Author + information
- Received June 19, 2007
- Revision received July 26, 2007
- Accepted July 31, 2007
- Published online November 13, 2007.
- Akshay S. Desai, MD⁎,⁎ (, )
- Karl Swedberg, MD, PhD, FACC†,
- John J.V. McMurray, MD, FACC‡,
- Christopher B. Granger, MD, FACC§,
- Salim Yusuf, MD, Dphil, FACC∥,
- James B. Young, MD, FACC¶,
- Mark E. Dunlap, MD, FACC#,
- Scott D. Solomon, MD⁎,
- James W. Hainer, MD⁎⁎,
- Bertil Olofsson, PhD††,
- Eric L. Michelson, MD, FACC⁎⁎,
- Marc A. Pfeffer, MD, PhD, FACC⁎,
- CHARM Program Investigators
- ↵⁎Reprint requests and correspondence:
Dr. Akshay S. Desai, PBB-A3, AB370, Cardiovascular Division, Brigham and Women’s Hospital, 75 Francis Street, Boston, Massachusetts 02115.
Objectives We explored the incidence and predictors of hyperkalemia in a broad population of heart failure patients.
Background When used in optimal doses to treat patients with heart failure, renin-angiotensin-aldosterone system (RAAS) inhibitors improve clinical outcomes but can cause hyperkalemia.
Methods Participants in the CHARM (Candesartan in Heart Failure-Assessment of Reduction in Mortality and Morbidity) (n = 7,599) Program were randomized to standard heart failure therapy plus candesartan or placebo, titrated as tolerated to a target of 32 mg once daily with recommended monitoring of serum potassium and creatinine. We assessed the incidence and predictors of hyperkalemia associated with dose reduction, study drug discontinuation, hospitalization, or death over the median 3.2 years of follow-up.
Results Independent of treatment assignment, the risk of hyperkalemia increased with age ≥75 years, male gender, diabetes, creatinine ≥2.0 mg/dl, K+≥5.0 mmol/l, and background use of angiotensin-converting enzyme inhibitors or spironolactone. Candesartan increased the rate of aggregate hyperkalemia from 1.8% to 5.2% (difference 3.4%, p < 0.0001) and serious hyperkalemia (associated with death or hospitalization) from 1.1% to 1.8% (difference 0.7%, p < 0.001), with hyperkalemia associated with death reported in 2 (0.05%) candesartan patients and 1 (0.03%) placebo patient. The benefit of candesartan in reducing cardiovascular death or heart failure hospitalization (relative risk reduction 16%, p < 0.0001) was uniform in these subgroups, as was the incremental risk of hyperkalemia.
Conclusions The risk of hyperkalemia is increased in symptomatic heart failure patients with advanced age, male gender, baseline hyperkalemia, renal failure, diabetes, or combined RAAS blockade. Although these groups derive incremental clinical benefit from candesartan, careful surveillance of serum potassium and creatinine is particularly important.
When used in effective doses to treat heart failure (HF), inhibitors of the renin-angiotensin-aldosterone system (RAAS), including angiotensin-converting enzyme (ACE) inhibitors, angiotensin-receptor blockers, and mineralocorticoid receptor antagonists, reduce morbidity and mortality, but may raise serum potassium despite the common concomitant use of potassium-depleting diuretic therapies. Heart failure patients are particularly susceptible to hyperkalemia, since the reduction in renal function associated with HF, older age, and comorbidities such as diabetes mellitus hampers baseline potassium excretion. Moreover, optimal medical therapy of patients with symptomatic HF requires simultaneous utilization of multiple neurohormonal antagonists that alone and, especially, in combination increase the risk of hyperkalemia.
The CHARM (Candesartan in Heart Failure-Assessment of Reduction in Mortality and Morbidity) Program investigated the impact of treatment with the angiotensin receptor blocker (ARB) candesartan, both alone and in combination with an ACE inhibitor, across a broad spectrum of symptomatic HF patients, including those with both depressed and preserved left ventricular ejection fraction (LVEF) and those treated with various combinations of neurohumoral antagonists. Protocol-directed, serial monitoring of serum potassium and creatinine, as well as surveillance for serious adverse events (including hyperkalemia) permitted a quantitative assessment of risk in both placebo- and candesartan-assigned patients. We utilized the CHARM Program to investigate the incidence and predictors of hyperkalemia in a contemporary HF population managed with modern medical therapy.
The CHARM Program
The design and overall results of the CHARM Program have been previously reported in detail (1). Eligible patients were women and men age 18 years or older who had symptomatic HF (New York Heart Association functional class II to IV) for at least 4 weeks’ duration. Major exclusion criteria included baseline serum creatinine ≥3 mg/dl (265 μmol/l); baseline serum potassium ≥5.5 mmol/l (mEq/l); history of marked ACE inhibitor-induced hyperkalemia resulting in either a serum potassium ≥6.0 mmol/l or a life-threatening adverse event; known bilateral renal artery stenosis, symptomatic hypotension, critical aortic or mitral stenosis; recent (within 4 weeks) myocardial infarction, stroke, or open heart surgery; and use of an ARB within 2 weeks of trial enrollment. Eligible, consented patients were enrolled into 1 of 3 trials according to LVEF higher than 40% (CHARM-Preserved, n = 3,023), 40% or lower and treated with an ACE inhibitor (CHARM-Added, n = 2,548), or 40% or lower and not treated with an ACE inhibitor due to prior intolerance (CHARM-Alternative, n = 2,028) (2–4). In CHARM-Added, investigators were instructed to use “individualized optimum doses” of ACE inhibitors based on doses proven effective in clinical outcomes trials and individual tolerability (2). Angiotensin-converting enzyme inhibitor use was left to the discretion of the investigator in CHARM-Preserved. The primary outcome for each of the component trials was cardiovascular (CV) death or unplanned admission to the hospital for management of worsening HF, which was also used as the primary outcome for this supplementary, retrospective analysis. Whereas serum creatinine and potassium were measured on the full cohort of 7,599 patients enrolled in CHARM-Overall, core laboratory values were recorded only for the subset of 2,675 patients enrolled from North America.
Study drug was dosed according to an incremental, biweekly, forced titration scheme. At randomization, study drug was initiated at 4 mg once daily (or at 8 mg once daily at the investigator’s discretion). The dose was then doubled every 2 weeks as tolerated up to a maximum target dose of 32 mg candesartan (or matching placebo) once daily. Visits were scheduled at 2, 4, and 6 weeks; 6 months; and then every 4 months until study end. Investigators were instructed to assess serum creatinine and potassium before drug initiation, within 2 weeks of dose escalation or completion of dose titration, yearly thereafter, and at their discretion. There were no prespecified mandates regarding the management of hyperkalemia. Follow-up was conducted over a minimum of 2 years, and a maximum of 4 years with median duration of 3.2 years.
For this analysis, the diagnosis of hyperkalemia was based not on a prespecified laboratory threshold, but on investigator assessments of when clinically important increases in serum potassium were present. Adverse events (serious or nonserious) that led to a dose reduction or discontinuation of study drug treatment, including hyperkalemia, were routinely reported on the case report forms. Narrative summaries of all hospitalization and death events in the CHARM Program were examined using a free text search to identify those that were associated with hyperkalemia, without regard to attribution of causality to treatment by the investigators. A post-hoc composite outcome of “clinically important hyperkalemia” was then defined to identify all hyperkalemia events requiring medical intervention, including hyperkalemia requiring study drug dose reduction, study drug discontinuation, or hospitalization or hyperkalemia causing death. This composite outcome was designed to reflect a more all-encompassing “worst-case” scenario definition for hyperkalemia in the absence of an effective monitoring strategy, making the conservative assumption that all dose reductions or discontinuations would have become serious events (hospitalization or death) had they not occurred.
We determined the incidence of clinically important hyperkalemia for both placebo- and candesartan-treated patients in CHARM-Overall and in each of the component CHARM trials. Odds ratios (ORs) and 95% confidence intervals (CIs) comparing treatments, stratified by trial and by subgroups, were derived using the Mantel-Haenszel method. Incidence rates (per 1,000 patient-years) were also calculated to account for variable exposure duration across the different CHARM trials. The composite hyperkalemia outcome and the clinical efficacy outcome of CV death or HF hospitalization were examined in subgroups according to baseline demographic factors and clinical features previously reported to increase risk for hyperkalemia (5) (age ≥75 years, male gender, diabetes, concurrent ACE inhibitor or spironolactone use, baseline creatinine ≥2.0 mg/dl, and baseline potassium ≥5.0 mmol/l), using a test for heterogeneity to assess for interactions between subgroup and treatment assignment. As baseline values of creatinine and potassium were available from the case report forms only for patients enrolled in North America, the analysis for these subgroups was confined to the North American study population. The time to development of clinically important hyperkalemia (and serious hyperkalemia) was analyzed using Cox proportional hazards models and displayed in standard Kaplan-Meier plots. Finally, Cox models including age ≥75 years, male gender, diabetes, ACE inhibitor or spironolactone use, and treatment assignment were utilized to examine the multivariate predictors of clinically important hyperkalemia. Models were extended to include baseline potassium and renal function (measured as both serum creatinine and as glomerular filtration rate estimated by the Modification of Diet in Renal Disease Method [eGFR] ) in the subset of patients for whom laboratory data were available. All analyses were conducted using SAS statistical software (Cary, North Carolina).
The breakdown of adverse events associated with hyperkalemia reported for the overall CHARM Program is displayed in Table 1.Overall, clinically important hyperkalemia was noted in 1.8% of placebo-treated (6.6 events/1,000 patient-years) and 5.2% of candesartan-treated patients (18.5 events/1,000 patient-years) in CHARM-Overall, reflecting an excess absolute risk of 3.4% with candesartan therapy (11.9 events/1,000 patient-years). The excess of serious hyperkalemia events (hospitalization or death) was 0.7% (1.8% vs. 1.1% for placebo), reflecting a difference of 2.7 events/1,000 patient-years. Hyperkalemia associated with death was rare in the CHARM Program, and was reported for only 3 (0.04%) patients in the overall program, including 2 (0.05%) patients assigned to candesartan in the CHARM-Added trial, and 1 patient (0.03%) assigned to placebo in the CHARM-Alternative study.
The incidence of clinically important hyperkalemia (hereafter, hyperkalemia) according to treatment assigned in the composite CHARM Program and each of the component trials is summarized in Table 2.Hyperkalemia was most common for both placebo- and candesartan-treated patients in CHARM-Added (2.9% placebo, 8.4% candesartan) in which patients were treated concurrently with an ACE inhibitor (p < 0.001). Candesartan increased the risk of hyperkalemia relative to placebo in each of the CHARM component trials and in CHARM-Overall (OR 2.9, 95% CI 2.2 to 3.9), but there was no heterogeneity in the incremental risk of hyperkalemia with candesartan across the different trial populations (p for interaction by study = 0.95). Hyperkalemia was, however, more common in the low LVEF patients as a whole (CHARM-Added + CHARM-Alternative) than in those with preserved LVEF (CHARM-Preserved) (4.3% vs. 2.3%, p < 0.001).
Patient subgroups at high risk for hyperkalemia in CHARM-Overall
Independent of treatment assignment in the CHARM-Overall study, the risk of hyperkalemia increased with male gender, age ≥75 years, diabetes, background use of ACE inhibitors or spironolactone, baseline creatinine ≥2.0 mg/dl, and baseline potassium ≥5.0 mmol/l (Table 3).The greatest relative risk increase for hyperkalemia was seen in those with baseline renal insufficiency, whether or not they received candesartan (OR = 4.1, 95% CI 2.4 to 7.3 for creatinine ≥2.0 vs. <2.0 mg/dl). Serious hyperkalemia events (associated with hospitalization or death) were also more common in this group (OR = 3.5, 95% CI 1.5 to 7.9 for creatinine ≥2.0 vs. <2.0 mg/dl). Notably, the incidence of hyperkalemia in patients with creatinine ≥2 mg/dl receiving placebo (10%) was more than twice that of patients with creatinine <2 mg/dl receiving candesartan (4.9%).
The incremental risk of hyperkalemia with candesartan therapy (roughly 3-fold) over placebo was statistically uniform in each of the examined subgroups (p for interaction >0.10) (Fig. 1)excepting those on spironolactone, for whom the odds of hyperkalemia with candesartan appeared to be substantially higher (p for interaction = 0.04). In multivariable Cox models containing treatment assignment, age ≥75 years, gender, blood pressure, diabetes, ACE inhibitor use, spironolactone use, nonsteroidal anti-inflammatory drug use, ejection fraction ≤40%, beta-blocker use, and loop diuretic use, assignment to candesartan, male gender, age ≥75 years, diabetes, and background use of ACE inhibitors or spironolactone remained clinically important predictors of hyperkalemia. The multivariable hazard ratios for hyperkalemia in the full CHARM cohort are outlined in Table 4.When the model was extended to include baseline creatinine and potassium for the subset of 2,675 patients in whom laboratory data were available, the effects of gender and diabetes were attenuated, but renal function, baseline potassium, age ≥75 years, baseline use of ACE inhibitors or spironolactone, and assignment to candesartan were still relevant predictors. Of note, substituting eGFR (in quintiles) for creatinine as a measure of renal function improved the overall model fit and accounted for age, but did not change the other relevant multivariable associations. Independent of treatment assignment, the risk of hyperkalemia in the CHARM Program steadily increased across eGFR quintiles with the greatest risk seen in the group with eGFR <45 cc/min (corresponding roughly to a creatinine of ≥2.0 mg/dl).
Despite increased risk, candesartan therapy was associated with a consistent, highly significant (p < 0.0001) 16% reduction in the primary end point of CV death or hospitalization for the management of HF across all subgroups (Fig. 2).Although the increased incidence of hyperkalemia related to candesartan appeared to be higher during the first 3 to 6 months (while drug titration was occurring and during which serum potassium values were being checked more frequently), hyperkalemia events (including serious events associated with hospitalization or death) accrued over the entire study period (Fig. 3).The overall, unadjusted hazard ratio for hyperkalemia with candesartan therapy was 2.8 (95% CI 2.1 to 3.7, p < 0.0001), and the hazard ratio for serious hyperkalemia events (hospitalization or death) was 1.7 (95% CI 1.2 to 2.5, p = 0.006).
Our analysis of the CHARM Program demonstrates that clinically important hyperkalemia occurred in a broad spectrum of symptomatic HF patients, whether or not they were treated with candesartan. Hyperkalemia events accrued throughout the study period, not merely during the dose titration phase of the study. Relative to placebo, candesartan therapy was associated with 34 excess hyperkalemia events per 1,000 patients over a median 3.2 years of follow-up (11.9 events/1,000 patient-years). By comparison, candesartan treatment was associated with 43 fewer CV death or HF hospitalization events per 1,000 patients (23 events/1,000 patient-years) in the CHARM-Overall study, as well as numerical reductions in all-cause (approximately 16 fewer events/1,000 patients) and CV mortality (approximately 21 fewer events/1,000 patients) (1). Due to an effective, protocol-directed monitoring strategy and the conduct of this study at sites committed to the care of patients in the context of a clinical trial, many hyperkalemia events prompted a dose change or study drug discontinuation without important consequence for the patient. Serious hyperkalemia events, however, remained more common in candesartan-treated patients, with an excess of 7 hyperkalemia-associated hospitalization or death events per 1,000 patients (2.7 events/1,000 patient-years) over the follow-up period and an apparent excess of only 1 hyperkalemia-associated death in the 3,803 patients treated with candesartan.
Subgroups at highest risk for hyperkalemia in the CHARM Program were those with advanced age, diabetes, male gender, high potassium at baseline, renal dysfunction (identified by creatinine ≥2 mg/dl or eGFR <45 cc/min/1.73 m2), and those receiving background therapy with ACE inhibitors or spironolactone. After adjusting for baseline potassium and renal function in multivariable models (for the subset of patients in whom data were available), the addition of candesartan to standard medical therapy for HF was associated with 2- to 3-fold increase in risk that was further amplified by cotreatment with ACE inhibitors or spironolactone. Combination RAAS blockade, therefore, should provoke special concern for hyperkalemia among clinicians, especially on the backdrop of renal dysfunction.
The ability of the kidney to maintain potassium homeostasis is critically dependent on adequate renal perfusion and sodium delivery to the distal nephron, normal aldosterone production, and normal function of aldosterone-sensitive potassium channels in the cortical collecting duct. Increases in dietary potassium load (e.g., through potassium supplementation), reduction in the filtered sodium load (as a consequence of intrinsic renal disease, nephrotoxins, hypovolemia, or reduced cardiac output), diminished aldosterone production (as a result of old age, diabetes, or drugs that inhibit renin release), or attenuated tissue responsiveness to aldosterone or decreased skeletal muscle uptake of potassium (as a consequence of aldosterone-blockade or beta-2 receptor blockade) may therefore precipitate hyperkalemia in the vulnerable patient. Since aldosterone production is decreased in the elderly, patients with diabetes, and in patients receiving drugs that block the production or action of renin (nonsteroidal anti-inflammatory drugs, renin inhibitors including beta-blockers) and angiotensin II (ACE inhibitors and ARBs), these groups are particularly vulnerable to the development of hyperkalemia with additional perturbations of potassium homeostasis. Progressive decline in glomerular filtration rate as a consequence of age-related nephron loss and comorbid medical illness further enhances the risk of hyperkalemia.
Hyperkalemia is a well-known complication of medical therapy for symptomatic HF (7). When used in optimal doses, several inhibitors of the RAAS reduce morbidity and mortality in this population, but also reduce potassium excretion, especially in those with pre-existing renal dysfunction (8). Fortunately, increases in serum potassium are typically modest (<1 mEq/l), and severe or life-threatening hyperkalemia is uncommon (9). Increasingly, however, optimal medical therapy for patients with HF due to systolic dysfunction relies heavily on simultaneous treatment with multiple neurohormonal antagonists that collectively amplify the risk of significant hyperkalemia when used in effective doses, particularly in patients with diabetes or underlying renal disease.
Enhanced prescription of spironolactone, an aldosterone receptor antagonist, spurred by the results of the landmark RALES (Randomized Aldactone Treatment Evaluation Study) (10), was associated with a dramatic increase in the rates of hospitalization for hyperkalemia in a recently published large, community-based sample (11). This experience highlights the potential difference between the carefully controlled environment of a clinical trial and the broader application of new therapies to a “real world” population (12). Although an excess of serious hyperkalemia events was observed in 2.7 per 1,000 patient-years in CHARM, the number may have been as high as 11.9 per 1,000 patient-years had an effective monitoring strategy not been in place. However, even this more all-encompassing “worst-case scenario” of hyperkalemia in the CHARM Program likely underestimates the incidence in an unselected population of patients with HF in the community who may be more vulnerable due to age, comorbid medical illness (particularly renal failure), and less careful monitoring than that offered in the context of a clinical trial. Buttressing this concern, it has been observed that monitoring for hyperkalemia in patients treated with RAAS blockade remains poor, with nearly 30% to 60% of patients failing to receive recommended surveillance of serum potassium (13,14). Nonetheless, the increased risk of hyperkalemia associated with candesartan (and, presumably, other ARB) therapy in HF must be interpreted in the context of its clear benefits with regard to CV morbidity and mortality, even among populations at special risk for elevated potassium (15,16), and as well in the context of potentially reduced serious adverse events related to hypokalemia (0.4% in candesartan-treated patients versus 0.9% in placebo-treated patients in the CHARM Program).
Our analysis must be viewed with attention to its limitations. This was a retrospective analysis using a post-hoc definition of “clinically important hyperkalemia.” The decision to report an adverse event of hyperkalemia and the management of hyperkalemic patients were left to the discretion of the individual investigators; as such, the actual serum potassium values and clinical context (such as electrocardiographic manifestations) at the time of adverse events were not available for analysis. Although guidelines for monitoring of serum potassium were provided to the CHARM investigators (particularly during dose titration), no regularly scheduled surveillance of serum potassium was mandated. It is possible therefore, that physicians may have intervened to manage potassium levels that would have been well tolerated by the patients, or alternatively, that hyperkalemia may have gone undetected because it was not associated with specific electrocardiographic or clinical manifestations. Finally, though hyperkalemia prompting study drug dose reduction or discontinuation was specifically reported in the case report forms for the trial, hospitalization or death associated with hyperkalemia was only identifiable from review of investigator-reported adverse events. Although we may as a consequence have included some hospitalizations that were not specifically due to hyperkalemia, we believe that our conservative assumption reflects the upper bound of the incidence of clinically important hyperkalemia in the CHARM study.
Additional studies of hyperkalemia in large community cohorts or registries are necessary to more accurately measure the incidence outside the context of a clinical trial. In the absence of such data, however, our data from the CHARM Program provide important reference points for practicing clinicians to gauge the risk for hyperkalemia in the management of patients with HF. Reassuringly, many hyperkalemia events in the CHARM Program could be adequately addressed by dose adjustment or discontinuation of the study drug, with the minority progressing to hospitalization or death. We conclude that periodic surveillance of serum potassium and creatinine (e.g., at baseline, within 1 to 2 weeks of a change in drug dosing, and at least annually thereafter) is critically important to limiting adverse events. The subset of older HF patients who have moderate or severe renal dysfunction (stage III chronic kidney disease or worse, many of whom were not eligible for the CHARM Program), those with high serum potassium at baseline, and those who receive combination RAAS antagonists are particularly vulnerable, necessitating even closer monitoring. As well, in all patients with HF, strategies to limit the toxicity of RAAS antagonists, such as restriction of potassium supplements (and in some cases, dietary potassium), elimination of drugs that may impair renal potassium excretion (e.g., nonsteroidal anti-inflammatory drugs, herbal remedies, licorice), and careful assessment of baseline renal function to identify those at especially high risk (e.g., eGFR <45 cc/min/1.73 m2), should be routinely considered (17). The addition of candesartan to standard HF therapy should be undertaken cautiously, if at all, where adequate monitoring of potassium and renal function is not possible. Although candesartan provides important incremental benefits with regard to CV morbidity and mortality over standard medical therapy for HF, a favorable balance of benefit and risk requires clinical vigilance and regular laboratory surveillance, particularly among patients at high baseline risk for hyperkalemia.
The authors wish to acknowledge Elizabet Hemlin, AstraZeneca, for her assistance in abstracting the data.
The CHARM Program was funded by AstraZeneca. Drs. Pfeffer, Swedberg, McMurray, Yusuf, Granger, Young, Solomon, and Dunlap have served as consultants to or received research grants and honoraria from AstraZeneca and/or other major pharmaceutical companies. Drs. Michelson, Hainer, and Olofsson are employees of AstraZeneca.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- angiotensin receptor blocker
- confidence interval
- glomerular filtration rate estimated by the Modification of Diet in Renal Disease method
- heart failure
- left ventricular ejection fraction
- odds ratio
- renin-angiotensin-aldosterone system
- Received June 19, 2007.
- Revision received July 26, 2007.
- Accepted July 31, 2007.
- American College of Cardiology Foundation
- Granger C.B.,
- McMurray J.J.,
- Yusuf S.,
- et al.,
- CHARM Investigators and Committees
- Schoolwerth A.C.,
- Sica D.A.,
- Ballerman B.J.,
- Wilcox C.S.
- Shah K.B.,
- Rao K.,
- Sawyer R.
- Hillege H.L.,
- Nitsch D.,
- Pfeffer M.A.,
- et al.
- McMurray J.J.,
- Young J.B.,
- Dunlap M.E.,
- et al.