Abstract
In established acute kidney injury (AKI), serum creatinine poorly differentiates prerenal and intrinsic AKI. A damage-associated nephron biomarker, neutrophil gelatinase-associated lipocalin (NGAL) could be a better discriminator. We tested the hypothesis that urinary NGAL distinguishes intrinsic and prerenal AKI, and tested its performance in the prediction of a composite outcome that included progression to a higher RIFLE (“risk, injury, failure, loss of function, end stage renal disease”) severity class, dialysis, or death. We measured urinary NGAL in 161 hospitalized patients with established AKI using a standardized clinical platform. We excluded 16 patients with postrenal obstruction or insufficient clinical information. Of the remaining 145 patients, 75 patients had intrinsic AKI, 32 patients had prerenal AKI, and 38 patients could not be classified. We found that urinary NGAL levels effectively discriminated intrinsic AKI from prerenal AKI (ROC 0.87, CI 0.81-0.94). An NGAL level >104 μg/L indicated intrinsic AKI (likelihood ratio 5.97), while an NGAL level <47 μg/L made intrinsic AKI unlikely (likelihood ratio 0.2). Patients experiencing the composite outcome had higher median urinary NGAL levels on inclusion (248.2 vs. 68.3 μg/L, p<0.001). In logistic regression analysis, NGAL independently predicted the composite outcome, when corrected for demographics, co-morbidities, creatinine, and RIFLE class. Hence, urinary NGAL is useful in classifying and stratifying patients with established AKI.
Introduction
Acute kidney injury (AKI) is common in hospitalized patients and is increasing in incidence (1, 2). Both the “risk, injury, failure, loss of function, end stage renal disease” (RIFLE) classification from the Acute Dialysis Quality Initiative (ADQI) and the Acute Kidney Injury network (AKIN) definition of AKI are based on serum creatinine levels and urinary output to define the severity of AKI (3, 4). Both classifications are effective in risk stratifying patients, since the more advanced RIFLE or AKIN stages predict poor clinical outcomes. However, serum creatinine levels do not always reflect the severity of renal damage. First, during early AKI even a large decrease in glomerular filtration rate (GFR) may cause only a modest increase in serum creatinine, because creatinine must first accumulate in the blood. Second, underlying chronic kidney disease (CKD) can cause a high serum creatinine level in the absence of AKI. Missing information about the baseline creatinine level often confounds interpretation of an elevated serum creatinine concentration. Third, substantial increases in serum creatinine can be observed in renal hypoperfusion even when the kidneys are structurally intact, resulting in prerenal AKI (5). Patients with prerenal and postrenal AKI must be identified as soon as possible, since their treatment is decidedly different than that of patients with intrinsic AKI (5, 6). Importantly, prerenal AKI is associated with a lower mortality than intrinsic AKI (2). Serum and urine diagnostic indices, including fractional excretion of sodium (FeNa), fractional excretion of urea (FeUrea), and urea creatinine (UC) ratio are commonly determined; however, these indices may be of limited utility in diagnosing intrinsic AKI (5, 7-9). Hence, the diagnostic workup and therapeutic management of patients with established AKI would greatly benefit from a clinical test that facilitates a differential diagnosis of intrinsic and prerenal AKI at an early time point and that helps in stratifying the patient at risk.
Novel biomarkers of renal tubular damage potentially fulfill the criteria of such a test. They include (but are not limited to) neutrophil gelatinase-associated lipocalin (NGAL), kidney injury molecule-1 (KIM-1), liver-type fatty acid binding protein (L-FABP) and interleukin 18 (IL-18) (10). These molecules are produced within the kidney in response to injury and can be detected in plasma or urine. In contrast to conventional markers, such as serum creatinine, blood urea nitrogen, or serum cystatin C, these markers do not reflect kidney function, but instead signify structural damage to cells of the kidney. Consequently, these markers are rapidly detectable in response to injury and their increased levels are independent of a functional deficit. Conversely, a purely functional drop of GFR, as is the case in pure prerenal AKI, is not expected to result in an upregulation of these markers. Hence, the detection of novel biomarkers in the setting of established AKI (by creatinine criteria) may faciliate a differential diagnosis between prerenal and intrinsic AKI and aid in risk-stratifying patients beyond currently available parameters (e. g. creatinine or RIFLE class).
One of the most promising novel biomarkers is NGAL, a 25 kDa protein that is produced in renal tubules in response to structural kidney injury (11, 12) and secreted into the urine (13). In previous studies, NGAL was effective in the early diagnosis of AKI in several clinical settings, including perioperative AKI, contrast-induced AKI, sepsis-associated AKI, and AKI following kidney transplantation (14-19). Importantly, NGAL effectively discriminated patients in the emergency room with intrinsic AKI from those with other diagnoses including prerenal AKI, chronic kidney disease and those who actually had normal renal function (20). Furthermore, several studies suggested that NGAL could be useful in the prediction of poor clinical outcomes in AKI (17, 20-22). In this study, we addressed the performance of urinary NGAL levels in established AKI based on RIFLE criteria. Specifically, we asked whether NGAL would aid in distinguishing intrinsic from prerenal causes and predict an unfavorable clinical course.
Results
Patient characteristics
Of 161 patients initially included into the study for established acute kidney injury (RIFLE-R, -I, or -F) 16 were excluded for postrenal obstruction (n=6) or insufficient clinical information (n=10). The remaining 145 patients were included into the outcome analysis and underwent diagnostic adjudication (Figure 1). At the time of inclusion into the study, 32 patients (22.1%) had RIFLE-R AKI, 65 (44.8%) had RIFLE-I AKI, and 48 (33.1%) had RIFLE-F AKI. 49 patients (33.8%) experienced the composite outcome. Of these, 19 patients (13.1%) displayed a step-up in RIFLE severity class, 18 patients (12.4%) required renal replacement therapy, and 28 patients (19.3%) died.
Figure 1. Study flow diagram.
Patients who experienced the composite outcome did not differ significantly from all other patients with respect to demographics or co-morbidities, although a previous history of CKD (stage 3 or up) tended to be more frequent in patients with an unfavorable clinical course (p=0.097) (Table 1). Serum creatinine and blood urea nitrogen levels at the time of inclusion were significantly higher in patients who experienced the composite outcome compared to all others, whereas RIFLE class, UC ratio, FeNa, and FeUrea were not significantly different.
Table 1.
Patient characteristics, by clinical course and diagnosis
| Characteristic | All patients (n=145) | Clinical course after inclusion: | Diagnosis on inclusion: | |||
|---|---|---|---|---|---|---|
| Composite outcome (n=49) | No aspect of composite outcome (n=94) | Intrinsic AKI (n=75) | Prerenal AKI (n=32) | Unclassifiable (n=38) | ||
|
Demographics
| ||||||
| Mean age (SD), y | 67.7 (14.4) | 65.6 (14.4) | 68.8 (14.4) | 64.3 (15.5)§ | 71.1 (14.2) | 71.5 (10.7) |
|
| ||||||
| Women, n (%) | 59 (40.7%) | 19 (38.8%) | 40 (41.7%) | 27 (36.0%) | 14 (43.8%) | 18 (47.4%) |
|
| ||||||
| Black race, n (%) | 1 (0.7%) | 0 (0%) | 1 (1%) | 0 (0%) | 1 (3.1%) | 0 (0%) |
|
| ||||||
|
Comorbidities
| ||||||
| Congestive heart failure (%) | 101 (69.7%) | 34 (69.4%) | 67 (69.8%) | 42 (56.0%)§ | 26 (81.3%) | 33 (86.8%) |
|
| ||||||
| Diabetes mellitus, n (%) | 42 (29%) | 12 (24.5%) | 30 (31.3%) | 19 (25.3%) | 11 (34.4%) | 12 (31.6%) |
|
| ||||||
| Hypertension, n (%) | 118 (81.4%) | 40 (81.6%) | 78 (81.3%) | 54 (72.0%) | 28 (87.5%) | 36 (94.7%) |
|
| ||||||
| Coronary artery disease, n (%) | 44 (30.3%) | 17 (34.7%) | 27 (28.1%) | 18 (24.0%) | 11 (34.4%) | 15 (39.5%) |
|
| ||||||
| Peripheral artery disease, n (%) | 21 (14.5%) | 9 (18.4%) | 12 (12.5%) | 10 (13.3%) | 6 (18.8%) | 5 (13.2%) |
|
| ||||||
| Cerebrovascular disease, n (%) | 13 (9%) | 4 (8.2%) | 9 (9.4%) | 7 (9.3%) | 2 (6.3%) | 4 (10.5%) |
|
| ||||||
| CKD stage at baseline (based on eGFR), n (%)b
| ||||||
| stage 1: ≥90 | 43 (29.7%) | 13 (26.5%) | 30 (31.3%) | 28 (37.3%) | 7 (21.9%) | 8 (21.1%) |
|
| ||||||
| stage 2: 60 to 89 | 64 (44.1%) | 19 (38.8%) | 45 (46.9%) | 28 (37.3%) | 11 (34.4%) | 25 (65.8%) |
|
| ||||||
| stage 3: 30 to 59 | 33 (22.8%) | 15 (30.6%) | 18 (18.8%) | 14 (18.7%)§§ | 14 (43.8%) | 5 (13.2%) |
|
| ||||||
| stage 4: 15 to 29 | 5 (3.4%) | 2 (4.1%) | 3 (3.1%) | 5 (6.7%) | 0 (0%) | 0 (0%) |
|
| ||||||
| CKD stage 3 or up at baseline, n (%) | 38 (26.2%) | 17 (34.7%) | 21 (21.9%) | 19 (25.3%) | 14 (43.8%) | 5 (13.2%) |
|
Kidney parameters on inclusiona
| ||||||
| Median serum creatinine baseline, μmol/l (IQR) | 87 (69-105.5) | 89 (68.5-121.5) | 84.5 (69.3-100) | 83 (69-108) | 95.5 (77-115.5) | 79 (66.8-95) |
|
| ||||||
| Median serum creatinine on inclusion, μmol/l (IQR) | 202 (157.5-263.5) | 225 (170-299)* | 188.5 (154.3-245.5) | 233 (190-357)§§§ | 175.5 (141.3-223) | 168 (151.8-220.3) |
|
| ||||||
| Median serum creatinine inclusion/baseline, ratio (IQR) | 2.3 (1.8-3.3) | 2.4 (1.9-3.4) | 2.2 (1.8-3.1) | 2.9 (2.1-3.8)§§§ | 2 (1.7-2.4) | 2 (1.7-2.7) |
|
| ||||||
| Median serum urea/serum creatinine ratio (IQR) | 81.9 (57.2-100.4) | 87.8 (57.4-102.2) | 81.2 (56.3-100) | 70 (47.6-91) | 95.5 (68.8-122.7) | 89.5 (75.4-107.5) |
|
| ||||||
| Median fractional excretion of sodium, % (IQR) | 0.9 (0.3-2.2) | 0.7 (0.2-2.1) | 1 (0.3-2.6) | 1 (0.4-2.7) | 1 (0.3-2.1) | 0.6 (0.2-2.2) |
|
| ||||||
| Median fractional excretion of urea, % (IQR) | 24.9 (15.4-43.3) | 23.9 (12.8-44.5) | 25.1 (16.1-41) | 33.5 (14.9-49.7) | 23.8 (16.3-31.7) | 20.3 (13.9-30.6) |
|
| ||||||
| Median NGAL on inclusion, μg/L (IQR) | 95.4 (37.6-362.2) | 248.2 (78-1010.6)*** | 68.3 (26.9-214) | 255.6 (98.5-872.9)§§§ | 31.3 (15.9-75.5) | 49.3 (29.8-112.1) |
|
| ||||||
| Median NGAL on inclusion, μg/g creatinine (IQR) | 115.7 (35.9-370.9) | 235.4 (103.8-900.6)** | 71.8 (28-220.7) | 273.5 (112.9-890.8)§§ | 36.5 (18.5-75.7) | 51.4 (27-166.2) |
|
Development of kidney parameters after inclusion
| ||||||
| Median serum creatinine 2 days after inclusion, μmol/l (IQR) | 148 (102-242) | 205 (149-366.5)** | 121.5 (95-168.5) | 199 (139-389)§§§ | 116.5 (87.5-152.3) | 111 (84.3-164.3) |
|
| ||||||
| Median peak serum creatinine within 7 days, μmol/L (IQR) | 190 (133.8-300.3) | 273 (204.8-436.3)*** | 159 (127-233.3) | 277 (191-422.8)§§§ | 128.5 (113.5-162.8) | 154 (126.3-227.8) |
|
| ||||||
| Median NGAL 2 days after inclusion, μg/L (IQR) | 102.2 (35.1-326.5) | 474.2 (119.4-1250.4)*** | 64.1 (27.1-139.3) | 187.2 (99.1-891.4)§ | 22.1 (10.2-91.8) | 41.2 (28.6-136) |
|
| ||||||
| Median NGAL 2 days after inclusion, μg/g creatinine (IQR) | 113.4 (39.2-482.2) | 483.1 (113.4-1576.8)** | 78.3 (27.9-260.6) | 330.4 (103.1-1517.2)§ | 30.6 (11.4-81.7) | 61.3 (32.2-184.4) |
|
Outcomes
| ||||||
| Step-up in RIFLE class after inclusion | 19 (13.4%) | 14 (19.4%)§§ | 0 (0%) | 5 (13.2%) | ||
|
| ||||||
| Initiation of renal replacement therapy, n (%) | 18 (12.4%) | 14 (18.7%)§ | 1 (3.1%) | 3 (7.9%) | ||
|
| ||||||
| In-hospital mortality, n (%) | 28 (19.3%) | 23 (30.7%)§§ | 1 (3.1%) | 4 (10.5%) | ||
|
| ||||||
| Composite outcome (Unfavorable clinical course), n (%) | 49 (33.8%) | 38 (50.7%)§§§ | 2 (6.3%) | 9 (23.7%) | ||
Units for serum creatinine are in μmol/L. To convert serum creatinine from μmol/l to mg/dl, devide by 88.4
eGFR was calculated using the Modification of Diet in Renal Disease (MDRD) formula
p<0.05,
p<0.01,
p<0.001 between patients with and without outcomes using t-test or chi square test as appropriate
p<0.05,
p<0.01,
p<0.001 between the intrinsic and prerenal AKI groups using t-test or chi square test as appropriate
107 patients were ultimately diagnosed with prerenal AKI (n=32) or intrinsic AKI (n=75). Of patients with intrinsic AKI, 46 patients (61.3%) had a clinical diagnosis of acute tubular necrosis, whereas 29 patients (38.7%) had intrinsic AKI caused by other renal diseases, mainly nephrotoxic AKI (n=9) and acute glomerulonephritis (n=7). 38 patients were unclassifiable, because they could not be clearly attributed to one of the two categories even after thorough inspection of all clinical information. In most of these cases, we could not differentiate prerenal AKI from acute tubular necrosis based on the clinical information (n=24). In other cases, a prerenal event coincided with contrast agent administration (n=10) or other potential nephron-damaging events (n=4).
Patients with prerenal AKI were significantly older than patients with intrinsic AKI (Table 1). In addition, congestive heart failure was significantly more frequent in patients with prerenal AKI. Serum creatinine and RIFLE severity class on inclusion were significantly higher in patients with intrinsic AKI when compared to prerenal AKI. The composite outcome (step-up in RIFLE severity class, dialysis initiation, or mortality) was experienced by 38 patients with intrinsic AKI (50.7%), but only by 2 patients with prerenal AKI (6.3%, p<0.001; Table 1). Of these 2 patients with prerenal AKI and adverse outcomes, one received a single session of hemodialysis in the setting of diuretics-induced prerenal AKI in chronic congestive heart failure, but made a quick recovery of creatinine thereafter. The other patient had diarrhea-induced prerenal AKI and initially recovered kidney function, but died later during hospitalization from an unrelated event. Unclassifiable cases resembled prerenal AKI cases with respect to demographics and comorbidities, but displayed higher peak creatinine levels during follow-up and more unfavorable clinical outcomes, which reflects the fact that they represent neither purely prerenal nor purely intrinsic AKI by clinical criteria (Table 1).
Urinary NGAL levels in the diagnosis of intrinsic vs. prerenal AKI
NGAL levels were significantly higher in patients with a clinical diagnosis of intrinsic AKI when compared with prerenal AKI (p=0.007) (Table 1, Fig. 2). Unclassifiable patients displayed intermediate NGAL levels consistent with the fact that they represent marginal cases. To determine NGAL test characteristics in the diagnosis of intrinsic AKI vs. prerenal AKI, we performed a receiver-operating characteristic (ROC) analysis. Table 2 shows the areas under the ROC curves (AUC-ROC), for different biomarkers of intrinsic AKI, including NGAL levels, serum creatinine at the time of inclusion, RIFLE class at the time of inclusion, UC ratio, FeNa and FeUrea. The AUC of NGAL (0.87, CI 0.81-0.94) was significantly higher than that of any other marker tested (p<0.05) (Fig. 3). We also determined the sensitivities, specificities, positive and negative predictive values and positive and negative likelihood ratios for NGAL and other biomarkers at different cut-off levels (Table 3). We had recently conducted a large multicenter study of patients admitted to emergency rooms, which had indicated that, in the absence of additional clinical information, a urinary NGAL > 104 μg/L was suggestive of intrinsic AKI, while a urinary NGAL level < 47 μg/L was suggestive of normal kidney function, stable chronic kidney disease, or prerenal AKI (24). To validate these results in our cohort with established AKI, we tested these same cutoffs on the current study population. An NGAL level cutoff level at 104 μg/L provided a high specificity (0.88) and a high positive likelihood ratio (5.97) for a diagnosis of intrinsic AKI. This finding indicates that a urinary NGAL > 104 μg/L in a patient with established AKI is suggestive of intrinsic AKI. Conversely, an NGAL cutoff level of 47 μg/L provided a high sensitivity (0.89) and a negative likelihood ratio of 0.2 indicating that an NGAL level below this cutoff is useful in excluding intrinsic AKI. Correction of the NGAL levels for urinary creatinine yielded similar test characteristics to NGAL alone (Table 2, Table 3). Test characteristics of the other markers at percentile-adjusted cutoffs are shown in Table 3. Together, these data indicate that NGAL levels can discriminate intrinsic from prerenal AKI and validate independently derived cut-off levels. Fig. 2 shows boxplots for urinary NGAL and serum creatinine on inclusion and two days after inclusion, stratified by the diagnostic category (intrinsic AKI vs. prerenal AKI), demonstrating that NGAL levels displayed less overlap between the diagnostic groups when compared with serum creatinine.
Figure 2. Box plots of biomarker levels in differential diagnosis of AKI and prediction of outcomes.
Biomarker levels are presented on a log 10 scale. Boxes indicate median, lower and upper quartiles. Lower and upper whiskers represent data within 1.5 inter-quartile ranges of the lower quartile, and within 1.5 inter-quartile ranges of the upper quartile, respectively, of log 10-transformed biomarker levels. Circles represent outliers.
Table 2.
ROC analysis of different biomarkers in the differential diagnosis of AKI and in the prediction of outcomes.
| Testa | Diagnosis of intrinsic AKI (vs. prerenal AKI) | Prediction of composite outcome (unfavorable clinical course) |
|---|---|---|
| Urinary NGAL | 0.87 (0.81-0.94)*** | 0.71 (0.62-0.8)*** |
| Urinary NGAL/urinary creatinine | 0.89 (0.82-0.95)*** | 0.71 (0.62-0.8)*** |
| Serum creatinine | 0.74 (0.63-0.84)***,# | 0.61 (0.51-0.71)*,§ |
| RIFLE class | 0.72 (0.62-0.82)***,# | 0.56 (0.47-0.66)# |
| Fractional excretion of urea | 0.59 (0.48-0.71)# | 0.49 (0.38-0.6)# |
| Fractional excretion of sodium | 0.54 (0.42-0.65)# | 0.45 (0.34-0.55)# |
| Serum urea/serum creatinine ratiob | 0.71 (0.59-0.82)**,# | 0.48 (0.37-0.58)# |
Data represent areas under the receiver-operating characteristic curves (AUC-ROC) with 95% confidence intervals
Smaller test result indicates more positive test
p<0.05 vs null hypothesis: true area=0.5
p<0.01 vs null hypothesis: true area=0.5
p<0.001 vs null hypothesis: true area=0.5
p<0.05 vs. urinary NGAL
p=0.067 vs. urinary NGAL
Figure 3. Urinary NGAL levels and outcomes.
Patients were stratified by NGAL level. Percentages of the subgroups that experience aspects of the composite outcome are shown. p for trend is significant for composite and individual clinical outcomes.
Table 3. Test characteristics of biomarkers in the prediction of intrinsic AKI vs. prerenal AKI at different cutoff levels.
Optimal cutoff levels for NGAL (μg/L) were derived from an independent patient cohort in the emergency room (24). Cutoff levels of the remaining biomarkers represent adjusted percentiles in the current study population to ensure comparability of the results.
| Biomarker | Cutoff level | Sensitivity | Specificity | Positive predictive value | Negative predictive value | Positive likelihood ratio | Negative likelihood ratio |
|---|---|---|---|---|---|---|---|
| Urinary NGAL (absolute level) | >47 μg/L | 0.89 (0.8-0.95) | 0.53 (0.35-0.7) | 0.82 (0.71-0.89) | 0.68 (0.46-0.84) | 1.91 (1.31-2.78) | 0.2 (0.1-0.41) |
| >104 μg/L | 0.75 (0.63-0.84) | 0.88 (0.7-0.96) | 0.93 (0.83-0.98) | 0.6 (0.44-0.73) | 5.97 (2.37-15.08) | 0.29 (0.19-0.43) | |
|
| |||||||
| Urinary NGAL (per urinary creatinine) | >45 μg/g | 0.88 (0.78-0.94) | 0.55 (0.36-0.72) | 0.82 (0.72-0.9) | 0.65 (0.44-0.82) | 1.94 (1.31-2.89) | 0.22 (0.11-0.43) |
| >128 μg/g | 0.72 (0.6-0.81) | 0.87 (0.69-0.96) | 0.93 (0.82-0.98) | 0.56 (0.41-0.7) | 5.55 (2.2-14.01) | 0.33 (0.22-0.47) | |
|
| |||||||
| Serum creatinine | >167 μmol/L | 0.85 (0.75-0.92) | 0.44 (0.27-0.62) | 0.78 (0.67-0.86) | 0.56 (0.35-0.75) | 1.52 (1.1-2.09) | 0.34 (0.18-0.63) |
| >205 μmol/L | 0.63 (0.51-0.73) | 0.66 (0.47-0.81) | 0.81 (0.68-0.9) | 0.43 (0.29-0.58) | 1.82 (1.1-3.03) | 0.57 (0.41-0.79) | |
|
| |||||||
| RIFLE severity class (R=1, I=2, F=3) | 2 | 0.88 (0.78-0.94) | 0.34 (0.19-0.53) | 0.76 (0.65-0.84) | 0.55 (0.32-0.76) | 1.34 (1.03-1.75) | 0.35 (0.17-0.72) |
| 3 | 0.49 (0.38-0.61) | 0.88 (0.7-0.96) | 0.9 (0.76-0.97) | 0.42 (0.31-0.55) | 3.95 (1.53-10.15) | 0.58 (0.46-0.73) | |
|
| |||||||
| Serum urea/serum creatinine ratio (urea {mmol/l}/creatinine {mmol/L}) | <63 | 0.45 (0.34-0.57) | 0.89 (0.7-0.97) | 0.92 (0.76-0.98) | 0.38 (0.26-0.51) | 4.07 (1.36-12.18) | 0.62 (0.5-0.77) |
| <84 | 0.63 (0.51-0.74) | 0.63 (0.42-0.8) | 0.82 (0.69-0.91) | 0.39 (0.25-0.54) | 1.7 (1.01-2.87) | 0.59 (0.42-0.83) | |
NGAL levels in the prediction of an unfavorable clinical course
To determine the test characteristics of NGAL and other biomarkers in predicting the composite outcome (step-up in RIFLE severity class, dialysis initiation, or mortality), we analyzed biomarker levels, scatter, ROC curves, and performed logistic regression analysis. Median NGAL levels on inclusion and 2 days after inclusion were significantly higher in patients, who later experienced the composite clinical outcome, when compared to all others (Table 1, Fig. 2). We stratified patients into four groups of increasing NGAL level (<47 μg/L, 47 to 104 μg/L, 104 to 426 μg/L, >426 μg/L). Cutoffs were determined based on the 60th, 75th, and 90th percentiles of NGAL levels in the above-mentioned emergency room cohort (24). We examined the relationship between the NGAL level and clinical outcomes (step up in RIFLE severity class, dialysis initiation, mortality). We found a progressive increase in outcome frequency with increasing urinary NGAL levels. The associations between NGAL level and all outcomes (composite and individual AKI outcomes) were statistically significant (Figure 3). ROC analyses indicated that NGAL had an AUC-ROC of 0.71 in the prediction of the composite outcome (unfavorable clinical course), while serum creatinine and RIFLE severity class had AUC-ROCs of 0.61 and 0.56, respectively (Table 2).
To analyze the effect of a sustained versus transient elevation of NGAL on clinical outcomes, we analyzed the effect of NGAL levels on inclusion in relation to NGAL levels two days after inclusion. Within patients with an NGAL level >104 μg/l on inclusion, 38 patients still displayed an elevated NGAL >104 μg/L two days after inclusion (sustained elevation of urinary NGAL), while 12 patients had NGAL levels that normalized to <104 μg/L (transient elevation of urinary NGAL). Notably, only two patients with a transiently elevated NGAL excretion (16.7%), but 22 patients with a sustained elevation of NGAL (57.9%) experienced the composite outcome indicating an unfavorable clinical course (p<0.05). These data suggest that a quick normalization of an elevated NGAL level is predictive of a beneficial outcome, while a sustained high NGAL excretion is a predictor of a poor clinical course.
We used logistic regression analysis to analyze how urinary NGAL performed after an adjustment for other clinical predictors of an unfavorable clinical course or other established laboratory tests. We generated a conventional prediction model that corrected for demographics, co-morbidities, serum creatinine, and RIFLE class on inclusion and then added NGAL to the model to test whether it significantly added to the prediction of the composite outcome (Table 4). We found that age and serum creatinine were individual parameters that contributed to the prediction of the composite outcome. NGAL excretion significantly added to the predictive performance of the model (p<0.001). Three measures of the overall performance of the regression model (R square, AUC-ROC and diagnostic accuracy) increased when NGAL was added to the model (Table 4). The adjusted odds ratio of an NGAL > 104 μg/L in the prediction of the composite outcome was 4.2. These data indicate that NGAL levels significantly add to the prediction of an unfavorable clinical course when combined with conventional predictors.
Table 4. Multiple Logistic Regression Models for the predicton of the composite outcome (step-up in RIFLE severity class, dialysis initiation, or mortality) a.
Wald score, p value and odds ratio are parameters to describe the contribution of the individual covariate to the model. R square, AUC-ROC and diagnostic accuracy are measures of the overall performance of the model.
| Biomarker Model: | ||||||
|---|---|---|---|---|---|---|
|
| ||||||
| Parameters | Wald score | p value | Odds Ratiob | R square | AUC-ROCb | Diagnostic accuracy |
| Age (>70 years) | 3 | 0.083 | 0.5 (0.2-1.1) | 0.215 | 0.74 (0.64-0.83) | 74.50% |
|
|
||||||
| Serum creatinine (>205 μmol/L) | 4.6 | 0.031 | 2.3 (1.1-4.9) | |||
| Urinary NGAL (>104 μg/L) | 13.3 | <0.001 | 4.2 (1.9-9.0) | |||
|
| ||||||
| Conventional Model: | ||||||
|
| ||||||
| Parameters | Wald score | p value | Odds Ratiob | R square | AUC-ROCb | Diagnostic accuracy |
|
| ||||||
| Age (>70 years) | 3.3 | 0.068 | 0.5 (0.25-1.1) | 0.093 | 0.65 (0.56-0.75) | 68.30% |
|
|
||||||
| Serum creatinine (>205 μmol/L) | 6.3 | 0.012 | 2.5 (1.2-5.2) | |||
Both models were adjusted for gender, histories of CKD, congestive heart failure, vascular disease, hypertension, diabetes, and RIFLE class
Data in parentheses are 95% confidence intervals
Since these data suggested that urinary NGAL levels may add to the RIFLE severity classification in risk stratifying patients with established AKI, we conducted an exploratory analysis on our data set and stratified patients according to RIFLE severity class and NGAL level (Fig. 4). The analysis revealed that, within the RIFLE-I class, patients with an NGAL >104 μg/L displayed a markedly higher percentage of poor clinical outcomes, compared to those with an NGAL ≤104 μg/L (p=0.001) (Fig. 4). Within the RIFLE-R or RIFLE-F class an NGAL level >104 μg/L was also associated with higher percentages of poor clinical outcomes, but these differences did not reach statistical significance (Fig. 4). It should be noted that our study was originally not powered to analyze subsets of patients by RIFLE severity class. Yet, these data strongly suggest that urinary NGAL may be particularly useful in risk stratifying patients with RIFLE-I AKI.
Figure 4. Sub-stratification of RIFLE severity classes by NGAL.
Patients were stratified by RIFLE severity class (R, I, F) and by NGAL level (cutoff 104 μg/L). Percentages of in-hospital mortality, in-hospital dialysis initiation, or a composite outcome are plotted by RIFLE class and NGAL levels. * p<0.05 vs. NGAL ≤ 104 μg/L.
Since urinary NGAL levels diagnosed intrinsic versus prerenal AKI and were useful in the prediction of an unfavorable clinical course, we asked in a secondary outcome analysis whether urinary NGAL may also be useful in predicting whether established AKI will be transient or sustained. Therefore we constructed a logistic regression model that included conventional markers and NGAL in the prediction of sustained AKI, defined as RIFLE-AKI lasting more than three days. In a conventional prediction model that was adjusted for demographics, comorbidities and established laboratory tests, RIFLE severity class was the single significant predictor of sustained AKI. However, when urinary NGAL was added to RIFLE severity class in a biomarker-assisted model, it independently added to the predictive performance of the model (p<0.001). In addition, three measures of the overall performance of the regression model (R square, AUC-ROC and diagnostic accuracy) increased when NGAL was added to the model (Suppl. Table 1). These data indicate that RIFLE severity class and NGAL level are independent predictors of sustained AKI. This further supports the argument that RIFLE class and biomarker levels can be used in a complementary fashion to improve clinical risk prediction in established AKI.
Discussion
In a patient cohort with established AKI as defined by RIFLE criteria, urinary NGAL at the time of presentation distinguished intrinsic AKI from prerenal AKI and predicted an unfavorable clinical course. ROC analyses indicated that NGAL performed significantly better than conventional laboratory tests in diagnosing intrinsic AKI. Logistic regression analysis indicated that NGAL was an independent predictor of an unfavorable clinical course and markedly improved the performance of the prediction model. In addition, our study suggests that NGAL may help in stratifying subclasses within the RIFLE severity classification, in particular within the RIFLE-I severity class.
While urinary NGAL is well established as a biomarker for the early diagnosis of acute kidney injury, our knowledge about its role in established AKI is limited. Our study is the first to systematically group patients with established AKI into distinct diagnostic categories, namely prerenal and intrinsic AKI, and link these categories with biomarker levels. To avoid bias in the diagnostic adjudications, study physicians were strictly blinded to the results of biomarker analyses and followed predefined guidelines to assign patients to diagnostic categories. To ensure a high quality dataset, patients with diagnostic ambivalence were excluded from the analysis. Our analysis of test characteristics of NGAL excretion indicated a good performance of NGAL in discriminating intrinsic AKI and prerenal AKI according to ROC analyses (AUC-ROC 0.87). NGAL performed significantly better in diagnosing intrinsic AKI than serum creatinine, RIFLE class, UC ratio, FeNa, and FeUrea. We also validated cutoffs derived from an independent emergency room study (24). Analyses of these cutoffs on our patient cohort confirmed that an NGAL ≤47 μg/L made intrinsic AKI decidedly unlikely, while an NGAL >104 μg/L was indicative of intrinsic AKI. These data are also consistent with a previous study carried out in emergency room patients, which showed that NGAL effectively discriminated patients with intrinsic AKI from those with prerenal azotemia, stable chronic kidney disease, and normal renal function (20). Notably, urinary NGAL levels had similar test characteristics whether or not they were corrected for urinary creatinine concentrations.
We were surprised to note the poor performance of several conventional indices used to differentiate intrinsic and prerenal AKI. Most remarkably, FeNa did not significantly discriminate prerenal and intrinsic AKI and FeUrea had only marginal performance. The limited diagnostic utility of these urinary indices had been previously acknowledged (27) and, in our cohort, may be explained in part by the high prevalence of congestive heart failure (69.7%) and hypertension (81.4%), since urinary indices are affected by diuretic therapy.
Our study confirmed the known association of AKI etiology and outcomes. While 33.8% of all patients and 50.7% of patients in the intrinsic AKI group experienced the composite outcome indicating an unfavorable clinical course, only 6.3% of patients with prerenal AKI experienced aspects of the composite outcome. These data show that in our cohort intrinsic AKI is an unfavorable predictor.
We further explored the utility of urinary NGAL in predicting unfavorable clinical outcomes. Several studies have linked high NGAL levels with clinical outcomes, including hemodialysis initiation, progression of AKI, and mortality (20-22, 28, 29). We defined a composite outcome for our study, consisting of a step-up in RIFLE severity class, dialysis initiation, or mortality after inclusion. NGAL performed well in predicting this outcome with an ROC of 0.71. This result is similar to the diagnostic performance of NGAL in the prediction of poor clinical outcomes in other cohorts (30). Importantly, logistic regression analysis indicated that a conventional prediction model that included age and serum creatinine was markedly improved when urinary NGAL level was added to the model. In a secondary endpoint analysis, NGAL was also a major predictor of sustained AKI (RIFLE-AKI lasting more than 3 days) and independently added to a multiple regression model that also included the main conventional predictor of sustained AKI, RIFLE severity class. These results indicate that the addition of NGAL significantly improves conventional prediction models of unfavorable clinical courses even after AKI has been established. This predictive utility of urinary NGAL may be due - at least in part - to its ability to identify intrinsic damage to the nephron.
One of the strengths of our study in comparison to earlier trials is that urinary NGAL levels were determined on a standardized clinical platform that is, or shortly will be, widely available to clinicians. This state-of-affairs overcomes one of the main limitations of earlier studies, which derived their results based on experimental NGAL assays. However, our study also has limitations. First, this study was carried out in a single center. Second, the adjudicators assigned the diagnostic categories intrinsic AKI vs. prerenal AKI based on the clinical information they had available. Thus, these categories could not be based on gold standard measures such as renal biopsy results, but instead relied on creatinine dynamics and on the response to clinical intervention. This resulted in a group of “unclassifiable” patients. Inspection of baseline criteria, outcomes and biomarker levels indicates that this unclassifiable group represents an intermediate group of patients that displays neither “purely prerenal” nor “purely intrinsic AKI”. We also attempted to conduct an outcome analysis on unclassifiable patients only. However, we did not find a significant association of urinary NGAL or serum creatinine with composite or individual outcomes in these patients, which is likely explained by insufficient statistical power in this relatively small group (n=38) with only few clinical outcomes (n=9). Finally, the conclusion that patients of individual RIFLE severity classes can be sub-stratified using urinary NGAL is based on a posthoc analysis of our dataset and clearly requires confirmation by additional studies. Nonetheless, our study clearly indicates a value of urinary NGAL in established AKI beyond that of currently used laboratory tests.
Methods
Patient population
The Charité University Ethics Committee approved the study (EA3/011/08) and written informed consent was obtained. All hospitalized patients meeting the RIFLE classification (either a >50% increase in serum creatinine concentration or a >25% decrease in GFR compared to baseline) (3) were eligible for assessment and were generally enrolled at the time of nephrology consultation. Two study physicians collected urine and blood samples, and obtained a clinical follow-up from the medical records. The study physicians were not involved in managing the patients. Patients were initially included (n=161) when the study physician confirmed AKI. Patients were excluded from the study when an assignment of the baseline creatinine level was not possible, when sufficient follow-up was not available, or when postrenal obstruction was identified as a cause of AKI (n=16).
Sampling and measurement of renal biomarkers
We obtained initial urine and blood samples and a follow up urine sample 2 days after enrollment. We centrifuged urine samples at 3000 rpm for 10 min and stored the supernatants at -80 °C. Urinary NGAL was measured using ARCHITECT® (Abbott Laboratories) technology, which utilizes a non-competitive, two anti-analytical antibody sandwich (23). In this assay, the first antibody is covalently attached to paramagnetic microparticles and the second antibody is covalently attached to acridinium. NGAL measurement was performed using an automated sequence, where urine sample and microparticle reagents were incubated together for 18 min, then particles were washed, and the acridinium labeled antibody was added for 4 min. After a second wash, the acridinium label was triggered by peroxide and base. The chemiluminescent signal was calibrated using known quantities of recombinant NGAL. The maximum NGAL concentrations of standards were 1500 ng/ml and specimens with greater concentrations were diluted to read within the calibration range. Creatinine, sodium, urea, FeNa, FeUrea determinations were part of the routine clinical assessments of AKI patients at our institution and were determined by standard automated methods (Jaffé reaction for creatinine, ion selective electrode for sodium, urease/glutamate based kinetic test for urea). Estimated glomerular filtration rate (eGFR) was calculated with the Modification of Diet in Renal Disease (MDRD) formula.
Diagnosis and Outcomes
Diagnostic adjudication was performed by two clinical consultants, who were instructed to use all available data, including the precipitating events, the time course of serum creatinine concentration, the response to treatment including volume resuscitation, modification of diuretic dosage, the use of vasopressors, immunosuppressive drugs, and the result of renal biopsies when available. Consultants were blinded to NGAL levels, but not to other laboratory tests. Baseline serum creatinine was determined by review of the prior 12 months of the medical record, or if unavailable, baseline serum creatinine was assumed from the hospital course. The most likely causes for the increase in creatinine were recorded and patients were classified into the three categories: prerenal AKI, intrinsic AKI, or unclassifiable.
By the inclusion criterion, all patients displayed RIFLE-AKI on the day of enrollment. Patients were classified as “prerenal AKI”, when the increase in serum creatinine concentration had been caused by factors that compromise renal perfusion, and when creatinine rapidly improved to baseline with volume repletion or improvement in cardiac output within 3 days of directed therapy. In select cases (n=6), patients with a slower normalization of serum creatinine (within 7 days) were scored as prerenal, when the slow course was clearly due to the fact that a more careful hydration regimen was warranted (e. g. in congestive heart failure). In patients with prerenal AKI there had to be no pre-enrollment exposure to insults that result in intrinsic kidney damage (e. g. nephrotoxin exposure).
Patients were classified as “intrinsic AKI” when the increase in serum creatinine concentration had been caused by a potential acute tubular necrosis-inducing event, such as prolonged hypotension, sepsis, systemic inflammatory response syndrome, or other insults associated with structural damage to the kidney. The latter included glomerulonephritis, vasculitis, preeclampsia, interstitial nephritis, contrast nephropathy, multiple myeloma, and thrombotic microangiopathy. In patients with intrinsic AKI serum creatinine levels did not respond appropriately to fluid resuscitation and/or hemodynamic optimization.
When disagreements between the two clinical consultants occurred, they convened to discuss the case. If that resolved the disagreement, the diagnosis was revised according to the outcome of the discussion. However, when a level of disagreement or uncertainty remained, the patient was placed into the “unclassifiable” category. These cases included patients that were exposed to additional clinical events after inclusion that may have affected the serum creatinine time course. Additional ambiguous cases occurred when a potential nephron-damaging event had preceded the onset of RIFLE-AKI, but when creatinine normalized quickly or when the timing of recovery was not clearly discriminatory between prerenal AKI and intrinsic AKI.
The prospective composite outcome of the study was an unfavorable clinical course defined by a step-up in RIFLE severity class (RIFLE-R to RIFLE-I or RIFLE-I to RIFLE-F) within seven days from the time of inclusion, need for in-hospital renal replacement therapy, or in-hospital mortality. The secondary outcome of the study was “sustained AKI” defined as RIFLE-AKI that lasted for more than 3 days.
NGAL cutoff levels
Urinary NGAL cutoff levels analyzed in this study were derived from an independent multi-centric cohort of 1677 patients admitted to emergency departments in the United States and in Germany (24). In that study, we had tested whether urinary NGAL levels effectively discriminated patients with intrinsic AKI from those with prerenal AKI, chronic kidney disease or normal kidney function. We had found that urinary NGAL levels below 47 μg/L (corresponding to the 60th percentile of urinary NGAL levels across the entire cohort) made a diagnosis of intrinsic AKI unlikely, while urinary NGAL levels higher than 104 μg/L (corresponding to the 75th percentile of urinary NGAL levels across the entire cohort) were indicative of intrinsic AKI (24) To validate these cutoffs in an independent cohort, we analyzed test characteristics of these NGAL cutoffs in the current study population. Cutoff levels of the remaining tests (serum creatinine, RIFLE severity class, UC ratio) represent percentiles within the current cohort that correspond to these NGAL cutoffs to ensure comparability among different tests.
Statistical Analysis
For statistical analyses, we used PASW Statistics Version 18.0 or 19.0 (SPSS, Chicago, Illinois) and SAS, version 9.1 (SAS Institute, Cary, North Carolina). We compared continuous variables between two groups using Student’s t-test and categorical variables by using chi-square tests, rejecting the null hypothesis at P<0.05. Data are presented as mean (standard deviation, SD) or median (interquartile range, IQR). We determined ROC curves and calculated area under the curve including 95% confidence intervals. We compared areas under the ROC curves using nonparametric tests (25). We calculated sensitivity, specificity, positive predictive value, and negative predictive value, positive likelihood ratio and negative likelihood ratio including 95% confidence intervals using efficient-score method (corrected for continuity) (26). To determine the association of biomarkers (NGAL, FeNa, FeUrea, serum creatinine, UC ratio), demographic variables (age > 70 years, gender, race), co-morbid conditions (diabetes, hypertension, congestive heart failure, coronary artery disease, peripheral artery disease, cerebrovascular disease, preexisting chronic kidney disease), and RIFLE severity class (1=R, 2=I, 3=F), with the composite outcome of unfavorable clinical course or with the secondary outcome of sustained AKI, we used univariate logistic regression analysis followed by multivariate logistic regression with stepwise entry (forward) or exit (backward) of variables to minimize the number of covariates in the model. We then constructed final models based on variables that remained associated with the prediction of the composite outcome (age > 70 years, creatinine >205 μmol/l, and NGAL>104 μg/L) and with the prediction of sustained AKI (RIFLE severity class, NGAL >104 μg/L). The overall fit of the logistic regression models was tested using Hosmer and Lemeshow test.
Supplementary Material
Supplementary Table 1: Multiple logistic regression models for the prediction of sustained AKI (RIFLE-AKI lasting more than 3 days) a
Acknowledgments
The Deutsche Forschungsgemeinschaft (DFG, Schm1730/2-1 and Schm1730/3-1) supports KMSO with grants-in-aid. We thank Sudarshan Hebbar and Frank Quinn from Abbott diagnostics (Chicago, USA) for their help in sample shipment and for providing NGAL ARCHITECT assay reagents. We thank Sonia La’ulu from ARUP Laboratories (Salt Lake City, Utah, USA) for performing the NGAL assays. KMSO and TLN have consultation agreements with Abbott. Columbia University has licensed the use of NGAL to Abbott Labs and to Biosite/Inverness.
References
- 1.Waikar SS, Curhan GC, Wald R, McCarthy EP, et al. Declining mortality in patients with acute renal failure, 1988 to 2002. J Am Soc Nephrol. 2006;17:1143–1150. doi: 10.1681/ASN.2005091017. [DOI] [PubMed] [Google Scholar]
- 2.Liano F, Pascual J. Epidemiology of acute renal failure: a prospective, multicenter, community-based study. Madrid Acute Renal Failure Study Group. Kidney Int. 1996;50:811–818. doi: 10.1038/ki.1996.380. [DOI] [PubMed] [Google Scholar]
- 3.Bellomo R, Ronco C, Kellum JA, Mehta RL, et al. Acute renal failure - definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Critical care (London, England) 2004;8:R204–212. doi: 10.1186/cc2872. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Mehta RL, Kellum JA, Shah SV, Molitoris BA, et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Critical care (London, England) 2007;11:R31. doi: 10.1186/cc5713. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Lameire N, Van Biesen W, Vanholder R. Acute renal failure. Lancet. 2005;365:417–430. doi: 10.1016/S0140-6736(05)17831-3. [DOI] [PubMed] [Google Scholar]
- 6.Macedo E, Mehta RL. Prerenal failure: from old concepts to new paradigms. Curr Opin Crit Care. 2009;15:467–473. doi: 10.1097/MCC.0b013e328332f6e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Carvounis CP, Nisar S, Guro-Razuman S. Significance of the fractional excretion of urea in the differential diagnosis of acute renal failure. Kidney Int. 2002;62:2223–2229. doi: 10.1046/j.1523-1755.2002.00683.x. [DOI] [PubMed] [Google Scholar]
- 8.Pepin MN, Bouchard J, Legault L, Ethier J. Diagnostic performance of fractional excretion of urea and fractional excretion of sodium in the evaluations of patients with acute kidney injury with or without diuretic treatment. Am J Kidney Dis. 2007;50:566–573. doi: 10.1053/j.ajkd.2007.07.001. [DOI] [PubMed] [Google Scholar]
- 9.Jurado R, Mattix H. The decreased serum urea nitrogen-creatinine ratio. Arch Intern Med. 1998;158:2509–2511. doi: 10.1001/archinte.158.22.2509. [DOI] [PubMed] [Google Scholar]
- 10.Parikh CR, Lu JC, Coca SG, Devarajan P. Tubular proteinuria in acute kidney injury: a critical evaluation of current status and future promise. Ann Clin Biochem. 2010;47:301–312. doi: 10.1258/acb.2010.010076. [DOI] [PubMed] [Google Scholar]
- 11.Mishra J, Ma Q, Prada A, Mitsnefes M, et al. Identification of neutrophil gelatinase-associated lipocalin as a novel early urinary biomarker for ischemic renal injury. J Am Soc Nephrol. 2003;14:2534–2543. doi: 10.1097/01.asn.0000088027.54400.c6. [DOI] [PubMed] [Google Scholar]
- 12.Mori K, Lee HT, Rapoport D, Drexler IR, et al. Endocytic delivery of lipocalin-siderophore-iron complex rescues the kidney from ischemia-reperfusion injury. J Clin Invest. 2005;115:610–621. doi: 10.1172/JCI23056. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Schmidt-Ott KM, Mori K, Li JY, Kalandadze A, et al. Dual action of neutrophil gelatinase-associated lipocalin. J Am Soc Nephrol. 2007;18:407–413. doi: 10.1681/ASN.2006080882. [DOI] [PubMed] [Google Scholar]
- 14.Bagshaw SM, Bennett M, Haase M, Haase-Fielitz A, et al. Plasma and urine neutrophil gelatinase-associated lipocalin in septic versus non-septic acute kidney injury in critical illness. Intensive Care Med. 36:452–461. doi: 10.1007/s00134-009-1724-9. [DOI] [PubMed] [Google Scholar]
- 15.Martensson J, Bell M, Oldner A, Xu S, et al. Neutrophil gelatinase-associated lipocalin in adult septic patients with and without acute kidney injury. Intensive Care Med. 36:1333–1340. doi: 10.1007/s00134-010-1887-4. [DOI] [PubMed] [Google Scholar]
- 16.Mishra J, Dent C, Tarabishi R, Mitsnefes MM, et al. Neutrophil gelatinase-associated lipocalin (NGAL) as a biomarker for acute renal injury after cardiac surgery. Lancet. 2005;365:1231–1238. doi: 10.1016/S0140-6736(05)74811-X. [DOI] [PubMed] [Google Scholar]
- 17.Bennett M, Dent CL, Ma Q, Dastrala S, et al. Urine NGAL predicts severity of acute kidney injury after cardiac surgery: a prospective study. Clin J Am Soc Nephrol. 2008;3:665–673. doi: 10.2215/CJN.04010907. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hirsch R, Dent C, Pfriem H, Allen J, et al. NGAL is an early predictive biomarker of contrast-induced nephropathy in children. Pediatric nephrology (Berlin, Germany) 2007;22:2089–2095. doi: 10.1007/s00467-007-0601-4. [DOI] [PubMed] [Google Scholar]
- 19.Hall IE, Yarlagadda SG, Coca SG, Wang Z, et al. IL-18 and urinary NGAL predict dialysis and graft recovery after kidney transplantation. J Am Soc Nephrol. 2010;21:189–197. doi: 10.1681/ASN.2009030264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Nickolas TL, O’Rourke MJ, Yang J, Sise ME, et al. Sensitivity and specificity of a single emergency department measurement of urinary neutrophil gelatinase-associated lipocalin for diagnosing acute kidney injury. Annals of internal medicine. 2008;148:810–819. doi: 10.7326/0003-4819-148-11-200806030-00003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kumpers P, Hafer C, Lukasz A, Lichtinghagen R, et al. Serum neutrophil gelatinase-associated lipocalin at inception of renal replacement therapy predicts survival in critically ill patients with acute kidney injury. Critical care (London, England) 2010;14:R9. doi: 10.1186/cc8861. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Yang HN, Boo CS, Kim MG, Jo SK, et al. Urine neutrophil gelatinase-associated lipocalin: an independent predictor of adverse outcomes in acute kidney injury. Am J Nephrol. 2010;31:501–509. doi: 10.1159/000309841. [DOI] [PubMed] [Google Scholar]
- 23.Grenier FC, Ali S, Syed H, Workman R, et al. Evaluation of the ARCHITECT urine NGAL assay: assay performance, specimen handling requirements and biological variability. Clin Biochem. 2010;43:615–620. doi: 10.1016/j.clinbiochem.2009.12.008. [DOI] [PubMed] [Google Scholar]
- 24.Nickolas TL, Schmidt-Ott KM, Canetta PA, Maarouf OH, et al. Comparative Analysis of Urinary Biomarkers in the Diagnosis of Acute Kidney Injury and the Prediction of Renal Replacement and Mortality at the Time of Admission to the Hospital. A Multicenter-International Study. Renal Week Meeting of the American Society of Nephrology; Denver, CO, USA. 2010. Abstract TH-FC007. [Google Scholar]
- 25.DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44:837–845. [PubMed] [Google Scholar]
- 26.Newcombe RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med. 1998;17:857–872. doi: 10.1002/(sici)1097-0258(19980430)17:8<857::aid-sim777>3.0.co;2-e. [DOI] [PubMed] [Google Scholar]
- 27.Bagshaw SM, Langenberg C, Bellomo R. Urinary biochemistry and microscopy in septic acute renal failure: a systematic review. Am J Kidney Dis. 2006;48:695–705. doi: 10.1053/j.ajkd.2006.07.017. [DOI] [PubMed] [Google Scholar]
- 28.Haase M, Bellomo R, Devarajan P, Schlattmann P, et al. Accuracy of neutrophil gelatinase-associated lipocalin (NGAL) in diagnosis and prognosis in acute kidney injury: a systematic review and meta-analysis. Am J Kidney Dis. 2009;54:1012–1024. doi: 10.1053/j.ajkd.2009.07.020. [DOI] [PubMed] [Google Scholar]
- 29.Ferguson MA, Vaidya VS, Waikar SS, Collings FB, et al. Urinary liver-type fatty acid-binding protein predicts adverse outcomes in acute kidney injury. Kidney Int. 77:708–714. doi: 10.1038/ki.2009.422. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Devarajan P. Review: neutrophil gelatinase-associated lipocalin: a troponin-like biomarker for human acute kidney injury. Nephrology (Carlton) 2010;15:419–428. doi: 10.1111/j.1440-1797.2010.01317.x. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Table 1: Multiple logistic regression models for the prediction of sustained AKI (RIFLE-AKI lasting more than 3 days) a