Neutrophil Gelatinase–Associated Lipocalin Combined With Delta Serum Creatinine Provides Early Risk Stratification for Adverse Outcomes After Cardiac Surgery: A Prospective Observational Study

Abstract

Objectives:

Novel biomarkers of renal injury appear inconsistent in identifying a creatinine-based diagnosis of acute kidney injury. To be clinically useful, novel acute kidney injury biomarkers should identify patients at increased risk for adverse outcomes that are a consequence of acute kidney injury earlier and with greater utility than conventional creatinine-based metrics. We sought to determine the prognostic utility of both urinary neutrophil gelatinase–associated lipocalin and varying creatinine-based metrics of renal injury at multiple time points associated with cardiac surgery.

Design:

Prospective observational study.

Setting:

Academic medical center.

Patients:

Six hundred three adults undergoing cardiac surgery.

Interventions:

Nil.

Measurements and Main Results:

Urinary neutrophil gelatinase–associated lipocalin was measured at baseline and again less than 1 hour, 3 hours, and 18–24 hours after separation from cardiopulmonary bypass. Creatinine-based metrics included a Kidney Disease: Improving Global Outcomes definition of acute kidney injury through 7 days postoperatively as well as ΔS_Cr-initial, defined as the incremental change in S_Cr from baseline to first postoperative measure. Multivariable regression determined the prognostic utility of neutrophil gelatinase–associated lipocalin and creatinine, alone and in combination, for the primary composite outcome of hospital mortality or renal replacement therapy. The primary outcome occurred in 25 patients. Adjusted for covariates ΔS_Cr-initial greater than or equal to 0.0 mg/dL provided early prognostic utility for the primary outcome (odds ratio, 8.9; 95% CI, 3.0–26.6), the odds ratio comparable to a creatinine-based Kidney Disease: Improving Global Outcomes definition of acute kidney injury applied over 7 days postoperatively. The upper quartile of urinary neutrophil gelatinase–associated lipocalin best predicted the primary outcome when measured 18–24 hours post–cardiopulmonary bypass (odds ratio, 18.6; 95% CI, 5.1–68.4; p = 0.001) with earlier post–cardiopulmonary bypass measures of uncertain utility. Combining both ΔS_Cr-initial and neutrophil gelatinase–associated lipocalin measured 3 hours after cardiopulmonary bypass provided excellent early risk stratification for the primary outcome (odds ratio, 18.3; 95% CI, 4.5–75.0).

Conclusions:

Combining urinary neutrophil gelatinase–associated lipocalin with a novel creatinine-based metric, both available soon after completion of surgery, may provide previously unavailable early and effective risk stratification for serious adverse outcomes after cardiac surgery. (Crit Care Med 2015; 43:1043–1052)

A cute kidney injury (AKI) after cardiac surgery represents a major healthcare burden (1–4), with no preventive or therapeutic intervention yet proven effective. One suggested factor contributing to this lack of success has been an inability to accurately detect renal injury in a timely manner, limiting the opportunity for highly targeted early intervention (5). Serum creatinine (S_Cr), the current diagnostic standard for AKI, is widely held to be inadequate due to limited sensitivity, specificity, and a markedly delayed response to injury (6). Despite this, small increases in S_Cr within 48 hours after cardiac surgery have consistently demonstrated an association with adverse outcomes (7–9). Similarly, small increases measured within just 6 hours of surgery have shown prognostic utility for a subsequent creatinine-based diagnosis of AKI. However, the prognostic utility of such early changes in S_Cr to predict important adverse clinical outcomes after cardiac surgery remains unknown.

Neutrophil gelatinase–associated lipocalin (NGAL) is one of the maximally amplified genes within 3 hours of renal ischemia and reperfusion, with a strong biological rationale supporting its potential role as an early indicator of AKI (10–12). However, the diagnostic performance of NGAL for a creati-nine-based diagnosis of AKI has varied dramatically in clinical studies (13–20). Nevertheless, the same limitations that make S_Cr a poor diagnostic tool for AKI also render it a poor surrogate against which to test the validity of new diagnostic markers (21). Importantly, for a new biomarker to change the diagnostic paradigm for AKI, it must identify patients at increased risk for adverse outcomes that are a consequence of AKI with greater accuracy than existing strategies, earlier than existing strategies, or both (22).

We sought to establish the prognostic utility of urinary NGAL, both alone and in combination with S_Cr measured at similarly early postoperative time points, for important adverse clinical outcomes that reflect AKI in a heterogenous cohort of adult patients undergoing cardiac surgery.

MATERIALS AND METHODS

After institutional review board approval, we enrolled adult patients undergoing cardiac surgery between January 2010 and June 2011 in a prospective observational study. In view of the minimal risk represented by collection of urine and routine perioperative data, and in keeping with New York State and U.S. Federal regulations, the institutional review board waived the need for individual patient consent. All types of cardiac surgical procedure were eligible for inclusion with patient selection based on research personnel availability and thus typically restricted to first cases in each operating room, Monday through Thursday.

Following induction of general anesthesia, 10 mL of fresh urine was collected from a urimeter attached to the patient’s Foley catheter for baseline measurement. Collected urine was centrifuged for 5–10 minutes at 2,000g removing cellular debris prior to storage at −80°C. Subsequent urine samples were typically collected within 60 minutes of separation from cardiopulmonary bypass (CPB) or completion of coronary anastomoses where off-pump coronary artery graft surgery was performed, again 3 hours after separation from CPB, and finally 18–24 hours after separation from CPB. Data collection included baseline characteristics as well as perioperative variables potentially associated with AKI or mortality.

At study completion, urinary NGAL was determined by batch analysis using commercially available enzyme-linked immunosorbent assay (ELISA) (R & D Diagnostics, Minneapolis, MN; Avidan HRP from eBioscience, San Diego, CA) by study personnel without formal blinding. Samples in which NGAL was undetectable were assigned a value of 0.5 ng/mL. Where a urine sample was missing, no attempt was made to impute biomarker values. Urinary creatinine was measured by enzymatic endpoint assay (Fisher Diagnostics, Middletown, VA). S_Cr was measured preoperatively and then typically daily for 4 days postoperatively by the central hospital laboratory with additional measures according to clinician judgment. The most recent S_Cr recorded prior to 7 AM on the morning of surgery was considered baseline. All aspects of clinical care were at the discretion of the treating physician blinded to NGAL values.

We hypothesized that perioperative urinary NGAL would identify patients at increased risk of adverse outcomes reflecting AKI, doing so earlier than varying creatinine-based metrics of renal injury. The primary outcome was the composite of hospital mortality or initiation of renal replacement therapy during the postoperative inpatient period. Secondary outcomes included individual components of the primary outcome as well as AKI defined according to recent Kidney Disease: Improving Global Outcomes (KDIGO) practice guidelines for AKI (23), omitting oliguric criteria.

Analyses were conducted using Stata 12 (Stata, College Station, TX) and R version 3.1.1 (R Foundation for Statistical Computing, Vienna, Austria). Creatinine-based metrics of renal injury included KDIGO guidelines applied over 7 days postoperatively as well as at earlier time points comparable to those at which NGAL was measured. We further defined two study-specific creatinine-based metrics available at time points comparable to urinary NGAL. ΔS_Cr-initial was calculated as the difference in S_Cr from baseline to first measure after surgery, whereas ΔS_{Cr-peak day 1} was calculated as the difference in S_Cr from baseline to peak measure within 24 hours after surgery. Receiver operator characteristic (ROC) curves were generated for each of these metrics to identify the primary outcome, selecting thresholds for both ΔS_Cr-initial and ΔS_{Cr-peak day 1} against which the prognostic utility of urinary NGAL at each time point could be pragmatically compared.

The study cohort was then split into quartiles according to NGAL values at each time point. This was based on a number of important facts: threshold NGAL values to best identify renal injury are not well established and may be both context-and assay-specific. Furthermore, optimal threshold values may vary with timing of measurement relative to injury while creatinine-based epidemiologic data support an AKI prevalence of around 25% that consistently portends an increased risk of adverse clinical outcomes (24, 25). Multivariable logistic regression determined the association between the upper quartile of urinary NGAL and the primary outcome, whereas both the c-statistic and integrated discrimination improvement (IDI) (26) evaluated the incremental benefit from adding urinary NGAL to existing clinical risk measures. Similar analyses were repeated for creatinine-based metrics of renal injury.

Patients were then categorized as NGAL+/−, defined by the upper quartile of NGAL at each time point, and S_Cr+/− defined by KDIGO AKI status within 7 days postoperatively determining the association between these two markers of renal injury when either discordant or concordant and the primary composite outcome. To better evaluate pragmatic utility, a similar analysis was repeated, again using the upper quartile of NGAL to define NGAL+/− status at each time point, but redefining S_Cr+/− status according to previously identified thresholds for ΔS_Cr-initial and ΔS_{Cr-day 1}, thereby simulating the potential value of combining these two biomarkers were they available in clinical practice at comparable times.

Based on existing data from within our institution, we assumed a 25% prevalence of AKI and 1.5% hospital mortality or RRT among patients without AKI. Using a two-tailed α error of 0.05, a sample size of 600 patients provided power of approximately 90% to detect an 8.0% prevalence of hospital mortality or RRT among patients with urinary NGAL values in the upper quartile (AKI equivalent) compared to 1.5% in patients from the combined lower three NGAL quartiles (non-AKI equivalent). Reporting of results was guided by the STrengthening the Reporting of OBservational studies in Epidemiology statement for reporting observational studies (27).

RESULTS

The study cohort comprised 603 patients for analysis (Fig. 1). The primary outcome occurred in 25 patients (4.1%), 14 of whom received postoperative RRT while 15 patients died in hospital, including 4 after RRT had been initiated. Median time to RRT initiation was 4 days (interquartile range, 2–6 d), with median time to death 12 days (interquartile range, 10–24 d). Outcome (+) patients had poorer baseline renal function, higher Euroscore, and were more likely to undergo surgery deemed nonelective and with longer duration of CPB (Table 1).

Figure 1.

Flow diagram of study participants.

Table 1.

Baseline Characteristics and Risk Factors for the Primary Composite Outcome of Hospital Mortality or Initiation of Renal Replacement Therapy

	Outcome (–)	Outcome (+)
Baseline Characteristics and Outcomes	n = 578	n = 25	P
Demographics
Age	65 (16)	67 (14)	0.63
Male gender	367 (63.5)	13 (52.0)	0.24
Body mass index, kg/m2	27.8 (5.3)	28.4 (5.9)	0.66
Preoperative risk factors
Left ventricular ejection fraction %	53 (12)	44 (19)	0.03
Unstable angina or recent myocardial infarction	143 (24.7)	6 (24.0)	0.9
Diabetes	129 (22.3)	10 (40.0)	0.04
Hypertension	419 (72.5)	18 (72.0)	0.9
Chronic obstructive pulmonary disease	57 (9.9)	2 (8.0)	0.9
Hyperlipidemia	379 (65.6)	13 (52.0)	0.16
Statin use	310 (53.6)	12 (48.0)	0.58
Smoker	44 (7.6)	0 (0.0)	0.25
Euroscorea	6.8 (3.5)	10.5 (3.1)	< 0.001
Preoperative hemoglobin, g/dL	12.9 (1.9)	10.7 (2.1)	< 0.001
Preoperative serum creatinine, mg/dL	1.0 (0.3)	1.6 (0.7)	< 0.001
Preoperative eGFRb, mL/min/1.73 m2	76 (22)	49 (24)	< 0.001
Preoperative eGFR < 60 mL/min/1.73 m2	136 (23.5)	16 (64.0)	< 0.001
Intraoperative/postoperative risk factors
Reoperation	123 (21.3)	11 (44.0)	0.01
Surgical procedure
On-pump CABG	105 (18.2)	1 (4.0)	0.16
Off-pump CABG	23 (4.0)	0 (0.0)
Valve ± other	321 (55.5)	15 (60.0)
Aortic root ± other	84 (14.5)	5 (20.0)
Ventricular assist device/extracorporeal membrane oxygenation/transplant/other	45 (7.8)	4 (16.0)
Case urgency
Elective	445 (77.0)	11 (44.0)	< 0.001
Nonelective	133 (23.0)	14 (56.0)
Cardiopulmonary bypass duration	106 (48)	173 (78)	< 0.001
Nadir intraoperative hemoglobin, g/dL	8.3 (1.4)	7.6 (0.9)	< 0.001
Peak intraoperative glucose, mg/dL	189 (47)	226 (65)	0.01
pRBC transfusion, n (%)	224 (38.8)	20 (80.0)	< 0.001
Transfused volume of pRBCs (mL)
Intraoperatively	0 (0–250)	750 (0–1,250)	< 0.001
Within 48 hr postoperatively	0 (0–0)	0 (0–1,300)	< 0.001
Outcomes
Acute kidney injuryc
Total	134 (23.2)	21 (84.0)	< 0.001
Stage 1	108 (18.7)	6 (24.0)
Stage 2	18 (3.1)	2 (8.0)
Stage 3	8 (1.4)	13 (52.0)
Length of stay (d)
ICU	2.3 (1.3–4.3)	10.1 (6.1–21.5)	< 0.001
Hospital	8.5 (6.6–11.6)	19.6 (10.8–23.6)	< 0.001

eGFR = estimated glomerular filtration rate, CABG = coronary artery bypass graft, pRBC = packed RBC.

^aEuroscore calculated from modified variables defined according to data available within the medical record and institutional usage to reasonably reflect Euroscore definitions.

^beGFR calculated according to the Chronic Kidney Disease Epidemiology Collaboration equation.

^cAcute kidney injury defined over 7 days postoperatively according to creatinine-based Kidney Disease: Improving Global Outcomes practice guidelines.

Urine collection was near complete with 2,400 of a possible 2,412 samples (99.5%) available for analysis. At all time points, urinary NGAL was higher in outcome (+) patients compared with outcome (−) patients (Fig. 2; Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/CCM/B221). Thresholds for the upper quartile of NGAL concentration were serially identified at 14, 85, 34, and 41 ng/mL, respectively, for each of the four time points of measurement. The first postoperative S_Cr measurement was available at a known time in 602 patients, measured within 3 hours of ICU admission in 90% of patients. This value was greater than or equal to preoperative S_Cr in 201 patients (33.4%), the incremental change in S_Cr greater in outcome (+) compared with outcome (−) patients even at this early postoperative time point (Fig. 3). ΔS_Cr-initial greater than or equal to 0.0 mg/dL and ΔS_{Cr-peak day 1} greater than or equal to 0.2 mg/dL were identified as novel thresholds against which urinary NGAL within 3 hours of CPB and urinary NGAL measured 18–24 hours after CPB, respectively, could be pragmatically compared or combined.

Figure 2.

Urinary neutrophil gelatinase–associated lipocalin (NGAL) values, absolute (ng/mL) and adjusted for urinary creatinine (Ucr; ng/mg), according to patient outcome of hospital mortality or renal replacement therapy. p ≤ 0.001 at all time points. CPB = cardiopulmonary bypass, IQR = interquartile range.

Figure 3.

Delta serum creatinine (S_Cr) values according to patient outcome of hospital mortality or renal replacement therapy. p ≤ 0.002 at all time points.

At all measured time points, the primary outcome occurred more frequently in patients with urinary NGAL levels in the upper quartile compared with the combined lower three quartiles (Supplemental Table 2, Supplemental Digital Content 1, http://links.lww.com/CCM/B221). Urinary NGAL provided fair-to-good discrimination for the primary composite outcome with area under the receiver operator characteristic curve (ROC_AUC) greatest when measured 18–24 hours post-CPB (ROC_AUC, 0.89; 95% CI, 0.81–0.97) (Supplemental Table 3, Supplemental Digital Content 1, http://links.lww.com/CCM/B221). Adjusted for Euroscore, baseline renal dysfunction (estimated glomerular filtration rate < 60 mL/min/1.73 m²), and CPB duration, NGAL provided its greatest prognostic value when measured 18–24 hours post-CPB (odds ratio [OR], 18.6; 95% CI, 5.1–68.4; p = 0.001). Adjusting for urinary creatinine concentration had an inconsistent effect on prognostic utility while likelihood testing found no interaction between NGAL and baseline renal dysfunction for predicting the primary outcome (Table 2). Adjusted for the same covariates, a KDIGO diagnosis of AKI within 7 days postoperatively was associated with a 10-fold increase in odds for hospital mortality or RRT, but this was attenuated and identified fewer patients when the same criteria were applied over shorter time periods after surgery. By contrast, ΔS_Cr-initial greater than or equal to 0.0 mg/dL provided good prognostic value for the primary outcome (OR, 8.9; 95% CI, 3.0–26.6). Added to established risk factors of baseline renal dysfunction, Euroscore, and duration of CPB, urinary NGAL measured 18–24 hours post-CPB raised the c-statistic of the multivariable model from 0.86 (95% CI, 0.79–0.92) to 0.93 (95% CI, 0.89–0.97; p for difference = 0.01), with IDI suggesting improved risk classification for patients in whom the primary outcome occurred (Supplemental Table 4, Supplemental Digital Content 1, http://links.lww.com/CCM/B221). Earlier postoperative NGAL measures were of less certain benefit.

TABLE 2.

Adjusted^a Odds Ratios for Composite Outcome of Hospital Mortality or Initiation of Renal Replacement Therapy Using Urinary Neutrophil Gelatinase–Associated Lipocalin^b and Varying Creatinine-Based Metrics of Renal Injury

	Composite Outcome				Hospital Mortality		Renal Replacement Therapy
Biomarker	n	OR (95% CI)	P	c-Statistic (95% CI)	OR (95% CI)	P	OR (95% CI)	P
NGAL: upper quartile at each time point
NGALbaseline	148	4.0 (1.5–10.3)	0.005	0.88 (0.81–0.95)	1.8 (0.5–6.1)	0.36	6.7 (1.7–25.6)c	0.006
NGAL< 1 hr post-CPB	150	2.8 (1.1–6.9)	0.03	0.88 (0.82–0.94)	1.8 (0.6–5.8)	0.33	3.3 (1.0–10.7)	0.05
NGAL3hr post-CPB	146	5.2 (1.8–15.7)	0.003	0.88 (0.81–0.95)	3.6 (0.9–15.1)	0.08	10.6 (2.1–52.9)	0.004
NGAL18–24 hr post-CPB	147	18.6 (5.1–68.4)	< 0.001	0.93 (0.89–0.97)d	21.6 (3.7–125)	0.001	10.9 (2.3–51.1)	0.003
NGAL (corrected for urinary creatinine): upper quartile at each time point
NGALbaseline	150	5.9 (2.2–16.2)	0.001	0.89 (0.82–0.95)	5.3 (1.4–19.8)	0.01	12.5 (2.6–59.9)	0.002
NGAL< 1 hr post-CPB	150	3.2 (1.3–8.2)	0.02	0.88 (0.82–0.94)	2.4 (0.7–7.9)	0.16	3.2 (1.0–10.4)	0.05
NGAL3 hr post-CPB	148	4.9 (1.6–14.6)	0.004	0.88 (0.81–0.95)	1.9 (0.5–7.4)	0.35	9.8 (2.0–48.3)	0.005
NGAL18–24 hr post-CPB	150	27.8 (6.0–129)	< 0.001	0.94 (0.90–0.97)d	57.8 (5.3–636)	0.001	22.9 (2.9–183)	0.003
Varying serum creatinine-based metrics of renal injury
AKIKDIGO-ICU admite	16	2.7 (0.6–13.1)	0.21	0.85 (0.79–0.92)	4.4 (0.8–24.5)	0.09	1.7 (0.2–17.4)	0.64
ΔSCr-initialf ≥ 0.0 mg/dL	201	8.9 (3.0–26.6)	< 0.001	0.90 (0.84–0.95)	22.5 (3.8–133)	0.001	5.7 (1.6–20.1)	0.007
AKI KDIGO-24 hrg	67	7.5 (3.0–19.1)	< 0.001	0.89 (0.84–0.95)	5.2 (1.5–17.3)	0.008	9.9 (3.0–33.3)	0.001
ΔSCr-peak day 1h ≥ 0.2 mg/dL	129	5.3 (2.1–13.4)	0.001	0.88 (0.82–0.94)	5.4 (1.5–19.4)	0.009	7.1 (2.1–24.6)	0.002
AKIKDIGOi	155	10.0 (3.2–31.2)	< 0.001	0.92 (0.88–0.96)d	3.4 (0.9–12.1)	0.06	-	-

OR = odds ratio, NGAL = neutrophil gelatinase–associated lipocalin, CPB = cardiopulmonary bypass, AKI = acute kidney injury, KDIGO = Kidney Disease: Improving Global Outcomes, S_Cr = serum creatinine.

^aAdjusted for Euroscore, baseline estimated glomerular filtration rate < 60 mL/min/1.73 m², and duration of CPB. Euroscore calculated from modified variables defined according to data available within the medical record and institutional usage to reasonably reflect Euroscore definitions. c-Statistic for base model: 0.86 (95% CI, 0.79–0.92).

^bUpper quartile at each time point compared to lower three quartiles combined.

^cHosmer-Lemeshow goodness-of-fit testing p < 0.05. For all other models tested, p > 0.05.

^dp < 0.05 when compared with c-statistic of multivariable base model without renal injury marker.

^eAKI defined by creatinine-based KDIGO guidelines using baseline and initial postoperative S_Cr only.

^fΔS_Cr-initial defined as initial postoperative S_Cr – preoperative S_Cr.

^gAKI defined by creatinine-based KDIGO guidelines applied over 24 hr postoperatively only.

^hΔS_{Cr-peak day 1} defined as peak S_Cr within 24 hr postoperatively – preoperative S_Cr.

ⁱAKI defined by creatinine-based KDIGO guidelines over 7 days postoperatively.

Dashes indicate the data was not calculable. No patient receiving renal replacement therapy while in hospital failed to meet creatinine-based criteria for AKI within 7 days postoperatively.

Adjusted for the same covariates, patients designated NGAL+ within 3 hours of CPB but who remained S_Cr–through 7 postoperative days by KDIGO criteria demonstrated no increase in odds for the primary composite outcome compared with NGAL–/S_Cr– patients (Table 3). By contrast, patients identified as both NGAL+ and S_Cr+ at any time post-CPB had a consistent and dramatic increase in odds for in-hospital mortality or RRT.

TABLE 3.

Adjusted^a Odds Ratios for Composite Outcome of Hospital Mortality or Initiation of Renal Replacement Therapy Associated With Neutrophil Gelatinase–Associated Lipocalin and Serum Creatinine Status^b

OR (95% CI)
Biomarker (Sampling Time)	NGAL-/SCr-	NGAL+/SCr-	NGAL-/SCr+	NGAL+/SCr+	c-Statistic (95% CI)
Uncorrected (raw) NGAL
NGAL< 1 hr post-CPB	1.00 (reference)	1.9 (0.2–13.9)	7.6 (1.5–38.8)	24.5 (5.0–122)	0.93 (0.90–0.97)
	n = 344	n = 104	n = 107	n = 46
NGAL3hr post-CPB	1.00 (reference)	4.3 (0.4–49.4)	12.3 (1.3–112)	58.6 (7.0–489)	0.94 (0.90–0.97)
	n = 351	n = 95	n = 97	n = 51
NGAL18–24 hr post-CPB	1.00 (reference)	9.9 (1.0–100)	4.7 (0.4–55.2)	74.5 (9.4–594)	0.95 (0.93–0.98)
	n = 365	n = 82	n = 90	n = 65
Corrected for urinary creatinine
NGAL< 1 hr post-CPB	1.00 (reference)	5.2 (0.5–51.8)	14.4 (1.7–122)	50.8 (6.1–424)	0.94 (0.91–0.97)
	n = 341	n = 107	n = 110	n = 43
NGAL3 hr post-CPB	1.00 (reference)	3.4 (0.3–40.6)	10.8 (1.2–99.9)	53.3 (6.5–440)	0.94 (0.90–0.97)
	n = 351	n = 95	n = 95	n = 53
NGAL18–24 hr post-CPB	1.00 (reference)	6.8 (0.7–69.9)	1.7 (0.1–29.5)	91.4 (11.3–737)	0.96 (0.93–0.99)
	n = 356	n = 91	n = 96	n = 59

OR = odds ratio, NGAL = neutrophil gelatinase–associated lipocalin, S_Cr = serum creatinine, CPB = cardiopulmonary bypass.

^aAdjusted for Euroscore, baseline estimated glomerular filtration rate < 60 mL/min/1.73 m², and duration of CPB. Euroscore calculated from modified variables defined according to data available within the medical record and institutional usage to reasonably reflect Euroscore definitions.

^bNGAL+ defined by upper quartile of urinary NGAL at each time point (NGAL > 85, > 34, and > 41 ng/mL at < 1 hr post-CPB, 3 hr post-CPB, and 18–24 hr post-CPB, respectively; corrected NGAL > 422, > 125, and > 44 ng/mg creatinine at < 1 hr post-CPB, 3 hr post-CPB, and 18–24 hr post-CPB, respectively). S_Cr + defined as acute kidney injury according to creatinine-based Kidney Disease: Improving Global Outcomes practice guidelines applied over 7 days postoperatively.Cr

A pragmatic analysis was then undertaken, redefining S_Cr+ as either ΔS_Cr-initial greater than or equal to 0.0 mg/dL (to combine with either NGAL_{< 1 hr} or NGAL_{3 hr post-CPB}) or ΔS_{Cr-peak day 1} greater than or equal to 0.2 mg/dL (to combine with NGAL_{18–24 hr post-CPB}) simulating what might occur if both biomarkers were readily available in the clinical environment. After adjusting for covariates and at all time points post-CPB, patients defined as NGAL+ and S_Cr+ demonstrated a marked increase in odds for in-hospital death or RRT, the prognostic value of the combination appearing greater than for either biomarker alone, improving the c-statistic of the base model from 0.86 to 0.90–0.94 (p for difference ≤ 0.02 at all time points) (Table 4).

TABLE 4.

Adjusted^a Odds Ratios for Composite Outcome of Hospital Mortality or Initiation of Renal Replacement Therapy Associated With Neutrophil Gelatinase–Associated Lipocalin/Serum Creatinine Status^b

		OR (95% CI)
Biomarker (Sampling Time)	NGAL-/SCr-	NGAL+/SCr-	NGAL-/SCr+	NGAL+/SCr+	c-Statistic (95% CI)
Uncorrected (raw) NGAL
NGAL< 1 hr post-CPB	1.00 (reference)	1.0 (0.2–4.5)	3.5 (0.7–16.5)	13.7 (3.8–48.7)	0.91 (0.85–0.96)c
	n = 313	n = 88	n = 138	n = 61
NGAL3 hr post-CPB	1.00 (reference)	2.3 (0.5–10.9)	2.9 (0.4–19.9)	18.3 (4.5–75.0)	0.90 (0.84–0.96)c
	n = 314	n = 86	n = 134	n = 59
NGAL18–24 hr post-CPB	1.00 (reference)	22.6 (2.6–194)	6.2 (0.5–73.9)	71.6 (8.7–588)	0.94 (0.89–0.98)c
	n = 379	n = 94	n = 76	n = 53
Corrected for urinary creatinine
NGAL< 1 hr post-CPB	1.00 (reference)	2.8 (0.6–12.9)	7.8 (1.5–40.1)	21.6 (4.9–94.5)	0.91 (0.85–0.96)c
	n = 311	n = 90	n = 140	n = 59
NGAL3 hr post-CPB	1.00 (reference)	3.8 (0.7–20.7)	5.8 (0.9–39.5)	27.1 (5.3–140)	0.91 (0.85–0.96)c
	n = 304	n = 96	n = 142	n = 51
NGAL18–24 hr post-CPB	1.00 (reference)	17.0 (1.9–148)	2.5 (0.1–43.6)	90.1 (10.9–743)	0.94 (0.90–0.98)c
	n = 369	n = 104	n = 83	n = 46

OR = odds ratio, NGAL = neutrophil gelatinase–associated lipocalin, S_Cr = serum creatinine, CPB = cardiopulmonary bypass.

^aAdjusted for Euroscore, baseline estimated glomerular filtration rate < 60 mL/min/1.73 m², and duration of CPB. Euroscore calculated from modified variables defined according to data available within the medical record and institutional usage to reasonably reflect Euroscore definitions.

^bNGAL+ defined by upper quartile of urinary NGAL at each time point (NGAL > 85, > 34, and > 41 ng/mL at < 1 hr post-CPB, 3 hr post-CPB, and 18–24 hr post-CPB, respectively; corrected NGAL > 422, > 125, and > 44 ng/mg creatinine at < 1 hr post-CPB, 3 hr post-CPB, and 18–24 hr post-CPB, respectively). S_Cr + defined according to study-specific creatinine-based definitions available at times comparable to urinary NGAL measurement (< 1 hr post-CPB and 3 hr post-CPB S_Cr+ defined as ΔS_Cr-initial ≥ 0.0 mg/dL; 18–24 hr post-CPB S_Cr+ defined as ΔS_{Cr-peak day 1} ≥ 0.2 mg/dL).

^cp < 0.05 when compared with c-statistic of multivariable base model without renal injury marker (0.86; 95% CI, 0.79–0.92).

S_Cr status defined by study-specific creatinine-based definitions available at times comparable to each urinary NGAL measurement.

Defined by creatinine-based KDIGO guidelines AKI occurred in 155 patients (25.7%), 138 of whom had an absolute increase in S_Cr greater than or equal to 0.3 mg/dL within 48 hours postoperatively while S_Cr increased by greater than or equal to 50% above baseline within 7 days postoperatively in 95 patients. Post-CPB urinary NGAL demonstrated limited ability to identify patients with a creatinine-based diagnosis of AKI (ROC_AUC’ 0.59–0.65). Performance did not differ according to baseline renal dysfunction. Adjusted for covariates it was only when measured 18–24 hours post-CPB that urinary NGAL provided any association with a conventional diagnosis of AKI (Supplemental Tables 5–7, Supplemental Digital Content 1 http://links.lww.com/CCM/B221). By contrast, ΔS_Cr-initial greater than or equal to 0.0 mg/dL provided independent prognostic utility for a subsequent KDIGO diagnosis of AKI within 7 days postoperatively (OR, 2.6; 95% CI, 1.7–4.1), the magnitude of which increased when combined with an elevated urinary NGAL measured 3 hours post-CPB (NGAL+/S_Cr+ OR, 3.7; 95% CI, 1.9–7.4).

DISCUSSION

Key Findings

In a heterogenous cohort of adult patients undergoing cardiac surgery, urinary NGAL measured 18–24 hours post-CPB provided independent prognostic utility for important adverse outcomes that may reflect AKI. However, used in isolation, its validity and prognostic utility prior to this time point remain uncertain, limiting its value as an early diagnostic tool. By contrast, when the first postoperative S_Cr measure failed to decline from baseline (ΔS_Cr-initial ≥ 0.0 mg/dL), the odds for hospital mortality or RRT were similar to that seen with an established creatinine-based definition of AKI applied over 7 days postoperatively. Simple combination of this early ΔS_Cr together with urinary NGAL, both measured around the time of postoperative ICU admission, demonstrated powerful prognostic utility for the composite outcome of hospital mortality or RRT. This may represent a previously unavailable opportunity for early identification of patients at highest risk of adverse outcomes, providing the opportunity to institute and evaluate highly targeted renoprotective strategies at a time when they may still be effective.

Relationship With Previous Studies

A small reduction in S_Cr commonly occurs early after cardiac surgery, reported in approximately 50% of patients (9, 28). However, despite small postoperative increases in S_Cr consistently associated with increased mortality (7–9), the 48-hour window typically used to assess these changes is well beyond the diagnostic time frame thought necessary for effective early intervention. More recently, small increases in S_Cr as early as 6 hours after cardiac surgery demonstrated prognostic utility for a creatinine-based diagnosis of AKI through 7 days postoperatively (28) although this highly selected cohort was underpowered to assess important clinical outcomes. The current study supports and extends the evidence for a low threshold ΔS_Cr measurement in the immediate postoperative period to enable early identification of clinically important renal injury in a heterogenous cohort of patients undergoing cardiac surgery.

The existence of an early and valid creatinine-based signal of renal injury within just a few hours of cardiac surgery challenges existing paradigms. However, any increase in S_Cr can only occur through a reduction in creatinine clearance (glomerular filtration rate [GFR]), an increase in creatinine production, or a reduction in apparent volume of distribution for creatinine (total body water). Sophisticated kinetic modeling suggests that the rise in S_Cr following an acute reduction in GFR is relatively slow (29), with any increase likely further delayed and limited by dilution of S_Cr occurring secondary to the increase in total body water typically associated with cardiac surgery or critical illness (30, 31). However, rate of change in S_Cr may also be subject to changes in creatinine production, a variable that remains largely unexplored in the context of cardiac surgery. Although speculative, an increase in creatinine production during cardiac surgery offers a plausible mechanism by which an early postoperative rise in S_Cr (or failure to decline) may reflect even mild reductions in GFR to provide unexpectedly early prognostic utility for AKI and associated adverse outcomes.

Despite numerous studies of NGAL as a potential early marker of AKI, few have been specifically designed and powered to assess the relationship between NGAL and clinically relevant adverse outcomes that are a consequence of AKI. In a pooled analysis of 10 studies, Haase et al (32) compared NGAL with a creatinine-based Risk, Injury, Failure, Loss and End-stage definition of AKI over 7 days and reported a 3.6-fold increase in hospital mortality or RRT for patients designated NGAL+/S_Cr–, suggesting that NGAL may facilitate early detection of subclinical AKI. However, lack of patient-level data together with single NGAL measurements at varied time points and on varying analytic platforms make cautious interpretation of this study essential. In a multicenter study of more than 1,200 adult patients undergoing cardiac surgery, increasing quintiles of both urinary and plasma NGAL were associated with a composite of hospital mortality or RRT (17). Nevertheless, prognostic utility of NGAL and creatinine-based measures of renal injury were not directly compared. The relationship between urinary NGAL and AKI, dialysis, and death was further explored in 591 patients within an ICU (33). However, a minority of these patients underwent cardiac surgery, and the only reported metric of biomarker performance was a c-statistic which, in isolation, does not adequately assess the relationship between a putative AKI biomarker and outcomes such as mortality or dialysis that are not expected to occur in every patient with AKI (34). More recently, Arthur et al (35) demonstrated enhanced prognostic utility for AKI progression by combining a ΔS_Cr term with NGAL and other novel biomarkers. However, in contrast to the current study, the ΔS_Cr term used by these investigators was identified only after S_Cr had already risen sufficiently to define the presence of AKI by conventional criteria with no analysis of the potential for early recognition of evolving injury.

Implications

Despite the typically early rise in NGAL following CPB in the current study, it was only 18–24 hours later that it provided clear and consistent prognostic utility for adverse outcomes with greater magnitude than established creatinine-based definitions of AKI. Such delayed recognition likely misses any window for effective institution of renoprotective strategies and is in stark contrast with the known rapid up-regulation of NGAL in response to injury (11, 16). Although our findings may indicate the importance of continued or additional renal insults beyond the intraoperative period, the remarkable increase in prognostic utility when combining earlier measures of both NGAL and ΔS_Cr may suggest an alternate explanation. NGAL is produced by multiple cell types, including inflammatory cells (10), and exists in at least three molecular formats that may vary with cellular origin (36). Current quantitative assays do not reliably differentiate these varying forms (37–39), and the typical surge in urinary NGAL soon after CPB may comprise significant amounts of extrarenal NGAL, with limited diagnostic utility for AKI. Combining these early measures of NGAL with similarly early ΔS_Cr may effectively select elevated NGAL that is predominantly renal in origin to better reflect true renal injury. However, although previous studies confirm potential elevation of urinary NGAL by leukocyturia (40, 41), a similar effect from the immediate post-CPB systemic inflammatory surge on “nonrenal” NGAL present in urine remains speculative. Alternatively, enhanced prognostic utility from combining NGAL with an early creatinine-based signal may simply reflect the expected improvement in predictive performance of a diagnostic test when pretest probability for the outcome of interest is increased.

Strengths and Limitations

Despite the relatively large size of our cohort, specifically addressing clinical endpoints, the limited number of outcomes and single-center setting preclude definitive conclusions and our findings should be confirmed in a larger multicenter cohort. Limited number of outcomes for analysis also precluded more detailed multivariable modeling or subgroup analyses to explore consistency of prognostic utility across a spectrum of baseline renal function. Nevertheless, covariates included in the multivariable model were selected on the basis of existing data supporting their prognostic relevance for mortality, RRT, or both after cardiac surgery while attempting to avoid model overfitting (42). Inclusion of all-cause mortality in our primary outcome may be criticized as nonspecific for renal injury. However, even mild forms of AKI are consistently associated with increased mortality for reasons that remain unclear, making this an essential endpoint for biomarker evaluation (22). Furthermore, enhanced prognostic utility for a creatinine-based diagnosis of AKI when combining urinary NGAL and ΔS_Cr strengthens the construct that this biomarker combination reflects true renal injury. Although the study cohort is nonconsecutive, the method of case selection is unlikely to have introduced systematic bias. Measurement of NGAL relied on a research-based ELISA technique with an inherent level of measurement error and imprecision. Nevertheless, high correlation between the commercially available ARCHITECT assay (Abbott Diagnostics, Abbot Park, IL) and other ELISA methodology has been demonstrated (17). Although the first postoperative S_Cr measurement occurred across a range of times not precisely matched with urinary NGAL measurement, the tested combinations represent a pragmatic assessment of typical, real-world practice likely achievable in a nonstudy context.

CONCLUSIONS

Urinary NGAL measured 18–24 hours after adult cardiac surgery provides important prognostic information for serious adverse outcomes. Limited prognostic utility prior to this time may be greatly enhanced by combining earlier NGAL measurement with a similarly early and novel ΔS_Cr-based metric of injury to enable highly effective early risk stratification. If confirmed, this novel biomarker combination may offer a previously unavailable early opportunity to institute and evaluate highly targeted renoprotective strategies after cardiac surgery.