We applied a multivariate analysis to a large series of serum biochemical tests in an attempt to identify a function that could efficiently discriminate cirrhosis from hepatocellular carcinoma (HC). We analyzed two successive temporal cohorts (1987-90; 1991-94) of HC and cirrhotic patients, all histologically classified (first cohort: 69 cirrhosis and 39 HC; second cohort: 66 cirrhosis and 38 HC). Using data from the first temporal cohort of patients, we obtained a discriminant function based on seven serum analytes: alpha-fetoprotein, the hepatic isoenzyme of alkaline phosphatase, lactate dehydrogenase isoenzyme 5, total gamma-glutamyltransferase (GGT), GGT isoforms complexed with low-density lipoprotein, aspartate aminotransferase, and copper. The same panel of analytes emerged when the second cohort was tested and also when both cohorts were tested together. In the two successive cohorts (total, 212 patients) with a prevalence of cirrhosis vs HC of approximately 2:1, the discriminant function correctly classified 93% of cases, the highest percentage of correct classification of the two diseases obtained so far by laboratory approaches. Validation with the jackknife reallocation statistical algorithm confirmed these results. In addition, of six patients with liver cirrhosis for whom we had the opportunity of following up and observing the evolution to HC, five were classified as HC at diagnosis by the multivariate discriminant analysis; i.e., discriminant analysis provided a diagnostic lead time of 6-12 months over histology. This discriminant function, based on easy-to-perform serum biochemical tests, may help solve a fundamental problem of differential diagnosis in the evolution of chronic liver diseases from cirrhosis to HC.