Clinical outcomes by serum potassium levels for patients hospitalized for heart failure: Secondary analysis of data from the China National Heart Failure Registry

Abstract Background Dyskalemia is a mortality risk factor in patients with heart failure (HF). Hypothesis We described the prevalence of dyskalemia, and clinical outcomes by serum potassium (sK) levels, in Chinese patients hospitalized for HF. Methods In this secondary analysis of the prospective China National Heart Failure Registry, adult patients hospitalized between January 1, 2013 and June 30, 2015 who had at least one baseline sK measurement were followed for up to 3 years after discharge. The use of renin–angiotensin–aldosterone system inhibitors at baseline and clinical outcomes during follow‐up were compared among sK groups. Results Among 6950 patients, 5529 (79.6%) had normokalemia (sK >3.5–5.0 mmol/L), 1113 (16.0%) had hypokalemia (sK 0–3.5 mmol/L), and 308 (4.4%) had hyperkalemia (sK >5.0 mmol/L). Baseline characteristics that were most common in patients with hyperkalemia than those with hypo‐ and normokalemia included older age, HF with reduced ejection fraction, New York Heart Association Class III/IV status, hypertension, and chronic kidney disease. Use of angiotensin‐converting enzyme inhibitors (ACEIs) or angiotensin receptor blockers (ARBs) differed across sK groups (p = .0001); reported in 64.1%, 63.4%, and 54.5% of patients with hypo‐, normo‐, and hyperkalemia, respectively. Overall, 26.6%, 28.6%, and 36.0% of patients with hypo‐, normo‐, and hyperkalemia had rehospitalization for worsened HF, or cardiovascular mortality; p = .0057 for between‐group comparison. Conclusions Patients with hyperkalemia received ACEIs or ARBs for HF treatment at baseline less frequently than those with hypo‐ or normokalemia, and had worse prognoses. This warrants further investigation into effective hyperkalemia management in HF.


| INTRODUCTION
Heart failure (HF) is a major and growing global health burden, being the end stage of many chronic diseases including cardiovascular (CV) diseases, hypertension, and diabetes. 1,2In China, over 12 million people were estimated to be living with HF in 2017, with an agestandardized prevalence rate of 1.10%. 3The burden of HF in China is expected to rise, with an aging population and an increase in chronic comorbidities as risk factors. 2,4e prognosis for Chinese patients hospitalized for HF remains poor, with high reported rates of rehospitalisation and mortality. 3,4In-hospital or 3-day postdischarge mortality rate in patients hospitalized for HF reached 3.2%, and more than 20% either died or were rehospitalized within a median of approximately 1-month postdischarge. 4 Over 40% of patients hospitalized for HF had at least three hospitalization episodes. 3is warrants strategies to improve the management of HF.
In patients with HF, dyskalemia (hypokalemia or hyperkalemia) is a common complication, arising due to pathophysiologic changes underlying HF itself, related comorbidities and medications used. 5Dyskalemia increases the risk of potentially fatal cardiac arrhythmias. 6Hyperkalemia may also dictate dose reduction or discontinuation of HF treatments, resulting in poor prognoses. 5,7While current Chinese guidelines recommend renin-angiotensin-aldosterone system inhibitors (RAASis) to treat HF, RAASi treatment is reportedly underutilized in China (received by 23%-48% of patients). 8,91][12][13][14][15][16] Patients with hyperkalemia are also more likely to discontinue RAASi treatment than others. 11To our knowledge, such data among Chinese patients with HF are lacking.
The China National Heart Failure (CN-HF) Registry (ClinicalTrials.gov, NCT02079428) was a nationwide, multicenter, prospective cohort study that aimed to understand the clinical characteristics, treatment, and prognosis of Chinese patients hospitalized for HF. 17 Here, in the SPLENDID study, we described the baseline characteristics, including serum potassium (sK) levels and RAASi use, and subsequent clinical outcomes in Chinese patients hospitalized for HF, across all ejection fractions, enrolled in the CN-HF Registry.We aimed to inform the risk of dyskalemia and the importance of improving sK management in clinical practice.

| Study design and patients
The CN-HF Registry enrolled adult patients hospitalized for HF between January 1, 2013 and June 30, 2015 from 45 hospitals across 20 provinces in China.Using a two-stage sampling method, candidate hospitals were first stratified by geographical regions and tiers, followed by randomized sampling according to the total population sizes in each stratum.Eligible study centers were second-and third-tier hospitals with cardiology and/or geriatrics departments that were representative of their respective geographical areas; had accessible medical records; had qualified medical staff performing accurate and reliable data extraction; and had passed the qualification review by relevant academic committees.Eligible patients were those from the selected hospitals who received an HF diagnosis, upon discharge from the cardiology or cardiovascular department or upon death during hospitalization.SPLENDID study was a secondary analysis of the CN-HF Registry, whereby patients who had at least one sK measurement during baseline hospitalization were included.
Prespecified RAASis included angiotensin-converting enzyme inhibitors (ACEIs), angiotensin receptor blockers (ARBs), and mineralocorticoid receptor antagonists (MRAs); there were no data on angiotensin receptor neprilysin inhibitors in the registry.Sodiumglucose cotransporter 2 inhibitors were not available at study enrollment.
Patients were followed up for clinical endpoints through outpatient visits or telephone calls every 3-6 months for up to 3 years after discharge or until death, whichever occurred earlier.The dates of occurrence (if any) of rehospitalization for worsened HF, CV mortality, and all-cause mortality for each patient were extracted from the registry.

| Endpoints
The primary endpoint was the proportion of patients in each sK category who experienced a composite of rehospitalization for worsened HF, or CV mortality, during follow-up.Secondary endpoints included the proportions of patients in each sK category who experienced the individual components of the primary endpoint, and all-cause mortality, during follow-up.Other secondary endpoints were the distribution of baseline sK levels; the proportions of patients with sK levels >5.0 mmol/L and >5.5 mmol/L at baseline; and the proportions of patients in each sK category who received RAASi at baseline.

| Statistical analysis
Among 7171 adult patients hospitalized for HF enrolled in the CN-HF Registry, 6950 patients had at least one sK measurement during baseline hospitalization, and were included in the full analysis set (FAS) of SPLENDID (Figure S1).Based on an estimated rate of hyperkalemia of 4.8% (i.e., approximately 333 patients in the FAS) at baseline, and assuming a primary endpoint event rate of 27.5% for this group, a precision level of 4.8% would be achieved.Assuming 251 patients (3.6%) had hypokalemia and 6366 (91.6%) had normokalemia with respective primary endpoint event rates of 26.5% and 23.0%, the precision levels achieved would be 5.5% and 1.0% for the corresponding groups. 13atistical analyses were performed using SAS, version 9.

| Clinical outcomes by sK categories
During follow-up, 1990 of 6950 patients (28.6%) experienced rehospitalization for worsened HF, or CV mortality, with event rates that differed across sK groups (p = .0057)(Table 4).A higher proportion of patients in the hyperkalemia group (36.0%) experienced this primary  In the FAS, hyperkalemia was associated with a statistically significant 37% increase in risk for the composite of rehospitalization for worsened HF, or CV mortality, compared with hypo-and normokalemia (Table 5).To account for the potential effect of different HF types on prognoses, the association of hyperkalemia with clinical outcomes was separately assessed in the HFrEF, HFmEF, and HFpEF subgroups.There was a trend of increased risk of the primary endpoint in patients with hyperkalemia versus those with hypo-or normokalemia across all HF types, although this did not reach statistical significance in the HFrEF subgroup.Similar trends of increased risk of the individual components of the primary endpoint with hyperkalemia were observed; these were statistically significant across all HF types except for rehospitalization for worsened HF in the HFrEF subgroup.The risk for all-cause mortality increased by approximately two-fold for hyperkalemia versus hypo-and normokalemia (p < .001) in the FAS and in all subgroups defined by HF type.

| DISCUSSION
Using data from the prospective real-world CN-HF Registry, we analyzed, by sK levels, the baseline characteristics and clinical outcomes of Chinese patients hospitalized for HF.Dyskalemia is common in patients with HF, with reported frequencies of 3.0%-25.8%for hypokalemia and 5.7%-39.0%1][22][23][24]25,26 Among HF subtypes, hyperkalemia versus hypo-and normokalemia was observed more often in patients with HFrEF and less often in those with HFpEF.This may be due to greater use of RAASi, which can potentially cause hyperkalemia, among patients with HFrEF as recommended by clinical guidelines, 27 although the use of RAASi by sK levels was not analyzed based on HF type in the present study.
Patients with hypokalemia appeared to have comparable baseline characteristics to those with normokalemia in this cohort, unlike in other studies. 13,15,16,28The identification of baseline characteristics associated with sK level may be useful for dyskalemia risk prediction in patients with HF.
0][31][32] Our findings indicated that ACEI or ARB use at baseline were lower in patients with hyperkalemia than those with hypo-or normokalemia.MRA use was not significantly different across sK categories, which may be due to the clinical practice in China where, rather than MRA, the use of ACEI or ARB is usually adjusted in patients upon the development of hyperkalemia.Real-world database studies in Western patients with HF who received RAASi have shown a direct association of hyperkalemia with treatment dose reduction or discontinuation. 30,31 observed a J-shaped relationship with hyperkalemia being associated with adverse clinical outcomes.Both hyperkalemia and hypokalemia were previously associated with increased risk for CV events and all-cause mortality in Western populations, suggesting a T A B L E 4 Clinical endpoint event rates by baseline serum potassium categories.4][35][36] The lack of an association between adverse clinical outcomes and hypokalemia in our study might be due to the easy correction of hypokalemia in the clinic, since subsequent normalization of sK was not considered.Despite differences in prognoses among HF types, 37,38 we consistently observed a higher risk for adverse clinical outcomes among patients with hyperkalemia across all HF subgroups.While we did not assess the association between RAASi use and clinical outcomes in this study, our data support real-world Western studies where patients who discontinued RAASis, or those on reduced doses, due to hyperkalemia, exhibited increased incidence of major adverse cardiac events and mortality than those on maximum dose. 30,31llectively, findings from SPLENDID and previous studies support the importance of long-term hyperkalemia management in HF, including in Chinese patients, which can potentially improve clinical outcomes by allowing optimal use of RAA-Sis. 30,31,39Current guidelines from the European Society of Cardiology suggest the use of novel potassium binders (sodium zirconium cyclosilicate [SZC] and patiromer) for the management of RAASi-associated hyperkalemia. 32According to a phase 3 study, correction of hyperkalemia and maintenance of normokalemia were observed with SZC treatment for up to 12 months among outpatients, without the need for RAASi medication restrictions; 87% of the study participants were able to maintain or increase their baseline RAASi treatment dosages. 40Further randomized controlled trials that directly evaluate the benefits of hyperkalemia management on RAASi optimization and clinical outcomes in Chinese patients with HF are warranted.
A strength of this study is the use of data from a prospective real-world registry, which included patients across all ranges of HF ejection fraction.A key limitation is the retrospective nature of the analysis.Several variables, including patients' diet, medications (e.g., diuretics), potassium supplementation, and the causes of hyperkalemia, were not available to adjust for confounding.Thus, the association between sK levels and clinical outcomes should be interpreted cautiously.Second, clinical outcomes were analyzed only based on sK at baseline, negating the effect of subsequent sK fluctuations.It was previously shown that persistent dyskalemia was linked to a higher mortality risk versus sK normalization, hence longitudinal monitoring of dynamic sK changes would provide a more accurate prediction of clinical outcomes. 13

| CONCLUSIONS
In this secondary analysis of a real-world registry, dyskalemia was observed in more than 20% of Chinese patients hospitalized for HF.
Several baseline characteristics, including older age, worse NYHA class function, HFrEF, hypertension, and CKD, were most common in patients with hyperkalemia.Compared with the hypo-and normokalemia groups, a lower proportion of patients with hyperkalemia were prescribed ACEIs or ARBs for the treatment of HF at baseline, but a higher proportion experienced rehospitalization for worsened HF, CV mortality, and all-cause mortality after discharge regardless of T A B L E 5 Association of clinical endpoint events with hyperkalemia (sK >5.0 mmol/L) versus normo-and hypokalemia (sK ≤5.0 mmol/L) analyzed by heart failure type.Abbreviations: CI, confidence interval; CV, cardiovascular; FAS, full analysis set; HFmEF, HF with mid-range ejection fraction; HFpEF, HF with preserved ejection fraction; HFrEF, HF with reduced ejection fraction; HR, hazard ratio; LVEF, left ventricular ejection fraction; sK, serum potassium.a Calculated using logistic regression analysis; based on unadjusted models.Reference group: sK >5.0 mmol/L.
3 (SAS Institute Inc.).All data were categorical, shown as frequencies and percentages (with 95% confidence intervals for clinical outcomes), and based on non-missing data unless otherwise specified.Statistical comparisons of baseline characteristics and clinical outcomes among sK categories were conducted post hoc.Differences in baseline characteristics were analyzed using the Pearson's Chi-square test (for binary variables), the Cochran-Mantel-Haenszel test (for variables with multiple groups), or the Cochran Armitage trend test (for RAASi use).Differences in the rates of clinical events were analyzed using the Cochran-Mantel-Haenszel test (row mean scores differ).The association of hyperkalemia (vs.hypo-and normokalemia) with clinical outcomes was evaluated using logistic regression analysis in the FAS and by HF type.

T A B L E 1
Distribution of patients across baseline serum potassium categories.
Baseline characteristics of patients grouped according to serum potassium levels.