Neutrophil Gelatinase-Associated Lipocalin for Assessment of Acute Kidney Injury in Cirrhosis: A Prospective Study

Study Population

This was a prospective study performed in consecutive patients with cirrhosis admitted at the Liver Unit of Hospital Clínic de Barcelona between December 2013 and July 2016 for an acute decompensation of the disease who either had AKI at admission or developed it during hospitalization. The diagnosis of cirrhosis was based on a combination of clinical, biochemical, ultrasonographic, endoscopic, and/or histological findings.

Exclusion criteria were as follows: age <18 or > 85 years, hepatocellular carcinoma beyond Milan criteria, extrahepatic malignancies, severe comorbidities (i.e., congestive heart failure with a New York Heart Association classification >2, chronic obstructive pulmonary disease with a Global Initiative For Chronic Obstructive Disease classification >2, patients requiring RRT), previous liver or kidney transplantation, human immunodeficiency virus infection, absence of AKI throughout the hospitalization, and lack of written informed consent. Patients admitted to the hospital for scheduled therapeutic or diagnostic procedures were not considered for participation.

Patients with a recurrent episode of AKI during the study period could be considered again as candidates for inclusion if the first episode of AKI had resolved completely. The disposition of patients in the study is shown in Supporting Fig. S1. There were a total of 1,074 admissions in 774 patients for acute decompensation. AKI was diagnosed in 356 admissions (45.2%). The remaining cases either didn't develop AKI (283 admissions) or had other exclusion criteria (435 admissions). Additionally, 36 admissions in 26 patients were excluded because a urine sample could not be collected at diagnosis of AKI. Characteristics of these patients were similar to those of the whole population (data not shown). Therefore, the study population consisted of 320 cases of AKI that occurred in 214 patients. Baseline demographic, clinical, and biochemical characteristics of the population are shown in Supporting Table S1. The protocol was approved by the institutional review board of the hospital, and all patients or relatives, if patients had hepatic encephalopathy, signed an informed consent.

The sample size was based on logistic and availability reasons, and it was expected to gather a cohort of around 300 admissions in this study. With this sample size, it was considered feasible to test a reasonable number of variables in multivariate models (around three to five) to assess the predictive clinical and laboratory data.23-26

Study Objectives and Endpoints

The main objective of the study was to investigate the accuracy of three urinary biomarkers (NGAL, monomeric NGAL [mNGAL], and IL-18) in the prediction of ATN versus other causes of AKI. Secondary objectives were correlations between urinary biomarkers and progression of AKI during hospitalization and 28-day mortality.

Study Protocol and Patient Assessment

On inclusion, data collected from the patients consisted of demographic and clinical data and an assessment of liver and kidney tests, which were then repeated every day in patients in the intensive care unit (ICU) and every 2-3 days in patients on the regular ward.

At inclusion, urinalysis with assessment of proteinuria, red blood cells, leukocytes, or casts was performed in all patients. Urine samples were collected at inclusion and at days 3, 7, and 14 (if patients were still hospitalized) to measure urinary biomarkers. Blood samples were collected to measure plasma renin activity, plasma norepinephrine concentration, and total serum bile acids (BAs) at inclusion. Six urinary biomarkers were investigated: three modern biomarkers (NGAL, mNGAL, and IL-18) and three conventional biomarkers (β2-microglobulin, FeNa, and albumin). Other modern biomarkers, such as kidney injury molecule 1 or liver-fatty acid binding protein 1, were not measured because previous studies in patients with cirrhosis have shown that they have lower accuracy compared with NGAL and IL-18 in the differential diagnosis between ATN and HRS-AKI.16-18 During hospitalization, patients were cared for by a team of nurses and attending physicians, as well as by the research team composed of two experienced hepatologists, a hepatology research fellow, and a research nurse. During hospitalization, particular attention was paid to changes in kidney function, as assessed by serum creatinine (sCr), and identification of the type of AKI. After discharge, patients were followed up until the end of the study period, liver transplantation, or death.

Diagnosis of AKI

AKI was defined according to internationally accepted criteria for patients with cirrhosis, as an increase in sCr ≥0.3 mg/dL with respect to baseline.9 The presence of AKI was assessed at hospital admission and throughout hospitalization. For the diagnosis of AKI at admission, the baseline sCr used was the most recent stable value available in the previous 3 months before admission, as reported elsewhere.9 In patients without sCr available within the previous 3 months, the last stable value available between month 4 and 1 year before admission was used as the baseline. If sCr was not available within 1 year before hospitalization, the first value at hospitalization was considered the baseline sCr. For the diagnosis of AKI during hospitalization, the baseline sCr used was that obtained at admission. Of the 320 cases included in the study, the majority (304, 95%) had baseline sCr available before AKI; in 267 cases, sCr was available within the previous 3 months (median time, 24 days); and in 37 cases, it was available between 3 months and 1 year. Only 16 cases (5%) did not have baseline sCr available, and the value used for AKI assessment was that obtained on admission to the hospital.

AKI was categorized in four stages (stage 1A, 1B, 2, or 3) according to a modified International Club of Ascites (ICA) classification, which divides patients with stage 1 in two categories according to sCr at diagnosis of AKI: 1A = <1.5 mg/dL and 1B = ≥1.5 mg/dL.3, 4, 6 This classification of AKI into four categories is recommended by the most recent European Association for the Study of the Liver clinical guidelines.27

AKI was also classified into four types: ATN, HRS-AKI, hypovolemia-induced AKI, or miscellaneous. The type of AKI was adjudicated by the investigators’ team at the time of diagnosis of AKI on the basis of preestablished definitions (see below). To ensure nonbiased adjudication of AKI classification, an experienced nephrologist (E.P.), unaware of the classifications given for each patient, repeated the process at the end of hospitalization. As for the analysis of the results, the adjudication used was that made by the investigators’ team.

Definitions

Resolution and progression of AKI are defined in the Supporting Information.

Type of AKI

AKI was classified into four types: (1) hypovolemia-induced AKI when there was history of excessive fluid losses (i.e., excessive diuresis due to diuretic therapy with loss of body weight >500 g/day or 1,000 g/day in patients without and with edema, respectively; severe diarrhea) or bleeding (i.e., gastrointestinal bleeding as defined by presence of hematemesis and/or melena) within the days before development of AKI3, 4; (2) HRS-AKI as defined according to the ICA criteria9; (3) ATN as defined according to a combination of clinical and laboratory data, with three out of four of the following criteria: FeNa > 2%, urinary osmolality <400 mOsm/L, urinary sodium > 40 mEq/L, and presence of shock or use of nephrotoxic drugs5-7; and (4) miscellaneous, when patients could not be classified in one of the three previous categories.

Management of AKI

The management of AKI was dependent on AKI stage, type of AKI, and associated conditions following a specific algorithm, which was a modification of the algorithm proposed by the ICA, which takes into consideration the categorization of patients into four stages (1A, 1B, 2, or 3), as mentioned before9 (Fig. 1).

Diuretics were stopped in all patients, and causes of AKI were intensively searched for within the first 24-48 hours of diagnosis of AKI, with particular emphasis on any potential nephrotoxic drugs that patients could have received (nonsteroidal anti-inflammatory drugs, aminoglycosides, vasodilators), the presence of volume depletion (gastrointestinal fluid losses, excessive weight loss during the previous days, or gastrointestinal bleeding), hemodynamic instability, and bacterial infections. Patients with signs of dehydration, gastrointestinal bleeding, or shock were treated with standard measures that included administration of plasma volume expanders (saline in dehydrated patients, packed red blood cells in patients with bleeding, and albumin and vasopressors in those with shock). Patients with spontaneous bacterial peritonitis were treated with antibiotics together with albumin.10 In all other patients, plasma volume expansion was given according to AKI stage; patients with AKI stage 1B or greater were treated with albumin (1 g/kg 20% albumin for 2 days), whereas those with AKI stage 1A did not receive albumin unless AKI progressed to greater stages (Fig. 1). In all patients, kidney function was reassessed at 48 hours after AKI diagnosis (day 3). In patients with AKI resolution or in those in whom AKI decreased to stage 1A, no further therapy for AKI was given. Patients who still had an AKI stage ≥1B at day 3 and who met the criteria of HRS-AKI with sCr greater than 2 mg/dL were treated with terlipressin and albumin.27, 28 No specific therapy was given to patients with HRS-AKI with sCr <2 mg/dL because there is no information about the efficacy and safety of terlipressin in this subset of patients. Patients with an AKI stage ≥1B who did not meet the criteria of HRS-AKI were treated with RRT, if required. Both therapies, terlipressin and RRT, were individualized according to patient status, presence of comorbidities, and indication for transplantation.9, 10

Analytical Methods

See the Supporting Information.

Statistical Analysis

Methods used for statistical analysis are reported in the Supporting Information. Significance was set at two-tailed 0.05 for all analyses, except for in the multivariate models in which variables between 0.05 and 0.1 were not excluded. The statistical analysis was performed using SAS version 9.4 software (SAS Institute, Cary, NC) or Statistical Package for Social Sciences version 20.0 (SPSS Inc., Chicago, IL).

Characteristics of the Study Population

Of the 320 cases, 153 were hypovolemia-induced AKI (48%), 93 were HRS-AKI (29%), 39 were ATN (12%), and 35 were due to miscellaneous causes (11%). There was high concordance between the first and second adjudication of types of AKI (kappa index, 0.82), indicating high reproducibility of classification used for AKI types. Baseline demographic, clinical, and biochemical characteristics of the 320 cases categorized according to the type of AKI are shown in Table 1. In addition, the evolution of renal function throughout hospitalization is shown in Supporting Fig. S2. Of note, there were no statistically significant differences in serum BAs levels among AKI types. The characteristics of patients with miscellaneous AKI are shown in Supporting Table S2. Bacterial infections were very common in the whole group of patients (62%). The types and characteristics of bacterial infections as well as their relationship with the type of AKI are shown in Supporting Table S3. Gastrointestinal bleeding occurred in 44 of the 320 cases (14%). The characteristics of these patients are shown in Supporting Table S4.

Urinary Biomarkers in the Differentiation Between ATN and Other Types of AKI

Table 2 shows the values of the three urinary biomarkers in all patients categorized according to the type of AKI, both at diagnosis (day 1) and at day 3. The urinary levels of NGAL, mNGAL, and IL-18 at day 1 were markedly different across AKI types; patients with ATN had significantly higher values of the three biomarkers compared with patients with hypovolemia-induced AKI and HRS-AKI. Patients with miscellaneous AKI had high values of all three biomarkers, yet these were slightly lower than those of patients with ATN. NGAL and mNGAL had good accuracy in differentiating ATN from other types of AKI (with an area under the receiver operating characteristic curve [AUROC] of 0.80 for both), whereas IL-18 had lower accuracy (AUROC, 0.70).

At day 3, resolution of AKI had occurred in 118 of the 320 cases (37%); only 4 of the 118 patients (3.4%) had ATN. When urinary biomarkers were assessed for differentiating ATN from other types of AKI, in the 202 patients in whom AKI still persisted at day 3, NGAL had the highest diagnostic accuracy, followed by mNGAL, whereas the accuracy of IL-18 was much lower (AUROC of 0.87, 0.80, and 0.66, respectively) (Table 2). Comparison of AUROC curves between day 3 and day 1 showed that the predictive accuracy of NGAL in differentiating ATN from other types of AKI was greater at day 3. In multivariate analysis including values of urinary biomarkers at days 1 and 3, NGAL at day 3 was independently associated with the presence of ATN. The cut-off value of NGAL with the highest diagnostic accuracy for ATN at day 3 was 220 µg/g of creatinine. Most patients with ATN (30 out of 35, 86%) had NGAL above this value, whereas the majority of patients with HRS-AKI or hypovolemia-induced AKI had values below the 220-µg/g threshold (52 of the 59 [88%] and 81 of the 87 [93%], respectively). Combination of the three biomarkers did not significantly improve the predictive accuracy of NGAL at day 3.

Changes in urinary biomarker levels throughout hospitalization were also assessed. Patients with ATN showed markedly higher levels of NGAL during the whole hospitalization, and levels were statistically significant from day 1 to day 7 compared with those of patients with HRS-AKI or hypovolemia-induced AKI (Fig. 2). Patients with miscellaneous AKI had intermediate values. By contrast, IL-18 levels were significantly different between ATN and other AKI types at day 1 but not at days 3, 7, and 14. Changes in mNGAL levels were similar to those of NGAL.

Conventional urinary biomarkers, including β2-microglobulin, FeNa, and albumin, were also assessed, showing significant differences among groups. Patients with ATN had higher values compared with the other groups (Table 3). The predictive accuracy of these biomarkers in differentiating ATN from other types of AKI was significantly lower than that of NGAL and mNGAL (AUROCs at day 3 of 0.63, 0.76, and 0.69, respectively). In multivariate analysis, conventional urinary biomarkers were not independent predictors of ATN. There was no significant correlation between serum BAs and NGAL or any other biomarkers.

Outcome of AKI

The outcome of AKI at the end of hospitalization in the 202 patients with persistent AKI at day 3 was as follows: AKI progressed in 28 (14%), remained in the same stage in 74 (37%), and resolved in 100 (49%) (in 3 of these patients, resolution occurred after a transient progression). Types of AKI in the 31 progressors were as follows: 12 had ATN, 7 had miscellaneous AKI, 6 had HRS-AKI, and 6 had hypovolemia-induced AKI. Factors related to progression are shown in Supporting Table S5. In multivariate analysis, independent predictive factors of progression were NGAL at day 3, plasma norepinephrine, and ATN. The cut-off value of NGAL that best predicted progression of AKI was 280 µg/g of creatinine (AUROC, 0.75). Figure 3 shows the evolution of sCr and time course of NGAL during hospitalization in patients in whom AKI progressed versus those in whom AKI did not progress. Twenty-six patients required RRT during hospitalization. Independent predictive factors of RRT were NGAL at day 3 and ATN. The cut-off value of NGAL that best predicted RRT was 173 µg/g of creatinine (AUROC, 0.77).

Overall, resolution of AKI at the end of hospitalization occurred in 213 (67%) of the 320 cases. There was a clear relationship between AKI type and resolution of AKI. In fact, AKI resolved more frequently in patients with hypovolemia-induced AKI and HRS-AKI (77% and 72%, respectively), compared with miscellaneous cases and ATN (43% and 33%, respectively). Fifteen of the 93 patients with HRS-AKI were treated with terlipressin, and a complete response was achieved in 9 of the 15 patients treated (60%). The remaining patients with HRS-AKI did not receive terlipressin because sCr was <2 mg/dL (n = 58), the patients had contraindications to therapy, or the patients were in terminal condition and were not candidates for transplant (n = 20). There was no relationship between NGAL levels and response to terlipressin (100 µg/g vs. 90 µg/g of creatinine in responders vs. nonresponders, respectively).

Urinary Biomarkers and Mortality

Considering only patients with the first episode of AKI, 62 of the 214 patients (29%) died during the 28-day follow-up period, 9 (4%) underwent transplant, and 6 (3%) were lost to follow-up. The remaining 137 (64%) were alive at the end of 28-day follow-up. Overall, the probability of transplant-free survival from the diagnosis of AKI was 72% at discharge from the hospital and 69% at 28 days.

There was a strong relationship between the 28-day mortality rate and AKI stage at diagnosis: 17%, 28%, 39%, and 42% for stages 1A, 1B, 2, and 3, respectively. As for the type of AKI, the highest 28-day mortality rates were observed in patients with ATN and miscellaneous type, and the lowest rates were observed in patients with HRS-AKI and hypovolemia-induced AKI (64%, 43%, 23%, and 17% respectively; P < 0.001). In patients with persistent AKI at day 3, 28-day mortality was associated with liver, kidney, and circulatory function; leukocyte count; serum BAs; Model for End-Stage Liver Disease (MELD) and Child-Pugh scores; and urinary biomarker levels (Supporting Table S6). Interestingly, mortality was also strongly correlated with AKI resolution at day 3 (27% and 50% with and without resolution of AKI at day 3; P < 0.001). In multivariate analysis, independent predictive factors of 28-day mortality for patients with persistent AKI at day 3 were MELD score and NGAL. The cut-off value of NGAL that best predicted 28-day mortality was 110 µg/g creatinine. The time course of NGAL levels in patients who survived or died during hospitalization is shown in Fig. 4.

In patients with HRS-AKI, NGAL levels were associated with prognostic significance. Patients who survived had significantly lower NGAL levels at diagnosis of AKI compared with patients who died (48 vs. 127 µg/g of creatinine, respectively; P < 0.002). No prognostic significance for NGAL was observed within the group of patients with ATN or hypovolemia-induced AKI.

Internal Validation

We conducted an internal validation using bootstrap simulations. Results reported in Supporting Table S7 show that odds ratio (OR) and AUROC estimates from the constructed models and when using bootstrap simulations were both very similar and led to the same conclusions.