Predicting Severity of Liver Disease: Twelve Laboratory Tests Evaluated by Multiple Regression

Abstract

To determine the predictive value of laboratory procedures for severity of liver disease, twelve laboratory tests were evaluated in seventy-two patients with various liver disease and in nine non-liver disease hospitalized cases. A numerical score based on the number and extent of abnormal findings was developed for grading clinical and histological severity. Multiple linear regression, utilizing a forward stepwise selection procedure, was used to find the best combination of laboratory tests for prediction of disease severity. The best predictive model for both clinical and histological severity was found for lecithin cholesterol acyltransferase (LCAT), total plasma cholesterol, alkaline phosphatase and bile acids. Of the routine tests prothrombin time and albumin/globulin were useful if the four-test combination listed above was not used. There was a significant correlation (p <| 0·001) between the clinical and the histological score, confirming validity of clinical scoring. In conclusion, this study shows LCAT to rank as first in predicting severity of liver disease. Cholesterol metabolism appears to be affected by liver disease even more than prothrombin time, albumin and globulins. LCAT and bile acids have a place in routine testing of severity of liver disease.