Assessment of prognostic factors in alcoholic liver disease: toward a global quantitative expression of severity

The prognostic significance of a battery of clinical, laboratory, and histological indicators was assessed in relation to mortality risk in a 1-year study of 253 patients with alcoholic liver disease, of whom 51 died within such time. The relative risk associated with each abnormality was calculated. A number of abnormalities was found to be statistically associated with a higher risk of death. Among the clinical abnormalities, these were: collateral circulation, edema, ascites, encephalopathy, spider nevi, anorexia, and weakness. Among the laboratory tests, these were: albumin, bilirubin, hemoglobin, abnormal prothrombin time, and alkaline phosphatase. Two hundred and sixteen of these patients had liver biopsies in which the quantifiable abnormalities were scored. Among the histological findings, the alterations significantly related to mortality were necrosis, Mallory, and inflammation, while the presence of cirrhosis per se did not influence the mortality risk. The relative risk factors for mortality associated with the histological alterations were lower than those derived from clinical or laboratory measurements. The advantage of using only clinical and laboratory items to derive a global, quantitative expression of severity is discussed. The relative mortality risks provided a means of calculating a "unit of severity" for each clinical and laboratory abnormality. A combined clinical and laboratory index (CCLI) results when these mortality-risk units are added. Such a combined index had a quasi-linear relationship with the risk of mortality for the complete population. This method compared well with severity scores derived from computerized, linear step-wise discriminant function (SDF) analysis and from a logistic regression (LR) analysis. The factors chosen to have independent prognostic significance by the SDF analysis were: encephalopathy, albumin, prothrombin time, and hemoglobin, while only encephalopathy, albumin, and hemoglobin were chosen by the LR analysis. Within a range of values, LR can provide a good discrimination in relation to mortality, similar to that observed for the CCLI in its complete range. However, there are some advantages to the CCLI method vs. the LR or SDF analyses. The CCLI is less susceptible to being unduly influenced by a nonspecific effect of treatment on the items chosen than the SDF and LR analyses, as the CCLI contains a large number of factors. Obtaining a single-severity score such as the CCLI is of value in: (a) assessing the effectiveness of treatment modalities; (b) analyzing the success of randomization; (c) separating cohorts of different severity, and (d) comparing new liver tests, histological abnormalities, or specific biological events with the severity of alcoholic liver disease.