Observational Study Medicine ®

OPEN

Electrolyte and mineral disturbances in septic

acute kidney injury patients undergoing
continuous renal replacement therapy
Su-Young Jung, MDa, Hyunwook Kim, PhDa, Seohyun Park, MDa, Jong Hyun Jhee, MDa,
Hae-Ryong Yun, MDa, Hyoungnae Kim, MDa, Youn Kyung Kee, MDa, Chang-Yun Yoon, MDa,
Hyung Jung Oh, MDa, Tae Ik Chang, PhDb, Jung Tak Park, PhDa, Tae-Hyun Yoo, PhDa,
∗
Shin-Wook Kang, PhDa, Hajeong Lee, MDc, Dong Ki Kim, PhDc, Seung Hyeok Han, PhDa,

Abstract
Electrolyte and mineral disturbances remain a major concern in patients undergoing continuous renal replacement therapy (CRRT);
however, it is not clear whether those imbalances are associated with adverse outcomes in patients with septic acute kidney injury
(AKI) undergoing CRRT. We conducted a post-hoc analysis of data from a prospective randomized controlled trial. A total of 210
patients with a mean age of 62.2 years (136 [64.8%] males) in 2 hospitals were enrolled. Levels of sodium, potassium, calcium, and
phosphate measured before (0 hour) and 24 hours after CRRT initiation. Before starting CRRT, at least 1 deficiency and excess in
electrolytes or minerals were observed in 126 (60.0%) and 188 (67.6%) patients, respectively. The excess in these parameters was
greatly improved, whereas hypokalemia and hypophosphatemia became more prevalent at 24 hours after CRRT. However, 1 and 2
or more deficiencies in those parameters at the 2 time points were not associated with mortality. However, during 28 days, 89
(71.2%) deaths occurred in patients with phosphate levels at 0 hour of ≥4.5 mg/dL as compared with 49 (57.6%) in patients with
phosphate levels <4.5 mg/dL. The 90-day mortality was also significantly higher in patients with hyperphosphatemia. Similarly, in 184
patients who survived at 24 hours after CRRT, hyperphosphatemia conferred a 2.2-fold and 2.6-fold increased risk of 28- and 90-day
mortality, respectively. The results remained unaltered when the serum phosphate level was analyzed as a continuous variable.
Electrolyte and mineral disturbances are common, and hyperphosphatemia may predict poor prognosis in septic AKI patients
undergoing CRRT.
Abbreviations: AKI = acute kidney injury, APACHE II = acute physiology and chronic health evaluation II, CCI = Charlson
comorbidity index, CI = confidence interval, CRP = C-reactive protein, CRRT = continuous renal replacement therapy, eGFR =
estimated glomerular filtration rate, FiO2 = fraction of inspired oxygen, HR = hazard ratio, ICU = intensive care unit, MAP = mean
arterial pressure, SOFA = sequential organ failure assessment.
Keywords: acute kidney injury (AKI), continuous renal replacement therapy (CRRT), electrolyte, mineral

1. Introduction AKI varies depending on the definition used and is reported to be

7% in all hospitalized patients[2] and up to 60% in ICU
Acute kidney injury (AKI) commonly occurs in critically ill
patients.[3] The importance of AKI has been highlighted because
patients in the intensive care unit (ICU).[1] The incidence rate of
it is associated with high morbidity and mortality rates.[4] It is
Editor: Muhammed Mubarak.
also associated with increased costs[5] and length of hospital
stay.[6] Moreover, patients with AKI are more likely to develop
S-YJ, HK, S-YJ, and HWK contributed equally to this work.
chronic kidney disease, including end-stage renal disease than
The authors have no funding and conflicts of interest to disclose.
those without AKI.[7] Sepsis is a devastating medical condition
Supplemental Digital Content is available for this article.
triggered by an immune response to infection, leading to multiple
a
Department of Internal Medicine, College of Medicine, Institute of Kidney complications. AKI frequently develops as a consequence of
Disease Research, Yonsei University, Seoul, Republic of South Korea.,
b
Department of Internal Medicine, NHIS Medical Center, Ilsan Hospital,
sepsis, and its incidence has been reported to be 19% in patients
Goyangshi, Gyeonggi-do, c Department of Internal Medicine, Seoul National with sepsis, 23% in those with severe sepsis, and 51% in those
University College of Medicine, Seoul, South Korea. with septic shock, when blood cultures are positive.[8] Notably,
∗
Correspondence: Seung Hyeok Han, Department of Internal Medicine, College septic AKI is one of the life-threatening manifestations of multiple
of Medicine, Institute of Kidney Disease Research, Yonsei University, Yonsei-ro, organ dysfunction syndrome and has been shown to be
Seodaemun-gu, Seoul, Republic of South Korea (e-mail: hansh@yuhs.ac). independently associated with adverse clinical outcomes.[9]
Copyright © 2016 the Author(s). Published by Wolters Kluwer Health, Inc. All Electrolyte or mineral imbalances are common findings in many
rights reserved.
serious conditions, including septic AKI. However, the incidence of
This is an open access article distributed under the Creative Commons
Attribution-No Derivatives License 4.0, which allows for redistribution, commercial these alterations in critically ill patients is unclear. Adekola et al[10]
and non-commercial, as long as it is passed along unchanged and in whole, with reported that >66.78% of patients in the ICU had multiple
credit to the author. electrolyte and acid-base abnormalities. It is important to maintain
Medicine (2016) 95:36(e4542) these parameters within the physiologic levels to prevent
Received: 29 April 2016 / Received in final form: 5 July 2016 / Accepted: 14 July complications and adverse outcomes because electrolytes and
2016 minerals play important roles in protecting cellular function, tissue
http://dx.doi.org/10.1097/MD.0000000000004542 perfusion, and acid–base homeostasis.[11,12] Not surprisingly,

1
Jung et al. Medicine (2016) 95:36 Medicine

continuous renal replacement therapy (CRRT) has become the collected at both time points: hemoglobin, white blood cell, serum
modality of choice in critically ill patients with AKI as it is more creatinine, albumin, lactate, aspartate aminotransferase, alanine
effective in correcting uremia, electrolyte and mineral disturbances, aminotransferase, and total bilirubin levels. The estimated
acid-base disorder, and volume overload. In addition, CRRT can glomerular filtration rate (eGFR) was calculated by using the
provide immune modulation through the convective and adsorp- chronic kidney disease epidemiology collaboration equation.[20]
tive removal of various immune mediators.[13] The data also included the age-adjusted Charlson comorbidity
Although electrolyte and mineral disturbances are largely index (CCI), Sequential Organ Failure Assessment (SOFA)
improved with dialysis treatment, this issue has not yet been score, and Acute Physiology and Chronic Health Evaluation II
completely resolved even in patients undergoing CRRT. In fact, (APACHE II) score. Although the APACHE II is the most well-
overcorrection often causes serious problems such as neurologic known system that provides prognostic information for ICU
dysfunction, conductance abnormalities, arrhythmia, and gastro- patients, the score does not include specific criteria for kidney
intestinal hypomotility.[14–16] On the other hand, insufficient function.[21] Thus, we used the SOFA score for this study analysis.
correction remains despite CRRT and is sometimes ignored in
clinical practice. When septic AKI occurs, the clinical outcomes are
2.3. Study end points
primarily determined by the proper management of primary illness
and the presence or absence of multiorgan involvement. However, The study end point was death that occurred within 28 and 90 days
it is not clear whether electrolyte and mineral disturbances are after CRRT initiation. In addition, the duration of CRRT and
associated with adverse outcomes in patients with septic AKI length of hospital stay after CRRT initiation were also analyzed.
undergoing CRRT. Therefore, we aimed to investigate the
incidence, clinical course, and clinical utility as prognostic markers 2.4. Statistical analysis
of the altered levels of electrolytes and minerals in these patients.
All statistical analyses were performed by using SPSS for Windows
version 23.0 (IBM Corp., Armonk, NY). Continuous variables are
2. Materials and methods expressed as the mean ± standard deviation or median
2.1. Patient selection (interquartile range), whereas categorical variables are expressed
as a number (percentage). Comparisons between 2 groups or
We performed a post-hoc analysis of data from our recent study, among 3 groups were done through an analysis of variance or
“Effects of High-volume Continuous Renal Replacement Student’s t test for continuous variables, and with the chi-squared
Therapy on Inflammatory Mediators and Outcomes in Patients test or Fisher’s exact test for categorical variables. The normality of
with Septic Acute Kidney Injury” (NCT01191905). Briefly, this distribution was confirmed by using the Kolmogorov–Smirnov
prospective randomized controlled open-label trial was con- test. Cumulative survival curves were derived by using the
ducted at Yonsei University Health System Severance Hospital Kaplan–Meier method, and differences between curves were
and Seoul National University Hospital between January 2011 compared through the log-rank test. Cox proportional hazards
and August 2014. The aim of the study was to compare high and analysis was performed to determine the relationship between
conventional doses of CRRT in patients with septic AKI who the abnormality in each electrolyte and mineral, and mortality.
underwent CRRT. The detailed treatment was described P-values <0.05 were considered statistically significant.
elsewhere.[17] Patients were included if they met the following
criteria: consensus criteria for sepsis,[18] injury stage of the RIFLE 3. Results
(risk, injury, failure, loss, end stage) criteria[19] or more (>2-fold
increase in serum creatinine or urine output of <0.5 mL/kg/h for 3.1. Patient characteristics
12 hours), and absence of other established nonsepsis-related
The baseline clinical characteristics and laboratory findings of the
cause of AKI. Patients were excluded if they met the following
study subjects are shown in Table 1. The mean age of the
criteria: age <20 years or >80 years, with a life expectancy of <3
participants was 62.2 ± 12.9 years, and 136 (64.8%) of them
months, with Child-Pugh class C liver cirrhosis, pregnant or
were men. Among the 210 patients, 57 (27.3%) had diabetes and
lactating, and with a history of dialysis before the study. Finally, a
17 (8.0%) had cardiovascular diseases. The mean baseline eGFR
total of 210 patients were enrolled excluding 2 patients with
at the time of CRRT initiation was 26.30 ± 22.39 mL/min 1.73
missing data. The study was approved by the institutional review
m 2. The mean age-adjusted CCI was 4.64 ± 2.62. The SOFA
board of the 2 participating hospitals and followed the provisions
score was 14.18 ± 3.10, and the APACHE II score was 28.72 ±
of the Declaration of Helsinki. We obtained informed written
7.30. The mean arterial pressure (MAP) at 0 hour was 78.81 ±
consent from all the enrolled participants.
14.49 mm Hg. A total of 175 (83.3%) patients received
mechanical ventilation, and the mean fraction of inspired oxygen
2.2. Clinical and biochemical data collection
(FiO2) was 0.58 ± 0.24. The mean white blood cell count and
Demographic and clinical data such as age, sex, body mass index, C-reactive protein levels were 11,350 mL (5,037–18,815 mL) and
and comorbidities were recorded at the time of CRRT initiation. 18.43 mg/L (9.35–39.80 mg/L), respectively.
Biochemical data for electrolyte and mineral levels, such as sodium,
potassium, calcium, and phosphate, were measured before starting
3.2. Electrolyte or mineral deficiencies and excess before
CRRT (0 hour) and at 24 hours after CRRT initiation. Calcium
levels were corrected by using serum albumin levels (corrected and 24 hour after CRRT initiation
calcium = calcium [mg/dL] + 0.8  [4 – serum albumin {g/dL}]). Table 2 shows the deficiencies and excess of each electrolyte and
The reference ranges of each electrolyte and mineral are as follows: mineral at 0 and 24 hours after CRRT application. Overall
sodium, 135–145 mEq/L; potassium, 3.5–5.5 mEq/L; calcium and deficiencies or excesses in a single electrolyte or mineral were
corrected calcium, 8.5–10.5 mg/dL; and phosphate, 2.5–4.5 mg/dL. observed in 126 (60.0%) and 188 (67.6%) patients, respectively,
In addition, the following biochemical laboratory data were before starting CRRT (0 hour). The most frequent disturbances

2
Jung et al. Medicine (2016) 95:36 www.md-journal.com

Table 1 deficiencies, hyponatremia, hypokalemia, hypocalcemia, and

Baseline characteristics of subjects. hypophosphatemia were found in 35 (19.4%), 51 (28.3%), 15
(8.5%), and 36 (19.6%) patients, respectively. In contrast, for
Total (n = 210)
excess, hypernatremia (n = 8, 4.4%), hyperkalemia (n = 2, 1.1%),
Age, y 62.2 ± 12.9 and hyperphosphatemia (n = 57, 31.0%) occurred at lesser rates
Male, % 136 (64.8) than those at 0 hour (Table 2).
Hypertension, % 100 (47.6)
Diabetes mellitus, % 57 (27.3)
Heart failure, % 18 (8.6) 3.3. CRRT duration, hospital stays, and mortality
Ischemic heart disease, % 13 (6.2)
Myocardial infarction, % 4 (1.9)
Among the 210 patients, 138 (65.7%) and 162 (77.1%) deaths
Cerebrovascular disease, % 11 (5.3) occurred during 28 and 90 days. Twenty-two (10.5%) patients
COPD, % 12 (5.7) died during 24 hours after CRRT initiation. Overall, the median
Mechanical ventilation, % 175 (83.3) duration of CRRT and the median length of hospital stay after
CCI 2.94 ± 2.15 CRRT initiation were 3.90 (2.20–7.24) and 10.45 (2.97–31.33)
Age-adjusted CCI 4.64 ± 2.62 days, respectively. The duration of CRRT (3.94 [3.21–5.18] vs
BMI (kg/m2) at ICU admission 23.26 ± 4.22 3.91 [1.81–8.99], P = 0.003) and hospital stay (34.53
eGFR (mL/min/1.73 m2) at admission 72.58 ± 45.10 [19.40–70.16] vs 6.20 [2.33–19.74], P < 0.001) were significant-
eGFR at CRRT initiation (mL/min/1.73 m2) 26.30 ± 22.39 ly longer in survivors than in nonsurvivors (Supplementary
SOFA score 14.18 ± 3.10
Table 1, http://links.lww.com/MD/B261).
APACHE II score 28.72 ± 7.30
Systolic blood pressure, mm Hg 111.72 ± 20.52
Diastolic blood pressure, mm Hg 62.30 ± 13.90 3.4. Clinical outcomes according to electrolyte or mineral
Mean arterial pressure, mm Hg 78.81 ± 14.49 disturbances before and at 24 hours after CRRT initiation
Hemoglobin, g/dL 9.42 ± 1.93
White blood cell, mL 11350 [5037 – 18815] We compared the mortality rates according to the deficiencies and
Albumin, g/dL 2.47 ± 0.53 excesses of electrolytes and minerals (Supplementary Table 2, http://
Blood urea nitrogen, mg/dL 62.48 ± 32.60 links.lww.com/MD/B261). Any alterations in sodium, calcium, and
Creatinine, mg/dL 3.09 ± 1.77 potassium levels at 0 and 24 hours were not associated with
Aspartate aminotransferase, IU/L 89.50 [38.00–356.00] mortality (data not shown). In addition, the serum levels of these
Alanine transaminase, IU/L 39.50 [16.00–185.00]
parameters at 0 hour did not differ between survivors and
Bilirubin, mg/dL 1.90 [0.90–5.65]
nonsurvivors (Supplementary Table 3, http://links.lww.com/MD/
CRP, mg/L 18.43 [9.35–39.80]
Lactate, mmol/L 6.60 ± 5.11 B261). However, 89 (71.2%) deaths occurred in hyperphosphate-
FiO2 0.58 ± 0.24 mic patients as compared with 2 (33.3%, P = 0.08) in hypophos-
pH 7.30 ± 0.11 phatemic patients and 47 (59.5%, P = 0.02) in normophosphatemic
PaO2, mm Hg 103.97 ± 81.88 patients during 28 days. During 90 days, 105 (84.0%) patients with
PaCO2, mm Hg 38.71 ± 16.27 hyperphosphatemia died as compared with 57 (67.1%) patients
Base excess, mmol/L –6.90 [–11.13 to –3.80] with phosphate levels <4.5 mEq/L (P = 0.03). The serum phosphate
Data are expressed as mean ± standard deviations, median (interquartile range), or numbers (%). levels were also higher in nonsurvivors than in survivors (4.97±
APACHE II = acute physiology and chronic health evaluation II, BMI = body mass index, CCI = Charlson 2.78 mg/dL vs 5.79± 2.53 mg/dL, P = 0.04). This pattern was
comorbidity index, COPD = chronic obstructive pulmonary disease, CRP = C-reactive protein, CRRT = consistent when analyzed by using phosphate levels at 24 hours
continuous renal replacement therapy, eGFR = estimated glomerular filtration rate, FiO2 = fraction of (3.42± 1.37 mg/dL vs 4.21± 1.84 mg/dL, P = 0.01) (Supplementary
inspired oxygen, ICU = intensive care unit, SOFA = sequential organ failure assessment score.
Table 3, http://links.lww.com/MD/B261). The length of hospital
stay after CRRT initiation was shorter in patients with hyper-
phosphatemia at 24 hours than those for the normal or low-level
(deficiency or excess) were hypocalcemia (37%) and hyper- group (19.63 [6.28–35.53] days vs 15.81 [6.39–33.58] days vs 3.94
phosphatemia (59.5%) (Table 2). [1.91–22.70] days, P = 0.02) (Supplementary Table 4, http://links.
CRRT during the first 24 hours greatly improved the excess in lww.com/MD/B261). This finding was not observed for sodium,
electrolytes and minerals. However, disturbances in these potassium, and calcium. We further analyzed whether 2 or more
parameters were still common at 24 hours (Fig. 1). For deficiencies in electrolytes and minerals could predict adverse

Table 2
Electrolyte or mineral deficiencies and excess before starting continuous renal replacement therapy and 24 h after continuous renal
replacement therapy initiation.
0h 24 h Deficiency Excess
Mean ± SD Mean ± SD P 0h 24 h 0h 24 h
Sodium 138.24 ± 8.25 137.54 ± 4.34 0.22 58 (27.6) 35 (19.4) 30 (14.3) 8 (4.4)
Potassium 4.39 ± 0.94 3.83 ± 0.73 <0.001 26 (12.4) 51 (28.3) 24 (11.4) 2 (1.1)
Calcium, Ca 7.61 ± 1.20 8.35 ± 0.78 <0.001 168 (80.8) 95 (54.0) 5 (2.4) 2 (1.1)
Corrected Ca 8.82 ± 1.16 9.46 ± 0.74 <0.001 77 (37.0) 15 (8.5) 12 (5.8) 11 (6.3)
Phosphate 5.66 ± 2.48 3.99 ± 1.76 <0.001 6 (2.9) 36 (19.6) 125 (59.5) 57 (31.0)
Data are expressed mean ± standard deviations and numbers (%).
Calcium (mg/dL); corrected Ca (mg/dL); phosphate (mg/dL); potassium (mEq/L); sodium (mEq/L).
SD = standard deviation.

3
Jung et al. Medicine (2016) 95:36 Medicine

Sodium (0 h) Sodium (24 h)

14.3%
4.4%
27.6% 19.4%
1 1
2 2
3 3
76.1%
58.1%

Calcium (0 h) Calcium (24 h)

2.4% 1.1%
16.8%
1 1
2 44.9% 2
3 54.0%
3
80.8%

Corrected Ca (0 h) Corrected Ca (24 h)

5.8% 6.3% 8.5%

37.0% 1
1
2 2
3 85.2%
3
57.2%

Potassium (24 h)
Potassium (0 h)
11.4% 1.1%
12.4%
1 28.3% 1
2 2
3 3
70.6%
76.2%

Phosphate (0 h) Phosphate (24 h)

2.9%
19.6%
1 31.0%
37.6% 1
2 2
3 3
59.5% 49.5%
Figure 1. Electrolyte or mineral deficiencies and excess before starting CRRT and 24 hours after CRRT initiation: 1, deficiency; 2, normal; 3, excess. CRRT =
continuous renal replacement therapy.

outcomes. Compared with a single-deficiency group, groups with 2 there were only 6 patients with phosphate levels <2.5 mg/dL at 0
or more deficiencies did not show an increased mortality rate (Fig. 2). hour; thus, the patients were divided into 2 groups: <4.5 mg/dL and
≥4.5 mg/dL. Most baseline characteristics were comparable
between groups before CRRT initiation (Supplementary Table 5,
3.5. Hyperphosphatemia predicts adverse outcomes
http://links.lww.com/MD/B261). However, eGFR was significant-
Because more deaths occurred in the hyperphosphatemic group ly lower in patients with hyperphosphatemia (31.5± 30.7 vs 22.7±
than in the other 2 groups, we further investigated whether 13.2 mL/min 1.73 m 2, P = 0.01). We constructed 3 different Cox
phosphate levels could predict clinical adverse outcomes. First, models, as presented in Table 3. The crude hazard ratios (HRs) for

4
Jung et al. Medicine (2016) 95:36 www.md-journal.com

Figure 2. Kaplan–Meier plots for 28- and 90-day mortality according to single and 2 or more deficiencies in electrolytes or minerals before starting CRRT (A and B)
and 24 hours after CRRT initiation (C and D), Group 1 (single deficiency); Group 2 (2 or more deficiencies), CRRT = continuous renal replacement therapy.

the 28- and 90-day mortality rates compared with normophos- normophosphatemia. When the phosphate level was treated as a
phatemia or hypophosphatemia were 1.44 and 1.47 (95% continuous variable, hyperphosphatemia was a significant predic-
confidence interval [CI], 1.01–2.03, P = 0.04; 95% CI, tor of 28-day mortality (HR, 1.36 per 1 mg/dL increase; 95% CI,
1.07–2.02, P = 0.02), respectively. In a model 2 adjusted for age, 1.20–1.54; P < 0.001) and 90-day mortality (HR, 1.32 per 1 mg/dL
sex, and body mass index (BMI), hyperphosphatemia was also increase; 95% CI, 1.17–1.48; P < 0.001) (Table 4 and Fig. 3).
associated with an increased risk of death. Finally, in a fully
adjusted model after additional adjustment for CCI, SOFA score,
4. Discussion
and residual kidney function, this association became more evident
(Table 3). Next, patients were classified into 3 groups according to Electrolyte, mineral, and acid–base disturbances commonly
deficient, normal, and excess phosphate levels at 24 hours after occur in AKI patients.[22] Much effort has been made to maintain
CRRT (Supplementary Table 6, http://links.lww.com/MD/B261). homeostasis in clinical practice. In particular, CRRT greatly
The hyperphosphatemic group was younger (P = 0.18), had higher improves the abnormalities of these parameters; however, it can
BMI (P = 0.01) and SOFA score (P < 0.001), and had lower MAP often cause overcorrection, and suboptimal correction remains a
(P = 0.002). After full adjustment of these factors, hyperphospha- concern. This study evaluated the incidence, clinical course, and
temia conferred a 2.2-fold and 2.6-fold increased risk of 28-day prognostic implication of electrolyte and mineral disturbances in
(HR, 2.25; 95% CI, 1.40–3.61; P < 0.001) and 90-day mortality septic AKI patients undergoing CRRT. We found that, as
(HR, 2.65; 95% CI, 1.71–4.12; P < 0.001) compared with expected, electrolyte and mineral disturbances were common and

5
Jung et al. Medicine (2016) 95:36 Medicine

Table 3
Cox proportional hazard regression analysis for 28- and 90-day mortality before starting continuous renal replacement therapy with all-
cause mortality.
∗
Phosphate as a categorical variable (≥ 4.5 vs <4.5 mg/dL) Phosphate as a continuous variable (per 1 mg/dL increase)
HR (95% CI) P HR (95% CI) P
28-day
Model 1 1.44 (1.01–2.03) 0.04 1.04 (0.98–1.10) 0.24
Model 2 1.46 (1.03–2.07) 0.04 1.05 (0.99–1.11) 0.14
Model 3 1.59 (1.10–2.29) 0.01 1.08 (1.00–1.15) 0.04
90-day
Model 1 1.47 (1.07–2.02) 0.02 1.05 (1.00–1.10) 0.07
Model 2 1.49 (1.08–2.06) 0.02 1.06 (1.00–1.11) 0.05
Model 3 1.64 (1.18–2.29) 0.003 1.09 (1.02–1.16) 0.01
∗
Phosphate level <4.5 mg/dL is a referent.
Model 1: unadjusted.
Model 2: Age, gender, BMI at ICU admission.
Model 3: Model 2 + CCI, SOFA score, eGFR (0 h).
BMI = body mass index, CCI = Charlson comorbidity index, CI = confidence interval, eGFR = estimated glomerular filtration rate, HR = hazard ratio, ICU = intensive care unit, SOFA = sequential organ failure assessment.

CRRT largely corrected these abnormalities in the analyzed containing potassium and phosphate have been commercially
patients. Notably, a single deficiency and 2 or more deficiencies of available. Unfortunately, our centers do not have these solutions
electrolytes or minerals were not associated with adverse and thus cannot use them in the ICU. Even though the samples
outcomes. However, hyperphosphatemia was significantly were collected and stored at 0 and 24 hours according to the
associated with an increased risk of 28- and 90-day mortality original study protocol, we additionally measured serum levels of
rates in septic AKI patients undergoing CRRT. electrolytes and minerals at 10–12 hours after CRRT initiation
It is well known that deficiencies in electrolytes and minerals for early detection of disturbances in these parameters in AKI
can increase morbidity, if not properly treated.[23] In addition, patients, particularly those on CRRT. Presumably, prompt
many studies have shown that critically ill patients with correction of potassium and phosphate deficiencies might be
hypokalemia[24] or hypophosphatemia[25,26] are more likely to helpful to avoid serious complications. Another possible
die than patients without these abnormalities. In contrast to these explanation is the unique patient inclusion of this study; we
previous studies, in this study, any deficiencies in sodium, only included patients with septic AKI requiring CRRT. In fact,
potassium, calcium, and phosphate were found not to be related septic AKI is apparently distinct from nonseptic AKI in
to increased mortality rates in septic AKI patients undergoing pathophysiology and severity.[27] In sepsis, the systemic inflam-
CRRT. Before initiating CRRT, deficiencies in sodium and matory response triggered by infectious insult deteriorates failure
calcium were more common than deficiencies in potassium and in multiple organs and systems, subsequently resulting in
phosphate. However, hyponatremia and hypocalcemia were catastrophic physiological derangements. Therefore, it can be
much improved after CRRT. In general, a CRRT solution presumed that the hemodynamic alteration and deterioration due
contains 140 mEq/L sodium and 2.5–3.5 mEq/L calcium; thus, to sepsis-induced systemic inflammatory response syndrome
sodium and calcium deficiencies can be corrected by diffusion. In might overwhelm the impact of electrolyte and mineral
contrast, hypokalemia and hypophosphatemia became more disturbances in terms of affecting mortality.
prevalent 24 hours after CRRT. This overcorrection can In this study, we showed that hyperphosphatemia was
commonly occur during CRRT. To prevent this, CRRT solutions significantly associated with an increased risk of death. In

Table 4
Cox proportional hazard regression analysis for 28- and 90-day mortality in 184 patients who survived 24 h after continuous renal
replacement therapy initiation.
∗
Phosphate as a categorical variable (≥4.5 vs 2.5 to 4.5 vs <2.5 mg/dL) Phosphate as a continuous variable (per 1 mg/dL increase)
HR (95% CI) P HR (95% CI) P
28-day
Model 1 2.11 (1.41–3.16) <0.001 1.33 (1.20–1.47) <0.001
Model 2 3.12 (2.01–4.86) <0.001 1.46 (1.31–1.62) <0.001
Model 3 2.25 (1.40–3.61) <0.001 1.36 (1.20–1.54) <0.001
90-day
Model 1 2.54 (1.76–3.67) <0.001 1.36 (1.21–1.47) <0.001
Model 2 3.52 (2.24–5.30) <0.001 1.43 (1.29–1.59) <0.001
Model 3 2.65 (1.71–4.12) <0.001 1.32 (1.17–1.48) <0.001
∗
Phosphate level of 2.5–4.5 mg/dL is a referent.
Model 1: unadjusted.
Model 2: Age, gender, BMI at ICU admission.
Model 3: Model 2 + CCI, SOFA score, urine output (24 h).
BMI = body mass index, CCI = Charlson comorbidity index, CI = confidence interval, HR = hazard ratio, ICU = intensive care unit, SOFA = sequential organ failure assessment.

6
Jung et al. Medicine (2016) 95:36 www.md-journal.com

0 h phosphate level groups and 28-day mortality 0 h phosphate level groups and 90-day mortality
P = 0.04 P = 0.02

100 Group 1 and 2 100 Group 1 and 2

Group 3 Group 3

Percent survival

Percent survival
50 50

0 0
0 10 20 30 0 20 40 60 80 100
Days Days
A B
24 h phosphate level groups and 28-day mortality 24 h phosphate level groups and 90-day mortality
*; P = 0.002 (Group 1 vs. 3) *; P < 0.001 (Group 1 vs. 3 and group 2 vs. 3)
**; P < 0.001 (Group 2 vs. 3)

100 Group 1 100 Group 1

Group 2 Group 2

Percent survival
Percent survival

Group 3 Group 3

**
50
*
50
*
*
0 0
0 10 20 30 0 20 40 60 80 100
Days Days
C D
Figure 3. Kaplan–Meier plots for 28- and 90-day mortality according to phosphate levels before starting CRRT (A and B) and 24 hours after CRRT initiation (C and D), Group
1 (below normal levels) <2.5 mg/dL; Group 2 (normal levels) 2.5–4.5 mg/dL; Group 3 (elevated levels) >4.5 mg/dL. CRRT = continuous renal replacement therapy.

contrast to our findings, several previous studies have shown that from damaged cells. Various types of cell death have been
hypophosphatemia was a significant predictor of adverse suggested in sepsis.[42] Besides phosphate, other danger signals and
outcomes in critically ill patients. In fact, phosphate is a key inflammatory cytokines are also released upon cell death and may
component of bone and cell membranes, is essential in all body play more important roles in the development of life-threatening
functions requiring energy (as adenosine triphosphate), and is conditions.[43–45] Direct phosphate toxicity is also plausible.
especially important in nerve and muscle function.[28] Hypo- Increased phosphate from dead cells subsequently promotes
phosphatemia is caused by various conditions in ICU patients, vascular inflammation[46,47] and triggers phosphate influx into
such as sepsis, diuretic use, total parenteral nutrition, and cells, resulting in increased mitochondrial membrane potential and
gastrointestinal wasting.[29–32] Its harmful effects can be reactive oxygen species production.[48] This can produce a vicious
exemplified by decreased myocardial contraction, increased cycle and further aggravate cell death. In fact, a number of studies
development of arrhythmia, impaired response to vasopressors, have shown that patients with increased phosphate levels are more
and decreased granulocyte phagocytic activity.[33–37] The reason likely to develop cardiovascular diseases[47,49,50] and chronic
for the discrepancy in findings concerning phosphate levels and kidney disease,[51] even end-stage renal disease.[51]
mortality among studies is unclear. Interestingly, most of the This study has some limitations. First, as this is a post-hoc
adverse effects caused by hypophosphatemia result from very analysis of our prior randomized controlled trial, causality is
severe phosphate deficiency. In many previous studies, severe uncertain and the results should be interpreted with caution. In
hypophosphatemia was defined as a serum phosphate level of addition, the small sample size is another limitation, as it may
<1.0 mg/dL.[38–40] By using this definition in our study, severe result in a lack of statistical power. Second, magnesium is also an
hypophosphatemia was not observed in our subjects at 24 hours important mineral and its deficiency is common in AKI patients;
after CRRT. In addition, most patients with phosphate levels however, we did not include magnesium imbalance in the
<2.5 mg/dL at 0 hour had phosphate levels of 2.0–2.5 mg/dL and reported data because magnesium levels were not measured in
only 10 patients had phosphate levels <2.0 mg/dL at 24 hours. >50% of the patients. Nevertheless, we performed further
We strictly followed a protocol concerning the supplementation analysis in this limited number of patients and found that
for deficiencies in electrolytes and minerals, and this can explain magnesium disturbances did not affect mortality (data not
the rare incidence of severe hypophosphatemia in our study. shown). Third, further analysis with extended follow-up data,
An acute decrease in phosphate levels during dialysis treatment such as phosphate levels measured at 72 hours, would be helpful
is also important in determining adverse outcomes. Demirjian to confirm our findings. However, septic AKI is a devastating
et al[41] showed that patients with a decline in phosphate to <2 medical condition and many patients have died while receiving
mg/dL during CRRT had prolonged respiratory failure requiring CRRT. We analyzed 146 patients who survived at 72 hours after
tracheostomy. In our study, there were only 10 patients whose CRRT, by collecting additional blood samples, and found that
phosphate levels declined to <2 mg/dL at 24 hours after CRRT. hyperphosphatemia still remained a significant risk factor of
These patients were not significantly associated with an increased death (HR, 2.25; 95% CI, 0.13–4.89; P = 0.04, data not shown).
mortality as compared with patients with phosphate levels >2 Finally, the original study aimed to evaluate effects of high (80
mg/dL (HR, 1.12; 95% CI, 0.80–1.57; P = 0.50, data not shown). mL/kg/h) versus conventional (40 mL/kg/h) volume on clinical
The underlying mechanism responsible for the high mortality in outcomes; thus, the greater deficiencies in electrolytes and
hyperphosphatemic patients is largely presumptive. It is possible minerals might be attributed to high-volume treatment. Howev-
that phosphate simply reflects disease severity because it is released er, there were no differences in the changes of these parameters

7
Jung et al. Medicine (2016) 95:36 Medicine

after 24-h CRRT between the 2 groups. This finding suggests that [24] Hessels L, Hoekstra M, Mijzen LJ, et al. The relationship between serum
potassium, potassium variability and in-hospital mortality in critically ill
disturbances in electrolytes and minerals are not further
patients and a before-after analysis on the impact of computer-assisted
improved by high-volume treatment compared with treatment potassium control. Crit Care 2015;19:4.
with the conventional volume. [25] Suzuki S, Egi M, Schneider AG, et al. Hypophosphatemia in critically ill
In conclusion, this study showed that electrolyte and mineral patients. J Crit Care 2013;28:536.e9–19.
disturbances are common and hyperphosphatemia may portend a [26] Bech A, Blans M, Raaijmakers M, et al. Hypophosphatemia on the
intensive care unit: individualized phosphate replacement based on
poor prognosis in septic AKI patients undergoing CRRT. If phosphate serum levels and distribution volume. J Crit Care 2013;28:838–43.
is a biomarker that can reflect disease severity, then more careful [27] Shah SR, Tunio SA, Arshad MH, et al. Acute kidney injury recognition
attention should be paid to patients with increased phosphate levels. and management: a review of the literature and current evidence. Glob J
Health Sci 2015;8:49202.
[28] Knochel JP. The pathophysiology and clinical characteristics of severe
References hypophosphatemia. Arch Intern Med 1977;137:203–20.
[29] Shor R, Halabe A, Rishver S, et al. Severe hypophosphatemia in sepsis as
[1] Majumdar A. Sepsis-induced acute kidney injury. Indian J Crit Care Med a mortality predictor. Ann Clin Lab Sci 2006;36:67–72.
2010;14:14–21. [30] Liamis G, Milionis HJ, Elisaf M. Medication-induced hypophosphate-
[2] Nash K, Hafeez A, Hou S. Hospital-acquired renal insufficiency. Am J mia: a review. QJM 2010;103:449–59.
Kidney Dis 2002;39:930–6. [31] Martinez MJ, Martinez MA, Montero M, et al. Hypophosphatemia in
[3] Case J, Khan S, Khalid R, et al. Epidemiology of acute kidney injury in the postoperative patients with total parenteral nutrition: influence of
intensive care unit. Crit Care Res Pract 2013;2013:479730. nutritional support teams. Nutr Hosp 2006;21:657–60.
[4] Samimagham HR, Kheirkhah S, Haghighi A, et al. Acute kidney injury in [32] Paterson CR, Naismith KI, Young JA. Severe unexplained hypophos-
intensive care unit: incidence, risk factors and mortality rate. Saudi J phatemia. Clin Chem 1992;38:104–7.
Kidney Dis Transpl 2011;22:464–70. [33] Davis SV, Olichwier KK, Chakko SC. Reversible depression of
[5] Kerr M, Bedford M, Matthews B, et al. The economic impact of acute myocardial performance in hypophosphatemia. Am J Med Sci 1988;
kidney injury in England. Nephrol Dial Transplant 2014;29:1362–8. 295:183–7.
[6] Chertow GM, Burdick E, Honour M, et al. Acute kidney injury, [34] Bollaert PE, Levy B, Nace L, et al. Hemodynamic and metabolic effects of
mortality, length of stay, and costs in hospitalized patients. J Am Soc rapid correction of hypophosphatemia in patients with septic shock.
Nephrol 2005;16:3365–70. Chest 1995;107:1698–701.
[7] Garg AX, Parikh CR. Yin and Yang: acute kidney injury and chronic [35] O’Connor LR, Wheeler WS, Bethune JE. Effect of hypophosphatemia on
kidney disease. J Am Soc Nephrol 2009;20:8–10. myocardial performance in man. N Engl J Med 1977;297:901–3.
[8] Rangel-Frausto MS, Pittet D, Costigan M, et al. The natural history of the [36] Saglikes Y, Massry SG, Iseki K, et al. Effect of phosphate depletion on
systemic inflammatory response syndrome (SIRS). A prospective study. blood pressure and vascular reactivity to norepinephrine and angiotensin
JAMA 1995;273:117–23. II in the rat. Am J Physiol 1985;248:F93–99.
[9] Ronco C, Kellum JA, Bellomo R, et al. Potential interventions in sepsis- [37] Craddock PR, Yawata Y, VanSanten L, et al. Acquired phagocyte
related acute kidney injury. Clin J Am Soc Nephrol 2008;3:531–44. dysfunction. A complication of the hypophosphatemia of parenteral
[10] Adekola OO, Soriyan OO, Meka I, et al. The incidence of electrolyte and hyperalimentation. N Engl J Med 1974;290:1403–7.
acid-base abnormalities in critically ill patients using point of care testing [38] Brunelli SM, Goldfarb S. Hypophosphatemia: clinical consequences and
(i-STAT portable analyser). Nig Q J Hosp Med 2012;22:103–8. management. J Am Soc Nephrol 2007;18:1999–2003.
[11] Lee CT, Guo HR, Chen JB. Hyponatremia in the emergency department. [39] Schiffl H, Lang SM. Severe acute hypophosphatemia during renal
Am J Emerg Med 2000;18:264–8. replacement therapy adversely affects outcome of critically ill patients
[12] Shiber JR, Mattu A. Serum phosphate abnormalities in the emergency with acute kidney injury. Int Urol Nephrol 2013;45:191–7.
department. J Emerg Med 2002;23:395–400. [40] Yang Y, Zhang P, Cui Y, et al. Hypophosphatemia during continuous
[13] Joannidis M. Continuous renal replacement therapy in sepsis and veno-venous hemofiltration is associated with mortality in critically ill
multisystem organ failure. Semin Dial 2009;22:160–4. patients with acute kidney injury. Crit Care 2013;17:R205.
[14] Giuliani C, Peri A. Effects of hyponatremia on the brain. J Clin Med [41] Demirjian S, Teo BW, Guzman JA, et al. Hypophosphatemia during
2014;3:1163–77. continuous hemodialysis is associated with prolonged respiratory failure
[15] Efstratiadis G, Sarigianni M, Gougourelas I. Hypomagnesemia and in patients with acute kidney injury. Nephrol Dial Transplant 2011;
cardiovascular system. Hippokratia 2006;10:147–52. 26:3508–14.
[16] Al-Shoha M, Klair JS, Girotra M, et al. Magnesium toxicity-induced [42] Aziz M, Jacob A, Wang P. Revisiting caspases in sepsis. Cell Death Dis
ileus in a postpartum patient treated for preeclampsia with magnesium 2014;5:e1526.
sulphate. ACG Case Rep J 2015;2:227–9. [43] Roderburg C, Benz F, Cardenas DV, et al. Persistently elevated
[17] Park JT, Lee H, Kee YK, et al. High-dose versus conventional-dose osteopontin serum levels predict mortality in critically ill patients. Crit
continuous venovenous hemodiafiltration and patient and kidney Care 2015;19:271.
survival and cytokine removal in sepsis-associated acute kidney injury: [44] Schulte W, Bernhagen J, Bucala R. Cytokines in sepsis: potent
a randomized controlled trial. Am J Kidney Dis 2016;doi: 10.1053/j. immunoregulators and potential therapeutic targets–an updated view.
ajkd.2016.02.049. Epub 2016 Apr 12. Mediators Inflamm 2013;2013:165974.
[18] Bone RC, Balk RA, Cerra FB, et al. Definitions for sepsis and organ failure [45] Hirsiger S, Simmen HP, Werner CM, et al. Danger signals activating the
and guidelines for the use of innovative therapies in sepsis. The ACCP/ immune response after trauma. Mediators Inflamm 2012;2012:315941.
SCCM Consensus Conference Committee. American College of Chest [46] Lau WL, Pai A, Moe SM, et al. Direct effects of phosphate on vascular
Physicians/Society of Critical Care Medicine. Chest 1992;101:1644–55. cell function. Adv Chronic Kidney Dis 2011;18:105–12.
[19] Bellomo R, Ronco C, Kellum JA, et al. Acute dialysis quality initiative w. [47] Dhingra R, Sullivan LM, Fox CS, et al. Relations of serum phosphorus
acute renal failure—definition, outcome measures, animal models, fluid and calcium levels to the incidence of cardiovascular disease in the
therapy and information technology needs: the Second International community. Arch Intern Med 2007;167:879–85.
Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) [48] Papa S, Skulachev VP. Reactive oxygen species, mitochondria, apoptosis
Group. Crit Care 2004;8:R204–212. and aging. Mol Cell Biochem 1997;174:305–19.
[20] Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate [49] McGovern AP, de Lusignan S, van Vlymen J, et al. Serum phosphate as a risk
glomerular filtration rate. Ann Intern Med 2009;150:604–12. factor for cardiovascular events in people with and without chronic kidney
[21] Tsagalis G. Update of acute kidney injury: intensive care nephrology. disease: a large community based cohort study. PLoS One 2013;8:e74996.
Hippokratia 2011;15:53–68. [50] Silva AP, Fragoso A, Pinho A, et al. Phosphorus as an early marker of
[22] Claure-Del Granado R, Bouchard J. Acid–base and electrolyte morbidity and mortality in type 2 chronic kidney disease diabetic
abnormalities during renal support for acute kidney injury: recognition patients. J Diabetes Complications 2013;27:328–32.
and management. Blood Purif 2012;34:186–93. [51] O’Seaghdha CM, Hwang SJ, Muntner P, et al. Serum phosphorus
[23] Statland H. A fluid and electrolyte balance service for clinical use. J Am predicts incident chronic kidney disease and end-stage renal disease.
Med Assoc 1952;150:771–2. Nephrol Dial Transplant 2011;26:2885–90.