that should be used as initial therapy for most patients with hypertension, either or in combination with one of the other demonstrated to be beneficial in randomized Diuretic-induced hypokalemia should no be a major concern, when lower doses of diuretics are these and the proven diuretics are The noise has to other side such as an increase in glucose and Thus there is little evidence that the of BP is to however, that BP reduction and BP control are CV The relationship of BP to CV events is and from mm Hg on increase of 20 mm Hg BP or mm Hg BP, the risk of stroke, MI, or small in BP mm in the can have a substantial on CV disease. Diuretics have a unique in the management of hypertension. are the only agents that lower BP by volume and the effects of other As shown by the VA Cooperative Study that a diuretic have a of BP than other In fact, of diuretic use is the most for Diuretics well with and but are in any The continued of diuretics is we are to in the rate of BP