Factors That Predict 1-Month Mortality in Patients With Pregnancy-Specific Liver Disease

Abstract

Pregnancy-specific liver diseases such as acute fatty liver of pregnancy; hemolysis, elevated liver enzymes and low platelet syndrome; and preeclampsia-associated liver disease are associated with considerable morbidity and mortality. We investigated the ability of the model for end-stage liver disease (MELD) to predict 1-month mortality among patients with pregnancy-specific liver diseases. We also developed and tested a model to predict mortality based on features of pregnancy-specific liver diseases. We performed a retrospective study, analyzing hospital admission, clinical, hematologic, and biochemical data collected from 130 patients with pregnancy-specific liver diseases admitted to the St. John's Medical College Hospital (Bangalore, India) from January 2000 through April 2011. Patients were followed up until 3 months after delivery or death. Logistic regression models were fitted using the MELD score and other variables identified as clinically or statistically significant. The predictive accuracy and calibration of the models were assessed by receiver operating characteristic curves and the Hosmer-Lemeshow goodness-of-fit test. Thirty-two patients (24.6%) died. Mortalities from pregnancy-specific liver diseases within 1 month of admission among patients with MELD scores of 20 to 29, 30 to 39, or 40 or greater were 24.2%, 45.45%, and 90.9%, respectively. Univariate analysis identified encephalopathy, ascites, serum total protein, bilirubin, platelet count, alkaline phosphatase, serum creatinine, and international normalized ratio (INR) as significant variables. Multivariate analysis identified total bilirubin (P < .001) and INR (P < .003) as significant predictors of mortality. MELD score and a model based on only 2 variables (bilirubin level and INR) accurately predicted mortality (C statistics, 0.83 and 0.86, respectively) and were well calibrated (Hosmer-Lemeshow χ(2) = 9.7 [P = .28] and 1.9 [P = .98], respectively). A new logistic model based on only 2 variables (INR and total bilirubin) was comparable with the MELD model in predicting mortality among women with pregnancy-specific liver diseases.