Admission serum potassium levels and prognosis of vasospastic angina

Hypokalemia is a common electrolyte disturbance and is related to poor prognosis in patients with cardiovascular disease. However, the role of hypokalemia in patients with vasospastic angina (VSA) has not yet been studied. The present study enrolled 1454 patients diagnosed with VSA according to ergonovine provocation test results and available admission serum potassium data. The primary outcome was a composite of cardiac death, acute coronary syndrome, and new-onset life-threatening arrhythmia. Based on a hypokalemia definition as serum potassium concentration ≤ 3.5 mEq/L, the hypokalaemia group included 70 patients (4.8%). The median potassium levels were 3.4 mEq/L [interquartile range (IQR) 3.3–3.5] in the hypokalemia group and 4.1 mEq/L (IQR 3.9–4.3) in the no-hypokalemia group. The median follow-up duration was 764 days. Primary outcomes occurred in seven patients (10.0%) in the hypokalemia group and 51 patients (3.7%) in the no-hypokalemia group. The Kaplan–Meier analysis showed a higher cumulative incidence of primary outcomes in the hypokalemia group compared to that in the no-hypokalemia group (log-rank P = 0.014). Multivariate Cox regression analysis also showed that hypokalemia was an independent predictor of primary outcomes. In conclusion, hypokalemia at admission was associated with adverse clinical outcomes in VSA.


Discussion
The results of this multicentre cohort study showed that the prevalence of hypokalemia was 4.8% in patients with VSA and that hypertension and prior thiazide diuretic use were risk factors for hypokalemia. Furthermore, hypokalemia at admission was related to adverse cardiovascular outcomes in these patients.
Hypokalemia is a common electrolyte disturbance along with hyponatremia and is observed in approximately 20% of hospitalised patients 1,9 . Although severe hypokalemia can lead to life-threatening arrhythmia, rhabdomyolysis, respiratory failure, or death, most events in hospitalised patients are mild to moderate in degree. In this study, 4.8% of patients with VSA had hypokalemia and only one patient had severe degree (serum potassium < 2.5 mEq/L). The prevalence was relatively lower than those in other medical illnesses. Patients with VSA were less likely to have an electrolyte disturbance than those with other medical illnesses because they had fewer comorbidities, and VSA did not cause gastrointestinal or renal disturbances. In addition, while stress-induced β 2 -adrenergic stimulation related to intracellular potassium shifting has been reported in patients with severe medical illness 10,11 , patients with VSA are less affected by these mechanisms. Hypertension and prior thiazide diuretic use have been well-known risk factors for hypokalemia and were also consistently observed in the present study. Diuretics dose-dependently increase potassium excretion in distal tubule 12 . Cohen et al. reported that hypokalemia is related to an increased risk of sudden cardiac death in hypertensive patients only when taking diuretics 13 . Diuretics not only aggravate hypokalemia but also promote neurohormonal activation 14 . www.nature.com/scientificreports/ This neurohormonal activation can also promote the severity of hypokalemia and may be related to a synergic effect between hypokalemia and diuretics to increase the risk of sudden cardiac death. In the present study, unfortunately, the number of patients with hypokalemia was too small to allow further subgroup stratifications according to diuretic use; thus, we could not evaluate the synergic effect between hypokalemia and diuretics on VSA prognosis. However, only one patient with prior diuretic use in the hypokalemia group experienced adverse events during follow-up. Moreover, while multivariate analysis showed that hypokalemia independent of prior diuretic use predicted adverse clinical outcomes in patients with VSA, prior diuretic use was not associated with our study outcomes. Therefore, we can carefully assume that hypokalemia-associated poor prognosis in patients with VSA was independent of the prior use of diuretics. The present study analysed only the impact of admission potassium levels because we had no data on serial measurement or discharge potassium levels. However, potassium levels might not significantly change during hospitalisation without active supplementation, acute Although the present study showed that hypokalemia at admission was related to adverse outcomes in patients with VSA, future studies should investigate prolonged hypokalemia during hospitalisation or hypokalemia with or without diuretics and the associated adverse outcomes. The most important question regarding the use of hypokalemia as a poor prognostic factor in patients with acute medical illness is whether hypokalemia is causally related to poor prognosis or is a surrogate marker of the severity of underlying diseases. Especially in patients with cardiovascular disease, the results of previous studies have suggested that potassium homeostasis may have a causal relationship with poor prognosis. The maintenance of serum potassium level decreases blood pressure through several mechanisms and hypokalemia  www.nature.com/scientificreports/ is related to the incidence and outcome of hypertension 2,15,16 . These observations are also supported by several epidemiologic studies 17,18 . Furthermore, hypokalemia is related to life-threatening arrhythmia or death in patients with acute myocardial infarction or heart failure 4,19 . However, there have been no reports of a relationship between hypokalemia and VSA prognosis. In this study, we demonstrated that hypokalemia at admission was related to adverse cardiovascular outcomes in patients with VSA. Although the mechanism for these results is unclear, there are several possible explanations. Serum potassium has a vasodilator function by stimulation of membrane Na + , K + -adenosine triphosphatase activity, resulting in hyper-polarisation and relaxation of the vascular smooth muscle cell 20 . There was also report about potassium is involved in nitric oxide production 21 .
In addition, increasing potassium levels had a beneficial effect on vascular endothelial cells through inhibition of oxygen-free radial formation 22 . Impaired vascular smooth muscle cell function, nitric oxide production, and vascular endothelial function were crucial factors associated with VSA aggravation 7 . Accordingly, hypokalemia may mediate vascular or endothelial dysfunction as the main mechanism of the relationship between hypokalemia and adverse cardiovascular outcome in patients with VSA. Magnesium deficiency is a well-known electrolyte imbalance involved in VSA pathogenesis 23 . It is also related to VSA severity 24,25 . In general, habitual alcohol consumption in patients with VSA is assumed to promote the urinary excretion of magnesium, leading to tissue magnesium deficiency and coronary spasm 26 . Hypokalemia is a common electrolyte imbalance in patients with magnesium deficiency, occurring in 40-60% of patients 27 .  In the present study, although we had no data on magnesium levels, there is possibility that hypokalemia may have been accompanied by hypomagnesemia, which may have affected our results. A previous epidemiologic study showed that the magnesium supplementation was inversely correlated with mortality of ischemic heart disease, however, it did not establish a relationship between magnesium levels and VSA prognosis 28 . Magnesium deficiency is causally related to alcohol consumption. In this study, alcohol consumption was not associated with either the incidence of hypokalemia or prognosis of VSA. It might be indirect evidence that hypokalemia may be a poor prognostic factor for VSA independent of hypomagnesemia. In present study, hyponatremia was the strongest independent prognostic factor for the VSA outcome. Moreover, hypertension was an independent predictor of hyponatremia and thiazide diuretic use had insignificant trend, but hypokalemia was not a predictor of hyponatremia (see Supplementary Table S1 online). Only three patients had both hyponatremia and hypokalemia, and primary outcomes did not occur in these patients. It can be possible evidence that both hyponatremia and hypokalemia were independent predictors of primary outcomes. However, we do not have a clear explanation of the relationship between hyponatremia and the outcome of VSA. Although there have been several reports about hyponatremia as an independent predictor of clinical outcomes in hospitalized patients, we could not find any report to provide evidence of an association between hyponatremia and the outcome of VSA 29,30 . There is a possibility that hyponatremia might be a strong surrogate marker of the severity of underlying medical illness, and it has a causal reason of relationship between hyponatremia and clinical outcomes. On the other hand, we should be carefully considered that the relationship between hyponatremia and diuretic use or hypertension has some impact on the prognosis of VSA. However, there were only 29 patients with hyponatremia (2.0%) in the present study which obviously limited the possibilities of in-depth analysis to clarify these hypotheses. Therefore, further studies with a large sample size focused on the association between hyponatremia and the prognosis of patients with VAS is needed. Previous studies showed U-shape relationships between serum potassium level and mortality in acute medical illness 1,4 . In addition, there was concern regarding the benefit of maintaining high normal potassium levels in patients with heart failure or myocardial infarction 31 . We categorised patients into five different potassium groups: ≤ 3.5; 3.6-≤ 4; 4.1-≤ 4.5; 4.6-≤ 5; and > 5 mEq/L and evaluated whether there was a U-shaped relationship between potassium levels and primary outcomes. Although the number of patients was small, we observed no event in patients with potassium levels > 5 mEq/L. Primary outcomes were most common in patients with potassium levels ≤ 3.5 mEq/L and least common in patients with potassium levels of 4.1-4.5 mEq/L. We observed no U-shaped relationship between potassium level and the incidence of primary outcomes or hazard ratio for the primary outcomes in the present study. www.nature.com/scientificreports/ This study has several limitations. First, owing to its observational nature, confounding factors or unmeasured variables may have influenced the results. In addition, there was a possibility that patients in the hypokalemia group may have been more likely to have medical deterioration or chronic medical illnesses. Second, the hypokalemia group included only 70 patients and there were differences baseline characteristics between the hypokalemia and no-hypokalemia groups. We tried to mitigate the bias by applying multivariate tests; however, the small sample sizes may have caused bias. Third, this study did not capture serial potassium levels during hospitalisation; thus, we cannot evaluate the relationship between serial or discharge potassium levels and primary outcomes. Furthermore, we had no data on magnesium levels or discharge information regarding diuretics, which are important co-variables in potassium physiology. Fourth, the factors associated with frequency or severity of VSA symptoms, such as episodes of angina, incidence of visiting the emergency department due to chest pain, frequency of use of nitroglycerin, or escalation of anti-angina medications, which are important indicators for the evaluation of VSA outcomes, were not assessed. A previous study that assessed using magnesium levels showed that hypomagnesemia was related to VSA severity; however, the present study only demonstrated a relationship between hypokalemia and clinical outcome of VSA 24,25 . Finally, we could not further evaluate the potential significant impacts from hyponatremia because of small sample size. Therefore, the results of the present study should be interpreted cautiously.

Conclusions
To our knowledge, this is the first study to show the association between hypokalemia and adverse cardiovascular outcomes of VSA in a large prospective cohort. The overall incidence of hypokalemia was 5% in patients with VSA and hypertension and prior thiazide diuretic use were risk factors for hypokalemia. Hypokalemia at admission was an independent predictor of adverse cardiovascular outcomes in VSA patients and hypokalemia may mediate vascular or endothelial dysfunction related to adverse outcomes. Therefore, further studies are warranted to determine whether maintaining adequate potassium levels by correcting for causal factors of hypokalemia or dietary supplementation affect the clinical outcomes of VSA. Coronary angiography and provocation test for VSA. An EG provocation test was performed and the diagnosis of VSA was confirmed based on VSA diagnosis and treatment guidelines 7 . The vasodilatory drugs including nitrates, calcium channel blockers, and nicorandil were withheld at least 48 h before coronary angiography. After performing baseline coronary angiography, a spasm provocation test was performed with intracoronary EG injection on the left coronary artery. The EG was mixed with saline and administered by intracoronary bolus injection. Incremental doses of 20 μg (E1), 40 μg (E2), and 60 μg (E3) of EG were injected into the left coronary artery. In cases with a negative result in the left coronary artery, incremental doses of 10 μg (E1), 20 μg (E2), and 40 μg (E3) of EG were injected into the right coronary artery. The definition of a positive result on the EG provocation test was ≥ 50% luminal diameter narrowing during the test with or without ischemic symptoms or electrocardiographic (ECG) changes. Among the positive results, total occlusion or > 90% luminal diameter narrowing of the coronary artery accompanied by ischemic symptoms or ECG changes was classified as definite VSA, while 50-90% luminal narrowing with or without ischemic symptoms or ECG changes was classified as possible VSA. Negative results were defined as < 50% luminal narrowing without ischemic symptoms and ECG changes. A focal spasm was defined as a discrete luminal narrowing in one coronary segment, whereas a diffuse spasm was defined as the luminal narrowing observed continuously from the proximal to the distal segments of the coronary arteries. Underlying atherosclerotic stenosis was measured after full dilatation of the coronary arteries via the intracoronary injection of nitroglycerine, in which ≥ 50% atherosclerotic stenosis of one of the major epicardial arteries was defined as indicative of obstructive coronary artery disease.

Methods
Definitions and endpoints. Hypokalemia was defined as serum potassium ≤ 3.5 mEq/L and hyponatremia as ≤ 135 mEq/L on admission laboratory testing. The patients were divided into hypokalemia and no-hypokalemia groups according to serum potassium level. We also categorised patients into five different potassium groups: ≤ 3.5; 3.6-≤ 4; 4.1-≤ 4.5; 4.6-≤ 5; and > 5 mEq/L. Hypertension was defined as blood pressure ≥ 140/90 mmHg or currently using anti-hypertensive medications. Diabetes mellitus was defined as glycated haemoglobin level ≥ 6.5% or currently using oral hypoglycaemic agents/insulin. Chronic kidney disease was defined as a glomerular filtration rate less than 60 mL/min per 1.73 m 2 . Habitual drinking was defined as drinking more than three times per week regardless of the quantity of alcohol consumed. The primary outcome was a composite of cardiac death, acute coronary syndrome (ACS) including myocardial infarction, and new-onset life-threatening arrhythmia such as ventricular tachycardia/fibrillation or complete atrioventricular block dur- www.nature.com/scientificreports/ ing the follow-up period. Cardiac death was defined as death with a demonstrated cardiac cause such as myocardial infarction, fatal arrhythmia, or heart failure or sudden unexplained death without evidence of a non-cardiac cause. ACS was defined as prolonged ischemic chest pain lasting for more than 20 min with evidence of ischemic ECG changes and/or elevated cardiac troponin I. All adverse events of interest were adjudicated by the Local Events Committee of Seoul St. Mary's Hospital through source document review.
Statistical analysis. Continuous variables were expressed as means ± standard deviation and mean differences between groups were estimated using t-tests. Categorical variables were expressed as numbers and their percentages and differences between groups were estimated using χ 2 tests. Logistic regression analysis was used to determine the predictors of hypokalemia. The cumulative event rate was assessed using the Kaplan-Meier method with log-rank analysis. Multivariable Cox regression analysis was used to evaluate the adjusted relative risks of the variables. The multivariable models included age; sex; hypertension; diabetes; chronic kidney disease; renin-angiotensin system inhibitors and thiazide diuretic use at admission; spasm characteristics; the prescription of aspirin, calcium channel blocker, nitrates, and statin at discharge; hyponatremia; and hypokalemia. These variables were chosen according to their clinical relevance and were based on the results of previous trials 7, 33 . All statistical analyses were performed using IBM SPSS Statistics for Windows, version 26.0 (IBM Corp., Armonk, NY, USA), with P values < 0.05 considered statistically significant.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.