JACC: HEART FAILURE VOL. -, NO.

-, 2018
ª 2018 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION

PUBLISHED BY ELSEVIER

Incidence, Predictors, and Outcome

Associations of Dyskalemia in Heart
Failure With Preserved, Mid-Range,
and Reduced Ejection Fraction
Gianluigi Savarese, MD, PHD,a,* Hong Xu, MD,b,* Marco Trevisan, MSC,b Ulf Dahlström, MD, PHD,c
Patrick Rossignol, MD, PHD,d Bertram Pitt, MD, PHD,e Lars H. Lund, MD, PHD,a,y Juan J. Carrero, PHARMD, PHDb,y

ABSTRACT

OBJECTIVES This study investigated 1-year incidence and predictors of dyskalemia (dysK) and its outcome associations
in heart failure with preserved ejection fraction (HFpEF), HF with mid-range EF (HFmrEF), and HF with reduced EF
(HFrEF).

BACKGROUND DysK in real-world HF is insufficiently characterized. Fear of dyskalemia may lead to underuse or
underdosing of renin-angiotensin-aldosterone system inhibitors.

METHODS Patients enrolled in the Swedish HF registry from 2006 to 2011 in Stockholm, Sweden were included in the
analyses. Multivariate Cox regression analysis identified independent predictors of dysK within 1 year. Time-dependent
Cox models assessed outcomes associated with incident dysK (all-cause death, HF, and other cardiovascular disease
[CVD] hospitalizations) within 1 year from baseline.

RESULTS Of 5,848 patients, 24.4% experienced hyperkalemia (hyperK [K >5.0 mmol/l]) at least once, and 10.2% had
moderate or severe hyperK (K >5.5 mmol/l). Adjusted risk of moderate or severe hyperK was highest in HFpEF and
HFmrEF. Similarly, 20.3% of patients had at least one episode of hypokalemia (hypoK [<3.5 mmol/l]), and 3.7% had
severe hypoK (<3.0 mmol/l). Adjusted risk of any hypoK was highest in HFpEF. Independent predictors of both hyper-
and hypoK were sex, baseline potassium and estimated glomerular filtration rate, low hemoglobin, chronic obstructive
pulmonary disease (COPD), inpatient status, and higher New York Heart Association functional class. Incident dysK was
associated with increased risk of mortality. Furthermore, hypoK was associated with increased CVD hospitalizations
(HF-related excluded). There was no association between dysK and HF hospitalization risk, regardless of EF.

CONCLUSIONS DysK is common in HF and is associated with increased mortality. Risk of moderate or severe hyperK was
highest in HFpEF and HFmrEF, whereas risk of hypoK was highest in HFpEF. HF severity, low hemoglobin, COPD,
baseline high and low potassium, and low eGFR were relevant predictors of dysK occurrence. (J Am Coll Cardiol HF 2018;-:-–-)
© 2018 by the American College of Cardiology Foundation.

From the aDivision of Cardiology, Department of Medicine, Karolinska Institutet, Stockholm, Sweden; bDepartment of Medical
Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; cDepartments of Cardiology Medical and Health Sci-
ences, Linkoping University, Linkoping, Sweden; dUniversité de Lorraine, INSERM CIC-P 1433, Centre Hospitalier Régional Uni-
versitaire de Nancy, and INSERM U1116, FCRIN INI-Cardiovascular and Renal Clinical Trialists, Nancy, France; and the
e
Department of Medicine, University of Michigan, Ann Arbor, Michigan. Supported by institutional grants from Vifor Fresenius
Medical Care Renal Pharma and Relypsa to Karolinska Institutet, Swedish Heart and Lung Foundation, Swedish Research Council,
Stockholm County Council (including ALF projects), AstraZeneca, and Martin Rind’s and Westman’s Foundations. Dr. Lund is
supported by Swedish Research Council grants 2013-23897-104604-23 and 523-2014-2336, and Swedish Heart Lung Foundation
grants 20120321 and 20150557; has received grants from Relypsa, Vifor Pharma, AstraZeneca, and Novartis through his institution;
and consults for Relypsa, Vifor Pharma, AstraZeneca, Novartis, and Sanofi. Dr. Rossignol is supported by French National Research
Agency public grant ANR-15-RHU-0004 as part of the second Investissements d’Avenir program FIGHT-HF and by French PIA
project Lorraine Université d’Excellence grant ANR-15-IDEX-04-LUE; consults for Novartis, Relypsa, AstraZeneca, Grünenthal,
Stealth Peptides, Fresenius, Idorsia, Vifor Fresenius Medical Care Renal Pharma, Vifor, and CTMA; has received lecture fees from
Bayer and CVRx; and is a cofounder of CardioRenal. Dr. Savarese has received research grants from Merck and Co. and Boehringer

ISSN 2213-1779/$36.00 https://doi.org/10.1016/j.jchf.2018.10.003

2 Savarese et al. JACC: HEART FAILURE VOL. -, NO. -, 2018
Dyskalemia in HF - 2018:-–-

H
ABBREVIATIONS eart failure (HF) is associated with dispensed drug registry, and the SCREAM (Stockholm
AND ACRONYMS comorbidities (e.g., older age, CREAtinine Measurements) project has been previ-
chronic kidney disease [CKD], dia- ously described (11). More details are reported in
ACE inhibitors = angiotensin-
converting enzyme inhibitors
betes) and is treated with pharmacological Online Methods.
regimens (e.g., angiotensin-converting The present study enrolled HF patients who were
ARB = angiotensin receptor
blockers enzyme [ACE] inhibitors, angiotensin recep- registered for the first time in SwedeHF and resided
ARNI = angiotensin-receptor tor blockers [ARB], mineralocorticoid recep- in the region of Stockholm between January 1, 2006,
neprilysin inhibitors tor antagonists [MRA], angiotensin-receptor and December 31, 2010. These participants underwent
COPD = chronic obstructive neprilysin [ARN] inhibitors, and diuretics), plasma potassium tests (recorded in SCREAM) on the
pulmonary disease which may interfere with potassium homeo- day of entry in the SwedeHF trial (index date). From a
CVD = cardiovascular disease stasis, leading to dyskalemia (dysK) (both total of 8,896 HF patients registered in Stockholm
EF = ejection fraction hyperkalemia [hyperK] and hypokalemia during that period, we excluded those who did not
eGFR = estimated glomerular [hypoK]) (1–4). have measurements of serum creatinine and plasma
filtration rate Patients with HF with preserved ejection potassium (n ¼ 1,638), or who showed abnormal
HF = heart failure fraction (HFpEF), HF with mid-range EF plasma potassium concentrations (i.e., K <3.5
HFrEF = heart failure with (HFmrEF), or HF with reduced EF (HFrEF) [n ¼ 412] and >5.0 mmol/l [n ¼ 141]) at registration, or
reduced ejection fraction
have different demographics, exhibit had not had EF assessed (n ¼ 849), or were under-
MRA = mineralocorticoid different clinical characteristics, and are going chronic dialysis (n ¼ 8). This resulted in 5,848
receptor antagonists
treated differently, possibly leading to patients included in the analysis (see patient selec-
NYHA = New York Heart
different rates of dysK (5). However, current tion flow chart in Online Figure 1).
Association
evidence for the incidence of dysK comes EXPOSURE AND BASELINE COVARIATES. Demo-
PAD = peripheral artery
disease mainly from selective, randomized clinical graphic and clinical variables are reported in Table 1
PAF = population-attributable
trials in HFrEF, and no detailed dysK quan- (see definitions in Online Table 1). Ongoing medica-
fraction tification exists to date in HFpEF and tions were defined by the presence of a pharmacy
RAAS = renin-angiotensin- HFmrEF (6,7). Furthermore, data from ran- dispensation within 120 days from the index date (see
aldosterone system domized clinical trials may not represent definitions in Online Table 2). We defined HFpEF as
real-world clinical practice (8). Thus, a EF $50%, HFmrEF as EF 40% to 49%, and HFrEF as
deeper understanding of risk of dysK in these EF <40% (12). Plasma potassium concentration was
settings is needed, given that fear of dysK may lead to measured by using potentiometric titration. We
underuse or underdosing of renin-angiotensin- categorized baseline potassium levels as 3.5 to 3.9,
aldosterone system (RAAS) inhibitors (9,10) and that 4.0 to 4.4, and 4.5 to 5.0 mmol/l, with the stratum 4.0
the incidence of dysK is relevant in the design of to 4.4 mmol/l chosen as reference. Plasma creatinine
future trials in different HF phenotypes. concentration was measured by using either the
This study assessed potential differences and enzymatic or corrected Jaffe method (alkaline picrate
similarities in the incidence of dysK, its predictors, reaction) and was used to estimate glomerular filtra-
and outcome associations in a large and unselected tion rate (eGFR). Creatinine values <25 and >1,500
populations of real-world HF patients with different mmol/l were considered implausible and discarded.
EF. eGFR was calculated by using the CKD-Epidemiology
METHODS Collaboration (CKD-EPI) equation, and eGFR strata
were categorized as $90, 60 to 89, 45 to 59, 30 to 44,
DATA SOURCES AND STUDY POPULATION. The and <30 ml/min per 1.73 m 2. eGFR $ 90 ml/min per
Swedish HF Registry (SwedeHF) and its linkage 1.73 m2 was considered the reference.
through the unique personal identification of Swed- STUDY OUTCOME. The first study outcome was the
ish citizens with general government registries, the 1-year incidence of hyperK and hypoK, separately;

Ingelheim; and has received honoraria from Vifor and AstraZeneca. Dr. Carrero has received research support from AstraZeneca.
Dr. Pitt has received fees from Bayer, Sanofi, Astrazeneca, kdp pharmaceuticals, relypsa/vifor, cpharmaceuticals, Tricida, Stealth
Peptides, Sarfez Pharmaceuticals, and G3 Pharmaceuticals; and is a patent pending holder for eplerenone delivery. Dr. Dahlström
has received research support from AstraZeneca through his institution; and has received honoraria from and consults for
AstraZeneca and Novartis. All other authors have reported that they have no relationships relevant to the contents of this paper to
disclose. *Drs. Savarese and Xu contributed equally to this work and are joint first authors. yDrs. Lund and Carrero contributed
equally to this work and are joint senior authors.

Manuscript received September 6, 2018; accepted October 4, 2018.

JACC: HEART FAILURE VOL. -, NO. -, 2018 Savarese et al. 3
- 2018:-–- Dyskalemia in HF

T A B L E 1 Baseline Characteristics Overall and by Ejection Fraction Categories

Overall HFpEF HFmrEF HFrEF

(N ¼ 5,848) (n ¼ 1,397, 24%) (n ¼ 1,304, 22%) (n ¼ 3,147, 54%) p Value

Demographics
Mean age, yrs 73.0  13.0 76.7  12.2 73.9  12.3 71.1  13.3 <0.001
Women 2,146 (36.7) 758 (54.3) 497 (38.1) 891 (28.3) <0.001
Current smokers 683 (11.7) 104 (7.4) 154 (11.8) 425 (13.5) <0.001
Education (n ¼ 5,646) <0.001
Primary school 2,186 (37.4) 565 (40.4) 480 (36.8) 1,141 (36.3)
Secondary school 2,259 (38.6) 501 (35.9) 482 (37.0) 1,276 (40.5)
University 1,201 (20.5) 266 (19.0) 293 (22.5) 642 (20.4)
Laboratory tests at registration
Potassium, mmol/l 4.1  0.4 4.1  0.3 4.1  0.4 4.1  0.4 <0.001
Potassium strata <0.001
3.5 to <4.0 mmol/l 2,137 (36.5) 572 (40.9) 507 (38.9) 1,058 (33.6)
4.0 to <4.5 mmol/l 2,744 (46.9) 654 (46.8) 577 (44.2) 1,513 (48.1)
4.5 to 5.0 mmol/l 967 (16.5) 171 (12.2) 220 (16.9) 576 (18.3)
eGFR, ml/min per 1.73m2 61.3  24.3 58.0  22.9 61.4  24.3 62.8  24.9 <0.001
eGFR strata <0.001
90þ ml/min per 1.73 m2 728 (12.4) 137 (9.8) 157 (12.0) 434 (13.8)
60–89 ml/min per 1.73 m2 2,243 (38.4) 477 (34.1) 523 (40.1) 1,243 (39.5)
45–59 ml/min per 1.73 m2 1,255 (21.5) 344 (24.6) 270 (20.7) 641 (20.4)
30–44 ml/min per 1.73 m2 1,063 (18.2) 290 (20.8) 225 (17.3) 548 (17.4)
<30 ml/min per 1.73 m2 559 (9.6) 149 (10.7) 129 (9.9) 281 (8.9)
Hemoglobin, g/l 131.9  17.8 127.2  17.2 131.2  18.2 134.3  17.5 <0.001
Hemoglobin, <120 g/l 1,508 (25.8) 496 (35.5) 353 (27.1) 659 (20.9) <0.001
Comorbidities
Hypertension 1,746 (29.9) 408 (29.2) 402 (30.8) 936 (29.7) 0.64
Diabetes mellitus 1,108 (18.9) 227 (16.2) 242 (18.6) 639 (20.3) 0.005
Myocardial infarction 1,455 (24.9) 214 (15.3) 300 (23.0) 941 (29.9) <0.001
PCI or CABG 924 (15.8) 117 (8.4) 192 (14.7) 615 (19.5) <0.001
Stroke or TIA 408 (7.0) 81 (5.8) 73 (5.6) 254 (8.1) 0.002
PAD 240 (4.1) 44 (3.1) 49 (3.8) 147 (4.7) 0.045
Aortic stenosis 325 (5.6) 89 (6.4) 67 (5.1) 169 (5.4) 0.30
Atrial fibrillation 2,166 (37.0) 504 (36.1) 498 (38.2) 1,164 (37.0) 0.52
COPD 952 (16.3) 233 (16.7) 222 (17.0) 497 (15.8) 0.54
Liver disease 51 (0.9) 8 (0.6) 9 (0.7) 34 (1.1) 0.17
Alcoholism 106 (1.8) 13 (0.9) 16 (1.2) 77 (2.4) <0.001
Cancer (3 yrs) 235 (4.0) 56 (4.0) 44 (3.4) 135 (4.3) 0.37
Characterization of HF index
Caregiver (inpatient vs. outpatient) 4,273 (73.1) 1,138 (81.5) 889 (68.2) 2,246 (71.4) <0.001
NYHA functional class (n ¼ 4,793) <0.001
I 742 (12.7) 264 (18.9) 224 (17.2) 254 (8.1)
II 2,293 (39.2) 547 (39.2) 578 (44.3) 1,168 (37.1)
III 1,598 (27.3) 259 (18.5) 251 (19.2) 1,088 (34.6)
IV 160 (2.7) 26 (1.9) 24 (1.8) 110 (3.5)
Continued on the next page

the second outcome was the 1-year risk of all-cause recorded in health care within 1 year from the index
death, HF hospitalization, or cardiovascular disease registration. We defined incident hyperK and hypoK
(CVD) hospitalization other than HF after an inci- as the first measurements of potassium, >5.0
dence of hyperK or hypoK. Outcome analysis was and <3.5 mmol/l, respectively. However, because
also separately performed in the different EF cate- patients with K <3.0 and >5.5 mmol/l are at the
gories, as risk of these outcomes in HFpEF versus highest risk of mortality (13), we also considered the
HFmrEF versus HFmrEF has not been previously categories of moderate or severe hyperkalemia
investigated. (>5.5 mmol/l) and severe hypokalemia (<3.0 mmol/l).
The 1-year incidence of hyperK and hypoK was In order to exclude incident dysK that might have
calculated from all plasma potassium measurements been the consequence of abnormal potassium
4 Savarese et al. JACC: HEART FAILURE VOL. -, NO. -, 2018
Dyskalemia in HF - 2018:-–-

T A B L E 1 Continued

Overall HFpEF HFmrEF HFrEF

(N ¼ 5,848) (n ¼ 1,397, 24%) (n ¼ 1,304, 22%) (n ¼ 3,147, 54%) p Value

EF strata
EF >50% 1,397 (23.9) 1,397 (100.0)
EF 40%–49% 1,304 (22.3) 1,304 (100.0)
EF <40% 3,147 (53.8) 3,147 (100.0)
Medication
ACE inhibitor or ARB 4,760 (81.4) 1,011 (72.4) 1,028 (78.8) 2,721 (86.5) <0.001
Beta-blocker 4,856 (83.0) 1,068 (76.4) 1,070 (82.1) 2,718 (86.4) <0.001
MRAs 2,130 (36.4) 425 (30.4) 399 (30.6) 1,306 (41.5) <0.001
Digoxin 1,168 (20.0) 257 (18.4) 235 (18.0) 676 (21.5) 0.008
Long-acting nitrates 929 (15.9) 254 (18.2) 221 (16.9) 454 (14.4) 0.003
Calcium channel blockers 821 (14.0) 336 (24.1) 217 (16.6) 268 (8.5) <0.001
Antiarrhythmics 197 (3.4) 28 (2.0) 32 (2.5) 137 (4.4) <0.001
Statins 2,434 (41.6) 510 (36.5) 584 (44.8) 1,340 (42.6) <0.001
Aspirin/anti-platelets 2,605 (44.5) 581 (41.6) 623 (47.8) 1,401 (44.5) 0.005
Any oral anticoagulant 2,129 (36.4) 484 (34.6) 464 (35.6) 1,181 (37.5) 0.14
Antidiabetes 1,140 (19.5) 270 (19.3) 284 (21.8) 586 (18.6) 0.053
Diuretics 4,402 (75.3) 1,115 (79.8) 903 (69.2) 2,384 (75.8) <0.001

Values are mean  SD or n (%).

ACE ¼ angiotensin-converting enzyme; ARB ¼ angiotensin receptor blocker; CABG ¼ coronary artery bypass graft; COPD ¼ chronic obstructive pulmonary disease;
EF ¼ ejection fraction; eGFR ¼ estimate glomerular filtration rate; HFmrEF ¼ mid-range ejection fraction; HFpEF ¼ preserved ejection fraction; HFrEF ¼ reduced ejection
fraction; MRAs ¼ mineralocorticoid receptor antagonists; NYHA ¼ New York Heart Association; PAD ¼ peripheral artery disease; PCI ¼ percutaneous coronary intervention.

baseline values, we disregarded events occurring
within the first 2 weeks from the index date.
Number of deaths were obtained by linkage with
the patient register, as well as HF hospitalizations or
CVD hospitalization other than HF, which were
identified by relevant International Classification of
Diseases 10th revision at discharge (Online Table 1).
STATISTICS. Continuous variables were reported as
mean  SD, and categorical variables as counts and
proportions (%). Baseline characteristics
compared across the different EF categories (HFpEF
vs. HFmrEF vs. HFrEF) using Pearson chi-square for
proportions and analysis of variance (ANOVA) for
continuous variables.
The crude 1-year incidence rates of hyperK and
hypoK were assessed by using the Kaplan-Meier
method. Multivariate Cox regression models with
dysK as dependent variable were performed in the
overall population and in the different EF categories
to identify the baseline factors associated with dysK
occurrence, reporting hazard ratios (HRs) and 95%
confidence intervals (CI). We also tested multiplica-
tive interactions between the a priori-selected eGFR
strata (eGFR $ vs. <60 ml/min per 1.73 m2); comor-
bidities including hypertension, diabetes, myocardial
infarction, atrial fibrillation, and chronic obstructive
pulmonary disease (COPD); and EF
Finally, the weighted contribution of selected risk
factors to dysK incidence was quantified using the
population-attributable fraction (PAF) adjusted for
potential confounders by Cox proportional hazards
model (14).
Multivariate time-dependent Cox regression
models were performed to investigate the association
between incident dysKs and study outcomes within
the 1-year follow-up (from baseline) in the overall HF
population and separately in patients with HFpEF,
HFmrEF, and HFrEF. All covariates were defined at
baseline, except for dysK status, which was considered
were
time-dependent. Patients experiencing dysK, thus,
contributed to the reference group during their
person-time free from dysK and to the dysK risk group
during the remaining person-time. A p value of <0.05
was considered statistically significant for all analyses.
All analyses were performed using STATA version 14.0
software (Stata Corp., College Station, Texas).
RESULTS
BASELINE CHARACTERISTICS. Baseline characteris-
tics are reported in Table 1. Of 5,848 HF patients, 37%
were women, 24% had HFpEF, 22% had HFmrEF, and
54% had HFrEF. Mean age was 73  13 years, mean
plasma potassium at inclusion was 4.1  0.4 mmol/l,
and mean eGFR 61  24 ml/min per 1.73 m 2. Older age,
female sex, worse kidney function, lower hemoglobin
categories. levels, and higher user of diuretics were more com-
mon with higher EF. In contrast, the proportions of
diabetes, history of myocardial infarction, coronary
JACC: HEART FAILURE VOL. -, NO. -, 2018 Savarese et al. 5
- 2018:-–- Dyskalemia in HF

F I G U R E 1 DysK Event Rates During 1-Year Follow-Up, Overall and Stratified by Ejection Fraction

DysK ¼ dyskalemia; HFmrEF ¼ heart failure with mid-range ejection fraction; HFpEF ¼ heart failure with preserved ejection fraction;
HFrEF ¼ heart failure with reduced ejection fraction; K ¼ potassium.

revascularization, peripheral artery disease, worse
New York Heart Association (NYHA) functional class,
and the use of HF treatments renin-angiotensin-
system inhibitors, MRA, and beta-blockers increased
with lower EF.
INCIDENCE OF HyperK AND HypoK. Overall, 24.4%
of patients experienced at least 1 hyperK occurrence
within 1 year, and 10.2% reported moderate or severe
hyperK (Figure 1, note that events are nonexclusive).
Crude and adjusted analyses showed no statistically
significant differences in risk of any hyperK across the
EF spectrum (Figure 2, Table 2). Crude 1-year risk of
moderate or severe hyperK was comparable across
the EF strata but, after adjustments, it was higher in
HFpEF and HFmrEF than in HFrEF (Online Table 3,
Online Figure 2).
Overall, 20.3% of patients experienced at least 1
occurrence of hypoK within 1 year, and 3.7% experi-
enced severe hypoK (Figure 1, nonexclusive events).
Crude and adjusted risks of any hypoK were highest
in HFpEF (Figure 2, Table 2). Crude and adjusted risks
of severe hypoK were similar across the EF spectrum
(Online Table 3, Online Figure 2).
Most dysK events, regardless of severity, occurred
within the first 3 months of observation (Figure 3,
Online Figure 3).
RISK MARKERS FOR HyperK AND HypoK. Table 2
shows the baseline risk factors associated with the
risk of dysK. Variables independently associated
with increased hyperK risk were male sex,
baseline potassium 4.5 to 5.0 mmol/l, lower eGFR,
hemoglobin <120 g/dl, history of diabetes, COPD,
alcoholism, cancer, inpatient status, NYHA functional
class $II, use of MRA, and nonuse of beta-blockers.
Conversely, baseline potassium 3.5 to 3.9 mmol/l
was associated with reduced hyperK risk. Neither ACE
inhibitors/ARBs nor diuretics were associated with
incident hyperK. PAF was calculated to estimate the
relative importance of selected risk predictors.
We observed that 37% of events were attributable to
use of MRA, 33% of events to hemoglobin
concentration <120 g/l, 26% to eGFR <60 ml/min per
6 Savarese et al.
Dyskalemia in HF
F I G U R E 2 Kaplan-Meier Curves
Kaplan-Meier curves show the incidence of (A) hyperK and (B) hypoK, according to ejection fraction strata. LVEF ¼ left ventricular ejection fraction. hyperK ¼
hyperkalemia; hypoK ¼ hypokalemia.
1.73 m 2, 25% to diabetes, and 16% to COPD and care-
givers (inpatient vs. outpatient) (Online Table 4).
Table 2 also reports baseline risk factors for hypoK.
Female sex, potassium 3.5 to 3.9
eGFR <30 ml/min per 1.73 m 2, hemoglobin <120 g/dl,
history of myocardial infarction, atrial fibrillation,
COPD, alcoholism, being an inpatient, NYHA func-
tional class $III, use of diuretics, nonuse of ACE
inhibitors/ARBs and beta-blockers independently
predicted hypoK. PAF analyses showed 25% of
hypoK events were attributable to COPD, 20% to
hemoglobin <120 g/dl, 18% to female sex, 14% to
history of myocardial infarction, 13% to caregiver
(inpatient vs. outpatient), and 11% to the nonuse of
ACE/ARBs (Online Table 4).
Sensitivity analyses showed overall similar pre-
dictors for moderate or severe dysK (Online Tables 5
to 7). No multiplicative interactions were observed
between EF strata (HFpEF, HFmrEF, and HFrEF),
eGFR strata, and presence versus absence of comor-
bidities (hypertension, diabetes, myocardial infarc-
tion, atrial fibrillation, and COPD) in predicting study
outcomes (p > 0.10 for all).
DysK AND SUBSEQUENT RISK OF MORTALITY AND
HOSPITALIZATION. During the 1-year
from baseline, 1.45 deaths per 1,000 patient-years
occurred after hyperK compared to 0.33 deaths that
occurred in periods without or preceding hyperK.
After adjustments, incident hyperK was associated
with increased risk of death (HR: 4.03; 95% CI: 3.42 to
4.75) (Figure 4). Similar results were observed across
the EF spectrum (Table 3). In contrast, incident
JACC: HEART FAILURE VOL. -, NO. -, 2018
- 2018:-–-
hyperK was not associated with the risk of hospitali-
zation (Figure 4, Table 3).
Similarly, 1.43 deaths per 1,000 patient-years
mmol/l, occurred after hypoK compared to 0.36 deaths dur-
ing periods without or preceding hypoK (Figure 4).
After adjustments, incident hypoK was significantly
associated with the risk of death (HR: 3.28; 95% CI:
2.79 to 3.86) (Figure 4), with a similar magnitude
across the different EF strata (Table 3). Incident
hypoK was also associated with the risk of CVD hos-
pitalization other than HF (HR: 1.83; 95% CI: 1.24 to
2.71) but not with the risk of a new HF hospitalization
(Figure 4, Table 3).
DISCUSSION
This study found that the incidence of dysK (both
hyperK and hypoK) was common and associated with
similar predictors, regardless of EF category, in a
large cohort of real-world HF patients. Incident dysK
was a strong predictor of mortality regardless of EF
category. Incident dysK was not associated with the
risk of a subsequent HF hospitalization, whereas
incident hypoK (but not hyperK) was associated with
the risk of CVD hospitalization other than HF.
follow-up
INCIDENCE OF HYPER- AND HYPOKALEMIA IN
REAL-WORD HF. Current evidence of the incidence
of hyperK and hypoK in HF patients came mainly
from randomized clinical trials. Whereas most infor-
mation comes from trials including HFrEF patients
(15–18), 1 trial provided estimates in HFpEF (19), and
the CHARM (Candesartan in Heart Failure: Assessment
JACC: HEART FAILURE VOL. -, NO. -, 2018 Savarese et al. 7
- 2018:-–- Dyskalemia in HF

of Reduction in Mortality and Morbidity) program

T A B L E 2 Baseline Characteristics Associated With 1-Year Risk of Hypokalemia and
analyzed hyperK risk throughout the EF spectrum (20). Hyperkalemia
In the present study, 24.4% of patients experi-
Any Hyperkalemia Any Hypokalemia
enced at least once incidence of hyperK, and 10.2% (Kþ >5.0 mmol/l) (Kþ <3.5 mmol/l)
experienced moderate or severe hyperK within 1 year HR 95% CI HR 95% CI
of follow-up, most of which occurred during the first 3 Age, per 5 yrs 1.01 0.98–1.04 1.02 0.99–1.06
months of follow-up. This is certainly a larger inci- Women 0.86* 0.77–0.97 1.33‡ 1.18–1.51
dence than that reported in trials with longer follow- Smoker 1.11 0.93–1.33 0.93 0.76–1.14
up (e.g., in the CHARM program, clinically relevant Baseline potassium strata

hyperK occurred in 3.5% of the population over a 3.5–<4.0 mmol/l 0.79‡ 0.70–0.89 1.52‡ 1.34–1.72
4.0–<4.5 mmol/l Ref. Ref.
median follow-up of 3.2 years) (20). Whereas we
4.5–5.0 mmol/l 1.49‡ 1.30–1.72 0.85 0.71–1.03
showed that in HFrEF 9.6% and 18.0% of patients
Baseline eGFR strata
were observed to have a potassium level of >5.5
90þ ml/min per 1.73 m2 Ref. Ref.
or <3.5 mmol/l, respectively, the corresponding pro- 60–89 ml/min per 1.73 m2 1.46† 1.13–1.87 1.11 0.87–1.42
portions in RALES (Randomized Aldactone Evalua- 45–59 ml/min per 1.73 m2 2.01‡ 1.53–2.65 0.96 0.74–1.26
tion Study) were 12.2% (patients with potassium ¼ 5.5 30–44 ml/min per 1.73 m2 2.68‡ 2.02–3.56 1.13 0.85–1.50
mmol/l were included) and 11.4% (mean follow-up at <30 ml/min per 1.73 m2 4.10‡ 3.04–5.53 1.64† 1.22–2.21

24 months), respectively (21), 13.4% and 10.7%, Hemoglobin, <120 g/l 1.43‡ 1.27–1.61 1.23† 1.08–1.40
Hypertension 1.00 0.89–1.13 0.97 0.85–1.11
respectively, in EPHESUS (Eplerenone Post-Acute
Diabetes mellitus 1.33‡ 1.16–1.51 0.96 0.82–1.11
Myocardial Infarction Heart Failure Efficacy and
Myocardial infarction 0.99 0.87–1.12 1.19* 1.03–1.37
Survival Study) (mean follow-up 16 months) (22), and Stroke or TIA 1.05 0.87–1.26 1.01 0.82–1.25
8.9% and 9.3%, respectively, in EMPHASIS-HF PAD 1.20 0.95–1.51 1.12 0.86–1.44
(Eplerenone in Mild Patients Hospitalization and Atrial fibrillation 1.04 0.93–1.16 1.13* 1.00–1.28
Survival Study in Heart Failure) (mean follow-up COPD 1.22† 1.06–1.40 1.34‡ 1.16–1.55

of 21 months) (23). Additionally, in PARADIGM-HF Liver disease 1.26 0.82–1.94 1.24 0.72–2.14
Alcoholism 1.46* 1.05–2.04 1.67† 1.16–2.42
(Prospective Comparison of ARNI with ACEI to
Cancer (3 yrs) 1.30* 1.03–1.64 1.22 0.94–1.58
Determine Impact on Global Mortality and Morbidity
Hospitalization at diagnosis 2.01‡ 1.72–2.36 2.04‡ 1.69–2.45
in Heart Failure Trial), the risk of hyperK was 16.7%
NYHA functional class
(median follow-up of 27 months) (18). In patients with I Ref. Ref.
HFpEF, potassium levels of >5.5 and <3.5 mmol/l II 1.33* 1.06–1.66 1.23 0.98–1.55
were observed in 11.4% and 25.6%, respectively, of III 1.72‡ 1.36–2.17 1.66‡ 1.31–2.10
patients within 1 year, whereas in TOPCAT (Treat- IV 2.05‡ 1.45–2.88 1.99‡ 1.38–2.88

ment of Preserved Cardiac Function Heart Failure EF strata

>50% 1.06 0.92–1.22 1.17* 1.00–1.36
with an Aldosterone Antagonist), the corresponding
40%–49% 0.99 0.86–1.14 1.10 0.94–1.28
proportions were 13.9% (patients with potassium ¼
<40% Ref. Ref.
5.5 mmol/l were included) and 17.5%, respectively, ACE/ARB 1.07 0.93–1.23 0.63‡ 0.54–0.72
over a mean follow-up of 3.3 years (19,24). Addition- Beta-blockers 0.81† 0.70–0.94 0.83* 0.71–0.96
ally, a higher hypoK incidence was found in the pre- MRAs 1.85‡ 1.66–2.07 1.08 0.95–1.22
sent study than in trials, which has received less Diuretics 1.10 0.94–1.29 1.38‡ 1.16–1.64
attention in this patient population but may be
*p < 0.05. †p < 0.01. ‡p < 0.001.
equally or more harmful. CI ¼ confidence interval; HR ¼ hazard ratio; other abbreviations are as in Table 1.
A novel finding in the present study is that the risk
of dysK across the different EF categories was able to
versus 12.2%, respectively, whereas risk of hypoK
be estimated and compared. HF phenotypes reported
(K <3.5 mmol/l) was 17.5% versus 11.4%, respectively,
differences in clinical characteristics and treatment
suggesting higher risk of dysK in HFpEF than in
patterns. After extensive adjustments, patients with
HFrEF (19,21,24), also in trial populations.
HFpEF and HFmrEF were found to have higher risk of
moderate or severe hyperK, whereas those with INDEPENDENT PREDICTORS OF HyperK AND
HFpEF had the highest risk of any hypoK. The present HypoK. Previous post-hoc analyses from trials iden-
study, thus, confirms and expands to real-world tified predictors of hyperK in HFrEF (21–23). Despite
setting the trends observed comparing rates from different definitions of hyperK in those trials, com-
individual trials. For example, in TOPCAT (HFpEF) mon predictors across the different studies were
versus RALES (HFrEF), the crude incidence of mod- baseline potassium, kidney function, and use of MRA
erate or severe hyperK (K >5.5 mmol/l) was 13.9% (21–23). Generally, predictors of hypoK have received
8 Savarese et al.
Dyskalemia in HF
F I G U R E 3 Number and Time Distribution of Dyskalemic Events Registered During 12 Months of Follow-Up
less attention and have only been assessed in RALES
(21) where low baseline potassium, no use of MRA and
ACE-I/ARB use, higher diastolic blood pressure, fe-
male sex and black race were independently associ-
ated with the risk of hypokalemia.
Findings from the present unselected real-world
population of Swedish HF patients collectively
agree with the trial evidence but also provide accu-
rate information for dysK predictors in the less-
characterized categories of patients with HFpEF
and HFmrEF. These specific clinical characteristics
and medications associated with the occurrence of
dysK across the EF spectrum may foster the identi-
fication of high-risk populations more likely to
benefit from dysK monitoring and prevention stra-
tegies, including uptitration of HF therapies. Perhaps
not unexpectedly, it was also observed that higher
baseline potassium levels (4.5 to 5.0 mmol/l) were
associated with increased risk of hyperK, whereas
lower potassium levels (3.5 to 3.9 mmol/l) at baseline
were associated with increased risk of hypoK but
reduced risk of hyperK. Furthermore, worse NYHA
functional class, COPD, low hemoglobin concentra-
tion, and being an inpatient were associated with
both hyperK and hypoK. Unmeasured confounders,
such as treatments for comorbidities not considered
in the current analysis, may explain COPD and low
hemoglobin as independent predictors of both po-
tassium disturbances. However, low hemoglobin
concentration is also a surrogate for increased
congestion (along with low serum protein, albumin,
and hematocrit, all of which have been proposed as
surrogate markers of (de)congestion and have been
found to be associated with cardiovascular end-
points), which is 1 of the main predictors of adverse
outcomes and may trigger changes in kidney func-
tion and in HF medications, with an associated risk
JACC: HEART FAILURE VOL. -, NO. -, 2018
- 2018:-–-
of dysK (25). In line with the importance of kidney
function in clearing circulating potassium, it was
observed in the present study that there was a
monotonic gradual association between hyperK risk
and lower eGFR strata. However, an unexpected
finding was that individuals with advanced or severe
CKD were at higher risk of both hyperK and hypoK.
This may perhaps reflect the more fragile fluid and
electrolyte balance and the more frequent medica-
tion changes, such as increases or variations of
diuretic doses, variable diuretic resistance and
gastrointestinal and renal congestion over time, and/
or reductions of doses or discontinuation of RAAS
inhibitor agents in these patients. The observation
that women are at lower risk of hyperK is consistent
with that in previous studies, but it was also
observed in the present study that women were at
higher risk of hypoK. This may reflect differences in
body composition, resulting in lower total
exchangeable body potassium in women than in
men, or sex differences in comorbidities and treat-
ments that were not captured in the present analysis
(unmeasured confounders). Finally, as previously
reported (26), diabetes was associated with the risk
of hyperK, which may be explained by its role in
decreasing aldosterone production or accelerating
kidney function decline.
Understanding the impact of medication in real-
world data is difficult, as these variables are
affected by both confounding by indication and
reverse causation. For instance, ACE inhibitors/ARBs
did not associate with the risk of hyperkalemia in the
present study, perhaps because their very high use
(80%) did not offer discrimination, but also perhaps
because the rich covariate adjustment covered the
indications for which ACE inhibitors/ARBs were pre-
scribed. On the other hand, use of MRA was
JACC: HEART FAILURE VOL. -, NO. -, 2018 Savarese et al. 9
- 2018:-–- Dyskalemia in HF

F I G U R E 4 Outcomes Associated With Incident Dyskalemic Events

Models were adjusted for age, sex, smoking status, baseline potassium strata, baseline estimated glomerular filtration rate, hemoglobin,
history of hypertension, diabetes mellitus, myocardial infarction, stroke or transient ischemic attack, peripheral artery disease, atrial fibril-
lation, chronic obstructive pulmonary disease, liver disease, alcoholism, cancer, caregiver (inpatient vs. outpatient), left ventricular ejection
fraction strata, New York Heart Association functional class, and medication (renin-angiotensin system inhibitors, b-blocker mineralocorti-
coid receptor antagonists, diuretics). CVD ¼ cardiovascular disease; HF ¼ heart failure; HR ¼ hazard ratio.

associated with higher risk of hyperK, and use of di-
uretics predicted hypoK, in agreement with general
understanding and use of those medications. Beta-
blockers were associated with lower risk of hypoK,
which is coherent with their role in promoting po-
tassium shift out of cells by inhibiting sodium-
potassium ATPase. However, beta-blockers were
also associated with reduced risk of hyperK, which
may be explained by reverse causation.
CLINICAL CONSEQUENCES OF DYSKALEMIA. Inci-
dent hypoK or hyperK was strongly associated with
subsequent risk of death in the present study. Similar
findings were recently shown in an HF Danish
registry-based study (13). Furthermore, we observed a
similar risk of death across the EF spectrum, which
agrees with a subgroup analysis of a post-hoc study of
the DIG (Digitalis Investigation Group) trial, where
EF, categorized as # versus >45%, had no significant
interaction with hypoK (K <4 mEq/l) for prediction of
mortality (27). In addition, TOPCAT trial data showed
increased CVD mortality risk in HFpEF patients
reporting incident hyperK (K >5.5 mmol/l) or hypoK
(K <3.5 mmol/l) (24). Associations between dysK and
subsequent hospitalizations have received even less
investigation. In DIG trial, hypoK (K <4 mEq/l) was
not associated with increased risk of all-cause hospi-
talization (27). Conversely, this present study showed
10 Savarese et al. JACC: HEART FAILURE VOL. -, NO. -, 2018
Dyskalemia in HF - 2018:-–-

T A B L E 3 Clinical Outcomes Associated With Incident Hyperkalemia and Hypokalemia in the Different Ejection Fraction Strata Within 1 Year of Follow-Up

HFpEF HFmrEF HFrEF

Events/Person ER per HR Events/Person ER per HR Events/Person ER per HR

Time 1,000-pys (95% CI)† Time 1,000-pys (95% CI)† Time 1,000-pys (95% CI)†

Incident hyperkalemia*
Death 128/396,533 0.32 Ref. 114/379,001 0.30 Ref. 313/882,810 0.35 Ref.
103/64,457 1.60 4.97§ (3.51–7.06) 86/54,751 1.57 4.79§ (3.36–6.82) 203/151,641 1.34 3.62§(2.89–4.55)
HF hospitalization 18/393,611 0.05 Ref. 21/374,768 0.06 – 262/832,622 0.31 Ref.
5/63,125 0.08 1.52 (0.56–4.10) 1/53,568 0.02 – 32/134,204 0.24 0.98 (0.65–1.47)
Other CVD hospitalization 18/393,611 0.05 Ref. 24/375,320 0.06 – 133/860,206 0.15 Ref.
5/63,125 0.08 1.52 (0.56–4.10) 3/53,849 0.05 – 25/143,799 0.17 1.18 (0.73–1.91)
Incident hypokalemia*
Death 145/390,204 0.37 Ref. 129/383,220 0.34 Ref. 345/926,293 0.37 Ref.
86/70,786 1.21 3.00§ (2.19–4.11) 71/50,532 1.41 3.62§ (2.57–3.10) 171/108,158 1.58 3.29§(2.62–4.13)
HF hospitalization 15/387,759 0.04 – 20/379,649 0.05 – 272/871,338 0.31 Ref.
1/69,156 0.01 – 2/48,687 0.04 – 22/95,488 0.23 0.96 (0.62–1.50)
Other CVD hospitalization 17/387,918 0.04 Ref. 23/379,821 0.06 – 134/903,472 0.15 Ref.
6/68,818 0.09 1.68 (0.48–5.85) 4/49,348 0.08 – 24/100,533 0.24 1.99‡ (1.26–3.18)

Negative HR could not be calculated because of the limited number of events. *Models were adjusted as in Figure 4. †p < 0.05. ‡p < 0.01. §p < 0.001.
CVD ¼ cardiovascular disease; ER ¼ event rate; HF ¼ heart failure; HR ¼ hazard ratio; pys ¼ patient-years.

increased risk of CVD (excluding HF) but not HF
hospitalizations in hypoK patients (K <3.5 mmol/l),
particularly in patients with HFrEF. This finding may
be explained by higher mortality associated with
hyperK than with hypoK, and thus, death more likely
to circumvent hospitalization after hyperK than
hypoK.
STUDY LIMITATIONS. The unselected inclusion and
rich patient characterization of the present study is
a strength. Additionally, plasma (and not serum)
potassium levels were measured, because plasma
levels are not affected by leukocytosis and throm-
bocytosis and therefore the likelihood of false
hyperK is low. However, all observational data are
subject to selection bias, and the influence of un-
measured confounders cannot be ruled out. Detec-
tion bias can also exist in outcome ascertainment, as
patients at higher risk of events may be more likely
to use health care and have potassium levels
measured. Patients were included in SwedeHF on
the basis of clinician-judged HF, thus it is possible
that some patients, particularly those with HFpEF,
might not have had HF. HF phenotype was defined
according to the EF reported in the SwedeHF regis-
tration recorded on the same day that potassium
plasma levels were assessed. The limited amount of
longitudinal data available in the overall SwedeHF
population prevented performing separate analyses
in patients with transitioning EF. Furthermore,
use of therapies was recorded at the time that
plasma potassium levels were
changes in treatments and doses affecting potassium
homeostasis over time cannot be ruled out. How-
ever, these changes might have had the same
chance to occur across the different HF phenotypes
and not had a major impact on comparative results.
Finally, analysis was limited to years 2006 to
2011, which prevented inclusion of data for
ARN inhibitors, which have been shown to be
associated with a lower risk of hyperK than ACE
inhibitors/ARB.
CONCLUSIONS
The present study offers estimates on the incidence,
predictors, and outcomes associated with dysK in
real-world HF patients with different EF. These real-
world data may be helpful to tailor RAAS inhibitor
therapy, fostering identification of patients in need
of stricter potassium monitoring while they are
receiving therapies that affect potassium homeosta-
sis and those who are more likely to develop dysK
following changes in therapies and doses. Finally,
present analyses may support the design of trials of,
for example, RAAS inhibitors and K-binder agents in
HF by providing important insights regarding po-
tential populations to be enrolled.
ADDRESS FOR CORRESPONDENCE: Dr. Lars H. Lund,
Department of Medicine, Karolinska Institutet, and
Heart and Vascular Theme, Karolinska University
Hospital, FoU Tema Hjärta Kärl, Norrbacka, S1:02,
measured; thus, 171 76 Stockholm, Sweden. E-mail: Lars.Lund@
alumni.duke.edu.
JACC: HEART FAILURE VOL. -, NO. -, 2018 Savarese et al. 11
- 2018:-–- Dyskalemia in HF

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE: Patients TRANSLATIONAL OUTLOOK: This study provides

with HFpEF, HFmrEF, and HFrEF exhibit different de-
mographics and clinical characteristics and are treated
differently, possibly leading to different rates of dysK.
Fear of dysK in these patients may lead to underuse or
underdosing of RAAS inhibitors. The present authors
reported that dysKs are common in HF and are associ-
ated with increased mortality. Risk of moderate or se-
vere hyperK was highest in HFpEF and HFmrEF, whereas
risk of hypoK was highest in HFpEF. HF severity, low
hemoglobin, COPD, high and low potassium levels at
baseline, and low eGFR were predictors of dysK
occurrence.
data for incidence, predictors, and outcomes associated
with dysK in real-world HFpEF, HFmrEF, and HFrEF
patients that may be helpful to tailor RAAS inhibitor
therapy, fostering identification of those who are more in
need of stricter potassium monitoring while receiving
therapies that affect potassium homeostasis and those
who are more likely to develop dysK following change in
therapies and doses. These data may also support the
design of trials of, for example, RAAS inhibitors and K-
binder agents in HF.

REFERENCES

1. Palmer BF. Regulation of potassium homeosta-
sis. Clin J Am Soc Nephrol 2015;10:1050–60.
2. Bielecka-Dabrowa A, Mikhailidis DP, Jones L,
Rysz J, Aronow WS, Banach M. The meaning of
hypokalemia in heart failure. Int J Cardiol 2012;
158:12–7.
3. Sarwar CM, Papadimitriou L, Pitt B, et al.
Hyperkalemia in heart failure. J Am Coll Cardiol
2016;68:1575–89.
4. Nilsson E, Gasparini A, Arnlov J, et al. Incidence
and determinants of hyperkalemia and hypokale-
mia in a large healthcare system. Int J Cardiol
2017;245:277–84.
5. Koh AS, Tay WT, Teng THK, et al.
A comprehensive population-based characteriza-
tion of heart failure with mid-range ejection frac-
tion. Eur J Heart Fail 2017;19:1624–34.
6. Nishihara T, Tokitsu T, Sueta D, et al. Serum
potassium and cardiovascular events in heart
failure with preserved left ventricular ejection
fraction patients. Am J Hypertens 2018;31:
1098–105.
7. Pitt B, Rossignol P. Relation of serum potassium
to cardiovascular events in patients with heart
failure and preserved ejection fraction: “mind the
gap.” Am J Hypertens 2018;31:1087–9.
8. Nallamothu BK, Hayward RA, Bates ER.
Beyond the randomized clinical trial: the role
of effectiveness studies in evaluating cardio-
vascular therapies. Circulation 2008;118:
1294–303.
9. Thorvaldsen T, Benson L, Dahlstrom U,
Edner M, Lund LH. Use of evidence-based therapy
and survival in heart failure in Sweden 2003-2012.
Eur J Heart Fail 2016;18:503–11.
10. Trevisan M, de Deco P, Xu H, et al. Incidence,
predictors and clinical management of hyper-
kalaemia in new users of mineralocorticoid re-
ceptor antagonists. Eur J Heart Fail 2018;20:
1217–26.
11. Jonsson A, Edner M, Alehagen U, Dahlstrom U.
Heart failure registry: a valuable tool for
improving the management of patients with heart
failure. Eur J Heart Fail 2010;12:25–31.
12. Ponikowski P, Voors AA, Anker SD, et al.
2016 ESC Guidelines for the diagnosis and
treatment of acute and chronic heart failure: the
task force for the diagnosis and treatment of
acute and chronic heart failure of the European
Society of Cardiology (ESC). Developed with the
special contribution of the Heart Failure Associ-
ation (HFA) of the ESC. Eur J Heart Fail 2016;18:
891–975.
13. Aldahl M, Jensen AC, Davidsen L, et al. Asso-
ciations of serum potassium levels with mortality
in chronic heart failure patients. Eur Heart J 2017;
38:2890–6.
14. Rockhill B, Newman B, Weinberg C. Use and
misuse of population attributable fractions. Am J
Public Health 1998;88:15–9.
15. Pitt B, Remme W, Zannad F, et al. Eplerenone,
a selective aldosterone blocker, in patients with
left ventricular dysfunction after myocardial
infarction. N Engl J Med 2003;348:1309–21.
16. Zannad F, McMurray JJ, Krum H, et al.
Eplerenone in patients with systolic heart failure
and mild symptoms. N Engl J Med 2011;364:
11–21.
17. Pitt B, Zannad F, Remme WJ, et al. The effect
of spironolactone on morbidity and mortality in
patients with severe heart failure. Randomized
Aldactone Evaluation Study Investigators. N Engl J
Med 1999;341:709–17.
18. McMurray JJ, Packer M, Desai AS, et al.
Angiotensin-neprilysin inhibition versus enalapril
in heart failure. N Engl J Med 2014;371:
993–1004.
19. Pitt B, Pfeffer MA, Assmann SF, et al. Spi-
ronolactone for heart failure with preserved
ejection fraction. N Engl J Med 2014;370:
1383–92.
20. Desai AS, Swedberg K, McMurray JJ, et al.
Incidence and predictors of hyperkalemia in pa-
tients with heart failure: an analysis of the
CHARM program. J Am Coll Cardiol 2007;50:
1959–66.
21. Vardeny O, Claggett B, Anand I, et al. Inci-
dence, predictors, and outcomes related to hypo-
and hyperkalemia in patients with severe heart
failure treated with a mineralocorticoid receptor
antagonist. Circ Heart Fail 2014;7:573–9.
22. Pitt B, Bakris G, Ruilope LM, DiCarlo L,
Mukherjee R, Investigators E. Serum potassium
and clinical outcomes in the Eplerenone Post-
Acute Myocardial Infarction Heart Failure Efficacy
and Survival Study (EPHESUS). Circulation 2008;
118:1643–50.
23. Rossignol P, Dobre D, McMurray JJ, et al.
Incidence, determinants, and prognostic signifi-
cance of hyperkalemia and worsening renal
function in patients with heart failure receiving
the mineralocorticoid receptor antagonist epler-
enone or placebo in addition to optimal medical
therapy: results from the Eplerenone in Mild
Patients Hospitalization and Survival Study in
Heart Failure (EMPHASIS-HF). Circ Heart Fail
2014;7:51–8.
12 Savarese et al.
Dyskalemia in HF
24. Desai AS, Liu J, Pfeffer MA, et al. Incident
hyperkalemia, hypokalemia, and clinical outcomes
during spironolactone treatment of heart failure
with preserved ejection fraction: analysis of the
TOPCAT trial. J Card Fail 2018;24:313–20.
25. Girerd N, Seronde MF, Coiro S, et al. Integra-
tive assessment of congestion in heart failure
throughout the patient journey. J Am Coll Cardiol
HF 2018;6:273–85.
JACC: HEART FAILURE VOL.
26. Sowers JR, Epstein M. Diabetes mellitus and
associated hypertension, vascular disease, and
nephropathy. An update. Hypertension 1995;26:
869–79.
27. Alper AB, Campbell RC, Anker SD, et al.
A propensity-matched study of low serum po-
tassium and mortality in older adults with
chronic heart failure. Int J Cardiol 2009;137:
1–8.
-, NO. -, 2018
- 2018:-–-
KEY WORDS dyskalemia, heart failure,
hyperkalemia, hypokalemia, mid-range
ejection fraction, preserved ejection fraction,
reduced ejection fraction, SCREAM, SwedeHF
A PPE NDI X For an expanded Methods as well
as supplemental figures and tables, please see
the online version of this paper.