Journal of the Formosan Medical Association (2017) 116, 844e851

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.jfma-online.com

ORIGINAL ARTICLE

Comparison of RIFLE, AKIN, and KDIGO

classifications for assessing prognosis of
patients on extracorporeal membrane
oxygenation
Tsung-Yu Tsai a,b,d, Hao Chien a,d, Feng-Chun Tsai a,c,
Heng-Chih Pan a,b, Huang-Yu Yang a,b, Shen-Yang Lee a,b,
Hsiang-Hao Hsu a,b, Ji-Tseng Fang a,b, Chih-Wei Yang a,b,
Yung-Chang Chen a,b,*

a
Chang Gung University College of Medicine, Taoyuan, Taiwan
b
Kidney Research Center, Department of Nephrology, Chang Gung Memorial Hospital, Taiwan
c
Division of Cardiovascular Surgery, Chang Gung Memorial Hospital, Taiwan

Received 24 May 2017; received in revised form 4 August 2017; accepted 10 August 2017

KEYWORDS Abstract Background/Purpose: Acute kidney injury (AKI) developing during extracorporeal
AKIN; membrane oxygenation (ECMO) is associated with very poor outcome. The Kidney Disease:
ECMO; Improving Global Outcomes (KDIGO) group published a new AKI definition in 2012. This study
KDIGO; analyzed the outcomes of patients treated with ECMO and identified the relationship between
Prognosis; the prognosis and the KDIGO classification.
RIFLE Methods: This study examined total 312 patients initially, and finally reviewed the medical
records of 167 patients on ECMO support at a tertiary care university hospital between March
2002 and November 2011. Demographic, clinical, and laboratory variables were retrospectively
collected as survival predicators.
Results: The overall mortality rate was 55.7%. In the analysis of the areas under the receiver
operating characteristic curves, the KDIGO classification showed relatively higher discrimina-
tory power (0.840
0.032) than the Risk of renal failure, Injury to the kidney, Failure of kidney
function, Loss of kidney function, and End-stage renal failure (RIFLE) (0.826
0.033) and
Acute Kidney Injury Network (AKIN) (0.836
0.032) criteria in predicting in-hospital mortality.
Furthermore, multiple logistic regression analysis showed that KDIGO, hemoglobin, and Glas-
gow Coma Scale score on the first day of patients on ECMO were independent predictors for

Conflicts of interest: The authors have no conflicts of interest relevant to this article.
* Corresponding author. Department of Nephrology, Chang Gung Memorial Hospital, Chang Gung University College of Medicine, No. 222,
Maijin Road, Anle District, Keelung, Taiwan. Fax: þ886 3 3282173.
E-mail addresses: cyc2356@adm.cgmh.org.tw, cyc2356@hotmail.com, cyc2356@gmail.com (Y.-C. Chen).
d
Tsung-Yu Tsai and Hao Chien contributed equally to this manuscript.

http://dx.doi.org/10.1016/j.jfma.2017.08.004
0929-6646/Copyright ª 2017, Formosan Medical Association. Published by Elsevier Taiwan LLC. This is an open access article under the
CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
KDIGO and ECMO 845

in-hospital mortality. Finally, cumulative survival rates at 6-month follow-up after hospital
discharge differed significantly for KDIGO stage 3 versus KDIGO stage 0, 1, and 2
(p < 0.001); and KDIGO stage 2 versus KDIGO stage 0 (p < 0.05).
Conclusion: For those patients with ECMO support, the KDIGO classification proved to be a
more reproducible evaluation tool with excellent prognostic abilities than RIFLE or AKIN clas-
sification.
Copyright ª 2017, Formosan Medical Association. Published by Elsevier Taiwan LLC. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).

Introduction 0.5 ml/kg/hr for 6 h. Comparing to the AKIN criteria, KDIGO

Extracorporeal membrane oxygenation (ECMO) is often
used in critical patients with severe, reversible myocardial
dysfunction (e.g., postoperative cardiogenic shock) and
provides a bridging therapy for further treatment. Howev-
er, acute kidney injury (AKI) occurs frequently during ECMO
and is usually associated with very poor outcome.1,2 AKI
during ECMO is often multifactorial and possibly results
from hypotension, damaged cardiac function, new-onset
infection, or inflammatory-like reaction triggered by
ECMO implementation. Fluid overload following decreased
urine amount usually impairs tissue oxygenation and pul-
monary oxygen transport, eventually leading to organ
dysfunction of the heart, lungs, and brain.3
The Kidney Disease: Improving Global Outcomes (KDIGO)
group classification was proposed on the basis of the Risk,
Injury, Failure, Loss of Kidney Function, and End-stage
Kidney Disease (RIFLE)4 and the Acute Kidney Injury
Network (AKIN)5 classifications in 2012. AKI was defined as
an increase of the serum creatinine (SCr) level exceeding
0.3 mg/dl within 48 h, or an increase in SCr to 1.5 times the
baseline value within 7 days, or a urine output less than
also classifies patients initiating renal replacement therapy
(RRT) into stage 3 AKI, but removes the threshold of a
0.5 mg/dl increment for SCr >4 mg/dl in the criteria of
stage 3 AKI.6 The comparison of the definitions in three
classifications is shown in Table 1. Up to now, this new AKI
classification has been assessed in several studies,7e11 but
the application in patients treated with ECMO has not been
thoroughly evaluated.
Since previous studies have proved that both AKIN and
RIFLE classifications have excellent prognostic abilities to
predict the in-hospital mortality in patients on ECMO,1,12
we aimed to identify the relationship between in-hospital
mortality and KDIGO classification in critically ill ECMO
patients in this study.
Material and methods
Patient information and data collection
This study was approved by the local Institutional Review
Board of Chang Gung Memorial Hospital with a waiver for

Table 1 Comparison of RIFLE, AKIN, and KDIGO classifications.

Serum creatinine Urine output
RIFLE AKIN KDIGO
Definition SCr increase S50% SCr increase S50% SCr increase S0.3 mg/dL UO < 0.5 mL/kg/h
within 7 days or S0.3 mg/dL within 48 h within 48 h or S50% for 6 h
within 7 days
Staging RIFLE AKIN KDIGO
RIFLE-Risk SCr increase S50% or SCr increase S50% SCr increase S0.3 mg/dL UO < 0.5 mL/kg/h
AKIN stage 1 GFR decrease >25% or S0.3 mg/dL within 48 h or S50% within for 6 h
KDIGO stage 1 7 days
RIFLE-Injury SCr increase S100% or SCr increase S100% SCr increase S100% UO < 0.5 mL/kg/h
AKIN stage 2 GFR decrease >50% for 12 h
RIFLE stage 2
RIFLE-Failure SCr increase S200% or SCr increase S200% SCr increase S200% or UO < 0.3 mL/kg/h
AKIN stage 3 GFR decrease >75% or SCr S 4 mg/dL SCr S 4 mg/dL or need RRT for 24 h or anuria
KDIGO stage 3 or SCr S 4 mg/dL (with (with an acute rise for 12 h
an acute rise S0.5 mg/dL) S0.5 mg/dL) or need RRT
RIFLE-Loss Need RRT for >4 weeks
RIFLE-End stage Need RRT for >3 months
Abbreviation: RIFLE, risk of renal failure, injury to the kidney, failure of kidney function, loss of kidney function, and end-stage renal
failure; AKIN, Acute Kidney Injury Network; KDIGO, Kidney Disease Improving Global Outcome; SCr, serum creatinine; UO, urine output;
GFR, glomerular filtration rate; RRT, renal replacement therapy.
846
informed consent. Medical records of 312 patients on ECMO
support in the intensive care units (ICU) between March
2002 and November 2011 were examined. Patients on ECMO
support who expired within 24 h (15 patients), patients with
end stage renal disease (ESRD) (14 patients), pediatric pa-
tients younger than 18-year-old (38 patients), and patients
with acute respiratory distress syndrome (ARDS) (78 pa-
tients) were excluded. Finally, total 167 patients were
examined in this study. For patients with repeated ECMO
support during hospitalization, we only collected the data
on the first ECMO support.
Clinical data were obtained retrospectively, including
demographic data, primary diagnosis for ICU admission,
RIFLE, AKIN, and KDIGO classifications on day 1 of ECMO
support, and outcome. The worst physiological values on
day 1 of ECMO support were recorded for physiological
calculations. Blood transfusions and pressor/inotrope
agents were administered and titrated at the discretion of
the treating clinicians to meet acceptable clinical targets.
The primary study outcome was in-hospital mortality.
Follow-up at 6 months after hospital discharge was per-
formed via chart record review.
Definitions
The occurrence and severity of AKI were judged by the defi-
nitions of RIFLE, AKIN, and KDIGO classifications. For the AKIN
criteria, the lowest SCr within 48 h before ECMO imple-
mentation was considered as the baseline SCr. For the RIFLE
and KDIGO criteria, the lowest SCr within 7 days before ECMO
implementation was used instead. In each classification, both
the SCr and urine output criteria were checked, and the
criteria corresponding to the worse stage were chosen. In
order to simplify the record of AKI stages in different classi-
fications, we set a simple model as follows: non-AKI (0 points);
RIFLE-R, AKIN stage 1, and KDIGO stage 1 (1 point); RIFLE-I,
AKIN stage 2, and KDIGO stage 2 (2 points); RIFLE-F, AKIN stage
3, and KDIGO stage 3 (3 points) on day 1 of ECMO support.
Clinical management
The ECMO device (Medtronic, Inc., Anaheim, CA) included a
centrifugal pump and a silicone oxygenator (Medtronics,
Minneapolis, MN, USA) with an integrated heater. All ECMO
circuits had a heparin-bound Carmeda bioactive surface.
Peripheral cutaneous cannulation with a closed sternum was
preferred, and cut-down procedures were necessary for some
obese patients. Due to relatively small body size of oriental
patients, we used a 17e19 Fr percutaneous arterial (infusion)
cannula and 19e21 Fr percutaneous venous (drainage) can-
nula (DLP; Medtronic Inc., Minneapolis, MN). Percutaneous
access was preferred through the common femoral vein
(drainage) and the common femoral artery (infusion) for
venoarterial ECMO. An 8 Fr distal perfusion catheter was
implanted into the ipsilateral superficial femoral artery if
cyanosis was noted in the cannulated limb.
Statistical analysis
Descriptive statistics were expressed as means
standard
deviation. Primary analysis compared hospital survivors
T.-Y. Tsai et al.
with non-survivors. All variables were tested for normal
distributions with the KolmogoroveSmirnov test. The Stu-
dent t-test was employed to compare means of continuous
variables and normally distributed data; otherwise, the
ManneWhitney U test was applied. This study utilized the
c2 test for trend to assess categorical data associated with
RIFLE, AKIN, and KDIGO classifications. Moreover, correla-
tions between paired variables within groups were assessed
by using linear regression and Pearson analysis. Risk factors
were assessed by the univariate analysis, and statistically
significant (p < 0.05) variables in the univariate analysis
were subjected to the multivariate analysis by applying a
multiple logistic regression based on forward elimination of
data (admission year, duration of ECMO support, weaning
from ECMO support, and dialysis were not included).
Calibration was assessed by using the Hos-
mereLemeshow goodness-of-fit test to compare the num-
ber of observed and predicted deaths in risk groups for the
entire range of death probabilities. Discrimination was
assessed by using the area under the receiver operating
characteristic curve (AUROC), and the AUROCs were
compared by a nonparametric approach. The AUROC anal-
ysis was also utilized to calculate cutoff values, sensitivity,
specificity, and overall correctness. Finally, cutoff points
were calculated by obtaining the best Youden index
(sensitivity þ specificity  1). Cumulative survival curves as
a function of time were plotted by utilizing the
KaplaneMeier approach, and compared by using the log
rank test. All statistical tests were two-tailed; a value of
p < 0.05 was considered statistically significant. Data were
analyzed with SPSS 19.0 for Windows (SPSS, Inc., Chicago,
IL, USA).
Results
Subject characteristics
In this study, we examined 167 patients on ECMO support in
the ICU between March 2002 and November 2011. Among
these patients, 102 patients (61.1%) were male and 65 pa-
tients (38.9%) were female. The average age was 56 years
old, and the overall in-hospital mortality rate was 55.7%.
Table 2 compares patients’ demographic data and clinical
characteristics between survivors and non-survivors. The
in-hospital mortality rate increased as AKI stage advanced,
and rose sharply in the highest stage of each classification
(e.g., 88% (66 of 75) for RIFLE-failure, 88.6% (70 of 79) for
AKIN stage 3, and 88.6% (70 of 79) for KDIGO stage 3). Table
3 presents the primary diagnosis for ECMO support. In this
study, the most frequent indication for ECMO support was
postcardiotomy cardiogenic shock (62.9%).
In-hospital mortality and short-term prognosis
Table 4 shows significantly prognostic variables identified
by univariate and multivariate analysis. Univariate analysis
was used to examine the 12 variables in Table 2, all of
which was identified as prognostically valuable. Then,
multivariate analysis further identified the following vari-
ables as of independent prognostic significance: KDIGO
classification, Glasgow coma scale, and hemoglobin. The
KDIGO and ECMO 847

Table 2 Patients’ demographic data and clinical characteristics.

All patients (n Z 167) Survivors (n Z 74) Non survivors (n Z 93) p value
Admission year 2002e05/2006e08/2009e11 52/62/53 19/24/31 33/38/22 0.041
Age (years) 56
15 54
15 58
15 NS (0.106)
Gender (M/F) 102/65 46/28 56/37 NS (0.798)
Duration of ECMO support (days) 6
5 5
4 7
6 0.002
Weaning from ECMO support (Yes/No) 99/68 74/0 25/68 <0.001
Dialysis (Yes/No) 89/78 14/60 75/18 <0.001
GCS on ECMO first day (points) 10
5 12
4 8
4 <0.001
MAP on ECMO first day (mmHg) 51
14 55
14 48
14 0.003
UO on ECMO first day (mL) 1438
1663 2385
1662 792
1327 <0.001
SCr baseline for AKIN (mg/dL) 1.4
1.1 1.3
0.7 1.6
1.2 NS (0.052)
SCr baseline for RIFLE/KDIGO (mg/dL) 1.3
0.8 1.2
0.7 1.4
0.9 NS (0.199)
SCr on ECMO first day (mg/dL) 2.6
1.7 2.0
1.5 3.2
1.7 <0.001
WBC count (cu/mm  1000) 13
8 14
7 13
8 NS (0.543)
Hemoglobin on ECMO first day (g/dL) 8.9
2.0 9.4
2.1 8.6
1.8 0.007
Sodium on ECMO first day (mmol/L) 148
9 147
6 149
11 NS (0.150)
Potassium on ECMO first day (mmol/L) 4.2
1.4 3.7
1.0 4.6
1.6 <0.001
Albumin on ECMO first day (g/L) 2.6
0.6 2.8
0.7 2.5
0.6 0.005
Bilirubin on ECMO first day (mg/dL) 2.8
4.6 2.3
5.0 3.2
4.3 NS (0.189)
Lactate on ECMO first day (mmol/L) 107
73 89
58 122
80 0.005
AaDO2 on ECMO first day 298
200 235
179 349
202 <0.001
RIFLE Non-AKI/Risk/Injury/Failure 41/20/31/75 33/14/18/9 8/6/13/66 <0.001a
AKIN stage 0/1/2/3 26/38/24/79 23/25/17/9 3/13/7/70 <0.001a
KDIGO stage 0/1/2/3 25/36/27/79 22/25/18/9 3/11/9/70 <0.001a
Abbreviation: M, male; F, female; ECMO, extracorporeal membrane oxygenation; GCS, Glasgow Coma Scale; MAP, mean arterial pres-
sure; UO, urine output; SCr, serum creatinine; AKIN, Acute Kidney Injury Network; RIFLE, risk of renal failure, injury to the kidney,
failure of kidney function, loss of kidney function, and end-stage renal failure; KDIGO, Kidney Disease Improving Global Outcome; WBC,
white blood cell; AaDO2, alveolarearterial oxygen difference; AKI, acute kidney injury; NS, not significant.
a
Chi-square test for trend.

logarithm of the odds of death could be calculated with
regression coefficients of these variables as follows:
The logarithm of the odds of death Z 3.615 þ 1.393  KDIGO
classification (stage 0 Z 0/stage 1 Z 1/stage 2 Z 2/stage
3 Z 3)  0.291  Glasgow coma scale  0.331  Hemoglobin.
Calibration, discrimination, and correlation for
illness scoring systems
Table 5 shows the goodness-of-fit, as measured by the
HosmereLemeshow c2 test for predicted mortality risk, and
the predictive accuracy of those classifications. Table 5 also
compares the discriminatory value of the RIFLE, AKIN, and
KDIGO classifications. The AUROC analysis proves that all
three classifications had excellent discriminatory power,
and KDIGO classification was relatively higher than the
other two classifications.
Table 6 compares the predictive ability of those AKI
classifications and significantly prognostic variables. In-
hospital mortality, sensitivity, specificity, and overall cor-
rectness of prediction were assessed, and both the KDIGO
and AKIN classifications had the best Youden index and
highest overall prediction correctness. Cumulative survival
rates differed significantly for KDIGO stage 3 versus KDIGO

Table 3 Primary diagnosis for ECMO support.

All patients n (%) Survivors n (%)a Non survivors n (%)a p-value
Primary diagnosis for ECMO support
Post-cardiotomy cardiogenic shock 105 (63) 47 (64) 58 (62) NS (0.879)
Acute myocardial infarction 19 (11) 6 (8) 13 (14) NS (0.235)
Myocarditis 15 (9) 12 (16) 3 (3) 0.004
Post-lung or heart transplantation 4 (2) 1 (1) 3 (3) NS (0.630)
Shock status simultaneous with hypoxic 5 (3) 0 (0) 5 (5) NS (0.067)
respiratory failure
Decompensated heart failure 9 (6) 6 (8) 3 (3) NS (0.165)
Sudden collapse status post CPCR 6 (4) 2 (3) 4 (5) NS (0.694)
Severe traumatic injury 4 (2) 0 (0) 4 (5) NS (0.130)
Abbreviation: ECMO, extracorporeal membrane oxygenation; CPCR, cardiopulmonary cerebral resuscitation.
a
Number (%) of patients with the condition who survived or died.
848 T.-Y. Tsai et al.

Table 4 Variables showing prognostic significance.

Variable Beta Coefficient Standard error Odds ratio p value
Univariate logistic regression
GCS on ECMO first day (points) 0.238 0.043 0.788 (0.725e0.857) <0.001
MAP on ECMO first day (mmHg) 0.035 0.012 0.966 (0.944e0.989) 0.003
UO on ECMO first day (mL) 0.001 0.000 0.999 (0.999e1.000) <0.001
SCr on ECMO first day (mg/dL) 0.576 0.142 1.778 (1.347e2.348) <0.001
Hemoglobin on ECMO first day (g/dL) 0.223 0.085 0.800 (0.677e0.945) 0.009
Potassium on ECMO first day (mmol/L) 0.630 0.159 1.879 (1.375e2.565) <0.001
Albumin on ECMO first day (g/L) 0.769 0.286 0.463 (0.265e0.811) 0.007
Lactate on ECMO first day (mmol/L) 0.007 0.003 1.007 (1.002e1.012) 0.008
AaDO2 on ECMO first day 0.003 0.001 1.003 (1.001e1.005) <0.001
RIFLE 1.130 0.171 3.095 (2.214e4.324) <0.001
AKIN 1.310 0.193 3.705 (2.540e5.405) <0.001
KDIGO 1.359 0.200 3.891 (2.629e5.759) <0.001
Multivariate logistic regression
KDIGO 1.393 0.271 4.028 (2.268e6.851) <0.001
GCS on ECMO first day (points) 0.291 0.075 0.747 (0.645e0.866) <0.001
Hemoglobin on ECMO first day (g/dL) 0.331 0.148 0.718 (0.538e0.960) 0.025
Constant 3.615 1.793 e 0.044
Abbreviation: GCS, Glasgow Coma Scale; ECMO, extracorporeal membrane oxygenation; MAP, mean arterial pressure; UO, urine output;
SCr, serum creatinine; AaDO2, alveolarearterial oxygen difference; RIFLE, risk of renal failure, injury to the kidney, failure of kidney
function, loss of kidney function, and end-stage renal failure; AKIN, Acute Kidney Injury Network; KDIGO, Kidney Disease Improving
Global Outcome.

Table 5 Comparison of calibration and discrimination of the scoring methods in predicting hospital mortality.
Calibration Discrimination
HosmereLemeshow c 2
df p value AUROC
SE 95% CI p value
KDIGO 8.573 2 0.014 0.840
0.032 0.778e0.903 <0.001
AKIN 10.763 2 0.005 0.836
0.032 0.774e0.899 <0.001
RIFLE 7.222 2 0.027 0.826
0.033 0.761e0.891 <0.001
Abbreviation: df, degree of freedom; AUROC, area under the receiver operating characteristic curve; SE, standard error; CI, confidence
interval; KDIGO, Kidney Disease Improving Global Outcome; AKIN, Acute Kidney Injury Network; RIFLE, risk of renal failure, injury to the
kidney, failure of kidney function, loss of kidney function, and end-stage renal failure.

stages 0, 1, and 2 (p < 0.001); and KDIGO stage 2 versus AKI is a common complication in patients on ECMO, and
KDIGO stage 0 (p < 0.05) (Fig. 1). often results in increased mortality and poor out-
come.12,15e17 Advanced AKI usually accompanied with fluid
overload due to decreased urine amount. Fluid overload
Discussion
impairs pulmonary oxygen transport and causes cardiac
dysfunction, leading to prolonged mechanic ventilator and
The overall in-hospital mortality rate in this study was
ECMO support. Besides, fluid overload is the most common
55.7%, which is similar to that in previous studies for pa-
indication of renal replacement therapy in patients on
tients on ECMO support2,12e14 and confirms poor outcome in
ECMO,18 and associated with increased mortality in criti-
this subgroup of patients. However, the cumulative survival
cally ill patients.16,19 Moreover, severe AKI may produce an
rate increased gradually in recent years, which may result
irreversible loss of nephron units with deleterious effects
from experience acumination and care quality improve-
on renal reserve, leading to chronic kidney disease.20 It is
ment in patients on ECMO (Table 2). KDIGO, AKIN, and RIFLE
important to treat AKI in patients on ECMO and avoid
classifications all had excellent performance in outcome
accompanied complications or further multiple organ
prediction, and the KDIGO had relatively higher discrimi-
failure.
natory power than the other two classifications (Table 5).
RIFLE and AKIN classification have been proved to have
This study also demonstrated that KDIGO classification,
good prediction ability in in-hospital mortality in patients
Glasgow coma scale, and hemoglobin on day 1 of ECMO
on ECMO,1,12,21,22 but KDIGO classification has not been well
support were significantly related to in-hospital mortality
studied yet. In this study, AKI is associated with higher
(Table 6). Besides, cumulative survival rates differed
mortality, and the mortality rate rises as the stage ad-
significantly for KDIGO stage 3 versus KDIGO stages 0, 1, and
vances in all three classifications, especially in the highest
2; and KDIGO stage 2 versus KDIGO stage 0 (Fig. 1).
KDIGO and ECMO 849

Table 6 Prediction of subsequent in-hospital mortality.

Predictive factors Cutoff point Youden index Sensitivity (%) Specificity (%) Overall correctness (%)
a
KDIGO Stage 2 0.63 75 88 82
AKIN Stage 2a 0.63 75 88 82
RIFLE Stage 2a 0.59 71 88 79
GCS 12 0.49 72 77 75
Hemoglobin 9.9 0.21 35 86 61
Abbreviation: KDIGO, Kidney Disease Improving Global Outcome; AKIN, Acute Kidney Injury Network; RIFLE, risk of renal failure, injury
to the kidney, failure of kidney function, loss of kidney function, and end-stage renal failure; GCS, Glasgow Coma Scale.
a
Value giving the best Youden index.

stages (RIFLE-F, AKIN-3, and KDIGO stage 3). Besides, KDIGO
classification also has better prognostic abilities than RIFLE
or AKIN classification. In view of the definitions, RIFLE
categorizes an increase in SCr to 1.5 times the baseline
value within 7 days into the AKI group, but KDIGO also in-
cludes an increase of SCr level exceeding 0.3 mg/dl within
48 h. In contrast, AKIN utilizes the SCr change within 48 h
for AKI diagnosis, but KDIGO extends the time frame to
7 days. The difference in AKI diagnosis may affect the stage
judgment and decrease AKI overlooking due to improved
sensitivity.6,23,24 In our study, KDIGO classification identi-
fied 16 AKI patients who was classified as non-AKI in RIFLE
classification. These 16 patients had lower survival rate
(68.8%, 11 of 16) than the other patients classified as non-
AKI by RIFLE classification (88%, 22 of 25). Since the
creatinine change on the first day of ECMO support may be

Figure 1 Cumulative survival rate for 167 critically ill patients based on their Kidney Disease Improving Global Outcome (KDIGO)
classification staging.
850
not obvious in patients with mild AKI, KDIGO classification
helped us identify these patients who were not screened by
RIFLE classification, reminding us to give proper treatment
and monitor renal function change. Therefore, KDIGO
classification obtains better outcome prediction than RIFLE
and AKIN classification.
Anemia is a risk factor for AKI and long-term mortality in
critically ill patients.25,26 In patients on ECMO, pre-ECMO
anemia may exist due to hospital-acquired anemia, iron
deficiency anemia, or anemia of chronic disease.27 Besides,
blood loss during ECMO procedure and hemolysis in the
extracorporeal circuit also cause anemia. Anemia induces
decreased oxygen delivery and organ hypoperfusion, which
in turn aggravates multi-organ dysfunction and increases in-
hospital mortality.25e28 Moreover, severe anemia increases
the need of blood transfusion, possibly followed by
numerous insult including lung injury, immune system
dysfunction, and fluid overload.29 In this study, patients
with hemoglobin less than 9.9 g/dL on the day 1 of ECMO
support has significantly worse in-hospital mortality, which
is similar to the result in previous studies30,31 and helpful to
identify patients with worse outcome.
Neurologic complications including intracranial hemor-
rhage, ischemic stroke, and seizure are often observed in
patients on ECMO and associated with increased mortal-
ity.13,32e34 Patients on ECMO usually have unstable and
fluctuating hemodynamic status, combined with comor-
bidities such as infection, bleeding, or heart failure. The
brain is very vulnerable to hypoxia and hypoperfusion, so
neurologic injury may occur before, during, or after ECMO
support once inadequate oxygen and blood supply happens.
Glasgow coma scale has provided a practical method for
assessment of conscious impairment, the hallmark of acute
brain injury for many years.35 It reflects the severity of
neurologic complications and acts as a convenient tool for
outcome prediction in patients on ECMO.
Despite the promising results obtained by this study,
several important limitations must be recognized. First,
this retrospective study was performed at a single tertiary
care medical center, thereby limiting the generalization of
the findings. Second, difficulties were encountered by this
retrospective study because some laboratory data were not
available for all patients. Third, the patient group con-
sisted of patients on ECMO and excluded those who died
within 24 h, so the results may not be directly extrapolated
to other patient populations. Finally, the predictive accu-
racy of logistic regression models has its own limitations.
Further validation is necessary in the future due to the
relatively small sample size at a single tertiary-care med-
ical center.
Financial support
The authors declare no financial or non-financial conflicts of
interest.
Acknowledgement
This study was partly supported by the Chang Gung Me-
morial Hospital Research Program (CMRPG3D0791).
T.-Y. Tsai et al.
References
1. Lin CY, Chen YC, Tsai FC, Tian YC, Jenq CC, Fang JT, et al.
RIFLE classification is predictive of short-term prognosis in
critically ill patients with acute renal failure supported by
extracorporeal membrane oxygenation. Nephrol Dial Transpl
2006;21:2867e73.
2. Lin CY, Tsai FC, Tian YC, Jenq CC, Chen YC, Fang JT, et al. Eval-
uation of outcome scoring systems for patients on extracorporeal
membrane oxygenation. Ann Thorac Surg 2007;84:1256e62.
3. Chen CY, Tsai FC, Fang JT, Yang CW. Acute kidney injury in
adults receiving extracorporeal membrane oxygenation. J
Formos Med Assoc 2014;113:778e85.
4. Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky P, Acute
Dialysis Quality Initiative workgroup. Acute renal failure e
definition, outcome measures, animal models, fluid therapy
and information technology needs: the Second International
Consensus Conference of the Acute Dialysis Quality Initiative
(ADQI) group. Crit Care 2004;8(4):R204e12.
5. Mehta RL, Kellum JA, Shah SV, Molitoris BA, Ronco C,
Warnock DG, et al. Acute Kidney Injury Network: report of an
initiative to improve outcomes in acute kidney injury. Crit Care
2007;11(2):R31.
6. Kidney Disease: Improving Global Outcomes (KDIGO) Acute
Kidney Injury Work Group. KDIGO clinical practice guideline for
acute kidney injury. Kidney Int Suppl 2012;2:1e138.
7. Bastin AJ, Ostermann M, Slack AJ, Diller GP, Finney SJ,
Evans TW. Acute kidney injury after cardiac surgery according
to Risk/Injury/Failure/Loss/End-stage, Acute Kidney Injury
Network, and Kidney Disease: Improving Global Outcomes
classifications. J Crit Care 2013;28:389e96.
8. Machado MN, Nakazone MA, Maia LN. Acute kidney injury
based on KDIGO (Kidney Disease Improving Global Outcomes)
criteria in patients with elevated baseline serum creatinine
undergoing cardiac surgery. Rev Bras Cir Cardiovasc 2014;29:
299e307.
9. Zeng X, McMahon GM, Brunelli SM, Bates DW, Waikar SS. Inci-
dence, outcomes, and comparisons across definitions of AKI in
hospitalized individuals. Clin J Am Soc Nephrol 2014;9:12e20.
10. Fujii T, Uchino S, Takinami M, Bellomo R. Validation of the
Kidney Disease Improving Global Outcomes criteria for AKI and
comparison of three criteria in hospitalized patients. Clin J Am
Soc Nephrol 2014;9:848e54.
11. Kuo G, Yang SY, Chuang SS, Fan PC, Chang CH, Hsiao YC, et al.
Using acute kidney injury severity and scoring systems to
predict outcome in patients with burn injury. J Formos Med
Assoc 2016;115:1046e52.
12. Chen YC, Tsai FC, Chang CH, Lin CY, Jenq CC, Juan KC, et al.
Prognosis of patients on extracorporeal membrane oxygena-
tion: the impact of acute kidney injury on mortality. Ann
Thorac Surg 2011;91:137e42.
13. Zangrillo A, Landoni G, Biondi-Zoccai G, Greco M, Greco T,
Frati G, et al. A meta-analysis of complications and mortality
of extracorporeal membrane oxygenation. Crit Care Resusc
2013;15(3):172e8.
14. McCarthy FH, McDermott KM, Kini V, Gutsche JT, Wald JW,
Xie D, et al. Trends in U.S. extracorporeal membrane oxygen-
ation use and outcomes: 2002e2012. Semin Thorac Cardiovasc
Surg 2015;27(2):81e8.
15. Tsai TY, Tsai FC, Chang CH, Jenq CC, Hsu HH, Chang MY, et al.
Prognosis of patients on extracorporeal membrane oxygenation
plus continuous arteriovenous hemofiltration. Chang Gung Med
J 2011;34(6):636e43.
16. Askenazi DJ, Selewski DT, Paden ML, Cooper DS, Bridges BC,
Zappitelli M, et al. Renal replacement therapy in critically ill
patients receiving extracorporeal membrane oxygenation. Clin
J Am Soc Nephrol 2012;7(8):1328e36.
KDIGO and ECMO
17. Kielstein JT, Heiden AM, Beutel G, Gottlieb J, Wiesner O,
Hafer C, et al. Renal function and survival in 200 patients un-
dergoing ECMO therapy. Nephrol Dial Transpl 2013;28(1):
86e90.
18. Fleming GM, Askenazi DJ, Bridges BC, Cooper DS, Paden ML,
Selewski DT, et al. A multicenter international survey of renal
supportive therapy during ECMO: the Kidney Intervention
During Extracorporeal Membrane Oxygenation (KIDMO) group.
ASAIO J 2012;58:407e14.
19. Chou YH, Chen YF, Pan SY, Huang TM, Yang FJ, Shen WC, et al.
The role of brain natriuretic peptide in predicting renal
outcome and fluid management in critically ill patients. J
Formos Med Assoc 2015;114:1187e96.
20. Villa G, Katz N, Ronco C. Extracorporeal membrane oxygena-
tion and the kidney. Cardiorenal Med 2015;6:50e60.
21. Yan X, Jia S, Meng X, Dong P, Jia M, Wan J, et al. Acute kidney
injury in adult postcardiotomy patients with extracorporeal
membrane oxygenation: evaluation of the RIFLE classification
and the Acute Kidney Injury Network criteria. Eur J Car-
diothorac Surg 2010;37(2):334e8.
22. Zwiers AJ, de Wildt SN, Hop WC, Dorresteijn EM, Gischler SJ,
Tibboel D, et al. Acute kidney injury is a frequent complica-
tion in critically ill neonates receiving extracorporeal mem-
brane oxygenation: a 14-year cohort study. Crit Care 2013;
17(4):R151.
23. Li Z, Cai L, Liang X, Du Z, Chen Y, An S, et al. Identification and
predicting short-term prognosis of early cardiorenal syndrome
type 1: KDIGO is superior to RIFLE or AKIN. PLoS One 2014;
9(12), e114369.
24. Shinjo H, Sato W, Imai E, Kosugi T, Hayashi H, Nishimura K,
et al. Comparison of kidney disease: improving global out-
comes and acute kidney injury network criteria for assessing
patients in intensive care units. Clin Exp Nephrol 2014;18(5):
737e45.
25. Han SS, Baek SH, Ahn SY, Chin HJ, Na KY, Chae DW, et al.
Anemia is a risk factor for acute kidney injury and long-term
mortality in critically ill patients. Tohoku J Exp Med 2015;
237(4):287e95.
851
26. Fowler AJ, Ahmad T, Phull MK, Allard S, Gillies MA, Pearse RM.
Meta-analysis of the association between preoperative anaemia
and mortality after surgery. Br J Surg 2015;102(11):1314e24.
27. Enger T, Philipp A, Videm V, Lubnow M, Wahba A, Fischer M,
et al. Prediction of mortality in adult patients with severe
acute lung failure receiving veno-venous extracorporeal
membrane oxygenation: a prospective observational study.
Crit Care 2014;18(2):R67.
28. Ryu JA, Cho YH, Sung K, Choi SH, Yang JH, Choi JH, et al.
Predictors of neurological outcomes after successful extra-
corporeal cardiopulmonary resuscitation. BMC Anesthesiol
2015;15:26.
29. Mazzeffi M, Greenwood J, Tanaka K, Menaker J, Rector R,
Herr D, et al. Bleeding, Transfusion, and Mortality on Extra-
corporeal Life Support: ECLS Working Group on Thrombosis and
Hemostasis. Ann Thorac Surg 2016;101(2):682e9.
30. Corwin HL, Gettinger A, Pearl RG, Fink MP, Levy MM,
Abraham E, et al. The CRIT Study: anemia and blood trans-
fusion in the critically ill e current clinical practice in the
United States. Crit Care Med 2004;32(1):39e52.
31. du Cheyron D, Parienti JJ, Fekih-Hassen M, Daubin C,
Charbonneau P. Impact of anemia on outcome in critically ill
patients with severe acute renal failure. Intensive Care Med
2005;31(11):1529e36.
32. Mateen FJ, Muralidharan R, Shinohara RT, Parisi JE, Schears GJ,
Wijdicks EF. Neurological injury in adults treated with extra-
corporeal membrane oxygenation. Arch Neurol 2011;68(12):
1543e9.
33. Polito A, Barrett CS, Wypij D, Rycus PT, Netto R, Cogo PE, et al.
Neurologic complications in neonates supported with extra-
corporeal membrane oxygenation. An analysis of ELSO registry
data. Intensive Care Med 2013;39(9):1594e601.
34. Nasr DM, Rabinstein AA. Neurologic complications of extra-
corporeal membrane oxygenation. J Clin Neurol 2015;11(4):
383e9.
35. Teasdale G, Maas A, Lecky F, Manley G, Stocchetti N, Murray G.
The Glasgow Coma Scale at 40 years: standing the test of time.
Lancet Neurol 2014;13(8):844e54.