Author + information
- Received February 2, 2005
- Revision received May 11, 2005
- Accepted May 15, 2005
- Published online September 6, 2005.
- Keyur B. Shah, MD⁎,
- Krishnamurti Rao, BS†,
- Robert Sawyer, MD‡ and
- Stephen S. Gottlieb, MD, FACC§,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Stephen S. Gottlieb, University of Maryland Hospital, Division of Cardiology, 22 South Greene Street, Baltimore, Maryland 21201.
Objectives This study was designed to determine the adequacy of monitoring patients receiving spironolactone as well as spironolactone’s relationship to hyperkalemia.
Background After the Randomized Aldactone Evaluation Study (RALES) demonstrated a 30% mortality benefit for treating severe heart failure patients with spironolactone, acceptance of this drug was overwhelming. Hyperkalemia and worsening renal function were rare in RALES, but laboratory monitoring was frequent. In clinical practice, the incidence of hyperkalemia and worsening renal function and adequacy of follow-up is unknown.
Methods We reviewed the monitoring of congestive heart failure (CHF) patients with spironolactone initiation after publication of RALES. All potassium and creatinine determinations at baseline and within three months following therapy initiation were assessed. Increased potassium was defined as any [K] ≥5.5 mEq/l and severe hyperkalemia as any [K] ≥6.0.
Results A total of 840 patients had new prescriptions for spironolactone. Of these, 91% had baseline laboratory values, and 34% did not have any serum potassium or creatinine determined within three months. Patients seen in the cardiology clinic were more likely to have appropriate follow-up (p ≤ 0.001). Of 551 patients with follow-up laboratory values determined, 15% developed hyperkalemia and 6% developed severe hyperkalemia. Fifty-one patients (9%) developed renal dysfunction, of whom 25 developed hyperkalemia within three months. Hyperkalemia developed in 48 of 138 (35%) patients with baseline creatinine ≥1.5 mg/dl and 12 of 19 (63%) with baseline creatinine ≥2.5 mg/dl.
Conclusions Many patients treated with spironolactone for CHF do not receive needed follow-up of potassium or creatinine concentrations, although hyperkalemia and renal dysfunction are common. Elevated baseline creatinine predicts patients at high risk. Physician education of the risks of spironolactone and the need for follow-up is essential.
In 1999, the Randomized Aldactone Evaluation Study (RALES) demonstrated a 30% decrease in mortality for treating congestive heart failure (CHF) patients with spironolactone (1). The benefit was seen in patients already receiving background therapy with angiotensin-converting enzyme (ACE) inhibitors. Previously used in heart failure to promote diuresis, spironolactone was readily accepted as an inexpensive agent to treat severe systolic dysfunction (2,3). However, the potassium-sparing effects of spironolactone pose a great risk for retaining potassium and subsequent fatal arrhythmias, especially in patients taking other medications affecting the renin-angiotensin-aldosterone system. Indeed, Juurlink et al. (4) observed that following publication of the RALES trial there was a marked increase in prescriptions for spironolactone, as well as rates of hyperkalemia-associated mortality and morbidity greater than those observed before the RALES trial.
There are many reasons that could explain more safety problems with spironolactone use in clinical practice than in controlled trials. For example, inappropriate patient selection based on systolic function, New York Heart Association heart failure classification, background medication, and underlying renal disease has been reported in a previous retrospective study (5). We hypothesized that inadequate monitoring of serum potassium concentrations and renal function might also contribute to an increased incidence of hyperkalemia after initiation of spironolactone. We reviewed the monitoring patterns for hyperkalemia and renal dysfunction in CHF patients receiving spironolactone after publication of the RALES trial.
This was a retrospective study using the Veterans Affairs Information System Technology and Architecture (VISTA) database. We reviewed in-patient and outpatient electronic records for patients with CHF who received outpatient prescriptions for spironolactone at the following Maryland VA hospitals or clinics: Baltimore, Fort Howard, Cambridge, Perry Point, and Glen Burnie.
Identification of data
We extracted patient information and records from VISTA or through the Computerized Patient Record System (CPRS) interface program. Health care providers at the VA hospital centers use VISTA and CPRS to access and update medical records for patients. The information in the database is available to health care providers when prescribing spironolactone. Any prescriptions written at the VA medical center are entered into the database before being dispensed from the pharmacy.
We identified a cohort with an in-patient or outpatient diagnosis of heart failure (International Classification of Diseases 428.x) in VISTA. Consecutive patients who received their initial prescription for spironolactone between September 2, 1999, and April 1, 2004, were included. The data collection for follow-up laboratory data was through July 7, 2004, to allow evaluation of a three-month follow-up period. A total of 898 patients fit these criteria.
We defined background medications to include prescriptions providing medications up to 30 days before or after prescribing spironolactone. We determined cardiovascular medication prescriptions, including ACE inhibitors, angiotensin receptor blockers, beta-adrenergic antagonists, loop diuretics, potassium supplements, and cardiac glycosides.
We obtained ejection fraction data from echocardiogram or multiple gated acquisition scan reports. For values reported in closed-end ranges, we calculated a mean ejection fraction. For values reported as open-end ranges, we used defined lower or upper limits in our calculations.
The database provided in-patient and outpatient lab data that could be used to analyze monitoring patterns. The most recent serum potassium and creatinine values, up to one year before prescribing spironolactone, defined baseline values. Peak serum creatinine and potassium concentrations were obtained for the first three months after spironolactone was dispensed. Identical to the RALES trial, serum potassium concentrations ≥5.5 mEq/l defined hyperkalemia, and levels ≥6.0 mEq/l defined severe hyperkalemia. Rising serum creatinine concentrations from baseline to a value ≥2.5 mg/dl defined renal dysfunction. We reviewed in-patient and outpatient electronic charts for patients without follow-up laboratory data in the three months after initiating therapy.
The University of Maryland Medical Center Institutional Review Board and the Veterans Affairs Research and Development Committee reviewed and approved this project.
The data are expressed as mean ± SD. A two-tailed Student ttest was used to compare continuous variables. Chi-square analysis was used to compare discrete variables. The package used was SPSS for Windows, version 11.5 (SPSS Inc., Chicago, Illinois).
We identified 898 patients with a diagnosis of CHF and a new prescription for spironolactone. Thirty-four patients were excluded from the monitoring analysis because they discontinued care at the study institutions during the three-month follow-up period. Twenty-four patients were excluded because of death within three months of starting spironolactone. A total of 840 patients were included in the monitoring analysis.
We found that 556 patients (66%) had serum potassium and creatinine values monitored within three months following initiation of spironolactone (Fig. 1).Five of these patients had values obtained outside of the VA system, with results noted by the prescribing VA physician. Thus, data from 551 patients were available for determination of hyperkalemia frequency. Of the 840 patients, 284 (34%) did not have any follow-up laboratory data. Of the 284 patients, the prescribing physician did not order follow-up laboratory data for 149, 41 failed to follow up for scheduled laboratory draws, and 94 had prescriptions filled at VA while also under the care of outside physicians. There is no evidence that the prescribing VA physician was aware of any laboratory values obtained elsewhere.
Patients with serum potassium concentrations measured within three months of initiating spironolactone were more likely to be receiving the following cardiovascular medications: ACE inhibitors or angiotensin receptor blockers, beta-adrenergic receptor blockers, and digoxin (all p < 0.05). These patients were younger (69 ± 11 years vs. 71 ± 11 years, p < 0.05) and had lower baseline serum potassium concentrations (4.3 ± 0.5 mEq/l vs. 4.4 ± 0.5 mEq/l, p < 0.05) (Table 1).
Of 551 patients with follow-up data, 83 patients (15%) developed hyperkalemia ([K] ≥5.5 mEq/l); 31 (6%) were classified as severe hyperkalemia ([K] ≥6.0 mEq/l). Patients that developed hyperkalemia had higher baseline potassium (4.7 ± 0.7 mEq/l vs. 4.3 ± 0.5 mEq/l, p < 0.001) and serum creatinine concentrations (1.8 ± 0.9 mg/dl vs. 1.2 ± 0.5 mg/dl, p < 0.001) (Table 2).
Fifty-one patients (9%) developed renal dysfunction; 25 of these patients (49%) also developed hyperkalemia within three months of initiating spironolactone. The incidence of hyperkalemia increased with increasing baseline serum creatinine (Fig. 2):hyperkalemia occurred in 35% (48 of 138) of patients with a baseline serum creatinine ≥1.5 mg/dl, 46% (22 of 48) if ≥2.0 mg/dl, and 63% (12 of 19) if ≥2.5 mg/dl (as compared with 9%, 13%, and 14% if the baseline creatinine was <1.5, 2.0, and 2.5 mg/dl, respectively [all p < 0.001]). The mean follow-up serum potassium concentration was 5.2 ± 0.8 mEq/l if the baseline serum creatinine concentration was ≥1.5 mg/dl and 4.7 ± 0.7 mEq/l if creatinine <1.5 mg/dl (p < 0.001); 5.5 ± 0.9 mEq/l if creatinine ≥2.0 mg/dl and 4.7 ± 0.7 mEq/l if creatinine <2.0 mg/dl (p < 0.001); and 5.8 ± 0.9 mEq/l if creatinine ≥2.5 mg/dl and 4.8 ± 0.7 mEq/l if creatinine <2.5 mg/dl (p < 0.001).
This study demonstrates marked deviation of monitoring for hyperkalemia in clinical practice as compared with that performed in clinical trials. This may explain the increased mortality secondary to hyperkalemia observed since publication of the RALES trial. In the RALES trial, patients initially treated with spironolactone had serum potassium and creatinine concentrations measured at 4, 8, and 12 weeks, and then every 3 months for up to 1 year. We observed in this study that only two-thirds of the patients had serum laboratory values obtained in the first three months after initiating spironolactone. The remaining 34% were prescribed spironolactone by a physician without any follow-up monitoring. Although a portion (14%) of this group failed to present for lab draws, a larger percentage (53%) were never scheduled for follow-up. Thirty-three percent were prescribed spironolactone by a doctor who knew the patient was also being followed outside of the VA medical system and may have assumed that appropriate monitoring was taking place. However, with no documentation of knowledge of laboratory values, it must be assumed that the prescription occurred without assurance of safety.
Inadequate monitoring for side effects when prescribing outpatient cardiovascular medication is not a unique occurrence. Physicians have previously exhibited variable (and inadequate) monitoring patterns with other medications, including antiarrhythmic medications (6). However, we are not aware of any reports of the frequency of monitoring for basic electrolytes when spironolactone is initiated.
Patients with follow-up were more likely to be receiving medications proven to be of benefit in patients with heart failure: ACE inhibitors or angiotensin receptor blockers, beta-blockers, and digoxin. This suggests that patients with close laboratory follow-up also had more appropriately treated heart failure. Although it is possible that doctors were more concerned that patients receiving ACE inhibitors or digoxin might develop hyperkalemia, the higher frequency of use of beta-blockers appears to reflect better care in patients receiving more frequent laboratory assessments.
Although qualities inherent to a retrospective design limit the accuracy of our observations, the results remain alarming. These data illustrate the difficult translation of controlled clinical trials into clinical practice. In the RALES trial, study participants were closely followed with frequent lab work and medication dosage adjustments. We did not observe vigilant monitoring in the clinical environment. Juurlink et al. (4) described a disturbing rise in incidences of fatal and non-fatal hyperkalemia since publication of the RALES trial, which we believe is due partly to relaxed outpatient surveillance of potassium concentrations.
Of 551 patients with three-month follow-up lab data, 15% developed hyperkalemia and 6% developed severe hyperkalemia. In contrast, the RALES trial reported severe hyperkalemia following spironolactone initiation in only 2% of patients. Several factors may contribute to the observed high rate of hyperkalemia. Most importantly, investigators in the RALES trial enforced a strict surveillance regimen over the cohort with appropriate medication adjustment when rising potassium levels were observed. Thus, potential episodes of severe hyperkalemia may have been precluded by early intervention. In addition, frequent physician visits allow management of related medical issues, including volume status that may contribute to the overall health of the individual. As previously described, we did not observe such strict monitoring in clinical practice.
The exclusion criteria used in the RALES trial may have also limited the prevalence of hyperkalemia. These criteria are not applied in clinical practice. Of the patients we studied, 105 would have been excluded from therapy in the RALES trial because of baseline serum creatinine concentrations of 2.5 mg/dl or greater and/or a baseline serum potassium concentration of 5.0 mEq/l or greater. Predictably, approximately one-half (28 of 60) of these individuals developed hyperkalemia. Also, patients receiving spironolactone in clinical practice tend to be older than those in the RALES trial. The mean age in our population was 70 ± 11 years as compared to 65 ± 12 years in the RALES trial. Older patients will have lower glomerular filtration rates and, therefore, an increased predisposition for potassium retention.
When compared with patients enrolled in the RALES trial, more patients in the present study were receiving beta-blocker therapy (68% vs. 11%). Beta-adrenergic receptor antagonism can suppress the renin-angiotensin-aldosterone system by inhibiting renin secretion from the juxtaglomerular apparatus, predisposing patients to potassium retention. This may explain the high rate of severe hyperkalemia (5.5% at one year) in the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (EPHESUS), a large trial showing the benefits of the selective mineralocorticoid receptor antagonist eplerenone in post-myocardial infarction patients with background therapy that included beta-blockers (7).
As a safety measure, researchers excluded from the RALES trial individuals with a serum creatinine concentration >2.5 mg/dl. The present study supports the limited use of spironolactone in patients with increased renal impairment. Patients with higher baseline serum creatinine concentration had higher rates of hyperkalemia and greater mean follow-up potassium concentrations.
Of great concern is our finding that a baseline serum creatinine concentration ranging as low as 1.5 to 1.9 mg/dl can still predict a high risk (35%) of developing hyperkalemia. After excluding patients who would not have qualified for therapy in the RALES trial, we found that individuals with serum creatinine concentrations between 1.5 and 2.5 mg/dl have a 30% rate of hyperkalemia, compared with 9% in patients with baseline creatinine concentrations below 1.5 mg/dl. These results suggest that patients with renal dysfunction (measured by elevated serum creatinine concentration) are at higher risks of developing hyperkalemia and require vigilant monitoring.
Because the present study is a retrospective chart review, there are factors that might influence the findings. Some patients may have discontinued medications without appropriate physician documentation in the database and not required further monitoring for elevated serum potassium concentrations. Conversely, inclusion of in-patient labs may overestimate the outpatient physicians’ effort to monitor serum potassium concentrations.
It is also possible that outside physicians were monitoring the patients. However, if the prescribing physician did not know these data, the risk to the patient would still be high. We believe it unlikely that many prescribing physicians would know outside laboratory values and not record this information.
Considering that we accessed only records of dispensed medications, without actual follow-up evaluating medication adherence, it is possible that some cases of hyperkalemia are falsely associated with spironolactone. Of greater concern is that poor medication adherence or discontinuation of spironolactone without documentation has erroneously minimized our findings of the number of medicated patients who develop hyperkalemia; the percentage of patients taking spironolactone that develop hyperkalemia may actually be higher than what is calculated in this study. Our exclusion of patients who died in the three-month follow-up period removes the “high-risk” individuals from analysis and may also lead to an underestimate of the risks of spironolactone.
We do not believe that the observations in this study concerning safety are limited to the VA hospitals, as recent clinical models describe the quality of care within the VA health care system as exceeding that of Medicare fee-for-service beneficiaries (8). Furthermore, most physicians at the Baltimore VA are staff physicians at the adjoining University of Maryland Hospital; thus, the behavior of physicians observed in this study likely reflects that of other institutions and is not specific for the VA. Indeed, the excellent database of the VA system not only permits quality improvement analyses such as the present report, but can also make interventions possible to ensure that physicians perform required monitoring of patients who receive high-risk medications. Electronic reminders or mandatory monitoring with prescriptions could be considered. Our findings point to the likelihood that the real-world environment adversely shifts the risk-benefit of spironolactone compared with that observed in the RALES trial. It is probable that by improving monitoring, the results of therapy would more closely approximate the outcome of the trial.
Although mineralocorticoid receptor antagonism has been shown to decrease cardiac remodeling and, in turn, improve cardiac function and mortality, spironolactone use produces a higher risk for hyperkalemia than the RALES trial predicts. Hyperkalemia after initiating spironolactone is a common occurrence, and surveillance by medical professionals is poor. In addition to appropriate patient selection, limiting hyperkalemia and potential deadly arrhythmias requires close monitoring after initiating therapy. Patients with increased renal impairment, even if only minor, require especially attentive monitoring and reliable follow-up. Systems should be implemented to ensure that patients receiving spironolactone are carefully selected and closely monitored.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- congestive heart failure
- Computerized Patient Record System
- Randomized Aldactone Evaluation Study
- Veterans Affairs Information System Technology and Architecture
- Received February 2, 2005.
- Revision received May 11, 2005.
- Accepted May 15, 2005.
- American College of Cardiology Foundation
- van Vliet A.A.,
- Donker A.M.,
- Nauta J.P.,
- Verheugt F.W.
- Juurlink D.N.,
- Mamdani M.M.,
- Lee D.S.,
- et al.
- Bozkurt B.,
- Agoston I.,
- Knowlton A.A.