Hyperkalemia in Outpatients Using Angiotensin-Converting Enzyme Inhibitors:  How Much Should We Worry? FREE

ANGIOTENSIN-converting enzyme (ACE) inhibitors are widely used in the treatment of several common clinical conditions. It has been estimated that ACE inhibitors accounted for 24% of the antihypertensive agents prescribed in 1993.¹ In other conditions, such as congestive heart failure² and diabetic nephropathy,³ they have been shown to improve important clinical outcomes.

Hyperkalemia is a known complication of the use of ACE inhibitors. The incidence of hyperkalemia appears to be relatively low in patients with normal renal function (0%-6%)^4- 25 but becomes increasingly common in those with renal insufficiency (5%-50%).^26- 42 Life-threatening hyperkalemia during use of ACE inhibitors, although rare, has been reported.^43- 46 Consequently, physicians often carefully monitor potassium levels when ACE inhibitors are prescribed. However, there are little data to guide the intensity of such monitoring. In addition, when mild hyperkalemia is identified during the use of ACE inhibitors, little data exist to guide the clinician in deciding how to subsequently monitor patients or whether to continue the use of ACE inhibitors.

In this study, we describe the prevalence and risk factors for hyperkalemia in medical outpatients who were taking ACE inhibitors. In addition, we describe the subsequent clinical course of hyperkalemia to determine the frequency and risk factors of progression to severe hyperkalemia (potassium level ≥6.0 mmol/L) and mortality.

A case-control study design was used to determine risk factors for hyperkalemia. Case patients with hyperkalemia and normokalemic controls were identified from patients using ACE inhibitors who were enrolled in the internal medicine clinic of the Pittsburgh Veterans Affairs Medical Center (Pittsburgh, Pa) during a 2-year study period (1992-1993). Subjects were identified using the hospital's computerized administrative, laboratory, and pharmacy databases. Case patients were defined as patients receiving an ACE inhibitor who had a potassium level of 5.1 mmol/L or higher on the same day as their internal medicine clinic appointment. Hemolyzed specimens were excluded. Case patients having more than 1 occurrence of hyperkalemia were analyzed using the initial occurrence only. A control was identified for each case. Controls were defined as patients receiving an ACE inhibitor who had a potassium level of 5.0 mmol/L or lower on the same day as their internal medicine clinic appointment. Furthermore, their potassium levels were not higher than 5.0 mmol/L during the entire study period. Controls were matched within 1 month of the date of the index potassium determination of the case patient with hyperkalemia. The use of an ACE inhibitor was confirmed for all case patients and controls by chart review.

Each case patient with hyperkalemia was followed up for 1 year following the index episode of hyperkalemia to describe subsequent potassium levels and changes in therapy with ACE inhibitors. This retrospective cohort design was used to determine the frequency and risk factors of the development of severe hyperkalemia (potassium level ≥6.0 mmol/L) following initial identification of hyperkalemia.

Several variables were recorded as potential predictor variables for hyperkalemia. Laboratory values recorded included the most recent preceding values of serum creatinine, bicarbonate, glucose, and serum urea nitrogen. The most recent preceding values were chosen as candidate predictor variables to reflect factors that would be known to the clinician at the time that the potassium determination was ordered. Results of these same chemistries were recorded at the time of the index case (or control). Comorbidities at the time of determination of the index potassium levels were determined by chart review using explicit definitions adapted from those originally defined by Charlson et al.⁴⁷ The Charlson comorbidity index uses weights for common comordibities associated with mortality using the digit values of 1, 2, 3, and 6. A summary comorbidity index is created by adding the weights for individual comorbidites. A random sample of charts was reviewed by a second rater to determine interrater reliability, and an additional random sample was rated by the primary rater at a later date to determine intrarater reliability. The κ values for the reliability of chart abstraction for each of the comorbidities ranged from 0.6 to 1.0 for both samples. The use of prescription drugs at the time of the index potassium level was determined by chart review. Drugs were classified according to pharmacological class. The type and dosage of ACE inhibitor were also determined by chart review. Age was determined using administrative records.

The case patients with hyperkalemia were followed up for 1 year after determination of their index potassium levels as a retrospective cohort to describe their subsequent clinical course. Variables measured for case patients for 1 year following the index episode of hyperkalemia included changes in therapy with ACE inhibitors (increase or decrease in dosage, change in type of ACE inhibitor, or discontinuation of ACE inhibitor therapy) and maximum potassium values for each type of change. Vital status as of December 31, 1994, was determined for case patients and controls using the Department of Veterans Affairs centralized administrative database at Austin, Tex, as well as by review of local administrative records.

Univariate comparisons of potential predictor variables for hyperkalemia between case patients and controls were performed using the Student t, Wilcoxon rank sum, and McNemar tests as appropriate. Multivariate analysis was performed using stepwise logistic regression to determine independent risk factors for the development of hyperkalemia. Cutoff points for continuous variables were identified that maximally discriminated between status of case patients and controls. These cutoff points were used to create categorical variables for use in the logistic regression models. Univariate and multivariate analyses were similarly performed to identify factors associated with the development of severe hyperkalemia (potassium level ≥6.0 mmol/L) during the 1-year follow-up period. One-way analysis of variance was used to compare mean potassium values between the different types of change in therapy with ACE inhibitors during follow-up. Life-table analysis was used to determine whether hyperkalemia was associated with increased mortality. Cox proportional hazards models were used to determine factors associated with increased mortality among the case patients with hyperkalemia. Specifically, modeling was used to determine whether a level of hyperkalemia existed that was independently associated with increased mortality.

During the 2-year study period, 194 case patients (11%) with hyperkalemia were identified among the 1818 patients who had received prescriptions for ACE inhibitors. Since not all patients had potassium determinations, the 11% prevalence represents an underestimate. Of the 194 case patients with hyperkalemia, 37 had an index potassium level of 5.6 mmol/L or higher; 3 had index potassium levels of 6.0 mmol/L or higher.

Table 1 compares the clinical characteristics of case patients and controls at the time of the index potassium level determination. The laboratory data presented are the most recent values preceding the index potassium determination. The mean (±SD) number of days between the preceding laboratory values and the index potassium level determination was 192±191 for the case patients and 182±172 for the controls. Case patients with hyperkalemia had a significantly higher preceding mean serum urea nitrogen level than normokalemic controls (8.0 mmol/L vs 6.2 mmol/L [22 mg/dL vs 17 mg/dL]; P<.001). Similarly, case patients had a significantly higher preceding mean creatinine level than controls (120 µmol/L vs 103 µmol/L [1.3 mg/dL vs 1.1 mg/dL]; P<.001). Hypertension was highly and equally prevalent in both case-patient and control groups (88% vs 93%; P=.10). Diabetes was similarly prevalent in case patients and controls, although case patients were more likely to have "severe" diabetes (16% vs 6%; P=.007) as defined by the presence of end-organ complications (creatinine level >136 µmol/L [1.5 mg/dL] or proliferative retinopathy). The majority of subjects classified as having severe diabetes had renal insufficiency: 28 (90%) of 31 case patients and 7 (64%) of 11 controls with severe diabetes had a creatinine level of higher than 136 µmol/L (1.5 mg/dL). Congestive heart failure, peripheral vascular disease, and cerebrovascular disease were also significantly more common in the group with hyperkalemia. Among prescription drugs used at the time of the index potassium level determination, the use of potassium supplements was no different in case patients with hyperkalemia than in normokalemic controls (13% vs 19%; P=.10). Only 3 case patients with hyperkalemia and 1 normokalemic control were using potassium-sparing diuretic agents. There was no significant difference in the average age for the case patients compared with the controls (67.4 years vs 66.5 years; P=.82).

Table 2 compares the type and dosage of ACE inhibitors used for case patients and controls. Case patients with hyperkalemia were more likely to be using long-acting ACE inhibitors (lisinopril and enalapril) than normokalemic controls (91% vs 83%; P=.02). The vast majority of subjects using a long-acting ACE inhibitor were prescribed lisinopril. The mean daily doses of each type of ACE inhibitor did not differ significantly between case patients and controls.

Independent factors predicting hyperkalemia using logistic regression analysis are listed in Table 3. The factors identified are listed in order of entry into the model. An elevated serum urea nitrogen level and an elevated creatinine level were both identified as independent risk factors for hyperkalemia. A combined variable of the serum urea nitrogen and creatinine ratio did not prove to be a better predictor than the individual factors. Likewise, when creatinine was modeled using a single cutoff point of 136 µmol/L (1.5 mg/dL), the factors and their order of entry in the model did not change. Congestive heart failure was identified as an independent predictor of hyperkalemia. In addition, use of long-acting ACE inhibitors (lisinopril or enalapril) was independently associated with an increased risk of hyperkalemia. The use of thiazide and loop diuretic agents were each associated with a reduced risk of hyperkalemia in multivariate analysis.

Table 4 summarizes the changes in therapy with ACE inhibitors that occurred during the subsequent year among the case patients with hyperkalemia. Data for the entire follow-up period were available for 171 (88%) of the 194 case patients. For at least part of the follow-up period, 155 patients remained on a regimen of an ACE inhibitor. Of these, 146 had a follow-up potassium test while receiving ACE inhibitors. Fifteen (10%) of the 146 developed severe hyperkalemia (potassium level ≥6.0 mmol/L). No significant differences were observed in the mean maximum potassium levels for those whose ACE inhibitor dosage was increased, decreased, remained the same, or was discontinued. However, only 2 (4%) of 46 patients who had at least 1 potassium determination while not receiving ACE inhibitors during follow-up developed severe hyperkalemia.

Univariate predictors for the development of severe hyperkalemia in patients remaining on a regimen of ACE inhibitors are shown in Table 5. Both elevated serum urea nitrogen and creatinine levels were associated with an increased risk of developing severe hyperkalemia during follow-up. Other factors associated with an increased risk of severe hyperkalemia during follow-up included age greater than 70 years and a glucose level higher than 10.5 mmol/L (190 mg/dL). A serum bicarbonate level higher than 28 mmol/L was associated with a reduced risk of hyperkalemia. In multivariate analysis, a serum urea nitrogen level of more than 8.9 mmol/L (25 mg/dL) (odds ratio [OR], 4.5; 95% confidence interval [CI], 1.3-15) and age older than 70 years (OR, 5.4; 95% CI, 1.5-19) were independent predictors for the development of severe hyperkalemia during follow-up. A serum bicarbonate level higher than 28 mmol/L was associated with a reduced risk (OR, 0.2; 95% CI, 0.06-0.8) with multivariate analysis.

Twenty deaths occurred among the case patients with hyperkalemia and 20 deaths among the normokalemic controls during the follow-up period. No significant difference between the survival curves was observed during this period (P=.63).

Cox proportional hazards models were used to determine whether a level of hyperkalemia during the follow-up period was associated with increased mortality among the cohort of case patients with hyperkalemia. These models demonstrated that severe hyperkalemia during follow-up was associated with increased mortality, although a potassium level of 6.3 mmol/L or higher (P=.001) was more discriminant than our definition of severe hyperkalemia as 6.0 mmol/L or higher (P=.05). Other univariate predictors of increased mortality in the case patients with hyperkalemia included an index creatinine level higher than 136 µmol/L (1.5 mg/dL) (P=.002); index serum urea nitrogen level higher than 10 mmol/L (28 mg/dL) (P=.003); index potassium level higher than 5.3 mmol/L (P=.04); peripheral vascular disease (P=.02); pulmonary disease (P=.03); and use of digoxin (P=.006) (Table 6). A diagnosis of hypertension in the case patients with hyperkalemia was associated with reduced mortality (P=.05).

Hyperkalemia during the use of ACE inhibitors was relatively frequent among our medical outpatients. Our prevalence of 11% during a 2-year period is consistent with prior clinical studies, given the frequency of renal insufficiency in our study population.

An elevated serum urea nitrogen level, an elevated creatinine level, and congestive heart failure were strongly and independently associated with hyperkalemia. These factors, while expected, accounted for the majority of cases of hyperkalemia. The sensitivity of having any one of these factors for predicting hyperkalemia was 0.69 (133/194) and the specificity was 0.52 (101/194). An additional increased risk of hyperkalemia was observed with the use of long-acting ACE inhibitors compared with a short-acting ACE inhibitor. Reduced risk associations for hyperkalemia were observed with the use of thiazide and loop diuretic agents by multivariate analysis.

Prior studies^26- 42 have documented the increasing incidence of hyperkalemia associated with the use of ACE inhibitors in patients who have renal insufficiency. Additional factors reported to increase the risk of hyperkalemia during the use of ACE inhibitors have included the use of potassium-sparing diuretic agents^28,41 and potassium supplements.^28,48 In other settings, the use of nonsteroidal anti-inflammatory drugs⁴⁹ and a combination product of trimethoprim-sulfamethoxazole⁵⁰ has been associated with an increased incidence of hyperkalemia. In our patients, these agents were infrequent explanations for hyperkalemia. Indeed, the proportion of patients using potassium supplements and nonsteroidal anti-inflammatory drugs was lower in the case patients with hyperkalemia than in the normokalemic controls. No patients in the case or control groups were using trimethoprim-sulfamethoxazole at the time of their index potassium level determination. Congestive heart failure per se has not been reported as a risk factor for hyperkalemia while using ACE inhibitors. Instead, hyperkalemia during the use of ACEs for congestive heart failure has usually been attributed to concomitant use of potassium supplementation or potassium-sparing diuretic agents.^28,41

Few studies^51- 54 have compared the relative risk of hyperkalemia between different types of ACE inhibitors. Although tissue models have been described that support the potential for differing degrees of hyperkalemia as a result of differing effects on suppression of the renin-aldosterone system,⁵⁵ our observation of differing relative risk between the use of long-acting and short-acting ACE inhibitors is a cautious one. All ACE inhibitors are primarily renally excreted. The potential for higher and more sustained suppression of aldosterone levels may be greater with the use of ACE inhibitors having longer half-lives, particularly in the presence of impaired renal function. However, our case-control study design cannot reliably exclude whether bias in patient selection for receiving these agents may have predisposed patients receiving long-acting ACE inhibitors to hyperkalemia through other factors.

An important component of this study was the observation of patients following the initial episode of hyperkalemia. We are unaware of any data describing the clinical course of such patients to determine how to subsequently monitor them or whether to continue using ACE inhibitors. A large number of our patients remained on a regimen of an ACE inhibitor for at least part of the follow-up period. This study identified factors predicting severe hyperkalemia (potassium level ≥6.0 mmol/L) among patients remaining on a regimen of ACE inhibitors following an episode of modest hyperkalemia. The 2 independent factors identified, serum urea nitrogen level higher than 8.9 mmol/L (25 mg/dL) and age older than 70 years, proved to be excellent predictors of severe hyperkalemia within our group. The sensitivity of having either of these factors for predicting subsequent severe hyperkalemia was 87% (13/15). The specificity of these same factors was 43% (56/131). The negative predictive value of the factors was 97% (56/58). While the utility of these factors in predicting severe hyperkalemia needs to be validated in other populations, they provide guidance in responding to hyperkalemia occurring during the use of ACE inhibitors. The relative infrequency of hyperkalemia in patients with hypertension and normal renal function in previous studies supports these guidelines.

We were interested in whether a level of hyperkalemia independently predicted increased mortality among our cases with hyperkalemia. Multivariate analysis showed a potassium level of 6.3 mmol/L or higher to be independently associated with increased mortality. This association must be interpreted cautiously, however. We did not systematically ascertain the cause of death in our patients. Although higher levels of potassium may portend an increased predisposition to critical levels of hyperkalemia and resultant dysrhythmias, the observed association is likely to be less reflective of cause and effect than to serve as a marker of more severe comorbidity. Nonetheless, our observation supports our arbitrary definition of severe hyperkalemia at a level that would raise concern for most clinicians.

Dietary intake of potassium was not ascertained in our patients. It is possible that higher levels of potassium intake may increase the risk of hyperkalemia during the use of ACE inhibitors. Conversely, reduced intake may lower the risk of subsequent hyperkalemia. Although it is of clinical interest whether reduced intake of potassium is independently associated with a lower risk of hypekalemia, accurate assessment of potassium intake in this retrospective study was not possible. Nevertheless, we believe it unlikely that knowledge of potassium intake would have appreciably altered the risk factors for hyperkalemia observed in this study.

The limitations of this study require comment. The study sample included largely elderly men; extrapolation to other populations should be done cautiously. Nonetheless, we believe it unlikely that younger or female patients with hyperkalemia differ substantially. The case-control methodological procedure is inherently a weak study design to determine causality largely because of recall bias. As our chart abstraction methodological procedure relied on notes written prior to the index potassium determination, and was reliable as determined by duplicate abstraction, we believe recall bias was minimized.

In summary, the natural history of hyperkalemia occurring in the context of the use of ACE inhibitors is described. The risk factors for hyperkalemia observed in this study are consistent with prior clinical studies, although the use of other potassium-elevating drugs was relatively unusual in our patients. In the absence of renal insufficiency, azotemia, or congestive heart failure, hyperkalemia is unusual, and less vigilant monitoring for its development is likely to be safe. More important, a serum urea nitrogen level higher than 8.9 mmol/L (25 mg/dL) and age older than 70 years predicted the development of severe hyperkalemia (potassium level ≥ 6.0 mmol/L) following identification of modest hyperkalemia during the use of ACE inhibitors. In the absence of these factors, we observed only 2 patients with severe hyperkalemia. From these data, the following is suggested. Serum potassium levels should be measured in all patients after beginning therapy with ACE inhibitors. Should mild hyperkalemia develop while receiving ACE inhibitors, patients younger than 70 years with normal renal function can safely continue to use the drug since the frequency of severe hyperkalemia is low and frequent monitoring is not required. Those patients with either of the factors predicting severe hyperkalemia warrant either judicious surveillance of potassium levels or reconsideration of the use of ACE inhibitor.

Accepted for publication May 29, 1997.

In September 1997 Lawrence C. Reardon, MD, MPH, died at the age of 31 years and only a short time after starting his career as an academic general internist. The study presented herein was conducted during his general internal medicine fellowship. This study met one of the requirements for a Master's of Public Health degree that was awarded to Dr Reardon posthumously from the Graduate School of Public Health, University of Pittsburgh. Despite his brief career, he was known for his excellence in clinical care of patients and teaching ability. In his quiet way, Dr Reardon provided thoughtful, compassionate care to his patients. Never arrogant or insensitive, he was admired by his colleagues and patients for his quiet competence and equanimity. His dedication to teaching won him high acclaim from his students and a major teaching award. His thoughtfulness, work ethic, and compassion for his patients and colleagues will be missed by the medical community.

Reprints: David S. Macpherson, MD, MPH, Department of Medicine, University of Pittsburgh c/o Veterans Affairs Pittsburgh Healthcare System, University Drive C, Pittsburgh, PA 15240.