Hypokalemia, Its Contributing Factors and Renal

Outcomes in Patients with Chronic Kidney Disease

Hsiao-Han Wang1,6., Chi-Chih Hung1,4., Daw-Yang Hwang1, Mei-Chuan Kuo1,3, Yi-Wen Chiu1,3, Jer-
Ming Chang1,2,3, Jer-Chia Tsai1,4, Shang-Jyh Hwang1,3, Julian L. Seifter5, Hung-Chun Chen1,3*
1 Division of Nephrology, Kaohsiung Medical University Hospital, Kaohsiung Medical University, Kaohsiung, Taiwan, 2 Department of Internal Medicine, Kaohsiung
Municipal Hsiao-Kang Hospital, Kaohsiung Medical University, Kaohsiung, Taiwan, 3 Faculty of Renal Care, College of Medicine, Kaohsiung Medical University, Kaohsiung,
Taiwan, 4 Faculty of Medicine, College of Medicine, Kaohsiung Medical University, Kaohsiung, Taiwan, 5 Renal Division, Department of Medicine, Brigham and Women’s
Hospital and Harvard Medical School, Boston, Massachusetts, United States of America, 6 Department of Dermatology, Wan Fang Hospital, Taipei Medical University,
Taipei, Taiwan

Abstract
Background: In the chronic kidney disease (CKD) population, the impact of serum potassium (sK) on renal outcomes has
been controversial. Moreover, the reasons for the potential prognostic value of hypokalemia have not been elucidated.

Design, Participants & Measurements: 2500 participants with CKD stage 1–4 in the Integrated CKD care program
Kaohsiung for delaying Dialysis (ICKD) prospective observational study were analyzed and followed up for 2.7 years.
Generalized additive model was fitted to determine the cutpoints and the U-shape association between sK and end-stage
renal disease (ESRD). sK was classified into five groups with the cutpoints of 3.5, 4, 4.5 and 5 mEq/L. Cox proportional hazard
regression models predicting the outcomes were used.

Results: The mean age was 62.4 years, mean sK level was 4.260.5 mEq/L and average eGFR was 40.6 ml/min per 1.73 m2.
Female vs male, diuretic use vs. non-use, hypertension, higher eGFR, bicarbonate, CRP and hemoglobin levels significantly
correlated with hypokalemia. In patients with lower sK, nephrotic range proteinuria, and hypoalbuminemia were more
prevalent but the use of RAS (renin-angiotensin system) inhibitors was less frequent. Hypokalemia was significantly
associated with ESRD with hazard ratios (HRs) of 1.82 (95% CI, 1.03–3.22) in sK ,3.5mEq/L and 1.67 (95% CI,1.19–2.35) in
sK = 3.5–4 mEq/L, respectively, compared with sK = 4.5–5 mEq/L. Hyperkalemia defined as sK .5 mEq/L conferred 1.6-fold
(95% CI,1.09–2.34) increased risk of ESRD compared with sK = 4.5–5 mEq/L. Hypokalemia was also associated with rapid
decline of renal function defined as eGFR slope below 20% of the distribution range.

Conclusion: In conclusion, both hypokalemia and hyperkalemia are associated with increased risk of ESRD in CKD
population. Hypokalemia is related to increased use of diuretics, decreased use of RAS blockade and malnutrition, all of
which may impose additive deleterious effects on renal outcomes.

Citation: Wang H-H, Hung C-C, Hwang D-Y, Kuo M-C, Chiu Y-W, et al. (2013) Hypokalemia, Its Contributing Factors and Renal Outcomes in Patients with Chronic
Kidney Disease. PLoS ONE 8(7): e67140. doi:10.1371/journal.pone.0067140
Editor: Emmanuel A. Burdmann, University of Sao Paulo Medical School, Brazil
Received November 1, 2012; Accepted May 14, 2013; Published July 2, 2013
Copyright: ß 2013 Wang et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: [email protected]
. These authors contributed equally to this work.

Introduction hypokalemia and death as well as ESRD have been proposed

[4,5,8]. One paper demonstrated hypokalemia was associated with
The kidney plays a major role in potassium homeostasis by ESRD but this effect was attenuated after adjusting nutritional
renal mechanisms that transport and regulate potassium secretion, indices in patients with CKD stage 3 to 5 [4]. Another study
reabsorption and excretion [1]. Hyperkalemia is a common suggested hypokalemia was associated with renal progression but
electrolyte disturbance in patients with chronic kidney disease the association with hard renal outcome, such as reaching ESRD
(CKD) [2]. As eGFR decreases from above 60 to below 20 ml/ was unclear [5].
min/1.73 m2, the prevalence of hyperkalemia increases from 2 to Recent studies have reported common predispositions such as
42% [3]. In CKD individuals, a few studies suggested weak links diabetes, high dietary potassium and renin-angiotensin-aldoste-
between hyperkalemia and ESRD [4,5]. rone system blockers use for the development of hyperkalemia in
Chronic hypokalemia, on the other hand, has been reported to pateints with impaired renal functions [9–12]. Kaliuretic diuretics
enhance renal cytogenesis and may lead to interstitial scarring and such as furosemide and thiazides are common causes of
renal insufficiency [6,7]. Recently, in CKD patients comorbid with hypokalemia [13,14]. Diarrhea, vomiting, hyperaldosteronism,
and without cardiovascular diseases, the associations between magnesium deficiency and potassium redistribution induced by

PLOS ONE | www.plosone.org 1 July 2013 | Volume 8 | Issue 7 | e67140

 Hypokalemia and Renal Outcomes in CKD Patients

insulin, alkalosis and/or b-adrenergic activation, are all possible Quantification of Renal Function and Progression
prerequisites for hypokalemia [15]. However, studies on hypoka- Kidney function was quantified by using the estimated
lemia and aforementioned associated factors were done mostly in glomerular filtration rate (eGFR) derived from the simplified
general population. Moreover, the possible complex interplay Modification of Diet in Renal Disease (MDRD) Study equation.
between hypokalemia, its associating factors and renal outcomes The equation was eGFR ml/min/1.73 m2 = 186 6 Serum
has not been investigated. creatinine 21.154 6 Age 20.203 6 0.742 (if female) 6 1.212 (if
In the CKD population, whether sK level is associated with black patient). The average eGFR slope (ml/min/1.73 m2/yr) for
greater risks of renal outcomes has not been clearly defined. each patient was calculated by linear regression with varying-
Moreover, the verification of the reasons for the potential intercept and varying-slope without co-variates for estimation of
prognostic value of hypokalemia remains to be elucidated. Thus, the annual change of eGFR.
we investigated the contributing factors of hypokalemia and
whether hyperkalemia or hypokalemia is a risk factor for adverse Outcomes
renal outcomes in patients with CKD stage 1 to 4. Renal outcomes were assessed: end stage renal disease (ESRD)
and rapid renal progression. ESRD was defined as the initiation of
Methods hemodialysis, peritoneal dialysis or renal transplantation. The
initiation of renal replacement therapy was discovered by review of
Participants and Measurement charts or catastrophic cards. Rapid renal progression was defined
Integrated CKD care program Kaohsiung for delaying Dialysis as an eGFR slope,26.88 ml/min/1.73 m2/yr which was the
(ICKD) study was designed as a prospective cohort to investigate 20th percentile. Other cut-off values were also applied in
the impact of integrated CKD care program on clinical outcomes sensitivity tests. The timing for ESRD was according to the
from a diverse group of CKD stage 1–5 patients. The included regulation of Bureau of the National Health Insurance of Taiwan
population was CKD patients not on renal replacement therapy. regarding the laboratory data, nutritional status, uremic status and
The exclusion criterion was acute kidney injury defined as more serum creatinine (.6 mg/dL). ESRD was ascertained by review-
than 50% decrease in eGFR in three months. The study recruited ing charts and by matching with the database from Taiwan
patients from the nephrology out-patient departments of two Society of Nephrology.
hospitals in southern Taiwan. Between November 11, 2002 and
May 31, 2009, 3749 patients received integrated CKD care
program and were followed up until July 31, 2010. 90 patients
Statistical Analysis
were lost for follow-ups in less than 3 months and 1159 stage Statistic results of baseline characteristics of all subjects were
5 CKD patients were excluded. Total 2500 subjects with CKD expressed as percentages for categorical variables, mean 6
stage 1 to 4 were analyzed. The study protocol was approved by standard deviation (SD) for continuous variables with approxi-
the institutional review board of the Kaohsiung Medical Univer- mately-normal distribution, and median and interquartile range
sity Hospital (KMUH-IRB-990198). Informed consents have been for continuous variables with skewed distribution.
obtained in written form from patients and all clinical investigation We conducted both linear regression and generalized additive
was conducted according to the principles expressed in the models (GAM) to investigate the variables associated with sK.
Declaration of Helsinki. The patients gave consent for the GAM were fitted to detect nonlinear effects of continuous
publication of the clinical details. covariates (VGAM function of the VGAM package) [16,17] for
At the baseline visit, socio-demographic characteristics, medical sK because GAM were developed for smoothing the effects of
history, lifestyle behaviors and current medications were recorded. covariates in generalized linear models. The selection of variables
The medical history was confirmed by doctors’ chart-review. was based on previous models and also the nutritional and
Diabetes mellitus (DM) and hypertension were defined by clinical medication variables that we hypothesized.
diagnosis. Cardiovascular disease (CVD) was defined as clinical Cox proportional hazards models were used to evaluate
diagnosis of heart failure, acute or chronic ischemic heart disease, association of variables with time to ESRD. Covariates included
or cerebrovascular disease. Body Mass Index (BMI) was calculated in these models were age, gender, DM, CVD, eGFR, Urine
from the baseline enrolled measurement using the formula: weight protein by dipstick, Angiotensin-converting enzyme inhibitor
in kilograms/(height in meters)2. Mean blood pressure (MBP) was (ACEI), Angiotensin II receptor blocker (ARB), diuretics, MBP,
calculated by the averages of systolic and diastolic BP measured 3 BMI, HbA1c, log-transformed cholesterol, log-transformed CRP,
months before and after the enrollment using the formula: 2/3 phosphate, hemoglobin and albumin. Generalized additive models
average systolic BP +1/3 average diastolic BP. with restricted cubic regression splines were used to explore the
Biochemistry measurements were done on screening visit, functional form of the relationship between sK and ESRD. The
baseline visit and then every 3 months as the protocol. The knots for sK were at 4.03 and 5.11 mEq/L and we thus divided sK
laboratory data from 3 months before baseline visit to 3 months accordingly in the tables. Binary logistic regression analysis was
after baseline visit were averaged and analyzed. Six-month used to assess the relationship between sK and rapid renal
averaged sK was used for analysis and was classified into five progression. Covariates included into these models were the same
groups as sK ,3.5; K = 3.5–4; K = 4–4.5, K = 4.5–5 and K as those in Cox regression.
.5 mEq/L. The cutpoints were defined by generalized additive Also, the statistical tools for regression diagnostics such as
regression. Serum creatinine, albumin, cholesterol, C-reactive verification of proportional hazards assumption, residual analysis,
protein (CRP), potassium and phosphate levels were determined detection of influential cases, and check for multicollinearity were
by means of the Toshiba TBA C16000 chemistry analyzer. Serum applied to discover model or data problems. Statistical analysis was
hemoglobin was determined by means of the Sysmex XE-2100 performed using the R 2.15.2 software (R Foundation for
automated hematology analyzer. HbA1C was measured by Statistical Computing, Vienna, Austria) and Statistical Package
Primus CLC 385 automated analyzer. Venous blood bicarbonate for Social Sciences version 18.0 for Windows (SPSS Inc., Chicago,
was measured by AVL Omni 3 blood gas analyzer. IL).

PLOS ONE | www.plosone.org 2 July 2013 | Volume 8 | Issue 7 | e67140

 Hypokalemia and Renal Outcomes in CKD Patients

Results (Table 3). Non-linear effects of continuous variables were tested by

generalized additive models (Table S1 and Figure S1). sK was
Baseline Characteristics of CKD Patients Divided by negatively correlated with eGFR, MBP, bicarbonate, CRP and
Serum Potassium Quintiles hemoglobin and positively correlated with HbA1c and albumin.
A total of 2500 participants in CKD stage 1–4 were analyzed The use of ACEI, ARB, and insulin vs those non-users were
during a median 2.7-year of follow-up. Figure 1 displayed the associated with higher sK. Female vs. male and diuretics users vs.
distribution of five sK groups by CKD stage 1–4. There was still non-users were associated with lower sK (Table 3 and Table S1).
5.6% of participants with sK ,3.5 mEq/L in those with CKD
stage 4, while there was less than 1% of patients with sK .5 mEq/ Potassium and Renal Replacement Therapy
L in those with CKD stage 1. The majority of the participants There were 299 patients (12%) reaching ESRD during a
were male (64%), had hypertension (61.5%) with a mean age of median 2.7-year of follow-up. To explore the non-linear associ-
62.4614.5 years, mean sK of 4.260.5 mEq/L and mean eGFR of ation of sK and ESRD, we fitted a generalized additive model with
40.6623.3 ml/min/1.73 m2 (Table 1). Patients with lower sK a restricted cubic spline regression (Figure 3). The result revealed a
were more likely to have hypertension, higher eGFR, higher U-shaped association, with increased risk of ESRD when sK
serum hemoglobin, cholesterol and bicarbonate levels, whereas ,4.03 mEq/L and sK .5.11 mEq/L. Consistent with these
they were less likely to have a history of diabetes, to have higher associations, when compared to sK = 4.5–5mEq/L, both sK
HbA1c and phosphate levels, in comparison with those with ,3.5 mEq/L and sK = 3.5–4 mEq/L were significantly associated
higher sK. Figure 2 further demonstrated the relationship between with 82% (HR 1.82; CI 95% 1.03–3.22; p = 0.041) and 67% (HR
sK and other important predictive covariates. There were higher 1.67; CI 95% 1.19–2.35; p = 0.003) excess risk of ESRD,
proportions of patients presenting with hypoalbuminemia respectively in fully-adjusted Cox regression (Table 4). sK
(,3.5 mg/dL) and lower BMI (20 kg/m2) in the groups with sK .5mEq/L was also associated with increased risk of becoming
,3.5 and .5mEq/L (Figure 2A). ESRD with HR of 1.60 (CI 95% 1.09–2.34).
In total participants, ESRD rate was 43 per 1000 person-years Other risk factors for ESRD revealed in the multivariate Cox
(Table 1). There was a trend of greater crude rates of ESRD and regression analysis included diabetes, decreased eGFR, diuretic
steeper decline in eGFR in groups with higher sK. There were no use, higher MBP, proteinuria (by dipstick or by urine protein-to-
significant univariate linear relationships between sK groups and creatinine ratio (UPCR)), hyperphosphatemia, higher HbA1c,
rapid renal progression and increment of proteinuria respectively. lower BMI, anemia and hypoalbuminemia (Table 4). We also
However, greater increase of proteinuria and higher proportion of tested the proportional hazard assumption and linearity assump-
participants with rapid renal progression were observed in the tion. The results were shown in Supplementary Table 2. Three
groups with sK ,3.5 mEq/L and sK .5 mEq/L with p for Chi- variables including gender, albumin and bicarbonate were against
Square as p = 0.009 and p,0.001 respectively. proportional hazard assumption and were treated by stratification.
The percentages of patients receiving ACEI, ARB, oral Four variables including serum potassium, BMI, mean blood
hypoglycemic agents and insulin were higher in those with more pressure and C-reactive protein had non-linear relationships with
elevated sK (Table 2). The proportions of other medication use ESRD (Table S2 and Figure S2). An alternative model with
such as furosemide, thiazide, phosphate binder and b-blocker did UPCR showed similar results (data not shown).
not show significant trends among groups. Despite similar MBP
across five groups of sK, the use of anti-hypertensive medication Subgroup Analysis for Hypokalemia and Renal
was less frequent while the use of diuretics was more prevalent as Replacement Therapy
sK decreased (Figure 2B). The prevalence of nephrotic range The association between the sK ,3.5mEq/L and ESRD was
proteinuria was the highest in the groups with sK ,3.5 and significant in patients with ARB use (HR 3.14; 1.36–7.24;
.5 mEq/L but the use of RAS inhibitors was the least in sK p,0.001), but not in those without (HR 1.32; 0.58–2.97;
,3.5 mEq/L (Figure 2C). p = 0.508; p for interaction, 0.041). The other pre-specified
All the univariate significant and non-significant but relevant subgroups did not show significant interaction.
covariates were tested in multivariate linear regression for sK
Potassium and Rapid Renal Progression
The median eGFR slope was 21.8 ml/min/1.73 m2/yr. There
were 492 participants (20%) experiencing eGFR slope less than
26.88 ml/min/1.73 m2/yr. Serum K lower than 3.5 mEq/L was
significantly associated with faster yearly decline in eGFR with
odds ratio (OR) of 1.69 (CI 95% 1.06–2.70; p = 0.027) compared
with sK = 4.5–5 mEq/L after full adjustment (Table 4). On the
other hand, sK .5mEq/L didn’t show significant association with
rapid progression of renal function. Additional risk factors for
rapid renal progression included diabetes, higher MBP, less ARB
use, proteinuria, hyperphosphatemia, higher HbA1c, anemia and
hypoalbuminemia.

Discussion
In our cohort, we examined the associations between sK, its
associating factors and renal outcomes in 2500 patients with CKD
Figure 1. Serum Potassium Distribution by Chronic Kidney stage 1–4. We demonstrated that diuretic use vs. non-use,
Disease (CKD) Stages. hypertension, higher bicarbonate, hemoglobin and CRP levels
doi:10.1371/journal.pone.0067140.g001 significantly correlated with hypokalemia. The use of ACEI, ARB,

PLOS ONE | www.plosone.org 3 July 2013 | Volume 8 | Issue 7 | e67140

 Table 1. Baseline Demographic and Clinical Characteristics.

sK ,3.5 mEq/L sK = 3.5–4 mEq/L sK = 4–4.5 mEq/L sK = 4.5–5 mEq/L

Variable All (n = 2500) (n = 194) (n = 683) (n = 974) (n = 474) sK .5 mEq/L (n = 175) P for trend

Demographics and Medical History

Age (years) 62.4614.5 62.1614.9 61.3615.0 62.7614.1 63.0614.3 63.8614.7 0.102
Gender (female %) 899 (36.0) 73 (37.6) 251 (36.7) 354 (36.3) 163 (34.4) 58 (33.1) 0.234
Hypertension (%) 1538 (61.5) 131 (67.5) 434 (63.6) 586 (60.2) 281 (59.3) 106 (60.6) 0.038
Diabetes mellitus (%) 1045 (41.8) 70 (36.1) 260 (38.1) 404 (41.5) 225 (47.5) 86 (49.1) ,0.001
Cardiovascular disease (%) 564 (22.6) 43 (22.2) 161 (23.6) 206 (21.1) 101 (21.3) 53 (30.3) 0.462
Physical Examination

PLOS ONE | www.plosone.org

BMI (kg/m2) 25.164.0 25.164.2 25.263.9 25.064.0 25.264.1 24.864.3 0.578
MBP (mmHg) 99.4613.4 100.5614.2 99.6613.9 98.9613.2 99.6613.5 98.9611.8 0.279
Renal Function Status
CKD stage ,0.001
Stage 1 (%) 118 (4.7) 14 (7.2) 61 (8.9) 38 (3.9) 4 (0.8) 1 (0.6)
Stage 2 (%) 238 (9.5) 35 (18.0) 96 (14.1) 90 (9.2) 14 (3.0) 90 (9.2)
Stage 3 (%) 1183 (47.3) 91 (46.9) 319 (46.7) 496 (50.9) 234 (49.4) 496 (50.9)
Stage 4 (%) 961 (38.4) 54 (27.8) 207 (30.3) 350 (35.9) 222 (46.8) 350 (35.9)
eGFR (mL/min/1.73 m2) 40.6623.3 47.9628.2 47.1628.2 40.5621.2 33.4614.5 27.2612.3 ,0.001
Proteinuria by dipstick (%)
none 920 (36.8) 77 (39.7) 267 (39.1) 360 (37.0) 163 (34.4) 360 (37.0) ,0.001
+ 652 (26.1) 55 (28.4) 185 (27.1) 262 (26.9) 117 (24.7) 262 (26.9)

4
++ 481 (19.2) 24 (12.4) 122 (17.9) 202 (20.7) 90 (19.0) 202 (20.7)
+++,++++ 446 (17.8) 38 (19.6) 108 (15.8) 150 (15.4) 104 (21.9) 150 (15.4)
Laboratory Data
Albumin (mg/dL) 3.960.5 3.860.7 3.960.6 3.960.5 3.960.5 3.860.5 0.595
Hemoglobin (mg/dL) 12.262.2 12.662.3 12.662.2 12.362.1 11.862.2 10.961.8 ,0.001
Total cholesterol (mg/dL) 201.4658.2 212.2684.2 204.5662.3 199.2650.3 197.5656.0 201.3652.1 0.022
CRP (mg/L) 1.0 (0.324.7) 1.5 (0.524.9) 1.1 (0.424.6) 0.9 (0.324.0) 1.0 (0.2 2 5.5) 0.9 (0.3 2 4.0) 0.205
HbA1c (%) 6.761.7 6.561.8 6.561.7 6.661.7 7.061.8 6.761.7 0.007
Potassium (mEq/L) 4.260.5 3.360.2 3.860.1 4.360.1 4.760.1 5.360.3 ,0.001
Bicarbonate (mEq/L) 23.763.7 25.363.6 24.463.8 23.763.4 23.163.7 21.363.7 ,0.001
Phosphate (mg/dL) 3.960.8 3.760.9 3.860.8 3.860.8 4.160.8 4.460.9 ,0.001
Surrogate endpoints
eGFR slope (mL/min/1.73 m2/yr) 21.8 (25.620.7) 21.0 (27.921.7) 21.5 (25.321.4) 21.8 (25.220.6) 21.9 (25.1 2 0.3) 23.5 (27.8 2 21.0) ,0.001
Rapid renal progression* 492 (20.0) 50 (26.2) 129 (19.3) 173 (18.1) 92 (19.5) 48 (27.9) 0.765
Proteinuria increment# 255.06489.7 357.86830.8 278.36537.3 216.06376.8 225.06349.5 347.86622.1 0.489
Primary endpoints
ESRD (%) 299 (12.0) 16 (8.2) 69 (10.1) 93 (9.5) 69 (14.6) 52 (29.7) ,0.001
Rate per 1000 person-Years 43 37 39 33 47 109

Data expressed as mean 6 standard deviation, median (interquartile range) or percentage.

Hypokalemia and Renal Outcomes in CKD Patients

July 2013 | Volume 8 | Issue 7 | e67140

BMI, body mass index; MBP, mean blood pressure; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; CRP, C-reactive protein; HbA1c, glycated hemoglobin; UPCR, Urine protein-to-creatinine ratio; ESRD, End
Stage Renal Disease.
Comparisons are made by ANOVA or the chi-square test.
*eGFR slope less than 26.88 mL/min/1.73 m2/yr (%).
#
Urine protein-to-creatinine (mg/g).
doi:10.1371/journal.pone.0067140.t001
 ated by chi-square tests.

PLOS ONE | www.plosone.org

doi:10.1371/journal.pone.0067140.g002
Other Predictive Covariates. Figure 2A. Percentages of malnutrition

across potassium groups. Figure 2C. Proportions of ACEI, ARB and

index (BMI) less than 20 kg/m2 across potassium groups. Figure 2B.

proteinuria with nephrotic range across potassium groups. ACEI,

hypertensive (anti-HTN) medication and mean blood pressure (MBP)
Proportions of diuretics use (Thiazide and/or Furosemide use), anti-
indices including serum albumin lower than 3.5 mg/dL and body mass
Figure 2. The Relationships between Serum Potassium and

blocker; UPCR, Urine protein-to-creatinine ratio. P-values were gener-

angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor

5
Table 2. Use of Medications.

sK ,3.5 mEq/L sK = 3.524 mEq/L sK = 424.5 mEq/L sK = 4.525 mEq/L sK .5 mEq/L

All (n = 2500) (n = 194) (n = 683) (n = 974) (n = 474) (n = 175) P for trend

ACEI (%) 726 (29.0%) 43 (22.2) 186 (27.2) 298 (30.6) 146 (30.8) 53 (30.3) 0.026
ARB (%) 1004 (40.2%) 54 (27.8) 263 (38.5) 388 (39.8) 222 (46.8) 77 (44.0) ,0.001
Furosemide (%) 313 (12.5%) 29 (14.9) 90 (13.2) 120 (12.3) 52 (11.0) 22 (12.6) 0.067
Thiazide (%) 237 (9.5%) 24 (12.4) 72 (10.5) 87 (8.9) 36 (7.6) 18 (10.3) 0.051
Oral hypoglycemic agents (%) 725 (29.0%) 42 (21.6) 165 (24.2) 271 (27.8) 180 (38.0) 67 (38.3) ,0.001
Insulin (%) 153 (6.1%) 7 (3.6) 17 (2.5) 61 (6.3) 50 (10.5) 18 (10.3) ,0.001
Phosphate binders (%) 310 (12.4%) 16 (8.2) 92 (13.5) 112 (11.5) 63 (13.3) 27 (15.4) 0.161
b-blocker (%) 571 (22.8%) 43 (22.2) 169 (24.7) 210 (21.6) 118 (24.9) 31 (17.7) 0.386

Data expressed as mean 6 standard deviation, median (interquartile range) or percentage.

ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker.
doi:10.1371/journal.pone.0067140.t002
Hypokalemia and Renal Outcomes in CKD Patients

July 2013 | Volume 8 | Issue 7 | e67140

 Hypokalemia and Renal Outcomes in CKD Patients

Table 3. Linear Regression of Serum Potassium.

Multivariate linear regression

Variables Beta coefficient 95% CI Beta coefficient p

constant 4.83
Gender (female) 20.13 20.18 to 20.09 ,0.001
GFR per 10 mL/min/1.73 m2 20.03 20.04 to 20.02 ,0.001
ACEI user vs non-user 0.05 0.005 to 0.09 0.026
ARB user vs non-user 0.06 0.02 to 0.10 0.007
Diuretics use vs non-user 20.13 20.18 to 20.08 ,0.001
Bicarbonate (mEq/L) 20.03 20.03 to 20.02 ,0.001
Phosphorus (mg/dL) 0.08 0.05 to 0.10 ,0.001
Log-transformed CRP 20.05 20.07 to 20.03 ,0.001
HbA1c (%)
In total population 0.02 0.01 to 0.04 0.001
In HbA1c ,10% group* 0.044 0.025 to 0.064 ,0.001
In HbA1c §10% group* 20.081 20.141 to 20.021 0.008
Hemoglobin (g/dL) 20.03 20.04 to 20.01 ,0.001
Albumin (g/dL)
In total population 0.09 0.05 to 0.14 ,0.001
In albumin ,4 g/dL group* 0.161 0.096 to 0.225 ,0.001
In albumin §4 g/dL group* 20.156 20.288 to 20.024 0.021
MBP (mmHg) 20.002 20.004 to 20.001 0.002
Insulin user vs non-user 0.15 0.07 to 0.24 ,0.001

ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker.

MBP, mean blood pressure; eGFR, estimated glomerular filtration rate; CRP, C-reactive protein; HbA1c, glycated hemoglobin; UPCR, Urine protein-to-creatinine ratio.
All relevant variables were tested. Variables with significance were presented. * segmental linear regression with the same variables.
doi:10.1371/journal.pone.0067140.t003

and insulin vs. those non-users, higher HbA1c and albumin levels damage not observed in states of equivalent hypokalemia in the
were associated with higher sK. The association of sK with rapid absence of diuretics [20].
renal progression and ESRD was U shape. Hypokalemia defined Second, inadequate RAS inhibition related to either insufficient
as ,4 mEq/L and hyperkalemia as .5 mEq/L in our cohort RAS blockade or resistance to RAS inhibitors may be related to
were significantly associated with greater risk of ESRD. hypokalemia. Insufficient RAS blockage was suggested by a
Our results found hypokalemia to be a significant indicator of greater extent of proteinuria and less use of ACEI/ARB in the
renal progression in CKD evidenced by 82% increased risk of hypokalemic population (Fig. 2C). There is well-established
reaching ESRD, rapid annual decline in eGFR and increase of evidence that RAS inhibitors (ACEI or ARB) have beneficial
proteinuria in CKD population. Our results are in accordance effects on renal outcomes in both diabetic kidney disease and non-
with a study of 1227 males with CKD by Kovesdy et al. They diabetic patients with proteinuria [21,22]. However, in the
found that a sK ,3.6 mEq/L was associated with more severe subgroup of ARB use in our cohort, hypokalemia was still
progression (–0.23 ml/min/1.73 m2/yr, p = 0.002) compared to associated with increased risk of ESRD. Previous studies had
sK = 3.6–5.5 [5]. On the contrary, Saran et al. reported that the shown that ACE insertion/deletion (I/D) polymorphism contrib-
association between hypokalemia (,4 mEq/L) and ESRD disap- uted to the variability in ACEI treatment response and sodium-
peared after adjustment for serum albumin. In our studies, we potassium balance [23–25]. The Antihypertensive and Lipid-
proposed that insufficient anti-hypertensive medication, increased Lowering Treatment to Prevent Heart Attack Trial (ALLHAT)
diuretic use, insufficient RAS inhibition and malnutrition associ- reported the associations of mortality with new-onset hypokalemia
ated with hypokalemia could be related with worse outcomes. (,3.5 mEq/L) and hyperkalemia (.5.4 mEq/L) determined at
We propose the following as plausible mechanisms: First, less year-1 of treatment with chlorthalidone, lisinopril or amlodipine
use of anti-hypertensive medication despite similar BP and greater [26]. The risk of mortality among participants who developed
diuretic use were associated with hypokalemia (Fig. 2B, and hypokalemia by year 1 in the Lisinopril group was the highest (HR
Table 3). Diuretics are often used in patients with heart and/or 3.825; CI: 2.26–6.44; P,0.001) than in either of the other 2
renal disease with volume overload. However, chronic diuretic use groups. From our result and previous reports, hypokalemia might
has been reported to be associated with higher long-term mortality be related to poor response to RAS blockade.
and hospitalization for HF [18]. Our study (Table 4) as well as Third, hypoalbuminemia and lower BMI were associated with
clinical studies and animal models have suggested a correlation hypokalemia (Fig. 2A, Table 3). Hypoalbuminemia, as a marker of
between diuretic use and progression of renal disease [19]. both nutritional and inflammatory status was associated with an
Thiazides have been proposed to associate with renal injury via increased rate of mortality in the population with renal disease
the induction of hypokalemia, metabolic abnormalities, and [27–29]. Hypokalemia might also be a surrogate marker of
volume depletion and to lead to additional focal glomerular malnutrition and correlated with hypoalbuminemia in both
peritoneal and hemodialysis patients [30–32]. Both hypoalbumin-

PLOS ONE | www.plosone.org 6 July 2013 | Volume 8 | Issue 7 | e67140

 Hypokalemia and Renal Outcomes in CKD Patients

Figure 3. Restricted Cubic Spline Regression Plot of the U-shape Association between sK and the Risk for End Stage Renal Disease.
Covariates included in the model were the same as the Cox regression in Table 4. Serum potassium (sK), body mass index, mean blood pressure and
C-reactive protein were treated as restricted cubic spline functions. The solid line represents the log transformed multivariable-adjusted hazard ratio
of ESRD. The dashed lines indicate the 95% confidence intervals. sK below 4.03 and above 5.11 mEq/L were associated with higher hazard (log hazard
ratio .0). Tick marks on the x-axis indicate individual observations at corresponding levels of sK.
doi:10.1371/journal.pone.0067140.g003

emia and hypokalemia remained significant predictors of worse diabetes mellitus, by De Zeeuw et al. reported that increased sK
renal outcomes when included simultaneously in the Cox $5 mEq/L after treatment with Losartan was associated with
regression model (Table 4). These findings may point out that increased risk of ESRD. One potential mechanism could be the
hypokalemia is an independent risk factor for CKD progression vicious cycle of increasing aldosterone levels in response to
and indicates additional prognostic importance of hypoalbumin- sustained hyperkalemia [41]. Aldosterone has been proposed to
emia. cause renal progression through both hemodynamic effects and
Significant correlations between hypokalemia and alkalosis and direct cellular actions as it promotes endothelial dysfunction,
hypertension respectively were observed (Table 3). Hypokalemia facilitates thrombosis, reduces vascular compliance and causes
related augmented ammoniagenesis and subsequent renal injury myocardial and vascular fibrosis [42,43].
has been documented [33]. Hypokalemia was also reported to Our study has limitations to be considered. As an observational
stimulate renin and angiotensin II despite direct suppression of cohort, our ability to elucidate the definite causal links was limited.
aldosterone synthesis leading to salt-sensitive hypertension, intra- The variation of sK levels affected by nutritional status or
renal vasoconstriction and ischemia [34–37]. Other mechanisms medication use during the follow-up period could confound the
mediating renal injury in hypokalemia such as increased results. However, the average sK level is a more precise
inflammatory mediators, oxidative stress [38] and impaired description of baseline status than the single measurement mostly
angiogenesis [39] have been reported. Human and animal used in published studies. Even with markers of nutritional status
pathologic studies also demonstrated that prolonged hypokalemia such as serum albumin and BMI, lack of records of dietary intake
was accompanied by development of multiple renal medullary might still confound the findings. Only a small group of patients in
cysts, tubular degeneration, marked interstitial fibrosis and intense the sK ,3.5 group had rapid progression (26.2%) and incident
macrophage infiltration eventually leading to the impairment of ESRD (8.2%). These patients also presented heavier proteinuria at
renal function [6,40]. baseline. We attempted to account for the differences in baseline
Hyperkalemia, beginning with serum values greater than proteinura in our adjusted models. However, residual confounding
.5 mEq//L in our cohort was associated with higher risk of by factors related to the underlying kidney diseases in these
ESRD. Recent post-hoc analysis of the RENAAL trial of Type 2 patients might still remain. This also limits our ability to make

PLOS ONE | www.plosone.org 7 July 2013 | Volume 8 | Issue 7 | e67140

 Hypokalemia and Renal Outcomes in CKD Patients

Table 4. Association of Categorical Potassium with ESRD and Rapid Renal Progression in Full-Adjusted Model.

ESRD Rapid renal progression*

Risk factors HR (95% CI) p OR (95% CI) p

constant 2.71 3.04

Diabetes mellitus 1.57 (1.19 to 2.08) 0.001 1.30 (1.00 to 1.69) 0.047
2
eGFR ml/min/1.73 m 0.92 (0.90 to 0.93) ,0.001 1.02 (1.01 to 1.02) ,0.001
Proteinuria by dipstick
2 1(reference) 1(reference)
+ 2.06 (1.26 to 3.35) 0.004 1.44 (1.02 to 2.02) 0.036
++ 4.11 (2.61 to 6.47) ,0.001 2.86 (2.04 to 3.99) ,0.001
+++,++++ 7.50 (4.73 to 11.88) ,0.001 4.83 (3.34 to 6.95) ,0.001
ARB user vs non-user 0.88 (0.69 to 1.12) 0.312 0.69 (0.55 to 0.89) 0.004
Diuretics user vs non-user 1.44 (1.07 to 1.93) 0.015 1.14 (0.85 to 1.53) 0.386
Potassium
sK ,3.5 mEq/L 1.82 (1.03 to 3.22) 0.041 1.69 (1.06 to 2.70) 0.027
sK = 3.524 mEq/L 1.67 (1.19 to 2.35) 0.003 1.18 (0.83 to 1.66) 0.364
sK = 424.5 mEq/L 1.23 (0.89 to 1.70) 0.205 1.12 (0.82 to 1.55) 0.470
sK = 4.525 mEq/L 1 (reference) 1 (reference)
sK .5 mEq/L 1.60 (1.09 to 2.34) 0.016 1.39 (0.89 to 2.20) 0.149
Body mass index (kg/m2) 0.96 (0.94 to 0.99) 0.016 0.97 (0.95 to 1.00) 0.072
MBP (mmHg) 1.01 (1.00 to 1.02) 0.019 1.01 (1.01 to 1.02) 0.001
Phosphorus (mg/dL) 1.36 (1.20 to 1.54) ,0.001 1.21 (1.05 to 1.39) 0.008
HbA1c (%) 1.09 (1.02 to 1.17) 0.013 1.10 (1.03 to 1.18) 0.004
Hemoglobin (g/dL) 0.85 (0.79 to 0.92) ,0.001 0.82 (0.77 to 0.88) ,0.001
Albumin (g/dL) 0.45 (0.36 to 0.57) ,0.001 0.61 (0.49 to 0.77) ,0.001

*eGFR slope less than 26.88 ml/min/1.73 m2/yr.

Values expressed as hazard ratio (95% confidence interval) [HR (95% CI)] and odds ratio (95% confidence interval) [OR (95% CI)].
Full-adjusted model included age, gender, DM, CVD, eGFR, Urine protein by dipstick, ACEI, ARB, diuretics, MBP, BMI, HbA1c, log-transformed cholesterol, log-
transformed CRP, phosphate, hemoglobin and albumin. Variables with significance were presented.
doi:10.1371/journal.pone.0067140.t004

generalizations about the relationships between hypokalemia and line represents the log transformed predictive value of serum
renal outcomes. Finally, the follow-up period of 2.7 years could be potassium (sK). Same annotations were used for Figure S1B and
relatively short for the assessment of progression to ESRD, S1C. B. HbA1c below 6.69 and above 12.79% were associated
especially for our study participants with higher eGFR. However, with higher sK. C. Albumin below 3.68 and above 4.56 mg/dL
other indicators of renal progression such as the eGFR slope and were associated with higher sK.
proteinuria were examined and all pointed to the associations (TIF)
between worse renal outcomes and hypo- and hyperkalemia.
Figure S2 Restricted Cubic Spline Regression Model of
In conclusion, both hypokalemia and hyperkalemia are
the Hazard Ratio of sK, BMI, MBP and CRP for End Stage
associated with elevated risk of developing ESRD in a CKD
Renal Disease. A. Serum potassium (sK) below 4.03 and above
population. Hypokalemia is related to increased use of diuretics,
5.11 mEq/L were associated with higher hazard. Tick marks on the
decreased use of RAS blockade and malnutrition, all of which may
x-axis indicate individual observations at corresponding levels of sK.
impose additive deleterious effects on renal outcomes. Clinical
The solid line represents the log transformed multivariable-adjusted
implication may be drawn from our study that sK should be
hazard ratio of ESRD. Same annotations were used for Figure S2B,
routinely evaluated and special attention given for sK between
S2C and S2D. B. Body Mass Index (BMI) below 23.36 and between
3.5–4 mEq/L, generally considered as low ‘‘normal’’. However,
29.68 and 39.49 kg/m2 were associated with higher hazard for
further clinical trials are required to determine if interventions
ESRD. C. Mean blood pressure (MBP) below 72.27 and between
aimed at achieving optimal sK level will slow the progression
101.16 and 127.66 mmHg were associated with higher hazard for
towards ESRD.
ESRD. D. C-reactive protein (CRP) below 39.36 mg/L was
associated with higher hazard for ESRD.
Supporting Information (TIF)
Figure S1 Restricted Cubic Spline Regression Model of Table S1
the Non-linear Effect of BMI, HbA1c and Albumin on sK. (DOC)
A. Body Mass Index (BMI) below 16.4 and above 36.1 kg/m2
were associated with higher sK. Tick marks on the x-axis indicate Table S2
individual observations at corresponding levels of BMI. The solid (DOC)

PLOS ONE | www.plosone.org 8 July 2013 | Volume 8 | Issue 7 | e67140

 Hypokalemia and Renal Outcomes in CKD Patients

Author Contributions Contributed reagents/materials/analysis tools: DYH MCK YWC JMC

JCT SJH JLS HCC. Wrote the paper: HHW CCH.
Conceived and designed the experiments: CCH HHW. Performed the
experiments: HHW CCH. Analyzed the data: HHW CCH SJH.

References
1. Giebisch G, Hebert SC, Wang WH (2003) New aspects of renal potassium 23. Baudin B (2000) Angiotensin I-converting enzyme gene polymorphism and drug
transport. Pflugers Arch 446: 289–297. response. Clin Chem Lab Med 38: 853–856.
2. Einhorn LM, Zhan M, Hsu VD, Walker LD, Moen MF, et al. (2009) The 24. Parving HH, de Zeeuw D, Cooper ME, Remuzzi G, Liu N, et al. (2008) ACE
frequency of hyperkalemia and its significance in chronic kidney disease. Arch gene polymorphism and losartan treatment in type 2 diabetic patients with
Intern Med 169: 1156–1162. nephropathy. J Am Soc Nephrol 19: 771–779.
3. Moranne O, Froissart M, Rossert J, Gauci C, Boffa JJ, et al. (2009) Timing of 25. Ruggenenti P, Bettinaglio P, Pinares F, Remuzzi G (2008) Angiotensin
onset of CKD-related metabolic complications. J Am Soc Nephrol 20: 164–171. converting enzyme insertion/deletion polymorphism and renoprotection in
4. Korgaonkar S, Tilea A, Gillespie BW, Kiser M, Eisele G, et al. (2010) Serum diabetic and nondiabetic nephropathies. Clin J Am Soc Nephrol 3: 1511–1525.
potassium and outcomes in CKD: insights from the RRI-CKD cohort study. 26. Alderman MH, Piller LB, Ford CE, Probstfield JL, Oparil S, et al. (2012)
Clin J Am Soc Nephrol 5: 762–769. Clinical significance of incident hypokalemia and hyperkalemia in treated
5. Hayes J, Kalantar-Zadeh K, Lu JL, Turban S, Anderson JE, et al. (2011) hypertensive patients in the antihypertensive and lipid-lowering treatment to
Association of Hypo- and Hyperkalemia with Disease Progression and Mortality prevent heart attack trial. Hypertension 59: 926–933.
in Males with Chronic Kidney Disease: The Role of Race. Nephron Clin Pract 27. Foley RN, Parfrey PS, Harnett JD, Kent GM, Murray DC, et al. (1996)
120: 8–16. Hypoalbuminemia, cardiac morbidity, and mortality in end-stage renal disease.
6. Torres VE, Young WF, Jr., Offord KP, Hattery RR (1990) Association of J Am Soc Nephrol 7: 728–736.
hypokalemia, aldosteronism, and renal cysts. N Engl J Med 322: 345–351. 28. Taal MW, Brenner BM (2006) Predicting initiation and progression of chronic
7. Menahem SA, Perry GJ, Dowling J, Thomson NM (1999) Hypokalaemia- kidney disease: Developing renal risk scores. Kidney Int 70: 1694–1705.
induced acute renal failure. Nephrol Dial Transplant 14: 2216–2218. 29. Menon V, Greene T, Wang X, Pereira AA, Marcovina SM, et al. (2005) C-
8. Bowling CB, Pitt B, Ahmed MI, Aban IB, Sanders PW, et al. (2010) reactive protein and albumin as predictors of all-cause and cardiovascular
Hypokalemia and outcomes in patients with chronic heart failure and chronic mortality in chronic kidney disease. Kidney Int 68: 766–772.
kidney disease: findings from propensity-matched studies. Circ Heart Fail 3: 30. Erlinger TP, Tarver-Carr ME, Powe NR, Appel LJ, Coresh J, et al. (2003)
253–260. Leukocytosis, hypoalbuminemia, and the risk for chronic kidney disease in US
9. Acker CG, Johnson JP, Palevsky PM, Greenberg A (1998) Hyperkalemia in adults. Am J Kidney Dis 42: 256–263.
hospitalized patients: causes, adequacy of treatment, and results of an attempt to 31. Kovesdy CP, Regidor DL, Mehrotra R, Jing J, McAllister CJ, et al. (2007)
improve physician compliance with published therapy guidelines. Arch Intern Serum and dialysate potassium concentrations and survival in hemodialysis
Med 158: 917–924. patients. Clin J Am Soc Nephrol 2: 999–1007.
10. Takaichi K, Takemoto F, Ubara Y, Mori Y (2007) Analysis of factors causing 32. Soriano S, Gonzalez L, Martin-Malo A, Rodriguez M, Aljama P (2007) C-
hyperkalemia. Intern Med 46: 823–829. reactive protein and low albumin are predictors of morbidity and cardiovascular
11. Maddirala S, Khan A, Vincent A, Lau K (2008) Effect of angiotensin converting events in chronic kidney disease (CKD) 3–5 patients. Clin Nephrol 67: 352–357.
enzyme inhibitors and angiotensin receptor blockers on serum potassium levels 33. Tolins JP, Hostetter MK, Hostetter TH (1987) Hypokalemic nephropathy in the
and renal function in ambulatory outpatients: risk factors analysis. Am J Med Sci rat. Role of ammonia in chronic tubular injury. J Clin Invest 79: 1447–1458.
336: 330–335. 34. Ray PE, Suga S, Liu XH, Huang X, Johnson RJ (2001) Chronic potassium
12. Sarafidis PA, Blacklock R, Wood E, Rumjon A, Simmonds S, et al. (2012) depletion induces renal injury, salt sensitivity, and hypertension in young rats.
Prevalence and factors associated with hyperkalemia in predialysis patients Kidney Int 59: 1850–1858.
followed in a low-clearance clinic. Clin J Am Soc Nephrol 7: 1234–1241. 35. Suga SI, Phillips MI, Ray PE, Raleigh JA, Vio CP, et al. (2001) Hypokalemia
13. Tannen RL (1985) Diuretic-induced hypokalemia. Kidney Int 28: 988–1000. induces renal injury and alterations in vasoactive mediators that favor salt
14. Greenberg A (2000) Diuretic complications. Am J Med Sci 319: 10–24. sensitivity. Am J Physiol Renal Physiol 281: F620–629.
15. Weiner ID, Wingo CS (1997) Hypokalemia–consequences, causes, and 36. Adrogue HJ, Madias NE (2007) Sodium and potassium in the pathogenesis of
correction. J Am Soc Nephrol 8: 1179–1188. hypertension. N Engl J Med 356: 1966–1978.
16. Yee TW (2012) VGAM: Vector generalized linear and additive models. 37. Suga S, Mazzali M, Ray PE, Kang DH, Johnson RJ (2002) Angiotensin II type 1
R package, version 0.9–0 (URL: http://CRAN.R-project.org/ receptor blockade ameliorates tubulointerstitial injury induced by chronic
package = VGAM. Accessed 2013 April 27.). potassium deficiency. Kidney Int 61: 951–958.
17. Yee TW, Wild CJ (1996) Vector generalized additive models. J R Stat Soc Ser C 38. Wang W, Soltero L, Zhang P, Huang XR, Lan HY, et al. (2007) Renal
58: 481–493. inflammation is modulated by potassium in chronic kidney disease: possible role
18. Ahmed A, Husain A, Love TE, Gambassi G, Dell’Italia LJ, et al. (2006) Heart of Smad7. Am J Physiol Renal Physiol 293: F1123–1130.
failure, chronic diuretic use, and increase in mortality and hospitalization: an 39. Reungjui S, Roncal CA, Sato W, Glushakova OY, Croker BP, et al. (2008)
observational study using propensity score methods. Eur Heart J 27: 1431–1439. Hypokalemic nephropathy is associated with impaired angiogenesis. J Am Soc
19. Reungjui S, Pratipanawatr T, Johnson RJ, Nakagawa T (2008) Do thiazides Nephrol 19: 125–134.
worsen metabolic syndrome and renal disease? The pivotal roles for 40. Novello M, Catena C, Nadalini E, Colussi GL, Baroselli S, et al. (2007) Renal
hyperuricemia and hypokalemia. Curr Opin Nephrol Hypertens 17: 470–476. cysts and hypokalemia in primary aldosteronism: results of long-term follow-up
20. Reungjui S, Hu H, Mu W, Roncal CA, Croker BP, et al. (2007) Thiazide- after treatment. J Hypertens 25: 1443–1450.
induced subtle renal injury not observed in states of equivalent hypokalemia. 41. Miao Y, Dobre D, Heerspink HJ, Brenner BM, Cooper ME, et al. (2011)
Kidney Int 72: 1483–1492. Increased serum potassium affects renal outcomes: a post hoc analysis of the
21. Strippoli GF, Bonifati C, Craig M, Navaneethan SD, Craig JC (2006) Reduction of Endpoints in NIDDM with the Angiotensin II Antagonist Losartan
Angiotensin converting enzyme inhibitors and angiotensin II receptor (RENAAL) trial. Diabetologia 54: 44–50.
antagonists for preventing the progression of diabetic kidney disease. Cochrane 42. Struthers AD, MacDonald TM (2004) Review of aldosterone- and angiotensin
Database Syst Rev: CD006257. II-induced target organ damage and prevention. Cardiovasc Res 61: 663–670.
22. Chiurchiu C, Remuzzi G, Ruggenenti P (2005) Angiotensin-converting enzyme 43. Epstein M (2001) Aldosterone and the hypertensive kidney: its emerging role as a
inhibition and renal protection in nondiabetic patients: the data of the meta- mediator of progressive renal dysfunction: a paradigm shift. J Hypertens 19:
analyses. J Am Soc Nephrol 16 Suppl 1: S58–63. 829–842.

PLOS ONE | www.plosone.org 9 July 2013 | Volume 8 | Issue 7 | e67140