Urinary Biomarkers and Renal Near-Infrared Spectroscopy Predict ICU Outcomes Following Cardiac Surgery in Infants Under 6 Months of Age

Abstract

Objective

To assess the ability of urinary acute kidney injury biomarkers and renal near-infrared spectroscopy (NIRS) to predict outcomes in infants following congenital heart surgery.

Methods

Urinary levels of neutrophil gelatinase-associated lipocalin (NGAL), interleukin-18 (IL-18), kidney injury molecule-1 (KIM-1), and cystatin C were measured pre- and post-operatively in 49 infants <6 months of age. Renal NIRS was monitored for the first 24 h following surgery. A composite poor outcome was defined as death, the need for renal replacement therapy, prolonged time to first extubation, or prolonged ICU length of stay.

Results

Forty-two patients (86%) developed acute kidney injury by meeting at least AKIN/KDIGO stage 1 criteria, and 17 (35%) patients experienced poor outcomes, including three deaths. With the exception of KIM-1, all biomarkers demonstrated significant increases within 24 h post-operatively among patients with poor outcomes. Low levels of NGAL and IL-18 demonstrated high negative predictive values (91%) within 2 h post-operatively. Poor outcome infants had greater cumulative time with NIRS saturations <50% (60 vs. 1.5 min, p=0.02) in the first 24 h.

Conclusions

Within the first 24 h following cardiopulmonary bypass, infants at increased risk for poor outcomes demonstrated elevated urinary NGAL, IL-18, and cystatin C, and increased time with low NIRS saturations. These findings suggest that urinary biomarkers and renal NIRS may differentiate patients with good vs. poor outcomes in the early post-operative period which could assist clinicians when counseling families and inform the development of future clinical trials.

Background

Acute kidney injury (AKI) is a common and potentially serious complication following congenital heart surgery. Depending on the study population and criteria used to define AKI, contemporary studies report an incidence ranging from 5.5 – 50%, with associated mortality ranging from 20 – 61% [1–3]. Well-established risk factors include young age, low weight, increased duration of cardiopulmonary bypass (CPB), and post-operative hemodynamic instability [1–3]. Although they represent a high risk population, infants often only comprise a small proportion of the study patients, and, because AKI is ill-defined in neonates, they are often entirely excluded [4].

In order to standardize the definition of AKI, expert consensus groups have created the RIFLE and AKIN scoring systems, which are based on changes in serum creatinine (S_Cr) and urine output [5, 6]. Modification and application of the RIFLE criteria to pediatrics (pRIFLE) has been shown to predict morbidity and mortality in critically ill children [7]. In addition, a modified version of the AKIN criteria has also been shown to portend a poor clinical outcome in infants following congenital heart surgery [8]. While these systems represent important advancements in the standardization of AKI diagnosis, they have limited clinical utility in the perioperative period due to delayed and unpredictable changes in S_Cr and urine output following CPB. Creatinine-based measures of AKI are particularly limited in neonates due to the influence of maternal S_Cr, low glomerular filtration rates, and effect of serum bilirubin on the assay itself [4].

Efforts to improve the sensitivity, accuracy, and timeliness of AKI diagnosis have resulted in the clinical testing of renal injury biomarkers. The molecules neutrophil gelatinase-associated lipocalin (NGAL), interleukin-18 (IL-18), kidney injury molecule-1 (KIM-1), and cystatin C (Cys C) are increased in the blood and/or urine of critically ill adults and children with AKI in a variety of clinical settings [9]. Following congenital heart surgery, these biomarkers are elevated within 12 h of CPB in patients who develop AKI by S_Cr [10–13].

Near-infrared spectroscopy (NIRS) uses an infrared light source to measure regional oxyhemoglobin saturation (rSO₂) continuously and non-invasively 1.5–2 cm into the tissue bed of interest [14]. It has the advantage of being non-invasive, real time, and specific to the organ of interest. Saturation values obtained from cerebral and flank NIRS probes are correlated with invasive mixed venous oxygen saturations following congenital heart surgery [15]. In addition, low rSO₂ values have been correlated with injury to the brain, gut, and kidney following CPB in children [16–18].

To date, urinary biomarkers and renal NIRS have not been simultaneously evaluated following congenital heart surgery and correlated with outcome. We hypothesized that infants with peri-operative kidney injury identified by elevated urinary biomarkers or low renal NIRS values, would have comparatively poorer clinical outcomes.

Methods

This prospective study was approved by the Institutional Review Board at the University of Michigan. Infants under 6 months of age undergoing cardiac surgery with CPB between July 2009 and July 2010 were eligible for enrollment. Fifty families were approached, and one family declined participation in the study. Premature infants less than 35 weeks gestation were excluded. After parental or guardian informed consent was obtained, patient demographic and surgical information were collected. Surgical complexity was ranked according to the Risk Adjustment for Congenital Heart Surgery 1 (RACHS-1) scoring system [19]. Post-operative hemodynamic status was estimated by calculation of a daily maximum vasoactive-inotropic score (VIS) for the first three days following surgery as per the equation below [20]:

• VIS = dopamine dose (µg/kg/min) + dobutamine dose (µg/kg/min) + 100*epinephrine dose (µg/kg/min) + 10*milrinone dose (µg/kg/min) + 10,000*vasopressin dose (units/kg/min) + 100*norepinephrine dose (µg/kg/min)

All patients received routine standard of care during the study period which included the use of dextrose-containing crystalloid solutions (75–100 cc/kg/day) during the first 24–48 h post-operatively, followed by the initiation of total parenteral nutrition. Patients were started on bolus furosemide (1 mg/kg/dose q6h) within the first 24 h post-operatively. One of the study patients had a peritoneal drain placed for mild abdominal compartment syndrome. Primary providers were aware that the patients were enrolled in a study looking at AKI and outcomes.

Biomarkers

Pre-operative urine samples were collected from a bag specimen or in-dwelling catheter placed in the operating room prior to initiation of CPB. Post-operative urine samples were obtained from the Foley catheter on admission to the ICU and at 2, 6, 12, and 24 h. The samples were centrifuged (2000 rpm, 5 min), divided into aliquots, and stored (−80 °C) until analysis. Urinary concentrations of NGAL, IL-18, KIM-1, and cystatin C were measured using commercially available enzyme-linked immunosorbent assay kits from R&D Systems (Minneapolis, MN).

NIRS

Following arrival in the ICU, a Somanetics (Troy, MI) INVOS Pediatric NIRS sensor was placed over the right flank and continuous rSO₂ was recorded in 30 sec intervals for the first 24 h following surgery. The primary caregivers were blinded to the data during the collection period and no clinical interventions were made based on the NIRS values. Faulty NIRS probes were discovered in two patients after completion of the observation period. Because of the impact of ECMO on oxygen saturation, in 6 patients NIRS data was censored from analysis.

AKI

Serum creatinine was measured pre- and post-operatively, then once daily as part of routine daily laboratory studies. AKI was defined using criteria proposed by the Acute Kidney Injury Network and Kidney Disease Improving Global Outcomes group, and recently validated in a study of infants with congenital heart disease [6, 8, 21]. Infants had AKI if they met AKIN stage 1 criteria, defined as an increase in S_Cr by either ≥ 0.3 mg/dL or a 50% rise from pre-operative baseline within the first three days post-operatively.

Outcome measures

Due to the relatively small size of our patient population and our low mortality and RRT rates, a composite poor outcome was used. A poor outcome was defined as any of the following: 1) death within 30 days of surgery; 2) need for RRT; 3) prolonged time to first extubation (upper quartile ≥ 6.5 days); or 4) prolonged ICU length of stay (upper quartile ≥ 9.9 days). Normative values for ventilator time and ICU length of stay were obtained from a contemporary internal database of infants < 6 months of age (N=395) who have undergone cardiac surgery.

Statistical Analysis

Demographic and clinical characteristics were compared using Student’s t-test for continuous variables, and Fisher’s exact test for categorical variables. Wilcoxon Rank Sum test was used to evaluate the association between each continuous biomarker or continuous NIRS saturation value and the risk of a poor outcome at each post-operative hour. In order to determine an optimal cutoff value for each biomarker for the diagnosis of a poor outcome over time, a receiver operating characteristic (ROC) curve was generated using a non-parametric approach under repeated measures design [22]. Unconditional (exact) odds ratios with their 95% confidence intervals were estimated at each time point using (exact) logistic regression to evaluate a risk of poor outcome for each biomarker above the optimal cutoff obtained from the ROC curve. Due to the small sample size that limited the ability to perform multivariable analyses, unadjusted odd ratios are presented. All analyses were performed using SAS Version 9.2 (SAS Institute Inc, Cary, NC), with statistical significance set at p-values less than 0.05 using 2-sided tests.

Results

A total of 49 infants were enrolled during the study period. With the exception of one infant who underwent a stage 2 hemi-Fontan operation, all patients underwent surgery for the first time to repair or palliate one of the following primary cardiac diagnoses: ventricular septal defect (n=12), Tetralogy of Fallot (n=9), Hypoplastic Left Heart Syndrome (n=9), atrioventricular septal defect (n=6), D-transposition of the great arteries (n=5), heterotaxy syndrome (n=3), severe coarctation or interrupted aortic arch with ventricular septal defect (n=3), truncus arteriosus (n=1), or total anomalous pulmonary venous return (n=1). The three heterotaxy patients underwent a Ross procedure, Blalock-Taussig shunt, and a hemi-Fontan operation. Based on our composite definition, seventeen patients (35%) experienced a poor outcome. Three infants died, 2 required RRT, 12 had a prolonged time to first extubation, and 16 had a prolonged ICU length of stay. Both patients requiring RRT were neonates who were placed on ECMO support following surgery due to myocardial stun. In both cases, RRT was initiated on post-operative day 2 due to severe fluid overload. Both of these patients died.

Patients with a poor outcome were more likely to be younger and undergo more complex surgery with a longer CPB time and use of hypothermic circulatory arrest (Table 1). Post-operatively, patients with a poor outcome had higher maximal VIS, were more likely to require ECMO support, and had a higher peak post-operative S_Cr. The majority of patients (86%) met AKIN/KDIGO stage I criteria (S_cr rise of 50% or ≥ 0.3 mg/dL), however, this AKI classification system did not distinguish between those infants with good and poor outcomes.

Table 1.

Demographics and Clinical Characteristics

	Poor Outcome
	Yes (n=17)	No (n=32)	P-value
Sex, N (%)
Female	5 (29)	9 (28)	1.00
Male	12 (71)	23 (72)
Age at surgery, days, median (IQR)	7 (5–10)	98 (33–150)	0.001
Pre-operative weight, kg	3.7 ± 0.7	4.9 ± 1.4	0.0002
RACHS-1 classification, N (%)
2 to 4	8 (47)	29 (91)	0.001
5 or6	9 (53)	3 (9)
Cardiopulmonary bypass time, min	149 ± 61	84 ± 35	0.001
Aortic cross-clamp time, min	52 ± 40	45 ± 25	0.55
Hypothermic circulatory arrest, N (%)	10 (59)	5 (16)	0.003
Post-operative ECMO, N (%)	6 (35)	0 (0)	0.001
Maximum Vasoactive Inotrope Score	23 ± 9	12 ± 7	< 0.0001
Peak serum creatinine, mg/dL	0.9 ± 0.3	0.6 ± 0.3	0.005
Estimated creatinine clearance	30 ± 10	49 ± 17	<0.0001
Acute kidney injury, N (%)
No	1 (6)	6 (19)	0.40
Yes	16 (94)	26 (81)
AKIN/KDIGO stage 1	11 (65)	17 (53)	0.81
AKIN/KDIGO stage2	3 (18)	8 (25)
AKIN/KDIGO stage3	2 (12)	1 (3)

Data are presented as mean ± SD unless otherwise indicated.

IQR, interquartile range (25th – 75th percentile)

Abbreviations; ECMO, extracorporeal membrane oxygenation; RACHS, risk adjustment for congenital heart surgery.

Prior to undergoing CPB, urinary concentrations of NGAL, IL-18, cystatin C, and KIM-1 were low in all patients (Figure 1). With the exception of KIM-1, all biomarkers demonstrated significant increases within 12 h post-bypass in patients with poor outcomes and only minimal increases among patients with a good outcome. The differences in urinary biomarker concentration was significant at all post-operative time points for NGAL and IL-18, but only within the first 2 h post-operatively for cystatin C. KIM-1 poorly differentiated patients with either good or poor outcomes and was, therefore, removed from further analysis.

Figure 1.

Biomarker profiles in infants undergoing cardiopulmonary bypass. Urinary concentrations of NGAL (panel A), IL-18 (panel B), Cystatin C (panel C), and KIM-1 (panel D) were determined by ELISA as described in the Methods. Data are presented as median values in the good outcome (grey dotted line) vs poor outcome (solid black line) groups. Error bars indicate interquartile range (25th–75th percentile). * = P <0.05

Cut-offs to identify patients with poor outcomes were determined using ROC curves for NGAL, IL-18, and cystatin C (Table 2). All three biomarkers performed well, with NGAL having the highest AUC of 0.79. The time points most predictive of a poor outcome occurred for NGAL at post-operative hour 12, IL-18 at post-operative hour 6, and cystatin C at post-operative hour 0 (Table 3). Of note, all three biomarkers showed better negative than positive predictive values. Specifically, only 3/32 (9%) infants with a good outcome had an elevated NGAL at post-op hours 6 and 12, 1/32 (3%) had an elevated IL-18 at post-op hour 6, and 4/32 (13%) had elevated cystatin C at post-op hour 2.

Table 2.

Biomarker Optimal Cut-off Values for Prediction of a Poor Outcome

Biomarker	Optimal Cut-off Value (ng/mL)	AUC	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
NGAL	76	0.79	0.64	0.87	0.72	0.83
IL-18	0.8	0.76	0.59	0.86	0.68	0.80
Cys C	179	0.69	0.51	0.79	0.55	0.76

Abbreviations: AUC, Area Under the Receiver Operating Characteristic (ROC) Curve.

Table 3.

Risk of Poor Outcome for Biomarker by Threshold

		Poor Outcome†
Biomarker	Post-Op Hour	Yes (n=17)	No (n=32)	P-value	Odds Ratio (OR)	95% CI
NGAL ≥ 76 ng/mL	0	11 (65)	7 (22)	0.001	9.4	2.2, 39.0
	2	10 (59)	4 (13)	0.002	9.4	2.0, 54.7
	6	9 (53)	3 (9)	0.001	10.2	2.0, 72.9
	12	11 (65)	3 (9)	<.0001	18.7	3.4, 144
IL-18 ≥ 0.8 ng/mL	0	9 (53)	4 (13)	0.004	9.1	1.8, 56.2
	2	11 (65)	5 (16)	0.0004	11.0	2.6, 45.9
	6	5 (29)	1 (3)	0.01	15.4	1.5, 813
	12	4 (24)	3 (9)	0.24	2.7	0.4, 21.3
Cys C ≥ 179 ng/mL	0	9 (53)	8 (25)	0.01	5.2	1.3, 20.1
	2	7 (41)	4 (13)	0.03	4.9	1.2, 20.4
	6	8 (47)	6 (19)	0.05	3.7	0.9, 17.3
	12	9 (53)	10 (31)	0.11	2.7	0.8, 9.3
NGAL ≥ 76 and IL-18 ≥ 0.8	0	9 (53)	2 (6)	0.0002	19.7	3.1, 233
	2	8 (47)	3 (9)	0.005	8.1	1.6, 58
	6	3 (18)	1 (3)	0.11	6.4	0.5, 360
	12	3 (18)	1 (3)	0.12	6.2	0.5, 348
NGAL ≥ 76, IL-18 ≥ 0.8, and Cys C ≥ 179	0	7 (41)	1 (3)	0.001	24.0	2.5, 1221
	2	6 (35)	2 (6)	0.02	7.8	1.2, 90
	6	2 (12)	1 (3)	0.27	4.0	0.2, 252
	12	3 (18)	1 (3)	0.12	6.2	0.5, 348

Abbreviation; CI, confidence interval.

^†Data are presented as N (%).

Having multiple elevated biomarkers on arrival to the ICU (post-operative hour 0) was highly predictive of a poor outcome (Table 3). Patients with both elevated NGAL and IL-18 had a nearly 20-fold increased risk for a poor outcome and patients with elevated NGAL, IL-18, and cystatin C had a 24-fold increased risk. As with the single biomarkers, the negative predictive value of having multiple biomarkers below threshold was high.

Analysis of the NIRS data indicated that infants with a poor outcome spent more cumulative time with rSO₂ below 50% compared to those with good outcomes (Table 4, supplemental Figure 1). There were significant differences in cumulative time below 50% starting at post-op hour 2, and after 24 h of monitoring, poor outcome infants spent nearly an hour more time below 50% compared to those infants who had a good outcome (1–2 min). Comparing patients with single- vs bi-ventricular physiology, we found no difference in the time spent with NIRS rSO₂ below 50% through the first 12 hours post-operatively, although a difference did appear by the end of the NIRS monitoring period (Supplemental Table 1). Examining the ability of NIRS to correlate with markers of kidney injury, we examined the distribution of low NIRS rSO₂ and the presence of AKI, as defined by meeting AKIN stage 1 (Supplemental Table 2), or urinary biomarkers (Supplemental Table 3). We found that there was no association between NIRS rSO₂ <50% and the presence of AKI, however, in some cases, a NIRS rSO₂ <50% did correlate with elevated urinary biomarkers. Due to the small sample size, we were unable to evaluate outcomes using a combination of cumulative time below 50% and biomarker levels at the individual time points.

Table 4.

Median Cumulative Time (in minutes) with NIRS Saturation Below 50%

NIRS saturation	Post-Op hours	Poor Outcome		P-value
		Yes (n=10)	No (n=31)
< 50%	2	1.0 (0.0–57.5)	0.0 (0.0–0.0)	0.02
	6	3.5 (0.0–285.0)	0.0 (0.0–1.5)	0.03
	12	17.2 (0.0–507.5)	0.0 (0.0–8.0)	0.08
	24	59.5 (4.0–507.5)	1.5 (0.0–20.5)	0.02

^*Data are presented as Median (IQR - 25^th percentile-75^th percentile) of cumulative time spent with NIRS saturation below 50% from the beginning of monitoring until the indicated post-operative hour.

Discussion

Biomarkers

Many recent studies have focused on the ability of renal injury biomarkers to predict AKI before a rise in S_Cr [10, 11, 13] in a mixed age pediatric population. We examined biomarker performance in relation to clinical outcomes in a very young population and found, similar to the findings in older patients, that within the first 12 hours of arrival to the ICU following cardiac surgery, neonates and infants with elevated urinary NGAL, IL-18, and cystatin C were more likely to have poor clinical outcomes. Importantly, to our knowledge, the negative predictive value of these biomarkers has not been previously tested and emphasized in infants. The majority of infants with good outcomes did not have biomarkers levels above the cohort-based calculated thresholds. In this study, an infant with a NGAL < 76 ng/mL and IL-18 < 0.8 ng/mL in the first 2 h following surgery had at least a 90% chance of a good clinical outcome. Despite our broad definition of poor outcome, the ability to predict a high likelihood of having a good outcome within only 2 h of surgery is not only useful clinically, but has implications for enrollment in interventional trials, such that patients at low risk for a poor outcome can be excluded, improving the risk/benefit ratio for an experimental intervention. This concept has also been proposed by Wong et al, who reported that the negative predictive value of serum IL-8 could be used for risk stratification prior to enrolment in pediatric septic shock trials [23].

We found little diagnostic utility for KIM-1 in this study. While Han et al found that in a cohort of children undergoing CPB, urinary KIM-1 detected AKI before S_Cr, they did not correlate these findings with the same clinical outcomes included in this study, other than the need for RRT [13]. Krawczeski et al found that KIM-1 at 12 h following CPB independently correlated with ventilator time after adjusting for patient age, CPB time, and RACHS-1 score [24]. While the median age of AKI patients in that study was 6 months, the mean age of the present study patients was 2 months, with a significant proportion of neonates. Further study will be needed to determine the clinical utility of KIM-1 in neonates and infants less than 6 months of age.

We also evaluated the performance of NGAL, IL-18, and cystatin C in combination, and found the highest risk for a poor outcome occurred in infants with elevations in all three biomarkers immediately following surgery. While Parikh et al found that urinary NGAL and IL-18 were both strong predictors of severe AKI, defined as a two-fold increase in S_Cr or need for dialysis, the combination of these two biomarkers minimally improved predictive power [12]. In contrast, Krawczeski et al found that adding combinations of the urinary biomarkers NGAL, IL-18, KIM-1, and liver fatty acid-binding protein (FABP) to a clinical model of AKI risk factors improved the predictive ability of the model in the first 24 h following CPB [24]. While the concept of a biomarker “panel” is appealing, more studies are needed to determine the optimal timing and combination that maximizes predictive power. Due to differences in organ maturity, it is possible that such a panel may be age-dependent (i.e. the optimal panel and marker thresholds for a child may be different from that which is most useful in a neonate or infant).

Using AKIN/KDIGO stage 1 criteria, the incidence of AKI in this study was 86%, which is higher than other contemporary infant cardiac surgical data [1, 8, 24]. In addition, the presence of AKI did not distinguish between those infants with good or poor outcomes. This result is in contrast with the findings of Blinder et al who found that the presence of AKI correlated with mortality, duration of mechanical ventilation, and prolonged ICU length of stay [8]. These differences may be due to the relatively small sample size on our cohort, the use of a composite outcome variable in this study, or the high proportion of neonates who underwent complex RACHS-1 category 5 and 6 surgeries.

NIRS

Our study found that increased time spent with NIRS rSO₂ <50% was associated with a poor outcome. Assuming that the NIRS probe is measuring the mixed oxygen saturation of blood in the renal parenchyma, low rSO₂ would be an indicator of poor renal oxygen delivery. Although the exact threshold below which renal injury occurs is not known, these results corroborate those of Owens et al who showed that rSO2 <50% for more than 2 h correlated with higher serum creatinine values and increased incidence of AKI in infants during the first 48 h following CPB [16].

Our study included both single and biventricular patients, however, there was no difference between these two groups in cumulative time with rSO₂ <50% during the first 12 post-operative hours (see Supplemental Table 1). This finding is similar to the results published by Uebing et al, who found depressed cerebral oxygen saturations immediately following CPB in neonates who underwent the Norwood and arterial switch operations, and may be attributable to a post-CPB effect on renal blood flow rather than the presence or absence of a parallel circulation [25].

Our study was not large enough to analyze the predictive power of the combination of renal NIRS and urinary biomarkers to predict outcomes. We did, however, observe that independent of outcome, higher urinary biomarker levels correlated with lower duration of time with renal NIRS rSO₂ <50% (see Supplemental Table 3). While a substantial degree of renal injury occurs during CPB, it is possible that the real time information obtained from NIRS may compliment the data obtained from urinary biomarkers, such that biomarkers would identify patients at risk for AKI upon arrival to the ICU. In doing so, NIRS would provide a method for monitoring the effectiveness of post-operative interventions by trending the rSO₂. Further study is needed to evaluate this hypothesis.

Conclusions

Infants under 6 months of age with elevated urinary NGAL, IL-18, and cystatin C immediately following CPB are at increased risk for poor ICU outcomes. Increased duration of renal NIRS rSO₂ <50% in the first 24 h following CPB also predicts a poor clinical outcome. Importantly, infants with low urinary biomarkers within the first two hours post bypass are likely to have a good ICU outcome, which can help clinicians counsel families. These findings also underscore the potential utility of urinary AKI biomarkers in identifying infants at risk for poor outcomes, which can aid in the development of prospective trials. Further studies are needed to determine if a panel of renal injury biomarkers and renal NIRS can be used together and in the post-operative management of infants following congenital heart surgery.

Supplementary Material

01

Supplemental Figure 1: Cumulative time with low renal NIRS rSO₂. Median cumulative time spent with rSO₂ <50% in the good outcome (grey squares) vs poor outcome (black circles) groups.