Study design and population
We conducted a two-stage prospective cohort study in which we first collected urine samples from a cohort to identify the best biomarker for prediction of AKI and PICU mortality among 10 candidate urinary biomarkers, including novel potential candidates and previously described biomarkers (derivation study). A separate independent cohort was used to validate the performance value of the best biomarker identified from the derivation (validation study).
The overall study design is shown in Figure 1. Both two cohorts were conducted in the PICU of a single tertiary children’s hospital and included critically ill children aged between 1 month and 16 years. The derivation cohort was conducted from September to December 2016 and the validation cohort was preformed from December 2017 to January 2018 and September to December 2019. Exclusion criteria were as follows: known congenital abnormality of the kidney and a failure to collect urine samples before discharge from the PICU or death. Children had multiple PICU admissions within a single hospital stay, only their last admission was included in the analysis. The study was approved by the Institutional Review Board at the Children’s Hospital of Soochow University and performed in accordance with the Declaration of Helsinki. Written consent forms were obtained from their parents involved in this study.
Clinical data collection
In both derivation and validation cohort studies, the medical records of eligible patients were reviewed. Demographic characteristics, including age, body weight and gender, admission diagnosis, clinical status as defined by illness severity, medication and therapeutic interventions were recorded daily until PICU discharge or death. Sepsis, multiple organ dysfunction syndrome (MODS), shock and disseminated intravascular coagulation (DIC) that developed during the PICU stay were diagnosed by the treating physicians, according to the criteria described previously [11].
Assessment of illness severity
The score of the pediatric risk of mortality III (PRISM III), which was calculated on the day of PICU admission, was used to assess illness severity of critically ill children In both derivation and validation cohorts, according to methods described in the original study [12] and in accordance with our previous studies [11, 13].
Diagnosis of AKI
The diagnosis of AKI was based on the increase of serum creatinine (sCr) and/or the reduction of the urine output within the first 7 days after PICU admission, according to the criteria of Kidney Disease: Improving Global Outcome (KDIGO) [14]. When the baseline sCr measurement was unavailable, the sCr value at hospital or PICU admission was used. For children with increased sCr ≥1.2 mg/dL (106.1 μmol/L) at admission, the lowest sCr value within 2 weeks while in the PICU was considered as the baseline sCr, Sin accordance with our previous studies [11, 13]. The sCr level was measured daily during the first week after PICU admission, followed by routine measurement every 48-72 hours during the PICU stay. Severity of AKI was characterized by KDIGO staging, and KDIGO stages 2 and 3 were defined as severe AKI.
Clinical outcomes
The PICU mortality, as the primary outcome, was defined as all-cause mortality occurring during the PICU stay, including death resulting from withdrawal of therapy.
Urine sample collection
In the derivation cohort, urine samples were collected within the first 24 h after PICU admission and followed by every 48-72 h during the first 7 days of PICU stay. In the validation cohort, the urine samples were only collected within 24 h after PICU admission. All acquired urine samples were collected using a plastic bag and immediately frozen and stored at -80°C. The samples were centrifuged at 1,500 g at 4°C for 10 min and the supernatants were aliquoted for the measurement.
Measurement of urinary biomarkers
In the derivation cohort study, six biomarkers (KIM-1, FABP-1, TIMP-1, renin, IP-10 and TFF-3) in urine were measured using multiplex bead assays incorporated in human kidney injury panel 1 (HKI1MAG-99K, MILLIPLEX MAP kit, Millipore, Billerica, USA) run on the Luminex FlexMAP 3D instrument according to manufacturer’s instructions. The calibration curve was calculated using a five-parameter logistic fit and the concentration of urinary biomarkers was determined. The Human Lipocalin 2/NGAL (ab113326, Abcam, USA), TIMP-2 (DY971, R&D Systems, USA) and IGFBP7 (DY1334-05, R&D Systems, USA) ELISA kits were used for the measurement of NGAL, TIMP-2 and IGFBP7 in urine, respectively. In the ELISA assays, the samples were diluted 10-fold to 1000-fold in Reagent Diluent to ensure that the enzymatic reaction was maintained within the linear range. The coefficient of variation of intra-assay and inter-assay within and between ELISA tests were less than 10%. In the validation cohort, the concentration of uTIMP-1 was measured by means of ELISA (DTM100, R&D Systems, USA). The minimum detectable level of TIMP-1 was <0.08 ng/mL, and the coefficient of variation of intra-assay and inter-assay were less than 5% and 4.9%, respectively.
In both derivation and validation cohort studies, the concentration of urinary biomarkers was expressed in nanograms per milligram of urinary Cr (ng/mg uCr). The uCr level from the aliquoted sample was measured automatically on an automatic biochemical analyzer (Hitachi 7600, Tokyo, Japan) by using the sarcosine oxidase method. For urinary [TIMP-2]•[IGFBP7], the concentrations of TIMP-2 and IGFBP-7 in the urine were multiplied and then divided by 1000 to convert them into international general units, (ng/mL)2/1000, in accordance with our previous study [15] and study by others [16].
In addition, the initial and the peak values of urinary biomarkers were used for data analysis in the derivation study. For each child, the level of urinary biomarkers from the sample collected in the first 24 h after PICU admission was denoted as the initial value. The highest level among collected samples during the first 7 days after PICU admission was denoted the peak value.
Statistical analysis
SPSS statistics software Version 22 and GraphPad software Inc. Prism Version 8 were used for statistical analyses. Continuous data were presented as median and interquartile range (IQR), as they were not-normally distributed. Categorical data were presented as counts and percentage. Continuous variables among groups were compared using the Mann-Whitney U test or Kruskal-Wallis H test, and categorical variables using the chi-square test or Fisher’ s exact test, as appropriate. Univariate and stepwise multivariate linear regression analyses were performed to investigate factors potentially associated with the levels of uTIMP-1 in the validation cohort. Multicollinearity of variables was evaluated via tolerance and variance inflation factor (VIF), and tolerance ≤0.5 and the VIF value ≥2 indicate the presence of multicollinearity. In both derivation and validation cohorts, the multivariate logistic regression analyses were performed to investigate the associations between urinary biomarker and AKI and PICU mortality after adjustment for potential confounders. The Hosmer-Lemeshow goodness-of-fit test was used to evaluate the model fit. Subsequently, the predictive values of urinary biomarkers for AKI and PICU mortality were assessed by the receiver operating characteristic (ROC) curves. The area under the ROC curves (AUC) with the corresponding 95% confidence interval (CI) was recorded. In the validation cohort, the predictive accuracy was further assessed by sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) at the optimal cutoff values, which were determined by the maximum Youden index. For all analyses, a 2-tailed P<0.05 was considered significant.