Traitement statistique des donnees biochimiques par la methode d'analyse discriminante. Selection des variables biochimiques discriminantes. Essai de discrimination biochimique de la cholostase intraheatique de l'obstacle extrahepatique et du cancer du foie

An original method of statistical treatment of biological data is proposed. It permits satisfactory biochemical classification of 322 patients divided up into 3 groups : intrahepatic cholestasis (235 patients), extrahepatic obstruction (44 patients) and carcinoma of the liver (43 patients). On the basis of 32 tests, it was possible to define discriminating areas permitting satisfactory diagnosis in 95 per cent of published cases. The reduction in the number of tests necessary for diagnosis was considered. The selection technic used was original to the extent that it dose not require, like most methods used today, the determination of better individual discriminators, but the establishment of a better discriminating subunit, obtained from the initial subunit composed of a group of variables. From the 32 parameters contained in the standard liver function tests, a search for a better discriminating subunit consisting of the best four tests, permitted the authors to select a group of 10 tests : bilirubin, alkaline phosphatase, 5-nucleotidase, Thymolturbidity, Cetavlon test, serum albumin, total LDH, TGP (ALAT), OCT, GLDH, of which the discriminating value remains very satisfactory.