Serum Albumin Can Identify Patients With Compensated Cirrhosis With a Good Prognosis

Abstract

Background: Many prognostic studies in cirrhosis were performed without distinguishing between compensated and decompensated patients and/or have evaluated the prognostic role of variables that are not routinely used. The aim was to evaluate predictors of survival in compensated and decompensated cirrhosis separately but in a concurrent cohort and focused on routine clinical variables.

Methods: Secondary analysis of a prospective cohort with cirrhosis collected in a tertiary center between August 2000 and May 2002 and followed until death or April 2006. Univariate, stratified univariate analysis, and multivariate Cox regression analysis were performed. Receiving operating characteristics curves were used to identify the best cutoff of variables predictive of death.

Results: A total of 242 patients were included (122 compensated, 120 decompensated). In a median follow-up of 30 months (range, 6 to 50 mo), 62 (26%) deaths occurred, 24 (20%) in the compensated and 38 (32%) in the decompensated group. In the whole cohort, decompensation was the strongest predictor of death. In the compensated group, age, albumin, and platelets and in the decompensated group model for end-stage liver disease, platelets, and albumin were identified as independent predictors of death. A serum albumin of 4 g/dL was the best cutoff to identify patients at risk for death in the compensated group with a hazard ratio of 13.3 [95% confidence interval, 1.8-98.8] in those with an albumin of <4.0 g/dL.

Conclusions: Albumin is a predictor of death in compensated and decompensated cirrhosis. In compensated cirrhosis a subset patients with particularly good prognosis can be identified. Different predictors were observed in both stages, confirming that compensated and decompensated cirrhosis are 2 separate disease stages.

Figure 1

Kaplan-Meier curves of patients in compensated (continuous line) and decompensated cirrhosis (discontinuous line). Log rank test 19.113, p value <0.001.

Figure 2

ROC curves of albumin for compensated (A) (c-statistic 0.732) and decompensated (B) (c-statistic 0.594) patients. ROC curves of MELD for compensated (C) (c-statistic 0.715) and decompensated (D) (c-statistic 0.717) patients.

Figure 2

ROC curves of albumin for compensated (A) (c-statistic 0.732) and decompensated (B) (c-statistic 0.594) patients. ROC curves of MELD for compensated (C) (c-statistic 0.715) and decompensated (D) (c-statistic 0.717) patients.

Figure 2

ROC curves of albumin for compensated (A) (c-statistic 0.732) and decompensated (B) (c-statistic 0.594) patients. ROC curves of MELD for compensated (C) (c-statistic 0.715) and decompensated (D) (c-statistic 0.717) patients.

Figure 2

ROC curves of albumin for compensated (A) (c-statistic 0.732) and decompensated (B) (c-statistic 0.594) patients. ROC curves of MELD for compensated (C) (c-statistic 0.715) and decompensated (D) (c-statistic 0.717) patients.

Figure 3

Kaplan-Meier curves of compensated patients according to the cut-off albumin (≥ 4 g/dL discontinuous line, < 4 g/dL continuous line). Log rank test 10.904, p value = 0.001.