Original Article. Defining reduced urine output in neonatal ICU: importance for mortality and acute kidney injury classification

Nephrol Dial Transplant (2013) 28: 901 909 doi: 10.1093/ndt/gfs604 Advance Access publication 24 January 2013 Original Article Defining reduced urine output in neonatal ICU: importance for mortality and acute kidney injury classification Candice Torres de Melo Bezerra 1, Lara Cavalcante Vaz Cunha 2 and Alexandre Braga Libório 1,2 Correspondence and offprint requests to: Alexandre Braga Libório; E-mail: alexandreliborio@yahoo.com.br ABSTRACT Background. Acute kidney injury (AKI) is an independent risk factor for mortality in adults and children. Generally, urine output (UO) < 1 ml/kg/h is accepted as oliguria in neonates, although it has not been systematically studied. prifle criteria suggest UO cut-offs similar to those of the adult population (0.3 and 0.5 ml/kg/h). The aim of the present study was to investigate UO in correlation with mortality in critically ill neonates and suggest changes in the prifle definition of reduced diuresis. Methods. A retrospective cohort study was performed in an eight-bed neonatal intensive care unit (NICU). UO was systematically measured by diaper weight each 3 h. Discriminatory capacity to predict mortality of UO was measured and patients were divided according to UO ranges: G1 > 1.5 ml/ kg/h; G2 1.0 1.5 ml/kg/h; G3 0.7 1.0 ml/kg/h and G4 < 0.7 ml/kg/h. These ranges were incorporated to prifle GFR criteria and its performance was evaluated. Results. Of 384 patients admitted at the NICU during the study period, 72 were excluded and overall mortality was 12.8%. UO showed good performance for mortality prediction (area under the curve 0.789, P < 0.001). There was a stepwise increase in hospital mortality according to UO groups after controlling for SNAPPE-II and diuretic use. Using these UO ranges with prifle improves its discriminatory capacity (area under the receiver operating characteristic curve 0.882 versus 0.693, P < 0.05). Conclusions. UO is a predictor of mortality in NICU. An association between a UO threshold < 1.5 ml/kg/h and 1 Public Health Post-graduate Program, University of Fortaleza, UNIFOR, Fortaleza, Ceará, Brazil and 2 Internal Medicine Department, Faculdade de Medicina, Universidade Federal do Ceará, Fortaleza, Ceará, Brazil Keywords: acute kidney injury, mortality risk, paediatric intensive care mortality was observed, which is higher than the previously published prifle thresholds. Adopting higher values of UO in prifle criteria can improve its capacity to detect AKI severity in neonates. INTRODUCTION Acute kidney injury (AKI) is a common event in critically ill patients and it is an independent risk factor for mortality in adults and children [1, 2]. Until recently, AKI definition was broad and largely different, making it difficul to compare study results. Recently, Akcan-Arikan et al. [3] adopted and validated RIFLE (Risk, Injury, Failure, Loss and End-stage renal disease) criteria for critically ill children prifle. In this study, the authors have adapted glomerular filtration rate (GFR) decline criteria from adults and maintained the same UO definition. In neonates, however, studies are still scarce and AKI actual incidence is unknown, as there is a lack of a classification system to define AKI in this population [4, 5]. In the aforementioned study that validated prifle, a newborn population was not included. Otherwise, the use of prifle criteria in a newborn population with the same definition used in older children can be cumbersome, especially regarding the UO classification. As stated by Jetton and Askenazi [6], UO < 0.5 ml/kg/h is a non-sensitive marker of AKI in the newborn population and concludes that these patients generally have non-oliguric AKI. In general, diuresis of <1 ml/kg/h is accepted as UO reduction in neonates [5], although no study has evaluated its association with clinical outcomes. The Author 2013. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. 901

Some studies have demonstrated the usefulness of prifle or similar criteria in predicting hospital mortality in the newborn population [7, 8]. During a review of these studies, we found only two studies that used UO criteria in this population [9]. First, it was a large prospective study of paediatric critically ill patients, including 154 (32.6%) newborns. AKI was considered present with GFR decline or UO < 0.5 ml/kg/ h. Mortality was higher in neonates with AKI, but no information was provided about how many neonates were classified as having developed AKI only by using UO criteria. The other is a study about AKI after cardiac surgery in term infants [10]. In this study, AKI occurred in 52% of the patients and oliguria was present in 40% of those developing AKI. The following were the aims of the current study, evaluating critically ill neonates: (i) to determine whether UO has any correlation with clinical outcomes; (ii) to determine the prevalence of reduced UO as defined by prifle, i.e. <0.5 ml/kg/h for 8 h, and its capacity to predict hospital mortality, and (iii) assess UO values that could be used to classify at-risk neonates, and propose values to be incorporated into prifle criteria. METHODS This is a retrospective cohort study that was performed in a reference eight-bed neonatal intensive care unit (NICU) in Fortaleza, Brazil. All newborns admitted from January 2010 to July 2011 with at least two serum creatinine (SCr) measurements with a maximum interval of 7 days were included. Medical records were examined daily, including UO, clinical and laboratory data. Patients with <24-h NICU stay and those with complex congenital cardiac disease were excluded as well as those receiving diuretics during the selected UO period (see below). The Institutional Ethical Committee approved this study under protocol number 73.674. Clinical and laboratory parameters At NICU admission, data on gestational age, birth weight and APGAR in the first and fifth minutes were collected. At NICU admission, body temperature, blood glucose, mean blood pressure, PaO 2 /FiO 2 ratio, ph, standard base excess, haematocrit, white blood cell and platelet count and potassium levels were recorded. During NICU stay, perinatal asphyxia, neonatal infection, seizures, distress respiratory syndrome, necrotizing enterocolitis, antibiotics and diuretic use was observed. SCr was requested according to the attending physician s decision. The highest urea level was recorded. All Scr collected after 48 h of life were used to estimate GFR and stage AKI as described below. The scores for neonatal acute physiology-perinatal extension II (SNAPPE-II) was calculated as previously described. Hyperkalaemia was defined as serum potassium > 5.5 meq/l and low platelet count as < 150.000/mm 3. Preterm neonates were considered when birth had occurred at <37 weeks of gestation and low birth weight (LBW) was <2.500 g. Septic shock was defined as neonatal infection and need for vasoactive drugs. Urine output measurement UO was systematically measured by diaper weight every 3 h even at nighttime. As variations in bladder retention could lead to erroneous oliguria classification using shorter time intervals, we measured total diuresis in a 24-h interval and calculated UO in ml/kg/h. In the immediate neonatal period, newborns can be oliguric during the first 24 h of life; thus, UO reduction was not considered a criterion for the first 24 h of life. The lowest UO in a 24-h period was selected. Patients receiving diuretics during the lowest UO 24-h period or the previous 24 h were also excluded. AKI definition AKI was defined by the original prifle as a decrease in GFR (based on the Schwartz formula [11]) 25% from baseline, or a reduction in UO (<0.5 ml/kg/h during at least 9 h). The Pediatric-modified Risk, Injury, Failure, Loss, End-stage kidney disease ( prifle) criteria proposed by Akcan-Arikan et al. [3] was modified to adjust our 3-h UO measurement. Patients were classified according to AKI severity, using the maximum reduction comparing two GFRs within a maximum interval of 7 days. The patients were assigned to the appropriate prifle strata if they met either the UO or SCr criterion or both: risk (R) was defined as an estimated GFR decrease of 25% and/or UO < 0.5 ml/kg/h in 6 h; injury (I) was defined as an egfr decrease of 50% and/or UO < 0.5 ml/kg/h in 15 h; failure (F) was defined as an egfr decrease of 75% and/or UO < 0.3 ml/kg/h in 24 h or anuria for 12 h. Loss (L), defined as renal failure > 4 weeks and end-stage renal disease (E), defined as renal failure > 3 months, were not necessary in our study. Only SCr collected after the first 48 h of life were considered to prevent maternal SCr influence. After that, patients were reclassified according to UO cut-off points as described below. Statistical analysis Descriptive statistics are expressed as mean ± SD. All variables were tested for normal distribution using the Kolmogorov Smirnov test. Primary analysis compared patients according to mortality and UO classification. The unpaired Student s t-test was applied to compare continuous variables and normal-distribution data when appropriate. Categorical data were tested using the χ 2 test. Association between UO groups of patients and in-hospital mortality was performed after adjusting for SNAPPE-II score and diuretic use and association measures were calculated [adjusted odds ratio (OR)], with 95% confidence interval. For the investigation of independent risk factors for in-hospital mortality and reduced UO, a stepwise backward elimination multivariate analysis was performed, and it included the factors that had a significance level < 20% in the univariate analysis. Kaplan Meier curves, censored for death and NICU discharge, obtained with the log-rank test were plotted to demonstrate the differences in patient survival between patients with or without reduced UO. Receiver operator characteristic curves (area under the curve, AUC) were applied to evaluate accuracy of the lowest 24-h UO during NICU stay in ml/kg/h 902 C.B. Torres de Melo et al.

and prifle classifications to predict mortality. A good discriminatory capacity was defined as an area under the curve > 0.7. Regarding UO, cut-off points were chosen according to the highest Youden index, which was calculated as [1 (1 sensitivity) + (1 specificity)]. P values < 0.05 were considered statistically significant. The statistical analysis was performed using SPSS 19.0 for Windows. Data are presented as mean ± SD. RESULTS Population characteristics Of the 384 patients admitted in neonatal ICU during the study period, 54 were excluded due to lack of data, 10 due to complex cardiac disease, 7 stayed at the NICU for <24 h and 11 because they received diuretics during their lowest UO period or during the previous 24 h. A total of 312 patients remained in the final analysis. Of these, 150 (48.1%) were preterm and 151 (48.3%) were classified as having LBW. Most newborns (n = 278, 89.1%) had a diagnosis of neonatal infection, 101 (32.4%) had perinatal asphyxia and 139 (44.5%) had neonatal respiratory distress syndrome. Diuretics were administered to 27 (8.6%) patients. Mean length of hospital stay was 21.4 ± 17.0 days and overall mortality was 12.8%. Complete data are shown in Table 1. Factors associated with in-hospital mortality A complete comparison between survivors and non-survivors is shown in Table 1. In addition to other variables, UO was associated with mortality in the univariate analysis. At the multivariate analysis, UO remained independently Table 1. Characteristics of patients according to hospital survival General population (n = 312) Non-survivors (n = 40) Survivors (n = 272) Pre-term neonates, n (%) 150 (48.1%) 21 (52.5%) 129 (47.4%) 0.649 5 min-apgar < 7 74 (23.7%) 16 (40.0%) 58 (21.3%) 0.015 Low birth weight, n (%) 151 (48.3%) 20 (52.6%) 131 (49.6%) 0.729 Seizures, n (%) 55 (17.6%) 9 (30.0%) 46 (20.1%) 0.212 Neonatal infection, n (%) 278 (89.1%) 40 (100%) 238 (87.5%) 0.018 SNAPPE-II 14.9 ± 8.9 30.5 ± 21.7 12.6 ± 8.6 <0.0001 Umbilical catheter, n (%) 218 (69.9%) 31 (77.5%) 187 (68.8%) 0.260 Necrotizing enterocolitis, n (%) 07 (2.2%) 2 (5.0%) 5 (1.8%) 0.207 Septic shock, n (%) 65 (20.8%) 33 (86.8%) 32 (12.1%) <0.0001 Perinatal asphyxia, n (%) 101 (30.4%) 20 (50.0%) 81 (29.8%) 0.011 Indomethacin use, n (%) 07 (2.2%) 1 (2.6%) 6 (2.2%) 0.891 Amphotericin, n (%) 09 (2.9%) 1 (2.5%) 8 (2.9%) 0.917 Amikacin, n (%) 249 (75.0%) 37 (92.5%) 212 (77.9%) 0.032 Parenteral nutrition time (days) 3.7 ± 2.4 4.6 ± 3.7 3.6 ± 2.6 0.264 Haematocrit (%) 40.3 ± 23.2 40.8 ± 18.8 39.9 ± 18.2 0.439 White blood cells ( 10 3 cells/ 14.1 ± 6.7 16.3 ± 7.2 13.8 ± 6.9 0.295 mm 3 ) Low platelets count, n (%) 75 (24.0%) 16 (40.0%) 59 (21.7%) 0.006 Maximum SCr (mg/dl) 0.92 ± 0.61 2.28 ± 1.26 0.73 ± 0.42 <0.0001 SBE (meq/l) 5.4 ± 4.8 8.6 ± 4.3 4.90 ± 4.3 <0.0001 Hyperkalaemia, n (%) 39 (12.5%) 17 (42.5%) 22 (8.1%) <0.0001 Reduced UO, n (%) 64 (20.5%) 23 (57.5%) 41 (15.1%) <0.0001 Mechanical ventilation time (days) 2.9 ± 1.9 7.2 ± 5.6 2,36 ± 2.0 <0.0001 Reduced UO, UO < 1.5 ml/kg/h. RDS, respiratory distress syndrome; SNAPPE-II, scores for neonatal acute physiology-perinatal extension II; SCr, serum creatinine; SBE, standard base excess. P 903 Defi ning reduced urine output in neonatal ICU

associated with in-hospital mortality, in addition to septic shock, perinatal asphyxia and low platelet count (Table 2). Urine output as a predictor of in-hospital mortality UO showed good prediction performance for in-hospital mortality [AUC 0.789, 95% confidence interval (CI) 0.702 0.877, P < 0.001; Figure 1]. Patients were divided according to UO values using Youden index. The groups were divided according to the following UO ranges: (1) Group 1: >1.5 ml/kg/h; (2) Group 2: 1.0 1.5 ml/kg/h; (3) Group 3: 0.7 1.0 ml/kg/h; (4) Group 4: <0.7 ml/kg/h. We also classified patients according to diuresis used in the original prifle criteria. Only 11 (23.5%) patients had reduced UO (3 had UO between 0.3 and 0.5 ml/kg/h for more than 15 h; 5 <0.3 ml/kg/h for 24 h and 3 were anuric for at least 12 h). Mortality according to the original prifle UO was elevated, but there was no difference between the two UO ranges (100 and 75%, for original prifle UO I and F, respectively). Urine output in LBW neonates When analysing only neonates with LBW (n = 151), patients with UO between 1.5 and 2.0 ml/kg/h (n = 26) had higher in-hospital mortality than neonates with UO > 2.0 ml/kg/h (6.4 versus 19.2%, P = 0.039). After adjusting for SNAPPE-II score and diuretic use, the difference remained (OR 4.132; 95% CI: 1.558 10.957, P = 0.004). Overall, 64 (20.5%) patients had a reduced UO, defined as UO < 1.5 ml/kg/h. In 34 of 64 (53.1%) cases, reduced UO occurred between 24 and 96 h of life. Patients with reduced UO had higher SNAPPE-II score at NICU admission and more perinatal asphyxia. Furthermore, these patients had more metabolic complications (acidosis and hyperkalaemia). The duration of mechanical ventilation was greater in patients with reduced UO (Table 3). Diuretics were administered to 8 (12.5%) patients with reduced UO and 19 (7.7%) with normal UO (P = 0.328). At the multivariate analysis, risk factors associated with reduced UO were septic shock, perinatal asphyxia and prolonged hospital stay (Table 4). LBW was a protective factor against reduced UO (see Discussion). Patients had stepwise increasing in-hospital mortality according to UO groups (see Figure 2). Survival probabilities are shown in Figure 3. Log-rank test of Kaplan Meier curves showed a significant difference between groups (log rank P < 0.001). After controlling for SNAPPE-II and diuretic use, the adjusted OR for mortality in each UO group confirmed this association (Table 5). Table 2. Predictors of in-hospital mortality at multivariate analysis Association between UO ranges and prifle GFR criteria Using the original prifle criteria, 54 (16.3%) patients had AKI. Of these 54 patients, 11 were diagnosed by both UO and GFR criteria; 43 only by GFR and no patients were included only by UO criteria. All patients with reduced UO by the original prifle criteria were also classified by prifle GFR criteria. In-hospital mortality in AKI newborns according to the original prifle stage is demonstrated in Figure 4. Although there was an increase in the mortality rate according to the original prifle criteria, there was no significant difference when comparing classes R and I (P = 0.768). Its capacity to predict in-hospital mortality was moderate at best (AUC 0.689; 95% CI: 0.587 0.791, P < 0.001; Figure 5). When adding new UO classifications to prifle criteria proposed prifle see Table 6, AKI incidence increased to 24.3% (n = 76). Of these, 42 were included by both UO < 1.5 ml/kg/h and reduced GFR; 22 only by UO < 1.5 OR 95% CI P 2.720 0.897 8.247 0.077 Perinatal asphyxia Septic shock 39.232 12.076 127.449 <0.0001 Reduced UO 3.773 1.348 10.560 0.011 (<1.5 ml/kg/ h) Low platelets 2.395 0.850 6.744 0.098 count(<150 000 mm 3 ) Constant 0.003 <0.0001 Out-of-model: 5 min-apgar < 7, neonatal infection, Amikacin use, base excess and hyperkalaemia. UO, urine output. FIGURE 1: Receiver operating characteristic (ROC) curve for UO. ROC curve is shown for UO to predict hospital mortality. 904 C.B. Torres de Melo et al.

Table 3. Characteristics of patients according to UO Reduced UO < 1.5 ml/kg/h (n = 64) Normal UO 1.5 ml/kg/h (n = 248) Pre-term neonates, n (%) 13 (20.3%) 137 (55.2%) <0.0001 5 min-apgar < 7 20 (31.2%) 47 (18.9%) 0.049 Low birth weight, n (%) 16 (25.0%) 135 (54.4%) <0.0001 Seizures, n (%) 21 (32.8%) 34 (13.7%) 0.001 Neonatal infection, n (%) 56 (87.5%) 222 (89.5%) 0.644 SNAPPE-II 20.6 ± 12.5 13.4 ± 7.7 0.002 Umbilical catheter, n (%) 50 (78.1%) 168 (67.7%) 0.107 Necrotizing enterocolitis, n (%) 1 (1.6%) 6 (2.4%) 0.680 Septic shock, n (%) 28 (43.7%) 37 (14.9%) <0.0001 Perinatal asphyxia, n (%) 36 (56.2%) 65 (26.2%) <0.0001 Indomethacin use, n (%) 1 (1.6%) 6 (2.4%) 0.688 Amphotericin, n(%) 2 (3.1%) 7 (2.8%) 0.872 Amikacin, n (%) 54 (84.4%) 195 (78.6%) 0.307 Diuretic use, n (%) 08 (12.5%) 19 (7.7%) 0.328 Parenteral nutrition time (days) 3.6 ± 2.6 3.8 ± 2.5 0.767 Haematocrit (%) 44.5 ± 8.3 43.6 ± 7.8 0.473 White blood cells ( 10 3 cells/ 16.3 ± 5.3 13.5 ± 4.9 0.170 mm 3 ) Low platelets count, n (%) 18 (28.1%) 57 (23.0%) 0.295 Maximum SCr (mg/dl) 1.50 ± 0.82 0.78 ± 0.41 <0.0001 SBE (meq/l) 7.50 ± 5.01 4.91 ± 4.60 <0.0001 Hyperkalaemia, n (%) 15 (23.4%) 24 (9.7) 0.003 Mechanical ventilation time 4.4 ± 2.7 2.6 ± 1.6 0.007 (days) Mortality 23 (35.9%) 17 (6.9%) <0.0001 RDS, respiratory distress syndrome; SNAPPE-II, scores for neonatal acute physiology-perinatal extension II; SBE, standard base excess; SCr, serum creatinine. P ml/kg/h and 12 only by reduced GFR. Most patients (59.2%) developed AKI in the first week post-partum. In-hospital mortality showed a stepwise increment with the proposed prifle criteria (Figure 4). The main difference between the two classifications was the better discriminatory capacity of the proposed prifle to differentiate mortality between prifle classes R and I. Moreover, the proposed prifle AUC (0.885; 95% CI: 0.819 0.957, P < 0.001) was greater when compared with the original prifle P for comparison between AUC < 0.05 (Figure 5). DISCUSSION To the best of our knowledge, this is the first study to evaluate prifle criteria in a general neonatal ICU. Previous studies that evaluated these criteria in newborns were performed only recently in patients submitted to complex cardiac surgery [8, 10], in very LBW neonates [7] and in patients with congenital diaphragmatic hernia requiring extracorporeal life support [12]. In the present study, we demonstrate that the UO criterion used in adults and older children is extremely restrictive when used in a neonatal population and only a few patients had an UO < 0.5 ml/kg/h during a minimum period of 9 h. Moreover, even with a UO > 0.5 ml/kg/h, reduced UO was associated with stepwise increased in-hospital mortality. Using SCr in neonatal AKI diagnosis is cumbersome. Some reasons include: the mother s creatinine influence, daily changes in normal GFR due to continuing nephrogenesis after birth and a wide distribution of normal GFR range [5, 6]. In spite of all these drawbacks, an increment in SCr has 905 Defi ning reduced urine output in neonatal ICU

been associated with worsening mortality in neonatal ICU. A retrospective matched case control study of premature infants showed that an increase in SCr of 1.0 mg/dl doubled the odds of death [13]. However, using the Acute Kidney Injury Network (AKIN) definition, Koralkar et al. [7] demonstrated that AKI was associated with mortality, but this association was not maintained after adjusting for confounding variables. New AKI biomarkers are emerging as possible tools to diagnose AKI earlier, but they are still expensive and not completely standardized to be used in daily practice [14, 15]. Diuresis can be a suitable alternative to identify patients at high risk in the neonatal ICU. Table 4. Predictors of reduced UO (<1.5 ml/ kg/h) at multivariate analysis Perinatal asphyxia OR 95% CI P 2.977 1.289 6.874 0.011 Septic shock 4.466 2.059 9.687 <0.0001 5 min- 1.268 1.011 1.590 0.040 APGAR < 7 Low birth weight 0.392 0.191 0.806 0.012 (<2.500 g) Constant 20.607 0.001 Out-of-model: low platelet count, mechanical ventilation time, base excess and hyperkalaemia, gestational age, seizures, SNAPPE-II. UO was an independent predictor of mortality in the multivariate analysis and there was a stepwise increase in mortality at the range of 1.5 ml/kg/h. It is believed that neonatal AKI is generally non-oliguric [16, 17] and normal diuresis is believed to be > 1 ml/kg/h [5], although no previous study has systematically evaluated the importance of UO in the neonatal population. In critically ill adult patients, oliguria (UO < 0.5 ml/kg/h) without any change in SCr was associated with an increase in mortality [18]. In newborns, our data support the same conclusion, but with higher cut-offs than those used in adults. When oliguria was considered, with an UO < 1.5 ml/kg/h, patients were more severely ill (higher SNAPPE-II score), had more traditional risk factors for AKI ( perinatal asphyxia) and more metabolic complications (metabolic acidosis and hyperkalaemia). These factors suggest that patients with reduced UO essentially had a renal dysfunction and not only a physiological variation. It is noteworthy that in all patients with a fatal outcome, the minimum UO rhythm was recorded >24 h before death, making it a useful marker of severity. The total body water content is greater in newborns than in adult patients. Especially in preterm infants, the total body water can be as high as 80% of body weight [19]. This difference in water content, in addition to immature tubular development, can explain why UO in newborns is normally greater than in other populations. The general belief that neonatal AKI is usually non-oliguric can be a misconception due to lack of knowledge about normal UO in critical newborns. In our series, there were more neonates with LBW in the group with normal UO (Table 2). We supposed that LBW infants had an even higher normal UO than other neonates, which was confirmed by diuresis as high as 1.5 2.0 ml/kg/h being associated with a fatal outcome. FIGURE 2: Hospital mortality according to UO ranges. P < 0.05 for all ranges, when comparing with subsequent range. C.B. Torres de Melo et al. 906

In the present study, we used prifle instead of the adapted AKIN criteria, as the first one maintains the UO criteria. Because our standard care includes a UO measurement each 3 h, we have adapted prifle UO criteria: instead of using 8- and 16-h periods, we chose less restrictive criteria 6- and 15-h for prifle R and I, respectively. It is unlikely that our adaptation would have caused any significant bias in our results, as no patient was classified as original prifle R and only three as prifle I category. AKI incidence by the prifle GFR criteria was 16.3%. Previous studies [20, 21] reported an estimated incidence between 6 and 24% and studies using AKIN/pRIFLE classification reported incidence from 18 to 64%, in special populations (very LBW and cardiac surgery neonates, respectively) [8, 17]. The incidence was slightly reduced probably because the study was performed in a general neonatal ICU and patients with complex congenital cardiac diseases were excluded. After cardiac surgery, Blinder et al. [10] showed that 40% of patients reached maximum AKI stage by UO criteria. In our data, only 20.4% of patients diagnosed as AKI by the original prifle criteria were classified by UO. This difference can be explained by the severity of AKI lesion after cardiac surgery (68% of patients with AKI needed peritoneal FIGURE 3: Cumulative patient survival between patients with normal and reduced UO. Table 5. Adjusted odds-ratio for mortality according to UO range Reference UO 1.5 ml/kg/ h 1.0 UO < 1.5 ml/kg/h 0.7 UO < 1.0 ml/kg/h Adjusted OR 95% CI P 2.874 1.063 8.519 7.045 3.045 21.862 UO < 0.7 ml/kg/h 23.353 6.046 65.493 Adjusted OR for SANPPE-II and diuretics use. UO, urine output. 0.047 0.001 <0.0001 FIGURE 4: Hospital mortality according to original and proposed prifle. P < 0.05 for all ranges, when comparing with subsequent range in both original and proposed prifle criteria, except between original prifle classes R and I (P = 0.768). 907 Defi ning reduced urine output in neonatal ICU

dialysis versus only one patient in our sample). Interestingly, in this same study, peritoneal dialysis was performed in 37% of patients without AKI diagnosis, mainly for fluid and electrolyte management. It is probable that applying a higher UO cut-off would lead to a better classification of these patients in relation to AKI diagnosis and severity. When using a more liberal definition of oliguria (<1.5 ml/ kg/h) in prifle criteria, AKI was diagnosed in 21.8% of patients. This increment in incidence was associated with a better performance in predicting in-hospital mortality, demonstrated by AUC. Many studies in children and newborns have failed to demonstrate stepwise increased mortality across all AKI stages [3, 7, 8, 22, 23]. We were also unable to show increased mortality between prifle R and I stages when using original prifle, but when the proposed prifle was applied, there was a continuous increment in mortality rate. Our study has several limitations. First, it was retrospective and performed in a single center, and multicentre studies must be performed to define oliguria in the neonatal population. Second, we did not have access to daily SCr levels and FIGURE 5: ROC curve for original and proposed prifle. ROC curve is shown for prifle to predict hospital mortality. some patients did not have any SCr measurements available and were not included. We only had access to laboratory results of those infants whose clinician had ordered the test. Unlike critically ill children and adults who have SCr measured daily per routine care, there is a reluctance to measure SCr frequently in neonates because of concerns regarding blood loss with multiple blood samplings [6]. In the present study, the mean of available SCr values was one measurement for each 4.6 days of ICU stay. Certainly, more frequent SCr measurements would increase AKI diagnosis and possibly allow a better performance of the prifle GFR criteria. Finally, most patients were not using indwelling urinary catheters and UO measurement was performed through diaper weight every 3 h. Considering neonatal ICU routine practice and the risks associated with prolonged urinary catheterization, it is difficult to avoid this technical difficulty and access hourly UO. Consequently, we chose a 24-h period to characterize reduced UO. It would be troublesome to evaluate UO in a smaller time period because temporary bladder retention would lead to a misclassification of oliguria. In conclusion, we showed that UO is a predictor of mortality in neonatal ICU. Moreover, using diuresis values from adults to classify AKI severity in newborns appears inadequate. Stepwise increase in mortality begins with UO values as great as 1.5 ml/kg/h. Even higher values can predict mortality in LBW neonates. It is not our intention to modify criteria based on a retrospective study, but assuming more liberal values of diuresis in prifle definition can improve its performance to detect AKI severity in neonatal ICU. CONFLICT OF INTEREST STATEMENT None declared. REFERENCES 1. Plötz FB, Bouma AB, van Wijk JA et al. Pediatric acute kidney injury in the ICU: an independent evaluation of prifle criteria. Intensive Care Med 2008; 34: 1713 1717 Table 6. Comparison between original and proposed prifle criteria Estimated GFR criterion Original urine output criterion Proposed urine output criterion Risk egfr decrease by 25% <0.5 ml/kg/h for 8 h <1.5 ml/kg/h for 24 h Injury egfr decrease by 50% <0.5 ml/kg/h for 16 h <1.0 ml/kg/h for 24k Failure egfr decrease by 75% <0.3 ml/kg/h for 24 h or anuric for 12 h Loss Persistent failure > 4 weeks End stage End-stage renal disease (persistentfailure > 3 months) <0.7 ml/kg/h for 24 h or anuric for 12 h C.B. Torres de Melo et al. 908

2. Ali T, Khan I, Simpson W et al. Incidence and outcomes in acute kidney injury: a comprehensive population-based study. J Am Soc Nephrol 2007; 18: 1292 1298 3. Akcan-Arikan A, Zappitelli M, Loftis LL et al. Modified RIFLE criteria in critically ill children with acute kidney injury. Kidney Int 2007; 71: 1028 1035 4. Durkan AM, Alexander RT. Acute kidney injury post neonatal asphyxia. J Pediatr 2011; 158: e29 e33 5. Askenazi DJ, Ambalavanan N, Goldstein SL. Acute kidney injury in critically ill newborns: what do we know? What do we need to learn? Pediatr Nephrol 2009; 24: 265 274 6. Jetton JG, Askenazi DJ. Update on acute kidney injury in the neonate. Curr Opin Pediatr 2012; 24: 191 196 7. Koralkar R, Ambalavanan N, Levitan EB et al. Acute kidney injury reduces survival in very low birth weight infants. Pediatr Res 2011; 69: 354 358 8. Morgan CJ, Zappitelli M, Robertson CM et al. Risk factors for and outcomes of acute kidney injury in neonates undergoing complex cardiac surgery. J Pediatr 2013; 162: 120 127 9. Duzova A, Bakkaloglu A, Kalyoncu M et al. Etiology and outcome of acute kidney injury in children. Pediatr Nephrol 2010; 25: 1453 1461 10. Blinder JJ, Goldstein SL, Lee VV et al. Congenital heart surgery in infants: effects of acute kidney injury on outcomes. J Thorac Cardiovasc Surg 2012; 143: 368 374 11. Schwartz GJ, Brion LP, Spitzer A. The use of plasma creatinine concentration for estimating glomerular filtration rate in infants, children, and adolescents. Pediatr Clin North Am 1987; 34: 571 590 12. Gadepalli SK, Selewski DT, Drongowski RA et al. Acute kidney injury in congenital diaphragmatic hernia requiring extracorporeal life support: an insidious problem. J Pediatr Surg 2011; 46: 630 635 13. Askenazi DJ, Griffin R, McGwin G et al. Acute kidney injury is independently associated with mortality in very low birthweight infants: a matched case-control analysis. Pediatr Nephrol 2009; 24: 991 997 14. Askenazi DJ, Koralkar R, Hundley HE et al. Urine biomarkers predict acute kidney injury in newborns. J Pediatr 2012; 161: 270 275.e1 15. Argyri I, Xanthos T, Varsami M et al. The role of novel biomarkers in early diagnosis and prognosis of acute kidney injury in newborns. Am J Perinatol 2012 16. Karlowicz MG, Adelman RD. Nonoliguric and oliguric acute renal failure in asphyxiated term neonates. Pediatr Nephrol 1995; 9: 718 722 17. Viswanathan S, Manyam B, Azhibekov T et al. Risk factors associated with acute kidney injury in extremely low birth weight (ELBW) infants. Pediatr Nephrol 2012; 27: 303 311 18. Macedo E, Malhotra R, Bouchard J et al. Oliguria is an early predictor of higher mortality in critically ill patients. Kidney Int 2011; 80: 760 767 19. Hartnoll G, Bétrémieux P, Modi N. Body water content of extremely preterm infants at birth. Arch Dis Child Fetal Neonatal Ed 2000; 83: F56 F59 20. Andreoli SP. Acute renal failure in the newborn. Semin Perinatol 2004; 28: 112 123 21. Vachvanichsanong P, McNeil E, Dissaneevate S et al. Neonatal acute kidney injury in a tertiary center in a developing country. Nephrol Dial Transplant 2012; 27: 973 977 22. Palmieri T, Lavrentieva A, Greenhalgh D. An assessment of acute kidney injury with modified RIFLE criteria in pediatric patients with severe burns. Intensive Care Med 2009; 35: 2125 2129 23. Prodhan P, McCage LS, Stroud MH et al. Acute kidney injury is associated with increased in-hospital mortality in mechanically ventilated children with trauma. J Trauma Acute Care Surg 2012; 73: 832 837 Received for publication: 1.11.2012; Accepted in revised form: 13.12.2012 909 Defi ning reduced urine output in neonatal ICU