Main

Although outcomes in very LBW (VLBW) premature infants have improved over the past few decades, morbidity and mortality continue to be high (1). Acute kidney injury (AKI) is common and may be independently associated with mortality (2–4) in VLBW infants. Our ability to improve outcomes in those with AKI is hampered by the inability to detect AKI early in the disease process. Novel urine biomarkers of AKI have been discovered and promise to reliably detect AKI before an increase in serum creatinine (SCr) in different critically ill populations (5–14). Many of the biomarkers available have been tested in newborns undergoing cardiopulmonary bypass (15,16) but must be evaluated in VLBW infants who are born with underdeveloped glomeruli and tubules.

Our current methods to diagnose AKI using SCr-based definitions are problematic for the following reasons (17–19): a) changes in SCr represent a functional abnormality that occurs as a consequence of injury, not a marker of injury; b) SCr concentrations may not change until 25–50% of the kidney function has already been lost and thus it may take days before a significant increase in SCr is seen; c) SCr varies by muscle mass, hydration status, age and gender; d) SCr reflects maternal levels at birth and normally decreases to represent the infant's kidney function in the first few weeks of life (depending on the level of prematurity); and finally, e) blood analysis is not without consequences in premature infants as their total blood volume can be quite low (estimated blood volume of 500 g infant is 40 mL).

Recent advances in the field of clinical proteomics have greatly accelerated the discovery of novel urinary proteins which promptly increases in response to renal tubular injury. Some of the most promising urine AKI biomarkers include neutrophil gelatinase-associated lipocalin (NGAL), (20), IL-18 (16), kidney injury molecule-1 (KIM-1) (21,22), osteopontin (OPN), beta-2 microglobulin (B2mG), and Cystatin-C (Cys-C) (23). These and other biomarkers are being tested extensively in different critically ill populations, including children (3,6,14,18,24,25) and hold the promise to change our approach to AKI as they can detect AKI hours after an insult as opposed to SCr which may take days to increase after an injury. Development of these biomarkers has advanced such that point-of-care biomarker assessment kits are now becoming available for serum and urine NGAL (7,19) and KIM-1 (26).

Because glomerular and tubular development continues until 34 wk after GA, baseline levels of candidate biomarkers may be different depending on the degree of prematurity. Baseline evaluation of urine NGAL in premature infants has been performed and shows that urine biomarkers are inversely related to both GA and birth weight (27). Confirmation of these findings and exploration of the effect that patient demographics and clinical interventions have on other candidate biomarkers are needed. To determine the baseline levels of candidate AKI biomarkers in premature infants, we collected urine from premature infants during the first 7 postnatal days to test the hypothesis that urine levels of NGAL, KIM-1, IL-18, OPN, B2mG, and Cys-C are dependent on the degree of prematurity.

METHODS

VLBW infants with birth weight between 500 and 1500 g admitted to the regional quaternary care NICU at the University of Alabama at Birmingham (UAB) between February 2008 and July 2009 were enrolled in the study. Parental consent was obtained, and the study was approved by the Institutional Review Board at UAB. Infants were excluded if they did not survive to 48 h of life or if they had known congenital abnormality of the kidney.

Demographics.

Infant demographic data obtained consisted of GA, birth weight, race, small for GA, and sex. Clinical data included administration of aminoglycoside antibiotics and/or indomethacin, ventilator support, and Apgar scores at 1 and 5 min. Maternal demographic data collected included age, diabetes, pregnancy-induced hypertension, and preeclampsia. Maternal clinical data included administration of corticosteroids, aminoglycosides, vancomycin, and indomethacin.

Definition of AKI was made if an infant met criteria for Stage 1 definition (an increase in SCr by >0.3 mg/dL in 48 h or an increase of >50% from a previous value) of the AKIN (Acute Kidney Injury Network) classification, within the first 7 postnatal days (28). SCr was analyzed from laboratory results obtained as part of routine care, and if available from remnant samples that would have otherwise been discarded.

Biomarker analysis.

Urine was collected during the first 7 postnatal days with cotton balls placed near the perineum. Urine was extracted and centrifuged for 10 min to remove any cotton fibers or cellular elements and then frozen at −70°C until sample evaluation. Urine biomarker analysis was performed by Core A of the National Institutes of Health P30 O'Brien Core Center for AKI research (www.obrienaki.org) using the Meso Scale Discovery (MSD; Gaithersburg, MD).

NGAL, Cys-C, OPN, and B2mG were measured in urine with a prototype four value multiplex (4-plex) assay. IL-18 and KIM-1 were measured with a prototype duplex (2-plex) assay developed for this study.

Capture antibodies to the four antigens (NGAL, Cys-C, OPN, and B2mG) were arrayed by the manufacturer onto separate spots on the carbon electrodes of 4-spot 96-well MSD MULTI-SPOT plates. Capture antibodies against IL-18 and KIM-1 were arrayed by the manufacturer onto separate spots on the carbon electrodes of 4-spot 96-well MSD MULTI-SPOT plates, the other two spots were blocked with BSA. MSD supplied the calibrators for the 4-plex assay. Calibrators and biotinylated detection antibodies to IL-18 and KIM-1 were purchased from R&D Systems (Minneapolis, MN). Urine samples for 4-plex assay were prediluted 1 to 200 with sample dilution buffer (MSD Diluent 2). No dilution of samples was required for the 2-plex assay. Wells were blocked with MSD diluent 2 for 30 min, and 25 μL samples and calibrators were added to predestined wells. Plates were incubated for 2 h at room temperature on a shaker set to 600 rpm. Plates were then washed five times with PBS containing 0.05% Tween-20 (PBS-T) using SkanWasher-300 (Molecular Devices, Sunnyvale, CA).

For the 4-plex assay (NGAL/Cys-C/OPN/B2mG), detection antibodies coupled to a Ruthenium(II)-tris-bipyridine ([Ru(bpy)32+], SULFO-TAG) were blended together and added to wells. For the 2-plex assay (IL-18/KIM-1), biotinylated detection antibodies were blended with streptavidin-coupled SULFO-TAG and added to wells. Plates were incubated at room temperature for 2 h on a shaker (600 rpm). Plates were washed five times with PBS-T on SkanWasher-300 followed by addition of read buffer [buffered Tripropylamine (TPA)].

Detection is based on electrochemical oxidation of [Ru(bpy)32+] in the presence of TPA, an electrochemiluminescence coreactant, leading efficient generation of electrochemiluminescence glow via the high-energy electron transfer reaction between Ru(bpy)33+ and TPA radical. The electrochemical signal emitted as light was detected in a Sector Imager 2400 with a charge-coupled device, and the signal was analyzed using MSD Workbench and Data Analysis Toolbox v3.0 software. Sample concentrations were back fitted from standard curves generated with a 4-parameter logistic curve fit model with 1/y2 weighting. The intra- and interassay precisions were <3% and <5%, respectively, for both 4-plex and 2-plex assays. Standard back-calculated recoveries were 90–110%. Calibrators and sample duplicate correlation variability ranged from 0.12%–7.9%.

Statistical analysis.

Descriptive statistics were performed to determine differences between infant and maternal characteristics among four GA groups of VLBW infants (Table 1). Shapiro-Wilk test and normal probability plot were used to test for normality of data. Because biomarker values were not normally distributed, Kruskal-Wallis test was used to describe variation in biomarker values based on GA categories. Categorical variables were analyzed using χ2 or Fisher's exact test as appropriate. For all descriptive statistics, a p < 0.05 was considered statistically significant.

Table 1 Demographics by GA categories
Full size table

For the regression analysis, biomarkers were converted to natural log to gain normal distribution. For each urine sample obtained, the log of the urine biomarker/creatinine (cr) was calculated. A mixed model analysis, with one random intercept included per child, was performed to explore the impact of the measured variables on the biomarker/creatinine values. GA and AKI were forced into all models. All other maternal and infant demographic variables were explored and removed in a stepwise backward selection model keeping variables with p < 0.2. The formula (exp(beta) − 1) × 100% from the regression coefficients was used to express the % change. SAS 9.2 (SAS Institute Inc., Cary, NC) was used for the all statistical analysis.

RESULTS

The demographic variability for 123 premature infants categorized by GA (Table 1) shows differences in infant characteristics as expected (birth weight, exposure to indomethacin and aminoglycosides, respiratory support, Apgar scores at 1 and 5 min, AKI, and small for GA; all p < 0.01). Maternal preeclampsia (p < 0.02) and hypertension (p < 0.001) were different among groups (p < 0.01).

As GA increased, urine NGAL, KIM-1, Cys-C, and B2mG values progressively decreased, but no significant differences were seen for other biomarkers (Table 2). When corrected for urine cr, all biomarker/cr ratios progressively declined with GA except for IL-18/cr (Table 3).

Table 2 Variation of candidate urine AKI biomarker levels by GA
Full size table
Table 3 Variation of candidate urine AKI biomarker levels corrected for urine creatinine by GA
Full size table

After controlling for other potential confounders, GA independently impacted NGAL/cr, OPN/cr, and B2MG/cr but not KIM-1/cr, IL-18/ cr, or Cys-C/cr. Birth weight was independently associated with Cys-C/cr. The use of blood pressure support medication and indomethacin was independently associated with Kim-1/cr levels. Female gender was independently associated with higher NGAL/cr and IL-18/cr and lower Kim-1/cr. Vancomycin was independently associated with OPN/cr values. Race was independently associated with OPN/cr and B2mG/cr values (Table 4).

Table 4 Mixed model regression to predict log of urine biomarker/creatinine
Full size table

DISCUSSION

Baseline values of urine NGAL, KIM-1, Cys-C, and B2mG decrease with increasing GA. With correction to urine cr, these markers and OPN decreased with increasing GA. IL-18 (with or without correction for urine creatinine) did not differ across GA categories. Controlling for other potential clinical and demographic confounders with regression analysis shows that NGAL/cr, OPN/cr, and B2mG/cr are independently associated with GA.

Even though this study was not designed to determine whether these biomarkers predict AKI, we forced AKI into the models to determine whether variables such as GA were simply a common variable to explain high biomarker values and AKI. We found that GA continued to be associated with higher biomarker values, even when controlling for AKI, which suggests that these biomarkers are affected by GA and not simply that infants with higher GA have more AKI.

Our conclusions and values for the differences in urine NGAL values based on birth weight (data not shown) were similar to those documented by Lavery et al. (27). Besides confirming previous results on baseline values of urine NGAL in VLBW infants, our study provides baseline values and evidence that these biomarkers are affected by GA in other urine biomarkers (KIM-1, Cys-C, B2-MG, and OPN) but not IL-18. These normative data will help investigators design and analyze AKI biomarker studies.

In 2008, Huynh et al. (29) described urine NGAL levels of infants (birth weight >750 g) and showed differences between males and females but did not stratify them according to GA or birth weight categories. Interestingly, the baseline urine NGAL values in our cohort were similar to values found by Huynh et al. (29) for those whose birth weight is >750 g, but the overall values in our cohort were higher likely because our population included many infants with birth weight <750 g. Also, our sample showed higher variability among NGAL levels likely due to differences in our sample population (Huynh excluded those with clinical risk factors for AKI, whereas we included a more heterogeneous premature infant population). Our data show an independent association between urine biomarker values and female gender which confirms the findings by Huynh et al. which showed consistently higher urine NGAL values in premature females.

Compared with other critically ill pediatric patients who were on ventilator support and required ionotropic medications (14,30), baseline NGAL values are significantly higher in our population while urine IL-18 levels were similar, likely because their cohort was made up of children with multiorgan failure. Similarly baseline values of urine NGAL (but not IL-18, B2MG, or OPN) were higher in a cohort of children evaluated in a pediatric emergency department (31).

The strength of this study includes the heterogeneous group of VLBW infants and the evaluation of six candidate biomarkers. However, we acknowledge several limitations. Even though we report the number of infants with known AKI in each category, we acknowledge that some infants may have sustained AKI but were “missed” if SCr levels around the time of insult were not performed; thus, from our study, the impact of AKI on these biomarkers cannot be ascertained. Although we controlled for AKI in our regression model, it is possible that the reason higher levels were seen in those with lower GA is due to higher incidence of AKI. Second, we acknowledge that there could be other variables which we did not account for that could explain variations in biomarker levels. Therefore, inferences from this analysis should be taken in context of the above limitations. Future studies to determine the ability of these and other biomarkers to detect AKI and mortality are greatly needed.

In conclusion, the normative urine values of NGAL, KIM-1, OPN, Cys-C, and B2mG (but not IL-18) are higher in the most premature. One must acknowledge these differences when designing validation studies to find candidate biomarkers of AKI. Whether clinicians and researchers will need to adjust for GA, birth weight, and gender differences when using these biomarkers is yet to be determined.