|Home | About | Journals | Submit | Contact Us | Français|
Concerns exist that diuretic-induced changes in serum potassium may have adverse effects in hypertensive patients. ALLHAT, a large practice-based clinical trial made it possible to examine consequences of observed changes in potassium during care in conventional practice settings. Normokalemic participants randomized to chlorthalidone versus amlodipine or lisinopril as first-step drug were stratified by year-1 potassium. Post-year-1 outcomes among hypokalemics (potassium<3.5mmol/L) and hyperkalemics (potassium>5.4mmol/L) were compared to normokalemics (potassium 3.5–5.4 mmol/L). Year-1 hypokalemia incidence was 6.8%; incidence in chlorthalidone (12.9%) differed from amlodipine (2.1%; p<0.001) and lisinopril (1.0%; p<0.01). Hyperkalemia incidence (2.0%) was greater in lisinopril (3.6%) than chlorthalidone (1.2%; p<0.01) or amlodipine (1.9%; p<0.01). Coronary heart disease occurred in 8.1% with hypokalemia, 8.0% with normokalemia, and 11.1% with hyperkalemia. Overall, mortality was higher in hypokalemics than normokalemics (Cox hazard ratio =1.21; 95% confidence interval=1.02–1.44) with statistically significant (interaction p<0.01) disparity in hazard ratios for the three treatment arms (hazard ratios: chlorthalidone=1.21, amlodipine=1.60, lisinopril=3.82). Hyperkalemia was associated with increased risk of combined cardiovascular disease (hazard ratio=1.58; 1.15–2.18) without significant treatment interactions. In conventional practice settings, the uncommon appearance of hyperkalemia was associated with increased cardiovascular disease risk. Hypokalemia was associated with increased mortality; however, the statistically significant heterogeneity in hazard ratios across treatment groups strongly suggests that the observed increase in mortality is unrelated to the specific effects of chlorthalidone. Thus, for most patients, concerns about potassium levels should not influence clinician’s decision about initiating hypertension treatment with low-moderate doses of thiazide diuretics (12.5–25 mg of chlorthalidone).
Concerns have been raised that low and high serum potassium (K+) concentration may be associated with adverse cardiovascular effects in hypertensive patients.1 While variations in serum K+ have been implicated in development and progression of coronary heart disease (CHD), new-onset diabetes mellitus,2 and myocardial infarction,3 the main concern has been the potential of diuretic-induced hypokalemia to provoke cardiac arrhythmia and sudden death.4
Diuretics, an antihypertensive mainstay for over 50 years, produces hypokalemia more frequently than do other antihypertensive agents.5 Concerns have also been raised about the potentially adverse effects of hyperkalemia induced by angiotensin converting enzyme inhibitor (ACEI) therapy, such as increased risk of cardiovascular disease (CVD) mortality in hypertensive patients.1
The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), a double-blind, practice-based hypertension treatment trial randomized 42,418 high CVD risk patients to an initial treatment with chlorthalidone (C), lisinopril (L), amlodipine (A), or doxazosin.6, 7 The doxazosin arm was discontinued early and is not considered in this report.8 Chlorthalidone was at least as effective as comparator drugs in preventing cardiovascular events and all-cause mortality, and superior to all in preventing new-onset heart failure. It was also superior to the ACEI in preventing combined CVD and, in Black participants, stroke. With its large size and long follow-up, ALLHAT offers a unique opportunity to determine the relative effects of these treatments on year-1 K+ and the potential impact of these perturbations on long-term (3–7 years) cardiovascular morbidity and mortality in patients treated in diverse clinical practice settings. Data are not available to assess clinical behaviors that occurred in response to these findings. Thus, our data describe associations of hypo- and hyperkalemia with the CVD outcomes but do not detail clinical practice responses (potassium supplementation, change in therapy, etc) that may have contributed to these results. Specifically, we report the association of new-onset hypokalemia (<3.5 mmol/L) and hyperkalemia (>5.4 mmol/L) determined at year-1 of follow-up by a central laboratory with major clinical endpoints, overall and by randomized treatment assignment.
The design of ALLHAT has been reported.7 The primary outcome was non-fatal myocardial infarction (MI) or CHD mortality. Secondary endpoints included fatal and nonfatal stroke, heart failure, combined CVD, and total mortality.
ALLHAT enrolled men and women (47%) aged 55 years and older with hypertension and at least one additional CVD risk factor; 35% of participants were Black; 19% were Hispanic.6 Recruitment was accomplished between February 1994 and January 1998.9 Active follow-up ended March 2002 (average of 4.9 years; range 4–8 years).9 Included in this report are normokalemic participants assigned to chlorthalidone, amlodipine or lisinopril who had potassium measurements at baseline and year-1. Participants were grouped into 3 strata according to their year-1 K+ level, and post-year-1 outcomes over an average 3.9 years were compared. All participants gave written informed consent, all centers obtained institutional review board approval, and the trial was monitored by a National Heart, Lung, and Blood Institute-appointed data and safety monitoring board.
Unless the drug regimen required tapering for safety reasons, individuals continued any prior antihypertensive medications until they received their randomized study drug. Participants were randomly assigned to C (12.5–25 mg/day), A (2.5–10 mg/day), or L (10–40 mg/day) in a ratio of 1.7:1:1, respectively. The study supplied open-label atenolol, reserpine, or clonidine at step 2, and hydralazine at step 3, if needed for blood pressure (BP) control. The BP control goal was <140/90 mmHg in all treatment groups and was assessed at each follow-up visit using the average of two BP measurements. After an initial titration visit at 1 month, participants were seen every 3 months during year-1 and every 4 months thereafter.7
Blood samples were obtained at ALLHAT clinics and analyzed at the ALLHAT Central Laboratory. Serum potassium concentrations were measured at baseline and follow-up visits at 1 month and years 1, 2, 4, and 6. Additional potassium concentrations could be determined locally, but were not collected centrally. Potassium <3.2 mmol/L or >5.5 mmol/L resulted in urgent advisories to the clinics for immediate local rechecks of participants. The laboratory also notified clinics of any K+ increases ≥1.0 mmol/L from the prior measurement. Oral potassium supplementation was provided free of charge to any participant with a potassium level <3.5 mmol/L on local clinic recheck.
The primary outcome was a composite of fatal CHD and nonfatal MI.7 Prespecified secondary outcomes were (1) all-cause mortality; (2) fatal and nonfatal stroke; (3) heart failure (HF; fatal, hospitalized, or treated non-hospitalized); and (4) combined CVD (primary outcome, coronary revascularization, stroke, angina [hospitalized, or treated non-hospitalized], HF, and treated peripheral arterial disease). Additionally, all-cause mortality was classified into CVD and non-CVD causes.
Baseline characteristics were compared across year-1 potassium concentrations using the z-test for continuous variables and contingency table analysis for categorical variables. Cumulative event rates were calculated using the Kaplan-Meier method.10 Cox proportional hazards (PH) regression models (hazard ratios and 95% confidence intervals) were used to compare hypokalemia/normokalemia and hyperkalemia/normokalemia (overall and within randomized drug groups) while adjusting for age, race, sex, history of diabetes, CHD and atherosclerotic CVD, cigarette smoking, baseline systolic BP and K+, and estimated year-1 glomerular filtration rate.10, 11 To account for the possible differences in follow-up BP and K+, Cox PH regression analyses with time-dependent covariates were also performed.
The PH assumption was examined with log-log plots and Schoenfeld residual analysis;12 the assumption was violated once, for HF among C participants, and a logistic model was used to obtain odds ratios and CIs. Heterogeneity of effects in subgroups was examined by testing for treatment-covariate interaction with the Cox PH regression model using P<0.05, indicating statistical significance. However, given the many multivariate, subgroup, and interaction analyses performed, statistical significance at this level should be interpreted with caution.
The study cohort was derived from 33,357 ALLHAT participants randomized to C, A, or L (Supplemental Figure S1). It comprised participants (n=19,731, 59%) who had normal baseline K+ values (3.5–5.4 mmol/L) and valid year-1 values (2.5–7.0 mmol/L); of these, 1,351 (6.8%) had hypokalemia, 17,982 (91.1%) normokalemia, and 398 (2.0 %) hyperkalemia at year-1. Baseline characteristics were similar between this and the overall ALLHAT cohort (Supplemental Table S1). In comparison to normokalemic subjects, those who became hypokalemic were more likely to be black, to be women, and to have received antihypertensive medications prior to enrollment; whereas they were less likely to have a history of CHD and/or diabetes, be taking aspirin, or be a past smoker. In addition, persons with hypokalemia tended to have higher baseline SBP and DBP, lower fasting glucose, higher HDL-cholesterol levels and lower triglyceride concentrations than those with normokalemia. Persons who became hyperkalemic by year-1 tended to be older, have lower DBP, have modestly lower eGFR, and were less likely to be in the lipid-lowering trial component than those with normokalemia.
Mean levels of K+ and BP at baseline and by follow-up year are presented in Supplemental Table S1 by serum K+ group. During follow-up, SBP was similar among these groups, while DBP in hypokalemics was slightly higher than those with normal K+ and was lowest in hyperkalemics.
Randomization to C was associated with increased risk of hypokalemia (1185/9159, 12.9%) compared to A (113/5371, 2.1%) and L (53/5201, 1.0%) (Table 1 footnote). Severe hypokalemia (K+<3.2 mmol/L) occurred in 277 (3.5%) C participants, 17 (0.3%) A, and 8 (0.2%) L. Overall, participants who developed hypokalemia by year-1 did not experience greater CHD, stroke, or HF than those who remained normokalemic (Table 2). The rate for combined CVD was actually lower for hypokalemics compared with normokalemics (HR=0.88), and the result was significantly different for C versus A (p for interaction=0.02; HRC=0.86, HRA=1.48, Tables 1 and and2).2). Total death rates for all hypokalemics exceeded that of normokalemics (HR=1.21, p=0.03) with an absolute risk difference of 2.5% and with significantly different results for L compared with C (p for interaction <0.01; HRC=1.21, p=0.03; HRA=1.60, p=0.06; HRL=3.82, p<0.001). Adjustment for follow-up SBP, DBP, and K+ in Cox PH regression analyses with time-dependent covariates, including fixed covariates previously examined, did not appreciably alter these HRs, slightly increasing the adjusted HR for stroke (1.01 to 1.02), total deaths (1.21 to 1.22), and CVD deaths (1.18 to 1.19) in hypokalemics and, similarly in hyperkalemics, stroke (1.25 to 1.26), death (1.15 to 1.16), and CVD deaths (1.23 to 1.24). The interaction HRs, likewise, did not change appreciably.
Overall mortality in hypokalemics compared with normokalemics comprised an 18% higher risk of CVD death (p=0.20) and a 23% higher risk of non-CVD death (p=.08; Table 2). Mortality from CHD causes accounted for 54% of the CVD deaths but did not differ significantly between hypo- and normokalemic groups (3.99/100 versus 3.78/100; HR=1.32, p=0.11). Notably, mortality from cancer causes, which comprised 52% of non-CVD deaths, was significantly higher in hypokalemics compared with normokalemics (5.48/100 versus 3.74/100; HR=1.52, p<0.01).
There was also heterogeneity between drug groups in several CVD outcomes. Specifically, those assigned to amlodipine who developed hypokalemia, compared with those remaining normokalemic, had significantly increased risk for CHD (HR=2.41), HF (HR=2.19), combined CVD (HR=1.48), and CVD death (HR=2.10). These results for A were significantly different than C, with all interaction p values <0.03. For those assigned to L, there was a significantly increased risk for hypokalemics compared with normokalemics at year-1 for HF (HR=3.10) and CVD death (HR=3.93). These results for L were significantly different than C, with all interaction p values <0.01.
Development of hyperkalemia was far less frequent than hypokalemia (398 vs. 1351) and was more common among L participants (3.6%) than C (1.2%) or A (1.9%). In L participants, those developing hyperkalemia were at increased risk of death (HR=1.49, 1.05–2.12, p=.02) compared with normokalemics (Table 3). Overall (Table 4), hyperkalemics were at significantly increased risk of combined CVD compared to normokalemics (HR=1.58), but there were no significant interactions with treatment.
Potassium supplementation was available (Supplemental Table S2), and 36% of participants with K+<3.2 mmol/L at their first follow-up visit (1–3 months after randomization) were reported to be on supplementation at the next visit (within 3 months); 28% of those with a K+<3.5 mmol/L, and 2% of those with a K+≥3.5 mmol/L, were on supplementation. These percentages increased by year-4 to 71% of those with K+<3.2 mmol/L, and 62% of those <3.5 mmol/L. At the year-1, an open-label diuretic was prescribed for 3% of C hypokalemics, 19% of A hypokalemics, and 23% of L hypokalemics. For hyperkalemics, open label ACEI was prescribed for 4% of C, 6% of A, and 3% of L.
ALLHAT data show that in conventional practice settings, notice of incident hypokalemia was not associated with adverse cardiovascular outcomes, and while associated with increased total mortality, the observed increase showed heterogeneity across treatment groups, with hazard ratio in C significantly different from L. This large study with careful ascertainment of clinical outcomes provides further assurance that appearance of hypokalemia is not likely to compromise the proven cardiovascular benefit of diuretic therapy. In addition, ALLHAT data show that hyperkalemia, while relatively rare and most common in patients randomized to L, was associated with increased total CVD outcomes.
Our analysis was limited to ALLHAT participants, randomized to C, A, or L, who had normal baseline K+ concentrations (3.5–5.4 mmol/L) and repeat measurements between 10 and 14 months later. These participants did not markedly differ from the entire trial population in demographic and clinical characteristics. As expected, modest year-1 hypokalemia (3.2–3.4 mmol/L) was primarily an experience of participants randomized to chlorthalidone (9.5% compared with 1.7% in A and 0.8% in L). Development of severe hypokalemia (<3.2 mmol/L) was less frequent (3.5% in C). By contrast, hyperkalemia (K+>5.4) occurred primarily in L participants (3.6%) and was least common among C (1.2%). Potassium supplements were prescribed to about 10% of participants. The availability and use of potassium supplements suggest that clinical care, in ALLHAT, is likely to mirror conventional care, and thus the CVD outcomes of interest in regard to incident hypokalemia are likely to be generally applicable in settings where clinicians are free to respond appropriately to this laboratory finding.
It has been suggested that hypokalemia, by contributing to atherosclerosis, platelet aggregation, and cardiac arrhythmia, may offset the benefits of BP reduction--thus perhaps increasing cardiovascular morbidity--and help explain a putative deficit in coronary event prevention achieved in clinical trials of diuretic therapy compared with that predicted from epidemiological data.13 Thus, concerns have primarily been directed at diuretic-induced hypokalemia, and its CVD consequences.
Thirteen percent of C participants were hypokalemic at year-1. Their subsequent CVD morbidity and mortality did not exceed that of normokalemic participants in any cardiovascular outcome, and they actually had lower rates of cardiovascular events than the hypo-, normo-, or hyperkalemic subgroups of either L or A participants. For CVD outcomes, heart failure seems to be responsible for most of the heterogeneity as opposed to CHD or stroke. There was less HF in C compared with A or L. Overall mortality among hypokalemic C participants was significantly higher than normokalemics. However, when stratified by CVD and non-CVD causes, only the latter remained statistically significant. The risk of death in hypokalemics compared with normokalemics was the highest in L participants (HR=3.82, p<0.01), intermediate in A (HR=1.60, p=0.06), and lowest in C (HR=1.21, p=0.03). This heterogeneity across treatment groups suggests that hypokalemia likely represents chronic conditions associated with potassium loss and high mortality (as evidence by the increased cancer mortality [HR=1.52, p<0.01]) and transient conditions such as gastro-intestinal disturbances (not documented in ALLHAT). The lowest risk in the diuretic arm was possibly due to admixture of the hypokalemia directly related to the effects of the drugs which in some patients may be corrected by homeostatic mechanisms and in others addressed by potassium supplementation per usual clinical standards. Thus, it appears that non-CVD mortality, specifically cancer deaths, contributes significantly to the excess mortality in hypokalemia. This experience in ALLHAT exceeds, in magnitude and data quality, any similar observational data linking K+ concentrations to subsequent cardiovascular morbidity in a treated hypertensive population.
Of interest in the ALLHAT analyses of hypo- and hyperkalemia and their potential association with clinical events is the degree of persistence of these clinical states during the remainder of the trial. The administration of K+ supplements was encouraged for all participants who had K+<3.5 mmol/L persistently. In this large trial, incidental K+ measurements done as part of routine patient care were not recorded centrally. Of the 19,731 ALLHAT participants included in this report, 16,213 (82%) had central K+ measures at 3, 12, and 24 months. Only 9.6% of the 1117 who were hypokalemic at 12 months were hypokalemic at all three time points. The vast majority of participants with hypokalemia were assigned chlorthalidone. In like fashion only 1.3% of the 309 who were hyperkalemic at 12 months were hyperkalemic at all three time points. The largest proportion was assigned lisinopril. The available data do not permit an explanation for the association of potassium abnormalities with CVD events or total mortality. However, the data do imply that the clinical response to learning that abnormalities in K+ had occurred was both appropriate and sufficient and that hypokalemic or hyperkalemic states detected in ALLHAT participants were not allowed to persist.
Altogether, clinical studies linking K+ concentrations to subsequent events have yielded inconsistent results.4, 14, 15 The limitations of methodology and observational nature of most studies make this inconsistency understandable. The potassium-losing effects of diuretics, particularly when higher doses were in fashion, have been widely described, as have the reverse effects of agents blocking the renin angiotensin system.16–18
Meta-analyses of clinical trials consistently indicate that, as was the case in ALLHAT, no other antihypertensive agent produces cardiovascular protection superior to that achieved when therapy is initiated with a diuretic.19 Diuretic dosage in ALLHAT, 12.5–25 mg chlorthalidone/day, can be considered moderate. The incidence of hypokalemia here was consistent with that seen in placebo controlled clinical trials in which CVD prevention was achieved.19
This report of the ALLHAT is a post-hoc observational analysis of subjects’ experience not protected by randomization, and is therefore, despite robust multivariable analysis, subject to residual confounding. This applies particularly to the drug subgroup analyses. We lack precise interval information on the course of both therapy and potassium concentrations and thus cannot assess interval interventions nor potassium levels proximal to events. It should also be noted that the data here reflect the relatively short-term impact of treatment-induced variations in potassium in older hypertensive patients. At the same time, ALLHAT provides a very large experience to determine the mid-term CVD consequences in routine clinical practice when evidence of incident hypo- or hyperkalemia has been systematically made available to treating physicians.
In summary, these results reveal that severe drug-induced alterations of K+ affect a small minority of treated patients. Hyperkalemia, although infrequent, usually occurred among ACE-inhibitor treated patients, and signals increased cardiovascular risk. On the other hand, much more common hypokalemia, affecting about 13% of chlorthalidone participants at year-1, was not associated with adverse cardiovascular consequences and likely represents chronic conditions associated with potassium loss and high mortality and transient conditions.
Diuretics have been demonstrated in multiple clinical trials to provide low-cost clinical benefit in the treatment of hypertension. Yet, many physicians have been reticent about prescribing diuretics due to concerns of diuretic-induced hypokalemia. Until now, little has been known about clinical ramifications of this hypokalemia. Several pertinent lessons can be learned from ALLHAT data. First, while the majority of the participants with hypokalemia at some time point during ALLHAT were assigned to the diuretic chlorthalidone, the hypokalemia seldom persisted throughout the study, likely due in large part to potassium supplementation--a need easily detected and therapy commonly prescribed in general medical practice. Critically, the appearance of hypokalemia in the diuretic group was not associated with increased cardiovascular outcomes; to the contrary, the risk of adverse cardiovascular outcomes including mortality among hypokalemic participants was lower in the diuretic arm than in either of the other two arms. Regardless of treatment group, participants with hyperkalemia fared worse than those with hypokalemia. Treating hypertension is fundamental, and treatment should often include a thiazide-type diuretic. Clinicians should feel reassured that hypokalemia associated with low-to-moderate dose diuretics (12.5–25 mg of chlorthalidone a day) affected less than 13% of patients and was easily remedied. Hyperkalemia, though infrequent, may present a more alarming cardiovascular risk, and deserves additional study.20 What is clear, however, is that the cardioprotective actions of diuretic use are unaffected by consequent, but treatable, alterations in serum potassium.
Severe drug-induced alterations of K+ affect a small minority of treated patients. Hyperkalemia, although infrequent, usually occurred among ACE-inhibitor treated patients, and signals increased cardiovascular risk. On the other hand, much more common hypokalemia, which usually occurred among diuretic-treated patients, was not associated with adverse cardiovascular consequences and likely represents chronic conditions associated with potassium loss and high mortality and transient conditions. ALLHAT data support the cardioprotective benefit of diuretics, given in low-to-moderate doses, in the treatment of hypertension, including among patients who develop incident hypokalemia.
Sources of Funding
This research was supported by contract number N01-HC-35130 from the National Heart, Lung, and Blood Institute. The ALLHAT investigators acknowledge contributions of study medications supplied by Pfizer Inc (amplodipine), AstraZeneca (atenolol and lisinopril), and Brister-Myers Squibb (pravastatin), and financial support by Pfizer, Inc.
Clinical Trial Registration: www.clinicaltrials.gov, NCT00000542
Dr. Alderman has received research grants from Sankyo. Dr. Calhoun has consulted for Eli Lilly and Novartis. Dr. Cushman has consulted for Daiichi Sankyo, Novartis, Noven, Sanofi Aventis, Takeda, and Theravance; has received honoraria from Bristol-Meyer Squibb, Daiichi Sankyo, Novartis, and Sanofi-Aventis; and has received research grants from GlaxoSmithKline, and Novartis. Dr. Davis has consulted for Amgen and Takeda. Dr. Ong has received honoraria from Novartis; and has received research grants from Amarin, Amylin, Daiichi Sankyo, GlaxoSmithKline, Johnson and Johnson, Luitpold, Novartis, Pfizer, Roche, Sanofi Aventis, Takeda, and XOMA. Dr. Oparil has consulted for Boehringer Ingelheim, Daiichi Sankyo, Eli Lilly, Forest Laboratories, Forest Pharmaceuticals, NicOx, Novartis, Omron Healthcare, Pfizer, and Schering Plough; has received research grants from Amgen, Daiichi Sankyo, Gilead, Merck, and Takeda. Dr. Papademetriou has received honoraria from Astra-Zeneca and Forest Pharmaceuticals. Dr. Probstfield has received research grants from Abbott Laboratories, Boehringer Ingelheim, GlaxoSmithKline, and Sanofi Aventis. Drs. Eckfeldt, Einhorn, Ford, Franklin, Furberg, and Piller have no financial interests to report.