The findings of the current study suggest that in ambulatory patients with chronic HF and CKD receiving ACE inhibitors and non-potassium-sparing diuretics, hypokalemia (<4 mEq/L) was common and was associated with increased mortality and hospitalizations. Further, we demonstrate that hypokalemia was mild (3.5–3.9 mEq/L) in most patients, and that even mild hypokalemia was associated with poor outcomes. Additionally, hypokalemia also increased risk of death in those with more advanced CKD (eGFR <45 ml/min/1.73 m2). To the best of our knowledge this is the first report of an association between hypokalemia and poor outcomes in propensity-matched cohorts of HF patients with CKD. The findings are important as both CKD and hypokalemia are highly prevalent in HF. While the presence of CKD increases the risk of hyperkalemia and associated complications, these findings demonstrate that underestimating the presence and the risk of hypokalemia in HF patients with CKD is also a concern.
There are several potential explanations for the associations between hypokalemia and poor outcomes in patients with chronic HF and CKD: confounding by imbalances in measured baseline characteristics, confounding by unmeasured baseline characteristics, and/or an intrinsic effect of low serum potassium. Bivariate associations between hypokalemia and poor outcomes may potentially be explained by residual bias. However, all measured baseline characteristics were well-balanced among our propensity-matched patients with normokalemia and hypokalemia. Therefore, hypokalemia-associated poor outcomes observed in our study may not be explained by imbalances in any of the measured baseline characteristics.
Confounding by an unmeasured baseline characteristic may also potentially explain the poor outcomes associated with hypokalemia. For example, we had no data on diuretic doses, which may be a potential confounder, as sicker HF patients were more likely to receive larger doses of diuretics and develop more severe hypokalemia. Diuretic use is associated with poor outcomes, which has been shown to be dose dependent.16, 28, 29
Although the prevalence of diuretic use was similar, it is possible that those with hypokalemia were using higher doses of diuretics. However, this is unlikely to explain away the observed associations as the findings from our sensitivity analysis suggest that these associations were robust and rather insensitive to the potential confounding effect of an unmeasured covariate. Further, the potential effect of an unmeasured confounder can also be indirectly assessed by examining balance on variables that might be strongly correlated with that unmeasured confounder.23
For example, NYHA class and symptoms and signs of fluid volume overload would be strongly correlated with the diuretic doses. However, in our study, these markers of higher diuretic doses were balanced after matching, suggesting that any confounding effect by diuretic dose would likely be minimal. Finally, the observation that the associations between hypokalemia and poor outcomes were observed at various degrees of hypokalemia and at various stages of CKD also highlights the robustness of those associations.
The notion that the associations between hypokalemia and poor outcomes may be intrinsic in nature is biologically plausible. Hypokalemia is known to enhance membrane excitability, increase cardiac automaticity, delay ventricular repolarization and predispose patients to reentrant arrhythmias.30–33
Hypokalemia-associated deaths have often been attributed to cardiac arrhythmias and sudden cardiac death. We have previously demonstrated that in HF patients with and without CKD, hypokalemia was associated with increased risk of death without an increase in hospitalization suggesting sudden death may have precluded hospitalization in those patients.1, 2
However, in the current analysis, we observed that hypokalemia was associated with both increased death and hospitalization, suggesting that the effect of hypokalemia in HF patients with CKD may be both sudden and non-sudden in nature. The progressive deleterious effects of hypokalemia in HF patients with CKD may also be mediated by aldosterone, which has been shown to cause myocardial fibrosis, diastolic dysfunction and disease progression in HF.33–36
Although the effect of hypokalemia in the setting of acute myocardial infarction is well known,37–39
little is known about the effect of hypokalemia in patients with chronic IHD. Although the prevalence of hypokalemia was lower in patients with IHD (, pre-match), the effects of hypokalemia were worse in those with IHD (), suggesting that infarcted/ischemic myocardium may provide a more suitable substrate for the adverse effects of hypokalemia.
An interesting observation of our study is that the prevalence of hypokalemia in patients with HF and CKD was high (19%) and similar to that in HF patients in general.1, 2
Among the 3739 patients without
CKD and with valid serum potassium (excluded
from the current analysis), only 18% had potassium <4 mEq/L (data not shown
). This is important as hyperkalemia is often considered a more common problem of potassium homeostasis in patients with CKD. However, findings from our study suggest that hypokalemia is common in patients with HF and CKD receiving ACE inhibitors and that even a mild reduction in serum potassium level (3.5–3.9 mEq/L) was associated with poor outcomes. These findings are important because patients with HF and CKD often require larger doses of diuretics increasing their risk of hypokalemia. Yet, hypokalemia in these patients is less likely to be treated for fear of causing hyperkalemia. Therefore, taken together with our prior reports and expert opinions, it may be suggested that serum potassium should be routinely monitored in HF patients with CKD and carefully maintained between 4 and 5 mEq/L.1, 2, 6, 9, 40
There are a few limitations of our study. The MDRD formula may underestimate GFR in patients with GFR >60 ml/min/1.73 m2
However, all patients in our analysis had eGFR <60 ml/min/1.73 m2
. Further, we were able to replicate our key findings in more advanced CKD patients. As previously mentioned, diuretic dose was not available. B-type natriuretic peptide (BNP) levels were also not available and could have provided further data on HF severity. Findings of our study are based on predominantly white men in normal sinus rhythm. Data on beta-blocker use was not collected in the DIG trial as these drugs were not approved for use in HF at that time. The transfer of potassium from plasma into cells is facilitated by stimulation of beta-2 receptors.42–44
Therefore, the prevalence of hypokalemia may be somewhat lower in patients receiving carvedilol and metoprolol extended-release, the two most commonly used beta-blockers in HF.45
However, the effect of hypokalemia on outcomes is unlikely to be substantially different from that observed in our study. Future studies may examine the effect of hypokalemia in contemporary HF patients with CKD.
In conclusion, in ambulatory patients with chronic HF and CKD, hypokalemia (<4 mEq/L) is common and associated with increased mortality and hospitalization. Further, hypokalemia in these patients is mostly mild (3.5–3.9 mEq/L) but even the mild hypokalemia is associated with poor outcomes. Serum potassium should be routinely monitored in HF patients with CKD, and should be carefully maintained between 4 and 5 mEq/L.