Serum Angiopoietin-2 Predicts Mortality and Kidney Outcomes in Decompensated Cirrhosis

Population and Setting

From 2013 to 2017, adults (age 18 years or older) admitted to the Massachusetts General Hospital who met eligibility criteria were enrolled and followed prospectively for 90 days from the time of hospital admission. Patients were included if they had decompensated cirrhosis and AKI. Patients were excluded if they were pregnant, nursing, or had a prior kidney transplant. Three contemporaneous reference groups of patients were included: (1) hospitalized patients with decompensated cirrhosis without AKI, (2) outpatients with cirrhosis and refractory ascites referred for transjugular intrahepatic portosystemic shunt (TIPS) placement, and (3) outpatients with compensated cirrhosis.

Definitions

Diagnosis and cause of cirrhosis were determined by the treating hepatologist. Cause of cirrhosis was deemed multifactorial if there were ≥2 causes felt to be contributing to liver disease. A member of the research team (A.S.A.) performed an in-depth review of the medical records to ensure accuracy. The Acute Kidney Injury Network (AKIN) criteria were used to diagnose AKI.28 Acute kidney injury was subdivided by cause: (1) prerenal AKI (resolution of AKI within 48 hours of volume administration), (2) acute tubular necrosis (ATN; clinical history consistent with ischemic or nephrotoxic AKI and failure to respond to volume administration), (3) HRS (exclusion of other potential causes of AKI based on International Club of Ascites Criteria),1 or (4) other AKI (e.g., serologic or histologic evidence of glomerulonephritis, obstruction). Two members of the study team (A.S.A. and X.V.P.) adjudicated the cause of AKI, with a third member (S.K.) providing a tie-breaking diagnosis, if needed. Patients were classified as having an infection on a clinical basis, using culture and cell count data, or if they received antibiotics.

Data Collection

All data were abstracted from the medical records. MELD and Chronic Liver Failure Consortium Acute-on-Chronic Liver Failure (CLIF-C ACLF) scores were calculated at the time of admission.6, 7, 29, 30 Patients provided serum and urine samples on enrollment and day 5 and 30 after enrollment (as available). Laboratory data were recorded from values closest to enrollment sample collection. Members of the study team did not participate or intervene in the clinical care of patients. Outcomes for this study, including the primary outcome of all-cause mortality, were determined at 90 days following the day of index hospitalization.

Measurement of Serum Angiopoietin-2

Serum Angpt-2 levels were measured using a commercially available quantitative enzyme-linked immunosorbent assay (ELISA; Catalog number DANG20, Human Angiopoietin-2 Quantikine ELISA kit, R and D systems, MN) and reported in ng/mL. The coefficient of variation for this assay was <10 %. All study samples were processed within 4 hours of collection and stored at −80°C, with Angpt-2 assays performed in batch at the end of the study. Results of Angpt-2 testing were not reported to the clinical team or to patients.

Analyses

Patients were divided into tertiles by serum Angpt-2 level at enrollment. The primary outcome of 90-day all-cause mortality among inpatients with decompensated cirrhosis was visualized using a Kaplan-Meier curve and compared across tertiles using a log-rank test. Multiple prespecified Cox regression models were used to evaluate the association between Angpt-2 level as a continuous variable and mortality. These included unadjusted models and models adjusted for MELD score, CLIF-C ACLF score, AKIN stage, cause of AKI, and presence of infection. Independent unadjusted models evaluating MELD and CLIF-C ACLF scores prediction of 90-day mortality were reported as reference. Three sensitivity analyses were performed: (1) excluding prior liver transplant recipients (n = 5), (2) excluding patients with hepatocellular carcinoma (n = 20), and (3) using Fine and Gray regression for a composite outcome of liver transplant or death by 90 days to account for these competing risks.31 Results of Cox models were summarized with hazard ratios (HR) and Wald asymptotic 95% confidence intervals (CI). Continuous variables were presented as median [interquartile range] given nonparametric distribution and categorical variables as percentages. The data were compared using a chi-squared test or a Fisher's exact test (for categorical variables), a Wilcoxon rank sum test (for comparison of differences in continuous variables between two independent groups), or a Pearson correlation test (for two continuous variables, with log transformation for nonparametric data). The incremental improvement of adding Angpt-2 to the existing predictive models of cirrhosis (MELD score, MELD-Na score, and CLIF-C ACLF score) was measured by change in C statistic, category-free net reclassification index (NRI), and integrated discrimination index (IDI).32 SAS version 9.4 (Cary, NC) was used for all analyses. Two-tailed P value < 0.05 was considered statistically significant.

Ethics Statement

The Partners Healthcare Institutional Review Board approved this study. All procedures and practices abide by guidelines set forth by the Declarations of Helsinki and Istanbul. Patients (or their health care designee) provided a written informed consent.

Demographics and Clinical Characteristics

Our study included 191 inpatients. The median length of stay was 15 [8, 23] days. The median time from admission to initial sample collection was 4 [2,7] days. The median age was 57 [49, 64] years, 27% were female, 91% were of white race, and 88% were of non-Hispanic ethnicity (see Table 1). The median admission MELD score was 23 [17, 30] and CLIF-C ACLF score was 44 [39, 50]. Serum creatinine was 1.9 [1.3, 3.0] mg/dL at the time of initial Angpt-2 measurement. Twenty patients (10%) had hepatocellular carcinoma. Sixty-six patients (35%) had evidence of infection. Twenty-eight patients (15%) received a liver transplant during follow-up. Among those with AKI, 48 patients (28%) had stage I AKI, 40 patients (23%) had stage II AKI, 86 patients (49%) had stage III AKI, and 49 patients (26%) required renal replacement therapy during follow-up.

Serum Angiopoietin-2 Levels

Serum Angpt-2 levels were measured in 191 hospitalized patients (176 with cirrhosis and AKI and 15 reference patients with cirrhosis and no AKI). The median enrollment Angpt-2 level was 18.2 [11.2, 26.5] ng/mL. Ninety-three patients also had Angpt-2 measured on day 5, and 40 patients had Angpt-2 measured on day 30. Among the 93 patients with interval samples, Angpt-2 levels were stable between enrollment (19.3 [11.8, 26.4] ng/mL) and day 5 (18.1 [12.5, 27.9]; P = 0.94). Patients were divided into tertiles by Angpt-2 level at enrollment: low tertile (Angpt-2 < 13.5 ng/mL; n = 64), mid tertile (Angpt-2 between 13.5 and 23.0 ng/mL; n = 64), and high tertile (Angpt-2 > 23.0 ng/mL; n = 63). Patients with higher Angpt-2 levels had higher MELD scores (low Angpt-2 tertile 19 [16, 27] vs. mid Angpt-2 tertile 23 [17, 30] vs. high Angpt-2 tertile 26 [22, 33]; P < 0.001) and higher CLIF-C ACLF scores (low tertile 41 [33, 47] vs. mid tertile 44 [39, 50] vs. high tertile 48 [43, 55]; P < 0.001). There was no difference across Angpt-2 tertiles by the etiology of cirrhosis (P = 0.26). Although patients with cirrhosis and AKI had higher MELD scores (24 [18, 31] vs. 18 [15, 22]; P = 0.004) and higher CLIF-C ACLF scores (45 [40, 51] vs. 38 [30, 41]; P < 0.001), Angpt-2 levels were comparable between those with cirrhosis and AKI (17.9 [11.7, 27.5] ng/mL) and those with cirrhosis and no AKI (21.4 [12.6, 22.5] ng/mL; P = 0.89). In outpatient reference groups, Angpt-2 was 7.1 [4.4, 11.1] ng/mL among 18 patients with decompensated cirrhosis referred for TIPS for refractory ascites and was 2.3 [1.6, 3.3] ng/mL among 10 patients with compensated cirrhosis (see Supporting Table S1).

Evaluation of continuous Angpt-2 levels against clinical outcomes are presented in Fig. 1. Higher Angpt-2 was associated with higher AKI stage (stage I 13.4 [9.8, 20.1] ng/mL vs. stage II 20.0 [14.1, 26.2] ng/mL vs. stage III 21.9 [13.0, 29.5] ng/mL; P = 0.002) and the need for renal replacement therapy (no renal replacement 16.5 [11.3, 23.6] ng/mL vs. need for renal replacement 25.1 [13.3, 30.3] ng/mL; P = 0.005). Among those with AKI, 61 patients (35%) were diagnosed with prerenal AKI, 54 patients (31%) were diagnosed with HRS, 30 patients (29%) were diagnosed with ATN, and 10 patients (6%) were diagnosed with other AKI. Angpt-2 was lower in those diagnosed with prerenal AKI (14.3 [9.6, 25.5] ng/mL) compared with HRS (18.6 [13.2, 25.0] ng/mL) and ATN (22.0 [13.9, 31.8] ng/mL; P = 0.02). Angpt-2 was higher with presence of infection (no infection 16.7 [11.4, 24.8] ng/mL vs. infection 22.1 [13.2, 29.5] ng/mL; P = 0.03), and trended toward statistical association with presence of hepatocellular carcinoma (no hepatocellular carcinoma 17.9 [11.7, 25.5] ng/mL vs. hepatocellular carcinoma 20.4 [14.7, 51.1] ng/mL; P = 0.07). To account for heterogeneity in sample collection timing, these analyses were repeated after exclusion of samples in the top quartile of time from admission to sample collection (>7 days). Among the remaining 144 patients, all Fig. 1 analyses remained stable, except for Angpt-2's association with cause of AKI, which crossed the P value threshold to 0.055). Pearson correlation between continuous serum Angpt-2 and several clinical variables (with log transformation for skewed distributions) was reported in Fig. 2, including serum creatinine, total bilirubin, international normalized ratio (INR), aspartate aminotransferase, alanine aminotransferase, and white blood cell count.

During the study period, 66/176 (38%) patients had resolution of AKI (creatinine improved within 0.3 mg/dL of baseline). Patients with AKI resolution had lower initial Angpt-2 levels (14.5 [9.7, 24.8] ng/mL vs. 19.4 [13.0, 28.0] ng/mL; P = 0.001). Of 176 patients, 57 (32%) had progression of AKI (increase in AKI stage or presenting with AKI stage III and later requiring dialysis). Patients with AKI progression had higher initial Angpt-2 levels (21.6 [13.0, 30.0] ng/mL vs. 16.1 [11.3, 25.0] ng/mL; P = 0.03).

Angiopoietin-2 and Mortality

Of 191 patients, 78 (41%) died within 90 days of follow-up. Initial Angpt-2 levels were higher among those who died compared with those who survived (died 21.9 [13.9, 30.3] ng/mL vs. alive 15.2 [9.8, 23.0] ng/mL; P < 0.001). Kaplan-Meier analysis of survival over 90 days by tertiles of Angpt-2 suggested a graded association (Fig. 3; P < 0.001). Clinical characteristics by vital status at 90 days are presented in Table 2. In addition to serum Angpt-2, there were several significant predictors of 90-day mortality in univariate analyses. Comparing alive versus died by 90 days, these included age, presence of hepatocellular carcinoma, MELD score, CLIF-C ACLF score, AKIN stage, need for renal replacement therapy, and cause of AKI. Among 93 patients with interval samples collected, there was no significant change in Angpt-2 between enrollment and day 5 when stratifying by vital status at 90 days: for 51 patients alive at day 90 (enrollment 15.6 [11.4, 29.4] ng/mL vs. day 5 15.0 [8.4, 23.6] ng/mL; P = 0.22) or for 42 patients who died by day 90 (enrollment 20.7 [14.5, 26.1] ng/mL vs. day 5 19.7 [14.4, 37.6] ng/mL; P = 0.15).

Multivariable Cox regression models to evaluate Angpt-2's ability to predict 90-day mortality are presented in Fig. 4, with hazard ratios reflecting risk per 5 ng/mL change in Angpt-2. In unadjusted models, higher serum Angpt-2 was significantly associated with mortality (HR 1.13 [95% CI 1.08, 1.17]; P < 0.001). Similarly, all adjusted models demonstrated a stable association between Angpt-2 level and increased mortality, including adjustment for (1) MELD score (HR 1.12 [1.07, 1.17]; P < 0.001), (2) CLIF-C ACLF score (HR 1.12 [1.07, 1.17]; P < 0.001), (3) AKIN stage (HR 1.12 [1.07, 1.17]; P < 0.001), (4) cause of AKI (HR 1.13 [1.08, 1.18]; P < 0.001), (5) presence of infection (HR 1.13 [1.08, 1.17]; P < 0.001), (6) MELD score and age (HR 1.09 [1.05, 1.15]; P < 0.001), and (7) MELD score, age, and presence of infection (HR 1.09 [1.04, 1.15]; P = 0.001). In sensitivity analyses, Angpt-2 was associated with increased mortality after (1) excluding 5 patients with a history of prior liver transplant (HR 1.13 [1.08, 1.18]; P < 0.001), (2) excluding 20 patients with a history of hepatocellular carcinoma (HR 1.14 [1.09, 1.19]; P < 0.001), and (3) in a competing risk analysis for liver transplant or death (HR 1.13 [1.09, 1.17]; P < 0.001). Exclusion of samples collected >7 days from admission did not change results (HR 1.12 [1.08, 1.17]; P < 0.001), nor did adjusting for time from admission to sample collection (HR 1.13 [1.08, 1.17]; P < 0.001).

To compare the added usefulness of serum Angpt-2 to existing prediction models for 90-day mortality in this population, models containing Angpt-2, MELD, MELD-Na, and CLIF-C ACLF scores were compared using C statistics, category-free NRI, and IDI for their ability to predict 90-day mortality (see Table 3). As independent measures, C statistic for prediction of 90-day mortality was 0.65 for serum Angpt-2, 0.59 for MELD score, 0.57 for MELD-Na score, and 0.64 for CLIF-C ACLF score. Addition of serum Angpt-2 to MELD score significantly improved its discrimination for 90-day mortality by change in C statistic (0.59 to 0.66; P = 0.03), IDI (0.067; P = 0.001), but not category-free NRI (0.22; P = 0.15). Addition of serum Angpt-2 to MELD-Na score significantly improved its discrimination for 90-day mortality by change in C statistic (0.57 to 0.66; P = 0.04) and IDI (0.071; P < 0.001), but not category-free NRI (0.21; P = 0.17).

Angiopoietin-2 Before and After Liver Transplant

Forty patients had serum Angpt-2 levels measured 30 days after their initial sample. Twelve of those patients underwent liver transplantation between measurements. Angpt-2 decreased significantly from 28.2 [16.6, 33.1] ng/mL to 13.5 [6.1, 17.2] ng/mL after liver transplantation (P = 0.002). Among the remaining 28 patients who did not undergo liver transplantation, 9/28 (32%) died by 90 days, and Angpt-2 increased significantly from 13.5 [9.6, 19.7] ng/mL to 19.7 [11.1, 26.9] ng/mL (P = 0.001) over 30 days in this group (see Fig. 5).