IL-18 and Urinary NGAL Predict Dialysis and Graft Recovery after Kidney Transplantation

Abstract

Current methods for predicting graft recovery after kidney transplantation are not reliable. We performed a prospective, multicenter, observational cohort study of deceased-donor kidney transplant patients to evaluate urinary neutrophil gelatinase-associated lipocalin (NGAL), IL-18, and kidney injury molecule-1 (KIM-1) as biomarkers for predicting dialysis within 1 wk of transplant and subsequent graft recovery. We collected serial urine samples for 3 d after transplant and analyzed levels of these putative biomarkers. We classified graft recovery as delayed graft function (DGF), slow graft function (SGF), or immediate graft function (IGF). Of the 91 patients in the cohort, 34 had DGF, 33 had SGF, and 24 had IGF. Median NGAL and IL-18 levels, but not KIM-1 levels, were statistically different among these three groups at all time points. ROC curve analysis suggested that the abilities of NGAL or IL-18 to predict dialysis within 1 wk were moderately accurate when measured on the first postoperative day, whereas the fall in serum creatinine (Scr) was not predictive. In multivariate analysis, elevated levels of NGAL or IL-18 predicted the need for dialysis after adjusting for recipient and donor age, cold ischemia time, urine output, and Scr. NGAL and IL-18 quantiles also predicted graft recovery up to 3 mo later. In summary, urinary NGAL and IL-18 are early, noninvasive, accurate predictors of both the need for dialysis within the first week of kidney transplantation and 3-mo recovery of graft function.

Kidney allograft function after transplantation varies from a rapid increase in GFR, causing brisk reductions in serum creatinine (Scr), to primary allograft failure. Defined as the need for dialysis within 1 wk of transplantation, delayed graft function (DGF) occurs in 20 to 33% of deceased-donor kidney transplants (DDKTs).^1–4 Recent strategies for increasing the donor pool include using “extended-criteria donor” (ECD) and “donation after cardiac death” (DCD) kidneys. Both types are associated with higher rates of DGF compared with standard-criteria kidneys.⁵ Thus, with more ECD/DCD transplants, as a strategy to reduce waiting lists, physicians will encounter DGF more frequently.

DGF, predominantly caused by ischemia-reperfusion injury (IRI) from allograft procurement, occurs infrequently in living-donor kidney transplants (LDKTs).The role of IRI in graft survival was first highlighted by Terasaki et al.,⁶ who showed that graft survival was better in LDKTs than DDKTs, regardless of antigen mismatches. Moreover, they showed six-antigen mismatched grafts that functioned on day 1 had better 3-yr survival than perfectly matched grafts with delayed function. Investigators have actively sought preventative/therapeutic techniques aimed at reducing IRI and the need for post-transplant dialysis; however, apart from using hypothermic machine allograft perfusion and minimizing cold ischemia time,⁷ results have been disappointing.

The deleterious effects of DGF in the immediate post-transplant period include increased lengths of stay and total hospital costs primarily because of the need for dialysis.^8,9 We examined the long-term role that DGF plays in patient and graft survival in a recent meta-analysis and showed more than 40% increased risk of graft loss at 1 yr with DGF.¹⁰ Even in patients not dialyzed after transplant, studies have shown poorer long-term outcomes with “slow graft function (SGF)” compared with “immediate graft function (IGF).”^3,11,12

Although most studies agree that DGF is associated with poorer outcomes, multiple DGF definitions are used.^13–17 Current means of diagnosing DGF or SGF, using Scr and urine output (UOP), require knowledge of previous values to interpret, are affected by diuretic use, and can take a number of days to confirm.¹⁸ As in other forms of acute kidney injury (AKI) caused by IRI, this lag in diagnosis has greatly hampered efforts to prevent or treat renal injury in human trials.¹⁹ Additionally, with the exception of IGF, it is often impossible to predict recovery. Noninvasive measurement of accurate biomarkers near injury onset could make postoperative course prediction feasible and may ultimately promote advances in kidney transplantation and ischemic AKI.

Ongoing studies of kidney IRI mechanisms have identified many potential biomarkers. Multiple translational studies have subsequently evaluated over 21 serum and urine biomarkers.²⁰ Three of the most intensely studied include neutrophil gelatinase-associated lipocalin (NGAL), IL-18, and kidney injury molecule-1 (KIM-1). Transplantation involves organ ischemia with eventual reperfusion. This makes DDKT a reliable model to evaluate the role of biomarkers in detecting IRI leading to DGF of variable severity/duration. We therefore conducted a multicenter, prospective-cohort study of patients receiving DDKTs to evaluate the timing and efficacy of using urinary NGAL, IL-18, and KIM-1 for predicting recovery of graft function and the need for dialysis within 1 wk after transplantation.

Results

Study Cohort

Of 92 DDKT recipients enrolled, one had primary graft failure caused by renal artery occlusion immediately after surgery and was excluded. Baseline characteristics of the remaining 91 recipients and their donors are listed (Table 1). Dialysis was required in 34 patients. Uremia was the most frequent indication, with hyperkalemia in less than 10%, and none dialyzed exclusively for volume overload. Of those not dialyzed, 33 had SGF and 24 had IGF. There were no clinically suspected or biopsy-confirmed cases of acute rejection (AR) during the 7-d study period.

Table 1.

Summary of baseline and clinical characteristics in transplant recipients and donors

Variable	Recipient	Variable	Donora
Age (yr)	51.6 ± 11.9	Age (yr)	37.2 ± 17
Male	57 (63)	Male	56 (62)
Race/ethnicity		Race/ethnicity
AA	52 (57)	AA	12 (13)
White	35 (38)	White	71 (78)
Other	4 (4)	Other	0 (0)
BMI	29.5 ± 6.7	BMI	27.9 ± 6.1
Cold ischemia time (h)	17.3 ± 10.1	Warm time (min)	48.6 ± 15.3
Previous transplant	11 (12)	Donor cause of death
Months on dialysis	61 ± 42.9	Anoxia	12 (13)
Mode of dialysis		CVA	35 (38)
Hemodialysis	73 (80)	Head trauma	35 (38)
Peritoneal dialysis	18 (20)	Unknown/other	9 (10)
Cause of ESRD		ECD kidney	41 (45)
Hypertension	30 (33)	DCD kidney	9 (10)
Diabetes	23 (25)	Inotropic support	13 (14)
Polycystic kidney	8 (9)	Hypertension	28 (31)
Other	29 (32)	Diabetes	8 (9)
Induction regimen		HLA mismatches
Thymoglobulin	51 (56)	0	5 (4)
Basiliximab/daclizumab	40 (44)	2–4	38 (37)
Class I PRA%	8.8 ± 22.4	>4	48 (46)
Class II PRA%	5.6 ± 17.7	Any DR mismatchb	70 (77)

Continuous values are mean ± SD; dichotomous values are N (%).

^aSome donors gave two kidneys, resulting in the sum for race/ethnicity being <91.

^bEither one DR mismatch or complete DR mismatch.

AA, African American; BMI, body mass index; CVA, cerebrovascular accident; ESRD, end-stage renal disease; PRA, panel reactive antibody.

Novel Biomarkers: Early Function

There was significant separation of median NGAL and IL-18 levels at all time points between the three groups (Table 2, means shown in Figure 1). At no time point were median KIM-1 values statistically different between groups. Receiver-operating characteristic (ROC) curves showed NGAL and IL-18 on the first postoperative day (POD) were moderately accurate in predicting dialysis, whereas KIM-1 was not (Figure 2; Table 3). There were no significant changes in areas under the curve (AUCs) for NGAL, IL-18, or KIM-1 after normalizing for urine creatinine (Ucr, data not shown).

Table 2.

Median biomarker, urine output, serum creatinine, and estimated GFR results by level of allograft function after transplant

	Time after Transplant	DGF (n = 34)	SGF (n = 33)	IGF (n = 24)	P Value
Urine NGAL (ng/ml)	0 h	483 (169–1498)	268 (105.7–1179)	418.9 (141.8–814.9)	0.013
	6 h	639.8 (339.4–2845)	340 (69.5–1196)	217 (58.1–632.2)	<0.001
	12 h	887 (335.1–2171)	321.9 (53–2109)	180.5 (38.1–858.3)	<0.001
	18 h	1045 (367.1–2304)	242.5 (51.8–1719)	117 (34.1–917.2)	<0.001
	First POD	1035 (95–3143)	248 (22–756.1)	60.5 (15.3–249.2)	<0.001
	Second POD	748.4 (86.5–3995)	122.4 (41.3–906.4)	47 (18.6–107.8)	<0.001
Urine IL-18 (pg/ml)	0 h	272.2 (89.8–510.1)	141.4 (0–473.4)	100.8 (0–437.6)	0.033
	6 h	319.1 (73.2–862.8)	67.8 (0–329)	0 (0–496.9)	0.001
	12 h	229.7 (23.9–842.5)	87.4 (0–336.5)	0 (0–463.2)	0.005
	18 h	383.7 (60.3–835.8)	60.6 (0–362.9)	0 (0–458.6)	0.001
	First POD	176.5 (47.5–687.2)	59.4 (0–318.1)	0 (0–68.7)	<0.001
	Second POD	167.8 (0–816.6)	0 (0–281.3)	0 (0–82.6)	<0.001
Urine KIM-1 (ng/ml)	0 h	0.8 (0–1.75)	0.8 (0.09–5.95)	1.1 (0.4–3.74)	0.267
	6 h	1 (0.25–3.57)	1.5 (0.32–5.56)	1.3 (0.59–3.16)	0.361
	12 h	1.3 (0.5–5.67)	2.1 (1.14–6.61)	2.9 (0.93–5.8)	0.084
	18 h	3.4 (0.78–12.37)	2.9 (1.17–6.44)	3.1 (1.33–9.01)	0.85
	First POD	2.9 (0.98–15.34)	2.8 (1.07–13.82)	2.1 (1.42–6.8)	0.938
	Second POD	2.6 (1.14–13.68)	2.3 (0.4–8.37)	2.9 (1.61–7.49)	0.249
Absolute decrease in Scr	0 h to first POD	0.8 (−1.7–3.7)	0.0 (−0.8–1.3)	2.0 (0.1–3.7)	<0.001
Relative decrease in Scr	(0 h to first POD)/0 h	0.1 (−0.24–0.38)	0.0 (−0.13–0.22)	0.2 (0.01–0.39)	<0.001
UOP	By first POD	24 (71%)	11 (33%)	3 (13%)	<0.001
<1 L	By second POD	27 (79%)	8 (24%)	1 (4%)	<0.001
	By third POD	22 (65%)	3 (9%)	0 (0%)	<0.001
Scr (mg/dl)	3 mo	1.9 (1.2–2.7)	1.6 (1.2–2.3)	1.5 (1.0–2.0)	0.03
GFR (ml/min/1.73 m2)	3 mo	48.8 (27–83)	50.9 (38–77)	61.3 (38–77)	<0.001

Values are medians (10th to 90th percentile) or n (percent of total).

The total time for UOP by the first POD was typically less than 24 h depending on when surgery was completed.

Figure 1.

Mean (and SE) for urinary biomarker values after kidney transplant by level of allograft recovery: (A) urinary NGAL, (B) urinary IL-18, and (C) urinary KIM-1. DGF (solid black line) defined by dialysis within 1 wk of transplant. SGF (dashed red line) defined by creatinine reduction ratio less than 0.7 by day 7 without need for dialysis. IGF (dotted black line) defined by absence of SGF without need for dialysis. POD, postoperative day. *P < 0.05, ^†P < 0.01, and ^‡P < 0.001.

Figure 2.

Receiver-operating characteristic curves for urinary biomarkers and changes in serum creatinine on the first postoperative day for predicting dialysis within 1 wk of kidney transplant. Absolute change of Scr is the difference between Scr within 1 h postoperatively and the first postoperative day Scr. Relative change of Scr is the absolute change divided by Scr within 1 h postoperatively.

Table 3.

Area under the receiver-operating characteristic curve at each time point for urinary biomarkers for predicting dialysis within 1 wk of kidney transplant

	Time after Transplant	AUC (95% CI)
Urine NGAL	0 h	0.68 (0.55–0.81)
	6 h	0.81 (0.70–0.92)
	12 h	0.76 (0.64–0.88)
	18 h	0.78 (0.67–0.89)
	First POD	0.82 (0.72–0.92)
	Second POD	0.84 (0.74–0.94)
Urine IL-18	0 h	0.68 (0.55–0.81)
	6 h	0.76 (0.64–0.88)
	12 h	0.72 (0.59–0.85)
	18 h	0.77 (0.66–0.88)
	First POD	0.82 (0.72–0.92)
	Second POD	0.74 (0.62–0.86)
Urine KIM-1	0 h	0.61 (0.47–0.75)
	6 h	0.60 (0.46–0.74)
	12 h	0.67 (0.53–0.81)
	18 h	0.52 (0.36–0.67)
	First POD	0.50 (0.36–0.64)
	Second POD	0.57 (0.43–0.71)
Urine creatinine	0 h	0.71 (0.54–0.87)
	6 h	0.63 (0.45–0.81)
	12 h	0.65 (0.49–0.82)
	18 h	0.59 (0.42–0.77)
	First POD	0.53 (0.35–0.71)
	Second POD	0.61 (0.44–0.79)

Traditional Markers: Early Function

The absolute and relative fall in Scr on the first POD resulted in AUCs (95% confidence interval [CI]) for dialysis of 0.52 (0.39 to 0.66) and 0.52 (0.39 to 0.65), respectively (Figure 2). Irrespective of diuretic use, UOP <1 L by the first POD occurred more frequently in the DGF group, with a sensitivity of 0.71 and specificity of 0.75 (Tables 2 and 4). As a continuous variable, UOP by the first POD had an AUC of 0.83 (0.74 to 0.91). Ucr as a biomarker resulted in the AUCs listed in Table 3.

Table 4.

Sensitivity, specificity, and predictive values for dialysis within 1 wk of kidney transplant using specific urinary biomarker and urine output cut-off values

	Cut-off value	Sensitivity	Specificity	PPV	NPV
First POD NGAL (ng/ml)	45	0.97	0.26	0.43	0.93
	350	0.77	0.74	0.62	0.85
	800	0.65	0.94	0.86	0.82
First POD IL-18 (pg/ml)	53	0.90	0.58	0.54	0.91
	90	0.83	0.75	0.65	0.89
	390	0.30	0.96	0.80	0.70
UOP by first POD	<1 L	0.71	0.75	0.63	0.81

The total collection time for UOP by the first POD extends from the end of surgery to the start of the first POD, with a positive test indicated by values less than 1 L. Positive tests for urinary NGAL and IL-18 are indicated by values greater than the cut-off listed.

PPV, positive predictive value; NPV, negative predictive values.

Recovery and Follow-Up

Mean Scr was higher throughout the 7-d study period for the upper tertiles of NGAL on the first POD (Figure 3A; P < 0.01). Similarly, mean Scr was higher by the third POD (P = 0.04) in those above versus below the median IL-18 value from the first POD. The upper tertile for NGAL on the first POD corresponded with the highest mean 3-mo Scr (P = 0.04), and the follow-up Scr was higher for those above the median IL-18 value (P = 0.05; Figure 3C).

Figure 3.

(A) Mean serum creatinine over the first 3 d after transplant separated by tertiles of urinary NGAL on the first postoperative day: upper tertile of values (solid black line), middle tertile of values (dashed red line), and lower tertile of values (dotted black line). (B) Mean serum creatinine over the first 3 d after transplant separated by medians of urinary IL-18 on the first postoperative day: upper median (solid black line) and lower median (dotted black line). (C) Mean 3-mo serum creatinine separated by NGAL tertiles (left) and IL-18 medians (right) from the first postoperative day. *P < 0.05, ^†P < 0.01, and ^‡P < 0.001.

Mean Scr was highest and GFR lowest at 3-mo follow-up in those with DGF compared with SGF or IGF (Table 2). Tertiles of Scr from the first POD were not associated with mean Scr values at 3-mo follow-up (P = 0.12), nor were tertiles of UOP by the first POD (P = 0.40).

Multivariate and Subgroup Analyses

Logistic regression, using variables from the first POD to predict DGF, showed increased odds ratios (ORs) for elevated NGAL and IL-18 after adjusting for recipient and donor age, cold ischemia time, Scr, and UOP <1 L. The optimal cut-off points for the two biomarkers were determined by the largest sums of sensitivity and specificity (Table 4). Elevated NGAL and IL-18 levels yielded adjusted ORs (95% CI) of 5.1 (1.14 to 22.8) and 6.8 (1.42 to 32.2), respectively (Table 5). Relative risks are also listed given the outcome occurred in more than 10% of the cohort.²¹ The stepwise addition of NGAL and IL-18 information to the clinical dialysis prediction model improved its accuracy as seen by the improvement in AUC of the combined model. The net reclassification index (NRI) for the fully combined model was 1.1, suggesting an overall improvement in classification of events and nonevents by 110% after adding both biomarkers (P < 0.001).²²

Table 5.

Logistic regression analysis with variables available on the first postoperative day for predicting dialysis within 1 wk of kidney transplant and model accuracy after combining variables

Variable	Adjusted OR (95% CI)	Risk Ratioa (95% CI)	Model	AUC
A—Recipient age (yr)	1.0 (0.93–1.07)	1.0 (0.96–1.04)	A + B + C	0.66
B—Donor age (yr)	1.0 (0.96–1.05)	1.0 (0.97–1.03)
C—Cold ischemia time (h)	1.0 (0.97–1.11)	1.0 (0.98–1.07)
D—First POD Scr (mg/dl)	1.0 (0.81–1.32)	1.0 (0.87–1.18)	A + B + C + D + E	0.81
E—First POD UOP <1 L	2.8 (0.66–12.3)	1.7 (0.75–2.36)
F—First POD NGAL >350 nm/ml	5.1 (1.14–22.8)	2.0 (1.08–2.49)	A + B + C + D + E + F	0.86
G—First POD IL-18 >90 pg/ml	6.8 (1.42–32.2)	2.1 (1.23–2.54)	A + B + C + D + E + G	0.87
Fully combined model = A + B + C + D + E + F + G.				0.88

OR/[(1 − P₀) + (P₀ × OR)], where P₀ = incidence of the outcome in the nonexposed (dialysis within 1 wk of transplant, or 37.4% in this cohort).

^aRisk ratio approximated using the formula described by Zhang and Yu (JAMA 1998;280:1690–1691).

RR, risk ratio.

There were no significant differences in biomarker levels when stratified by DGF/non-DGF status for the following subgroups: recipient or donor race, ECD/DCD kidneys versus standard-criteria kidneys, head trauma as donor cause of death versus all other causes, and more than four HLA mismatches versus no more than four mismatches (not shown). NGAL levels on the first POD were higher in non-DGF recipients who received thymoglobulin induction (n = 30) compared with basiliximab (n = 27; 483.5 ± 762.4 versus 144.9 ± 186.4 ng/ml; P = 0.03) but were no different in recipients with DGF. No other biomarker differences were noted between induction regimens. There were no differences between biomarker levels in DGF patients who received one versus more than one dialysis session (data not shown).

Discussion

These findings expand on our previous observations of urinary biomarkers in DGF as defined by the need for dialysis within 1 wk of kidney transplantation.²³ ROC analysis showed NGAL and IL-18 performed better than the absolute or relative change in Scr, which are arguably the most widely used methods for detecting DGF in current clinical practice. NGAL levels also showed excellent separation at nearly every time point between recipients with DGF, SGF, and IGF. Stratifying by NGAL and IL-18 on the first POD showed lower mean Scr levels several days as well as 3 mo after transplant in the groups with lower biomarker levels. Adding NGAL and IL-18 also significantly improved the risk prediction of the current clinical model for the diagnosis of DGF. Thus, urinary biomarkers on or before the first POD not only predict who will need dialysis within 1 wk but also discriminate between more subtle allograft recovery patterns.

We showed that urinary NGAL and IL-18 are superior to Scr in detecting DGF because of earlier diagnosis. It is noteworthy to acknowledge the biologically plausible role these biomarkers have in the causal mechanisms of IRI. During IRI, the gene for NGAL is significantly upregulated with detectable protein in proliferating cell nuclear antigen-positive tubule cells.^24,25 As a probable iron transporting protein, NGAL may play a primary role in renal survival and recovery and has been used therapeutically in IRI animal models.²⁶ IL-18 is a known mediator of inflammation, with studies showing its role in many tissues including lung, heart, bowel, and cartilage.^27–30 IL-18 is activated by cleavage by caspase-1, and mice deficient in the enzyme are protected from AKI induced by IL-18 injection.³¹ Further study showed the neutrophil-independent role that IL-18 has in ischemic AKI.³² Within the field of AKI biomarker development, NGAL and IL-18 are among the most extensively studied biomarkers to date.^23,33–38

UOP on the first POD also reasonably predicted dialysis, but the commonly used cut-off value of <1 L was not statistically significant in multivariate analysis. UOP on the first POD can be unreliable given the variability in surgery completion times, with timed collections typically <24 h. UOP includes both graft and native kidney output and may fluctuate with intra- and postoperative diuretics. Notably, urinary biomarker concentrations can vary depending on urinary flow, which could make any biomarker appear useful^39–41; however, neither KIM-1 nor Ucr predicted DGF, which argues against urinary flow as the sole explanation for our results.

Despite its previously shown accuracy in diagnosing AKI because of acute tubular necrosis in other populations,^42,43 KIM-1 did not perform well in this immediate posttransplantation cohort. KIM-1 is a type-1 transmembrane protein that is normally undetectable in kidney tissue and urine except with renal cell carcinoma and both ischemic and toxic renal injury.^43–45 Han et al.⁴² showed good accuracy for detecting AKI using KIM-1 in a study of 40 pediatric patients undergoing cardiac surgery, with AUCs of about 0.84. In the kidney transplant population, KIM-1 staining intensity in allograft protocol biopsies up to 1 yr after transplant was shown to be very sensitive for tubular injury, occurring in all cases of injury noted by hematoxylin-eosin (H&E) staining and even 28% of biopsies without tubular injury by H&E stain.⁴⁶ In addition, a study of 24-h urinary excretion of KIM-1 in patients at a median of 6 yr after transplant showed an AUC of 0.71 for predicting graft loss within 5 yr.⁴⁷

Transplantation is a reliable model for IRI, because all transplants have some degree of ischemia caused by procurement, leading to variable levels of injury. Even in patients who did not require dialysis in our cohort, mean KIM-1 levels normalized by Ucr ranged from 1.8 to 4.3 ng/mg. This is similar to the median normalized KIM-1 level of 3.3 ng/mg reported by Han et al.⁴² for 29 patients with AKI after cardiac surgery. Given the apparent sensitivity of KIM-1 for detecting tubular injury, the high levels of KIM-1 seen in the majority of patients immediately after DDKT is understandable. The observed range in KIM-1 levels, however, does not differentiate between ischemic injury resulting in DGF and less severe ischemic injury that does not require dialysis immediately after transplant. Additionally, although the biologic function of KIM-1 is unknown, it could be involved in healing after injury. Thus, high KIM-1 levels after transplant may give mixed messages between the degree of injury and the injury response.

The steady improvement in mean 1- to 3-yr graft survival attributable to modern immunosuppressive regimens has leveled off over the past decade.⁴⁸ This is despite significant improvements in AR rates, which are associated with diminished allograft function and survival.⁴⁹ As the most common cause of DGF, IRI may be involved in this plateau. IRI is a known risk factor for AR, a likely primer of chronic recipient-immunologic response, and a cause of diminished nephron mass leading to hyperfiltration.^2,50 Effective treatments for IRI could reduce the consequences of early graft injury and have the potential to increase long-term graft survival. The importance of biomarker development is clearly seen in this setting because it is a prerequisite step for assessment (as surrogate outcomes) and ultimate utilization of future treatment strategies.

The value of these findings is both a clarification of the definition of DGF and an improvement in early, clinical recognition of kidney injury. Early graft function is the result of a complex interplay between variables related to the recipient, donor, and transplant procedure itself. We cannot reliably predict the course of allograft function after DDKT given these complexities. Nonetheless, few would argue the significant role that IRI plays in the immediate post-transplant course, although many factors certainly contribute. Data from this study showed that urinary biomarkers detect the extent of IRI in the kidney, evident by increasing levels of NGAL and IL-18 across the spectrum of IGF, SGF, and DGF. By measuring biomarkers on the first POD, clinicians could quantify the severity of allograft injury sustained during transplantation, estimate time to recovery, and consider less nephrotoxic immunosuppressive regimens while awaiting later recovery. Furthermore, the appropriate intensity of outpatient follow-up could be indicated by degree of initial allograft injury rather than blanket protocol follow-up. For example, degree of injury, determined by postoperative NGAL levels, could guide protocol-transplant-biopsy decisions in addition to the frequency of early follow-up visits.

A prospective-cohort design and good racial mix, including a large proportion of African-American patients, are the major study strengths. To limit biases, we included DDKT patients from four centers, used a rigorous protocol to prospectively collect urine samples, and performed a retrospective, blinded evaluation of the predetermined biomarkers NGAL, IL-18, and KIM-1. These are the primary components of the proposed PRoBE design for biomarker evaluation as outlined by Pepe et al.⁵¹ The primary limitations are the relatively small sample size, the lack of formal post-transplant dialysis criteria, and the lack of standardized Scr measurements between centers. The latter is less problematic given the fact that our primary endpoint was prediction of dialysis rather than a specific Scr level.

In summary, we showed that urinary NGAL and IL-18 are early, noninvasive, and accurate predictors of the need for dialysis within the first week of kidney transplantation. Further studies are needed to confirm our findings using larger transplant cohorts and perform timed urine collections for absolute biomarker excretion to better understand biomarker variation with induction therapy and other donor and recipient variables. Biomarker validation could be included as an integral part of ongoing intervention trials for DGF. Although animal studies have suggested effective therapies in IRI, human trials of DGF to date have been less promising, likely because of late detection.⁵² If biomarker levels in the immediate post-transplant period can be shown to accurately predict duration of injury and longer-term graft function, they could be very useful as surrogate markers. Ultimately, insight into this area of research may provide support for the early detection and stratification of IRI, which will facilitate future DGF/AKI intervention trials.⁵³

Concise Methods

The Yale-New Haven Hospital Human Investigation Committee and the institutional review boards of all other participating centers approved this study. We obtained informed consent from patients before enrollment and recruited patients that were at least 18 yr old and who were to receive DDKTs. We excluded patients who had primary graft failure related to surgical/anatomic causes.

We obtained baseline characteristics of donors and recipients and information on organ procurement. Variable definitions were consistent with United Network for Organ Sharing variables. Daily measurements of UOP and Scr began on the day of transplantation and continued for 7 d or until discharge if hospitalized less than 7 d. We collected follow-up Scr at 3 mo to evaluate later recovery. Absolute change in Scr was the post-transplant Scr at 0 h minus the first POD Scr. Relative change in Scr was the absolute change divided by Scr at 0 h. DGF was defined by at least one dialysis session within 7 d of transplant. In those who did not require dialysis, SGF was defined as a creatinine reduction ratio (difference between Scr at 0 h and the Scr on day 7 divided by Scr at 0 h) less than 0.7, and IGF was defined as a ratio greater than or equal to 0.7.¹² Need for dialysis, any potential diagnosis of AR within the first week of transplant, and 3-mo Scr were confirmed with chart review. Decisions regarding induction and immunosuppressive therapy, transplant biopsy, and use of dialysis were all made by the clinicians at each institution.

Sample Collection/Laboratory Measurements

We collected 10 ml of urine, every 6 h, immediately after surgery for a total of four samples. We collected urine samples on the first and second PODs. We centrifuged samples at 5000 × g to remove cellular debris, aliquoted supernatants into 1-ml cryovials, labeled each with a random computer-generated barcode, and stored samples at −80°C.

The ELISA for NGAL was performed as previously published.³³ IL-18 was measured using a human IL-18 ELISA kit (Medical and Biologic Laboratories, Nagoya, Japan).⁵⁴ KIM-1 was measured using a previously described ELISA method (MaxiSorp; Nunc, Naperville, IL).⁴³ Ucr was measured by a quantitative colorimetric assay kit (Sigma, St. Louis, MO). Biomarkers were measured by personnel blinded to patient information. The interassay variability for all biomarkers was less than 10%.

Statistical Analyses

All analyses were two-tailed with a significance level of 0.05. We compared biomarker levels between patients with DGF, SGF, and IGF using both parametric (ANOVA) and nonparametric (Kruskal-Wallis) tests. We performed ROC curve analysis to compare the accuracy of NGAL, IL-18, and KIM-1 at each time point for predicting dialysis. The time point with the largest AUC was selected as the optimal time for measuring that biomarker. We stratified the cohort into tertiles for NGAL, Scr, and UOP on the first POD to compare mean Scr concentrations over the study period between levels. We used t tests to compare mean Scr in those above versus below the median value for IL-18 rather than comparing tertiles because of the number of values below the detectable range for IL-18 on the first POD. We did not stratify by KIM-1 because of a lack of statistically significant differences between groups. We performed logistic regression for associations between biomarkers and the need for dialysis while adjusting for recipient and donor age, cold ischemia time, the first POD Scr, and UOP less than 1 L on the first POD. To these clinical variables, we added NGAL and IL-18 individually and then combined to produce dialysis prediction models, which we compared using AUCs. To assess risk reclassification after the addition of biomarkers, we calculated NRIs using the method suggested by Pencina et al.,²² which gives additional information about risk classification for each model and can help clarify the role played by each new marker. By adding a new marker to the model, a certain number of patients may be reclassified as having the outcome when the old model predicted they would not. The opposite may also be true—some classified by the old model as having the outcome would not be at risk using the model that includes the new marker. These reclassification events are scored and combined to produce an NRI with a specific P value. We also performed secondary analyses to investigate potential differences in biomarker excretion between several important subgroups after stratifying for DGF. Analyses were performed using R 2.8 (http://www.r-project.org).

Disclosures

Dr. Parikh is co-inventor on the IL-18 patent issued to university of Colorado. Dr. Devarajan is a co-inventor on the NGAL patent filed by University of Cincinnati.