Discussion

In this large, pooled analysis of 12 HF trials, we found that the association between serum potassium levels and clinical outcomes differed notably between HFrEF and HFpEF. In HFrEF, there was a reverse J-shaped association between potassium concentration and adverse clinical outcomes, with a strong association between hypokalaemia and higher rates of death and HF hospitalization, whereas potassium concentrations up to 5.5 mmol/L were not associated with worse outcomes after multivariable adjustment. The lowest event rates in people with HFrEF were observed over the potassium range 4.2–5.0 mmol/L. In HFpEF, event rates were lower than in HFrEF for any given potassium concentration, and the associations between potassium concentration and outcomes were much flatter, with a shallow U-shaped relationship observed. The lowest event rates in HFpEF were observed over the potassium range 4.2–5.1 mmol/L (Structured Graphical Abstract).

These results have several clinical implications. Concerning HFrEF, although our findings are observational in nature and do not necessarily represent cause and effect, they are robust, with the differential risks related to lower and higher potassium levels persisting through extensive adjustment for other prognostic variables and use of treatments, including diuretics (and dose of diuretics). This reduces the likelihood that the observed risk related to low potassium level was solely due to confounding, and our findings reinforce prior observations in other smaller HFrEF datasets with less complete multivariable adjustment. The lack of association between potassium concentration and non-cardiovascular death also argues against a non-specific finding. More importantly, a large trial, POTCAST (Targeted Potassium Levels to Decrease Arrhythmia Burden in High-Risk Patients with Cardiovascular Diseases) recently showed that elevating potassium (to a ‘high-normal’ plasma concentration of 4.5 to 5.0 mmol/L) reduced the risk of a composite of sustained ventricular tachycardia or appropriate implantable cardioverter–defibrillator therapy, hospitalization for arrhythmia or HF, or death from any cause, supporting the findings of an earlier trial showing that a potassium-enriched salt substitute reduced cardiovascular events compared to salt alone.7 However, it should be noted that POTCAST reported plasma potassium concentrations rather than serum potassium concentrations. As serum potassium concentrations are generally reported to be approximately 0.1–0.4 mmol/L higher than plasma potassium concentrations,40 we assumed an approximate difference of 0.2 mmol/L for conversion between the two measurements. Accordingly, the targeted plasma potassium range in POTCAST (4.5–5.0 mmol/L) would correspond to an approximate serum potassium range of 4.7–5.2 mmol/L. The targeted potassium range in POTCAST closely aligns with the range associated with a nadir in event rates in the present analyses. However, although patients with plasma potassium levels ≤4.3 mmol/L (approximately corresponding to serum potassium ≤4.5 mmol/L) were included in POTCAST, the lower end of the serum potassium range associated with the lowest event rates in our study (approximately 4.2–4.5 mmol/L, corresponding to plasma potassium levels of 4.0–4.3 mmol/L) was not the primary range targeted for potassium elevation or specifically evaluated as an optimal range in POTCAST. Therefore, differences in measurement modality (plasma vs serum) and in the potassium ranges actively targeted between studies should be considered when interpreting the apparent concordance between the two studies. While it could also be argued that the similar association between potassium concentration and HF hospitalization is mechanistically harder to explain than the association with death, potassium influences cardiac contractility, and arrhythmias may also precipitate worsening HF. Indeed, in another recent report, a retrospective analysis from the LOOP trial (Atrial Fibrillation Detected by Continuous Electrocardiogram Monitoring Using Implantable Loop Recorder to Prevent Stroke in High Risk Individuals) found that implantable loop recorder-documented atrial fibrillation was associated with lower potassium concentrations.42 Each mmol/L decrease in plasma potassium was associated with a five-fold increase in odds of atrial fibrillation, and the risk of atrial fibrillation was mainly associated with potassium concentrations below 4 mmol/L. Beyond the arrhythmic risk, hypokalaemia may also adversely influence the course of HF through several systemic mechanisms. Low serum potassium can stimulate renin–angiotensin–aldosterone system activity, contribute to sodium retention and increased arterial tone. Furthermore, hypokalaemia may serve as a marker of excess aldosterone and/or mineralocorticoid receptor activation, which could partly explain the observed association with adverse outcomes.43 Collectively, these observational and interventional data question the focus on hyperkalaemia in HFrEF, current concepts about what is a ‘normal’ potassium concentration in these patients, and whether patients are potentially being denied life-saving treatments because of unnecessary concerns about potassium concentration. The totality of the data argues that, from a safety perspective, the optimum serum potassium concentration is in the range of 4.5 to 5.0 mmol/L in these patients, and that consideration should be given to defining ‘hypokalaemia’ as a potassium concentration <4.0 mmol/L. In support of this, a recent study in patients with chronic kidney disease also reported that serum potassium concentration <4.0 mmol/L was strongly associated with adverse renal and cardiovascular outcomes.44 Therefore, practitioners should be at least as concerned about low potassium concentrations as high potassium in HFrEF.

The picture related to HFpEF appeared different, with a much weaker and U-shaped (rather than reverse J-shaped) association between potassium level and outcomes. The flatter risk gradients in HFpEF may reflect distinct underlying mechanisms compared to HFrEF, where neurohormonal activation is more pronounced and patients are more susceptible to ventricular arrhythmias. In contrast, patients with HFpEF may have a greater comorbidity burden and few cardiovascular deaths as a contribution to overall mortality. Potentially, serum potassium level may function more as a surrogate marker of comorbidity burden and concomitant treatments rather than a direct mediator of risk. Nonetheless, even in HFpEF, modest potassium deviations from the nadir range were associated with higher risk, indicating that dyskalaemia should not be entirely overlooked in this population, and the optimum serum potassium concentration range appears to be the same as for HFrEF, i.e. approximately 4.2–5.0 mmol/L. Here, it is worth noting that many patients in POTCAST had preserved rather than reduced LVEF.7

What are the practical implications of our findings for the management of patients with HF? First, they underscore the importance of avoiding hypokalaemia, particularly in HFrEF. A serum potassium concentration of 4.2–5.0 mmol/L appears optimal, from a safety perspective, for patients with both HFrEF and HFpEF. Second, mild hyperkalaemia should not reflexively trigger discontinuation of guideline-directed medical therapy, especially in HFrEF, where such treatments are potentially life-saving.5,45–47 We found that approximately 13% of patients with HFrEF had a serum potassium concentration between 5.0 and 5.5 mmol/L, which was not independently associated with higher mortality. A strategy of excluding other causes of hyperkalaemia and close monitoring of potassium level may be preferable to discontinuing essential therapies. The role of potassium-binding treatments is still unclear, with an increased risk of hypokalaemia and, in the case of sodium zirconium cyclosilicate, an elevated incidence of worsening HF, possibly due to the sodium content of the drug causing volume overload.5,48 In HFpEF, strict potassium targets may be less critical than in HFrEF, given the overall lower event rates and attenuated potassium–outcome associations in this HF phenotype, although general prudence remains appropriate.

Limitations

First, this study represents a post hoc analysis, which may be subject to residual confounding. The analysed data were derived from patients enrolled in randomized controlled trials with specific inclusion and exclusion criteria, and the findings may not be generalizable to the broader HF patients in the general population. Notably, several trials excluded patients with hyperkalaemia at baseline, potentially underestimating the prognostic impact of severe potassium disturbances. Most trials also excluded patients with severe kidney dysfunction, although, in practice, these patients generally do not receive therapies likely to increase potassium level. However, our findings are complemented by a ‘real-world’ analysis of over 6000 patients with HF, which showed findings with higher risk associated with hypokalaemia and no increase in risk associated with those with a mild hyperkalaemia (5.0–5.5 mmol/L).49 In addition, serum potassium levels may be affected by pre-analytical factors such as sample haemolysis, phlebotomy technique, or delays in processing, which were not uniformly controlled across studies. Furthermore, this analysis focused exclusively on patients with HFrEF (LVEF ≤40%) and HFpEF (LVEF ≥50%) and did not include those with HF with mildly reduced ejection fraction (HFmrEF), limiting the applicability of our findings to HFmrEF.