Comparative diagnostic value of NGAL and multibiomarker models in pediatric sepsis-associated acute kidney injury in Indonesia: a preliminary study

Abstract

Purpose

Sepsis is a leading cause of pediatric critical illness and a major contributor to acute kidney injury (AKI). Sepsis-associated AKI (SA-AKI) worsens outcomes, yet current diagnostic criteria based on serum creatinine and urine output lack sensitivity for early detection. This study evaluated the diagnostic performance of neutrophil gelatinase-associated lipocalin (NGAL), angiopoietin-2, and syndecan-1 for early identification of SA-AKI in children with sepsis.

Methods

A prospective nested case-control study was conducted in children aged 1–18 years with sepsis. Participants were classified into SA-AKI and non-SA-AKI groups using pediatric Risk, Injury, Failure, Loss, End-stage Renal Disease (pRIFLE) criteria based on estimated creatinine clearance index (eCCI). Blood and urine samples were collected at baseline, 12, 24, and 48 hours to measure NGAL, angiopoietin-2, and syndecan-1 levels by enzyme-linked immunosorbent assay. Diagnostic accuracy was assessed using receiver operating characteristic (ROC) analysis.

Results

Of 23 children enrolled, eight (34.8%) developed SA-AKI. Baseline creatinine, eCCI, urinary NGAL, and mean corpuscular volume (MCV) differed significantly between groups, whereas angiopoietin-2 and syndecan-1 did not. ROC analysis showed that 24-hour eCCI (area under the curve [AUC]=1.000; sensitivity 100%; specificity 100%) and 12- to 24-hour urinary NGAL (AUC=0.900; sensitivity 87.5%; specificity 86.7%) were the strongest predictors. MCV had moderate accuracy (AUC=0.767). Urinary NGAL correlated negatively with eCCI (r=−0.67).

Conclusions

Urinary NGAL measured within 12–24 hours is a reliable early biomarker for SA-AKI in pediatric sepsis, complementing eCCI and pRIFLE criteria to enhance early diagnosis and management.

Introduction

Sepsis is a leading cause of critical illness in children, with the SPROUT study reporting severe sepsis in 8.2% of pediatric intensive care unit (PICU) patients across 26 countries [1,2]. In addition, sepsis is the predominant cause of acute kidney injury (AKI) in critically ill children, accounting for 45% to 47% of pediatric AKI cases. Up to 42% of children with severe sepsis develop AKI, of which approximately 20% are severe [2-4]. Mortality is significantly higher in patients with sepsis and AKI, particularly in sepsis-associated AKI (SA-AKI), which has a worse prognosis than other AKI causes [5].

The pathophysiology of SA-AKI is multifactorial and involves tubular injury, systemic inflammation, apoptosis, and microvascular dysfunction. Unlike traditional ischemic models, SA-AKI can develop despite preserved renal blood flow, with microcirculatory disturbances, glycocalyx injury, and immune-mediated mechanisms contributing to renal dysfunction [6]. The Risk, Injury, Failure, Loss, End-stage Renal Disease (RIFLE), pediatric RIFLE (pRIFLE), AKI Network (AKIN), Kidney Disease: Improving Global Outcomes (KDIGO), and estimated creatinine clearance index (eCCI) classifications stratify AKI severity based on serum creatinine (sCr) and urine output [7-9]. However, urine output may decrease before creatinine level increases, and sCr is limited by its nonlinear relationship with glomerular filtration rate (GFR) and dependence on age, sex, and muscle mass. These limitations are more pronounced in children, highlighting the need for more sensitive biomarkers for the early detection of AKI [9-13]. Serum and urine biomarkers enable earlier detection of AKI in children with sepsis, as an ideal marker increases rapidly after injury, correlates with severity, and decreases with recovery. Considering the limited number of conventional markers, there has been a growing interest in novel biomarkers that reflect the pathophysiology of SA-AKI.

Neutrophil gelatinase-associated lipocalin (NGAL), a 25-kDa protein involved in innate immunity and renal protection following ischemic damage, increases rapidly within 3 hours of ischemic or toxic kidney damage, and reaches its peak value at 6 to 12 hours. Thus, NGAL is a sensitive biomarker for early AKI detection [14]. Angiopoietin-2 disrupts vascular stability by promoting inflammation, leakage, and impaired renal blood flow [15]. Syndecan-1, a glycocalyx component, serves as a marker of endothelial injury [16]. Together, NGAL captures tubular damage, whereas angiopoietin-2 and syndecan-1 indicate endothelial and microvascular dysfunction, reflecting the multifactorial pathophysiology of SA-AKI. These biomarkers are associated with renal dysfunction, adverse outcomes, and mortality. However, their specificity, optimal thresholds, and clinical validation remain limited, necessitating further studies to establish standardized cutoffs and prognostic value.

This study aimed to evaluate the utility of NGAL, angiopoietin-2, and syndecan-1 as early biomarkers of SA-AKI in children. Specifically, this study sought to determine their diagnostic accuracy, prognostic value, and complementary roles in capturing tubular and endothelial injury pathways, aiming to improve early detection and risk stratification in pediatric patients.

Methods

Ethical approval and informed consent

The institutional review board of the authors reviewed and approved the study protocol. Informed consent was obtained from the patient’s parents or legal representatives before the start of the study. The reporting of this study followed the Standards for Reporting Diagnostic Accuracy Studies (STARD) [17].

Inclusion and exclusion criteria

This study included children aged 1 to 18 years who were admitted to the emergency department, high care unit, or PICU of the authors’ institution between May and August 2017. Eligibility criteria were as follows: (1) a diagnosis of sepsis, (2) good nutritional status, and (3) written informed consent from legal guardians. Exclusion criteria were as follows: (1) a history of kidney disorders, (2) pre-existing peritoneal or maintenance hemodialysis, (3) urinary tract infection or obstruction, and (4) a history of renal transplantation. Sepsis was screened using the quick Sequential Organ Failure Assessment for children aged >12 years [18]. The quick Pediatric Logistic Organ Dysfunction Score-2 score was used in younger children [19].

Study design and patient recruitment

This study employed a prospective observational cohort design, incorporating a nested case-control component for a diagnostic study. Pediatric patients with sepsis admitted to the PICU were consecutively recruited and subsequently classified into two groups: SA-AKI (cases) and sepsis without AKI (controls). AKI was classified according to the pRIFLE criteria using eCCI calculated with the Schwartz equation (1976) [20] as follows:

$$eCCI \left(mL/min/1.73 m^2\right)= \frac{k \times height \left(cm\right)}{Serum creatinine \left(mg/dL\right)}$$

where eCCI is the estimated creatinine clearance index and k is a standardized constant (k=0.55 for children and adolescent girls aged >13 years and k=0.70 for adolescent boys aged >13 years). Patients were stratified into Risk (≥25% decrease in eCCI from baseline), Injury (≥50% decrease), and Failure (≥75% decrease or eCCI <35 mL/min/1.73 m²). Considering that pre-illness sCr values were unavailable, baseline creatinine was estimated by back-calculation from an assumed normal eCCI of 120 mL/min/1.73 m². Accordingly, a value below 90 mL/min/1.73 m², representing a 25% decrease from baseline, was classified as Risk according to the pRIFLE criteria. Patients who met the Risk or Injury criteria were classified as having SA-AKI. Board-certified pediatricians conducted the recruitment and assessment processes. All patients received fluid resuscitation therapy in accordance with the American College of Critical Care Medicine–Pediatric Advanced Life Support 2017 guidelines.

Expected number of participants

The expected number of participants was determined based on the estimated sample size required to compare the continuous outcome:

$$n= \frac{{\left(Z_{1-\alpha /2} + Z_{1-\beta }\right)}^2 \times \left(1+1/k\right) }{d^2},$$

where n is the expected number of participants; Z_1−α/2 is the z-score corresponding to the defined α (probability of a false positive), which was 1.96 for α=5% in the present study; Z_1−β is the z-score corresponding to the defined β (probability of a false negative), which was 0.84 for β=80%; d is the standardized effect size or the minimum clinically important difference between groups; and k is the proportion of participants in the case group.

The standardized effect size was calculated using the following equation:

$$d= \frac{\left(f-1\right)}{CV},$$

where f is the fold change and CV is the coefficient of variation. A 2-fold change was considered the minimum clinically meaningful difference for the biomarkers [16]. The CV was calculated based on the method reported by Abdelrazic et al. [21], who measured urinary NGAL levels within 24 hours in children with sepsis admitted to the PICU. In their study, the mean urinary NGAL value was 1017.9 with a standard deviation of 574 (n=54), yielding a CV of 0.564. Using these values in the equation resulted in an effect size of d=1.77. Epidemiological studies have reported that approximately 40%–45% of children with sepsis develop AKI [2-4]. For the sample size calculation, the proportion of the case group (k) was determined based on the upper bound of this incidence, set at 45% (k = 0.45). Applying these values to the sample size calculation yielded an estimated sample size of approximately eight patients in the SA-AKI group. Assuming that SA-AKI accounted for 45% of sepsis cases, approximately 10 patients should be recruited in the non-SA-AKI group.

Laboratory measurements

Specimen collection and processing

Venous blood (3–5 mL) sample was collected at baseline (within 6 hours of sepsis diagnosis), and sample collection was repeated at 12, 24, and 48 hours after enrolment. Samples were collected in serum separator tubes for syndecan-1 and in EDTA tubes for angiopoietin-2. Blood samples were allowed to clot (serum) for 30 minutes at room temperature, followed by centrifugation at 1,500 ×g for 10 minutes at 4 °C. Plasma and serum were aliquoted into polypropylene cryovials in 200–500 µL portions to minimize freeze–thaw cycles. A midstream or catheterized urine sample (5–10 mL) was collected at the same time points for NGAL measurement. The sample was placed on ice, centrifuged at 2,000 ×g for 10 minutes at 4 °C to remove debris, and then aliquoted. All samples were stored at −80 °C until batch analysis.

Biomarker assays (ELISA)

Plasma angiopoietin-2, urine NGAL, and serum syndecan-1 levels were measured using commercial enzyme-linked immunosorbent assay (ELISA) kits from R&D Systems (angiopoietin-2, Cat #DANG20; NGAL, Cat #DLCN20; syndecan-1, Cat #DY2780). All assays were performed according to the manufacturer’s instructions.

Routine laboratory assessment

At baseline, all patients underwent standard laboratory evaluations as part of routine clinical care. Blood samples were analyzed for sCr to assess renal function, hemoglobin levels, and complete blood counts, including mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), and MCH concentration (MCHC). Liver function was assessed by measuring aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels. The inflammatory status was evaluated using C-reactive protein (CRP). All laboratory analyses were performed in the central clinical laboratory of the institution using automated, standardized methods.

Blinding and data handling

All laboratory personnel performing the index tests (i.e., NGAL, angiopoietin-2, and syndecan-1 assays) were blinded to the participants’ clinical information and reference standard results (pRIFLE and eCCI) to minimize measurement bias. Conversely, clinicians and investigators determining the SA-AKI status using the reference standard were blinded to the biomarker (index test) results during classification. Clinical data routinely recorded during patient care were accessible to the treating team but remained unavailable to laboratory analysts and outcome assessors during data processing.

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics, version 20.0 (IBM Corp.), and data visualization was conducted using GraphPad Prism, version 10.0 (GraphPad Software).

Handling of indeterminate results and missing data

Indeterminate or inconclusive results for either the index test (biomarkers) or the reference standard (pRIFLE-based eCCI classification) were excluded from the final analysis. Participants with incomplete clinical or laboratory data that prevented accurate AKI status classification were also excluded. All samples with missing biomarker measurements at any time point (0, 12, 24, or 48 hours) because of insufficient volume, hemolysis, or assay failure were excluded from the corresponding time-point analysis. However, these samples were included in other available measurements to maximize data use.

Normality test

The distribution of continuous variables was evaluated for normality using the Shapiro-Wilk test. The outcomes of this test guided the selection of statistical analyses, with parametric or nonparametric tests applied as appropriate.

Comparison between groups and time points

Categorical variables were analyzed using Fisher exact test. Statistical analyses of continuous variables were conducted according to data distribution. Baseline characteristics were compared using the independent samples t test for normally distributed variables or the Mann-Whitney U test for non-normally distributed variables. Differences between time points were analyzed using the paired samples t test or the Wilcoxon signed-rank test, as appropriate. Normally distributed variables were compared between groups and across time points using two-way analysis of variance, followed by post hoc pairwise comparisons with Tukey’s honestly significant difference test. For non-normally distributed variables, group comparisons were performed using the Kruskal-Wallis test, with significant results further analyzed using the Mann-Whitney U test. A P-value <0.05 was considered statistically significant.

Data are presented as mean±standard deviation for normally distributed variables or as median (interquartile range) for non-normally distributed variables. Between-group differences are expressed as the mean difference with a 95% confidence interval (CI) for normally distributed data and as the median difference with a 95% CI, estimated using the Hodges–Lehmann method, for non-normally distributed data.

Diagnostic performance and correlations among predictors

Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance of biomarkers at specific time points (0, 12, 24, and 48 hours) and of baseline characteristics showing significant differences between patients with and without SA-AKI. The optimal cutoff value for each predictor was determined using the Youden index (sensitivity+specificity−1). Correlations among significant predictors, as identified by their area under the curve (AUC) P-values, were assessed using Spearman’s rank-order correlation test.

Results

A total of 38 children met the inclusion and exclusion criteria and provided consent to participate in the study. However, data from 15 children were incomplete, resulting in a final analysis cohort of 23 participants. Eight children were classified as having progressed to SA-AKI according to the pRIFLE criteria.

Baseline characteristics

Baseline characteristics of the included cohorts are presented in Table 1. The sex distribution was relatively balanced, with 47.8% males and 52.2% females. Children who developed AKI were more frequently female (71.4%) than those without AKI (43.7%), although this difference was not significant (P=0.371). The mean age was slightly higher in the AKI group (7.43±5.41 years) than in the non-AKI group (6.13±4.38 years), but this difference was not statistically significant (P=0.612). Pneumonia (34.8%), leukemia (21.7%), and rhabdomyosarcoma (13.0%) were the most common comorbidities, with no significant differences in comorbid conditions between the AKI and non-AKI groups. Significant differences were observed in baseline creatinine (P=0.008), eCCI (P=0.001), urine NGAL (P=0.028), and MCV (P=0.036). However, no significant differences were observed in the other laboratory parameters, including serum syndecan, serum angiopoietin-2, hemoglobin, hematocrit, leukocyte count, thrombocyte count, MCH, MCHC, ALT, AST, and CRP.

Table 1.

Baseline characteristics of the participants

Characteristics	All (n=23)	SA-AKI (n=8)	Non-SA-AKI (n=15)	P-valuea)
Sex, No. (%)				0.669
Male	11 (47.8)	3 (37.5)	8 (53.3)
Female	12 (52.2)	5 (62.5)	7 (46.7)
Age (yr), mean±SD	6.52±4.63	6.75±5.37	6.23±3.90	0.868
Diagnosis, No. (%)
Sepsis	23 (100.0)	8 (100.0)	15 (100.0)	1.000
Pneumonia	8 (34.8)	1 (12.5)	7 (46.7)	0.176
Tumors	12 (52.2)	4 (50.0)	8 (53.3)	1.000
Leukemia (AML/ALL)	5 (21.7)	1 (12.5)	4 (26.7)	0.621
Rhabdomyosarcoma	3 (13.0)	1 (12.5)	2 (13.3)	1.000
Retinoblastoma	1 (4.3)	0	1 (6.7)	1.000
Soft tissue tumor	1 (4.3)	1 (12.5)	0	0.348
Yolk sac/embryonal carcinoma	2 (8.7)	1 (12.5)	1 (6.7)	1.000
Aplastic anemia	2 (8.7)	0	2 (13.3)	0.526
GI tract bleeding	3 (13.0)	2 (25.0)	1 (6.7)	0.269
Infections	3 (13.0)	2 (25.0)	1 (6.7)	0.269
Cardiomyopathy	1 (4.3)	0	1 (6.7)	1.000
Guillain-Baire Syndrome	1 (4.3)	0	1 (6.7)	1.000
GFR markers, median (IQR)
Creatinine (mg/dL)	0.38 (0.51)	1.12 (1.45)	0.38 (0.12)	0.008
eCCI (mL/min/1.73 m2)	132.00 (77.37)	57.76 (70.59)	148.16 (30.00)	0.001
Urine NGAL (ng/mL)	25.50 (220.45)	231.35 (2,331.78)	16.10 (26.35)	0.028
Serum angiopoietin-2 (ng/mL)	21.83 (43.78)	15.30 (39.81)	21.83 (45.35)	0.636
Serum syndecan (ng/mL)	2.93 (2.84)	3.24 (28.41)	2.92 (2.64)	0.591
Blood parameters, median (IQR)
Hemoglobin (g/dL)	8.20 (2.40)	7.90 (2.86)	8.30 (2.40)	0.825
Hematocrit (%)	26.20 (6.40)	25.25 (6.50)	26.20 (7.60)	0.506
Leukocyte count (×103/µL)	10.30 (24.73)	8.34 (19.32)	13.16 (63.35)	0.681
Thrombocyte count (×103/µL)	110.00 (265.50)	151.00 (337.50)	110.00 (247.50)	0.169
MCV (fL)	81.10 (9.50)	76.90 (11.55)	82.60 (9.60)	0.036
MCH (pg)	26.80 (1.95)	25.70 (4.40)	27.10 (1.45)	0.076
MCHC (g/dL)	32.50 (2.55)	32.50 (2.20)	32.50 (3.45)	0.975
Others, median (IQR)
ALT (U/L)	48.00 (96.00)	69.00 (420.50)	48.00 (57.50)	0.500
AST (U/L)	28.00 (36.00)	23.50 (143.75)	28.00 (25.00)	0.860
C-reactive protein (mg/dL)	9.40 (12.35)	10.75 (14.78)	7.50 (11.00)	0.925

SA-AKI, sepsis-associated acute kidney injury; SD, standard deviation; AML, acute myeloid leukemia; ALL, acute lymphoid leukemia; GI, gastrointestinal; GFR, glomerular filtration rate; IQR, interquartile range; eCCI, estimated creatinine clearance index; NGAL, neutrophil gelatinase-associated lipocalin; MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; AST, aspartate aminotransferase; ALT, alanine aminotransferase.

^a)Statistical data were assessed utilizing: sex and diagnosis were analyzed using Fisher exact test; age using the independent sample t test; and GFR markers, blood parameters, and other variables using the Mann-Whitney U test.

Biomarker level within 48 hours of observation

The eCCI and NGAL levels were significantly lower in the SA-AKI group than in the non-SA-AKI group at all observation time points (0, 12, 24, and 48 hours; all P≤0.001) (Figs. 1 and 2). In contrast, no significant differences were observed between the groups for angiopoietin-2 (Fig. 3) or syndecan-1 (Fig. 4) at any time point (all P>0.05). Moreover, within-group comparisons showed no significant changes in biomarker levels across the 48-h observation period (all P>0.05, data not shown).

Fig. 1.

Estimated creatinine clearance index (eCCI) within 48 hours in children with sepsis-associated acute kidney injury (SA-AKI) and sepsis without AKI. The bar plot shows mean±SD, with individual data points indicated by dots. SD, standard deviation; IQR, interquartile range; CI, confidence interval. Statistical significance was assessed using the Kruskal-Wallis test followed by the Mann-Whitney U test and is denoted above the graphs with asterisks as follows: ^**P<0.01, ^****P<0.0001.

Fig. 2.

Neutrophil gelatinase-associated lipocalin (NGAL) levels within 48 hours in children with sepsis-associated acute kidney injury (SA-AKI) and those without AKI. The bar plot shows mean±SD, with individual data points represented by dots. SD, standard deviation; IQR, interquartile range; CI, confidence interval. Statistical significance was assessed using the Kruskal-Wallis test followed by the Mann-Whitney U test and is indicated above the graphs with asterisks as follows: ^*P<0.05, ^***P<0.001.

Fig. 3.

Angiopoietin-2 levels within 48 hours in children with sepsis-associated acute kidney injury (SA-AKI) and those without AKI. The bar plot shows mean±SD, with individual data points represented by dots. SD, standard deviation; IQR, interquartile range; CI, confidence interval. Statistical significance was assessed using the Kruskal-Wallis test followed by the Mann-Whitney U test and no statistically significant differences were found between groups or across time points.

Fig. 4.

Syndecan-1 levels within 48 hours in children with sepsis-associated acute kidney injury (SA-AKI) and those without AKI. The bar plot shows mean±SD, with individual data points represented by dots. SD, standard deviation; IQR, interquartile range; CI, confidence interval. Statistical significance was assessed using the Kruskal-Wallis test followed by the Mann-Whitney U test, and no statistically significant differences were found between groups or across time points.

Diagnostic performance of biomarkers and significant baseline characteristic

ROC curve analysis (Fig. 5) revealed that eCCI was a robust predictor for distinguishing SA-AKI from non-SA-AKI, with an AUC consistently >0.900 at all time points. The highest accuracy was observed at 24 hours, when the model achieved perfect discrimination (AUC=1.000). The optimal cutoff value for predicting AKI was a 24-hour eCCI of <96.95 mL/min/1.73 m², with 100.0% (67.6%–100.0%) sensitivity and 100.0% (79.6%–100.0%) specificity (Table 2). NGAL demonstrated strong diagnostic performance at baseline (AUC=0.800; 95% CI, 0.588–1.000; P=0.020) and reached its highest predictive accuracy at 12 and 24 hours (AUC >0.900). However, its predictive value declined at 48 hours (AUC=0.675; 95% CI, 0.414–0.919; P=0.175). The optimal cutoff values were >97.85 ng/mL for 12 hour NGAL (sensitivity, 87.5% [52.9%–99.4%] and specificity, 80.0% [54.8%–92.9%]) and >117.40 ng/mL for 24-hour NGAL (sensitivity, 87.5% [52.9%–99.4%] and specificity, 86.7% [62.1%–97.6%]). Baseline MCV demonstrated moderate predictive ability for SA-AKI (AUC=0.767; P=0.039) with a cutoff value of <73.90 fL (sensitivity, 50.0% [21.5%–78.5%] and specificity, 93.3% [70.8%–99.7%]). In contrast, angiopoietin-2 and syndecan-1 showed no predictive value at any time point (P>0.05).

Fig. 5.

Receiver operating characteristic (ROC) curves of candidate biomarkers for predicting sepsis-associated acute kidney injury (SA-AKI) at specific time points (0, 12, 24, and 48 hours). AUC, area under the curve; CI, confidence interval; eCCI, estimated creatinine clearance index; NGAL, neutrophil gelatinase-associated lipocalin; Ang-2, angiopoietin-2; Syn-1, syndecan-1; MCV, mean corpuscular volume.

Table 2.

Binary logistic regression analysis of significant markers cutoff for predicting sepsis-associated acute kidney injury

Model	AUC (95% CI)	Cutoff value	Sensitivity, % (95% CI)	Specificity, % (95% CI)
eCCI
eCCI 0-hour	0.917 (0.793–1.000)	126.00	87.5 (52.9–99.4)	86.7 (62.1–97.6)
eCCI 12-hour	0.942 (0.829–1.000)	110.00	100.0 (67.6–100.0)	93.3 (70.2–99.7)
eCCI 24-hour	1.000 (1.000–1.000)	96.95	100.0 (67.6–100.0)	100.0 (79.6–100.0)
eCCI 48-hour	0.950 (0.849–1.000)	82.14	87.5 (52.9–99.4)	100.0 (79.6–100.0)
NGAL
NGAL 0-hour	0.800 (0.588–1.000)	252.20	62.5 (30.6–86.3)	93.3 (70.2–99.7)
NGAL 12-hour	0.900 (0.771–1.000)	97.85	87.5 (52.9–99.4)	80.0 (54.8–92.9)
NGAL 24-hour	0.908 (0.780–1.000)	117.40	87.5 (52.9–99.4)	86.7 (62.1–97.6)
NGAL 48-hour	0.675 (0.432–0.919)	99.50	75.0 (41.9–95.6)	73.3 (48.0–89.1)
Baseline MCV	0.767 (0.564–0.969)	73.90	50.0 (21.5–78.5)	93.3 (70.2–99.7)

AUC, area under the curve; CI, confidence interval; eCCI, estimated creatinine clearance index; NGAL, neutrophil gelatinase-associated lipocalin; MCV, mean corpuscular volume.

Correlation between significant predictors

A weak to moderate negative correlation was observed between baseline (0 hour) NGAL and eCCI at all subsequent time points (weak: r=−0.15 to −0.26) and (moderate: r=−0.36 to −0.50), indicating that baseline NGAL has limited reliability as a predictive biomarker. In contrast, NGAL measured at 12 and 24 hours showed stronger correlations with early renal function. Specifically, 12- and 24-hour NGAL correlated moderately with baseline eCCI (r=−0.50 and r=−0.53, respectively) and with 12-h eCCI (r=−0.41 and r=−0.36, respectively). The strongest correlations were observed between 12- and 24-hour NGAL and 24-hour eCCI (r=−0.62 and r=−0.67, respectively). At 48 hours, 12-hour NGAL maintained a moderate-to-strong correlation with eCCI, whereas the correlation for 24-hour NGAL declined. Baseline MCV exhibited only weak correlations with eCCI across all time points (r=0.18–0.35) (Fig. 6).

Fig. 6.

Spearman’s rank correlation matrix among significant predictors of sepsis-associated acute kidney injury (SA-AKI). This heatmap depicts the pairwise correlation coefficients (ρ) between estimated creatinine clearance (eCCI) at 0, 12, 24, and 48 hours; urinary neutrophil gelatinase-associated lipocalin (NGAL) at 0, 12, 24, and 48 hours; and baseline mean corpuscular volume (MCV). This heatmap depicts the pairwise correlation coefficients (ρ) between eCCI at 0, 12, 24, and 48 hours; urinary NGAL at 0, 12, 24, and 48 hours; and baseline MCV.

Discussion

In this prospective cohort of critically ill children with sepsis, NGAL emerged as the most reliable biomarker following standard eCCI. This result consistently demonstrates the strong predictive performance of NGAL in distinguishing SA-AKI from non-SA-AKI, particularly at 12 and 24 hours post-admission. This observation was further supported by the strong negative correlation between NGAL and eCCI. Although baseline MCV reached statistical significance in the ROC analysis, it lacked predictive power, with its optimal cutoff yielding low sensitivity (50.0%). Moreover, correlation analysis revealed only weak associations between baseline MCV and eCCI, the reference standard for SA-AKI classification in this study. The other biomarkers, including angiopoietin-2 and syndecan-1, showed no discriminatory ability. Overall, these findings highlight NGAL as the most robust and clinically relevant biomarker for the early detection of SA-AKI in children with sepsis.

Our findings are consistent with those of the meta-analysis by Xie et al. [5], which included 49 studies encompassing 4,189 patients. The meta-analysis demonstrated that urinary NGAL is one of the most reliable biomarkers for SA-AKI, exhibiting strong diagnostic accuracy with a sensitivity of 80% (95% CI, 77%–82%), specificity of 70% (95% CI, 67%–72%), and an AUC of 0.72 to 0.92. However, the meta-analysis did not specify the time point at which NGAL reached its highest accuracy. In our study, urinary NGAL demonstrated comparable performance, with an AUC of 0.800 at baseline (sensitivity, 62.5% [95% CI, 30.6%–86.3%]; specificity, 93.3% [95% CI, 70.2%–99.7%]), consistent with the reported range. Notably, the AUC increased to 0.900 and 0.908 at 12-hour and 24-hour post-admission, respectively. At both time points, the sensitivity was 87.5% (95% CI, 52.9%–99.4%). The specificity was 80.0% (95% CI, 54.8%–92.9%) at 12 hours and 86.7% (95% CI, 62.1%–97.6%) at 24 hours. These results indicate its superior predictive performance at these time points. These findings are consistent with those of Abbas et al. [22], who reported the highest NGAL accuracy at 12-hour post-admission but not at later time points.

A cutoff value of 97.85–117.40 ng/mL was identified to differentiate SA-AKI from non-SA-AKI cases. This threshold is lower than the previously reported clinical cutoff of 125 ng/mL (Fig. 2), which was proposed to distinguish stage 2 from stage 3 AKI within 48 to 72 hours of PICU admission in pediatric patients aged >90 days to <22 years [23]. However, our identified cutoff remains above the normal urinary NGAL range of 50–96 ng/mL reported for cohorts aged >3 months to <22 years [24].

In this study, AKI was defined using the pRIFLE criteria. Although pRIFLE is less frequently used than the KDIGO classification system, it has several advantages in pediatric populations. pRIFLE determines the severity of AKI based on the percentage decline in eCCI, offering a more accurate reflection of renal function in children. In contrast, KDIGO relies on absolute increases in sCr and urine output thresholds derived primarily from adult populations, making it less sensitive in children who generally have lower baseline sCr levels because of smaller muscle mass. Furthermore, pRIFLE allows for the retrospective estimation of baseline kidney function, which is particularly valuable in pediatric settings where pre-illness sCr values are often unavailable. Previous studies have also demonstrated that pRIFLE is more sensitive criterion than KDIGO, AKIN, and pROCK (Pediatric Reference Change Value Optimized for AKI) for detecting stage 1 AKI [7,25-27].

In this study, the pRIFLE criteria were applied using the eCCI derived from the 1976 Schwartz equation, which assumes a baseline eCCI of 120 mL/min/1.73 m² when pre-illness creatinine values are unavailable. Because of this assumption, a substantial proportion of patients met the AKI threshold at the 0-hour time point (upon intensive care unit admission). Consequently, patients presenting with an eCCI <90 mL/min/1.73 m² at 0 hours were immediately classified as belonging to the AKI group. The 1976 Schwartz equation was used instead of the modified 2012 version because it is specifically calibrated for sCr measured by the Jaffe method, which was the assay technique employed in our laboratory. In contrast, the 2012 Schwartz equation is standardized for enzymatic creatinine assays traceable to isotope dilution mass spectrometry reference methods [28]. Nevertheless, this baseline assumption may not be universally accurate because GFR varies with ethnicity, body composition, and age—Asian children and those under 2 years of age typically have lower GFR values. Consequently, the pRIFLE criteria may overestimate or misclassify the severity of AKI, particularly in younger or non-Western pediatric populations [29]. Furthermore, the equation requires accurate height measurement, which may be difficult to obtain in critically ill children [26]. Despite these limitations, the pRIFLE criteria remain widely used in clinical and research settings owing to their simplicity, affordability, and rapid applicability.

In this study, we observed a progressive decrease in the eCCI cutoff values from baseline to 48-hour post-admission in differentiating the SA-AKI from non-SA-AKI cohorts. This case indicates a deterioration in renal function among patients with SA-AKI. Interestingly, the cutoff values at 0 and 12 hours remained within the normal physiological range (110.0–126.2 mL/min/1.73 m²), whereas the 24-hour eCCI—although demonstrating 100% accuracy in distinguishing SA-AKI—was slightly higher than the threshold of the “Risk” category (>90 mL/min/1.73 m²). Conversely, in the non-SA-AKI cohort, eCCI values exceeded the upper limit of the normal range. This increase likely reflects a state of hyperfiltration or augmented renal clearance (ARC), a well-recognized phenomenon in patients with critical illness and sepsis, including children. ARC is characterized by an increased GFR (typically >130–150 mL/min/1.73 m² in pediatric populations) because of systemic inflammatory and hemodynamic responses during the hyperdynamic phase of sepsis. Enhanced cardiac output, systemic vasodilation, and increased renal blood flow transiently augment glomerular perfusion, resulting in elevated creatinine clearance values. Importantly, the presence of ARC may elevate the eCCI reference range, masking early declines in renal function and delaying AKI recognition when using fixed thresholds. In this context, a patient may experience a clinically significant reduction in renal reserve while still demonstrating eCCI values within or above the conventional “normal” range. Consequently, standard eCCI cutoff values derived from non-ARC populations may underestimate the true incidence or early onset of AKI in patients with sepsis and hyperdynamic circulation [30-32]. We hypothesized that the higher eCCI cutoff values observed in our cohorts were likely attributable to ARC, in which hyperfiltration during the early hyperdynamic phase of sepsis elevates baseline creatinine clearance and consequently raises the threshold for AKI detection.

We found that the vascular biomarkers, including angiopoietin-2 and syndecan-1, lacked diagnostic significance. Vascular biomarkers have not been extensively studied. Yu et al. [33] reported significantly higher angiopoietin-2 and syndecan-1 levels in intensive care unit patients with severe AKI compared with those without. Unlike our study, their cohort included patients with stage 2 and 3 AKI based on the KDIGO criteria, corresponding to the Injury and Failure categories in the RIFLE classification [33]. Collectively, these findings indicate that angiopoietin-2 and syndecan-1 are more effective for detecting AKI in later stages or in more severe disease than for early prediction. In this study, we observed significantly lower baseline MCV in children with SA-AKI than in those without SA-AKI. However, MCV did not demonstrate predictive value for SA-AKI. This finding is consistent with that of Hu et al. [34], who reported no diagnostic significance of MCV in predicting AKI (OR, 0.909; 95% CI, 0.651–1.268; P=0.573).

In summary, incorporating 12- to 24-hour NGAL into eCCI and pRIFLE criteria provides a more comprehensive and clinically meaningful approach for detecting SA-AKI in critically ill children. eCCI-based definitions such as pRIFLE remain valuable for assessing functional renal decline. However, their diagnostic accuracy may be compromised in the presence of ARC, a common early response to sepsis characterized by transient hyperfiltration. This physiological state can mask early kidney injury by maintaining or even elevating eCCI values despite ongoing tubular damage. In contrast, NGAL reflects structural tubular injury independent of glomerular filtration changes, facilitating the detection of subclinical or evolving AKI before a measurable decline in renal function occurs.

Our findings demonstrated the promising diagnostic value of NGAL in predicting SA-AKI in the pediatric population. However, several limitations should be acknowledged. First, the sample size was relatively small, which may have limited the statistical power of the study. Future studies with larger cohorts, particularly in Asian populations, are crucial for validating these results. Second, the diagnosis of SA-AKI was based on the pRIFLE criteria, which assumed a normal baseline eCCI of 120 mL/min/1.73 m²; this approach may have led to over- or underestimation of the incidence of AKI. Third, the study cohort included children aged 1 to 18 years, which may limit the generalizability to infants younger than 1 year. Finally, the study was conducted at a single center, which may restrict the external validity of the findings across different healthcare settings.

In conclusion, NGAL exhibits strong predictive value for the early detection of SA-AKI in children with sepsis, consistently correlating with the pRIFLE criteria. Incorporating NGAL into existing functional criteria enhances the sensitivity and timeliness of AKI diagnosis, facilitates earlier therapeutic intervention, and provides a more accurate representation of renal pathophysiology in children with sepsis. Larger, multicenter studies are needed to validate these findings and to support the integration of NGAL into standardized pediatric sepsis management guidelines.

Funding Statement

None.

Data availability statement

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.