Evaluating Toxin-induced Hepatic Injury in Rats by Laboratory Results and Discriminant Analysis

The ability of 14 serum biochemical assays to predict the presence of hepatic necrosis induced by carbon tetrachloride (CCI,) (centrilobular necrosis), allyl alcohol (periportal necrosis), and 1 -napththylisothio- cyanate (ANIT) (biliary duct necrosis) was evaluated in rats. Results of these assays were analyzed using multivariate discriminant analysis to determine: which assays have the highest predictive value for discriminating between control and treated rats, and which assays would discriminate between rats in the three treatment groups. Individual assays with the highest predictive value for CC1,-induced lesions versus controls were glu- tamate dehydrogenase (GDH), sorbitol dehydrogenase (SDH), and alanine aminotransferase (ALT). Assays with the highest predictive value for ANIT-induced lesions were GDH, 5'-nucleotidase (5'NT), and ALT. Assays with the highest predictive value for allyl alcohol-induced lesions were an ALThocitrate dehydrogenase (ICD) ratio, GDH, and ALT. Canonical correlation coefficients for each assay ranged from 0.98 to 0.9 1 with 95-1009'0 correct group membership predictions (treated versus control) provided by each assay. Individual assays were not highly predictive for determining group membership among all three treatment groups. A two assay com- bination of 5'NT and an ALTIICD ratio provided 100% correct group membership predictions and had high canonical correlations (f, = 0.95, fi = 0.83). Serum assays have been used for detecting hepato- cellular necrosis and cholestasis in domestic animals for years.' The correlation between serum activities of enzymes with different intracellular locations and the morphologic evaluation of hepatic lesions has been poor.8J2 One reason may be the incorrect assumption that cytosol enzymes always leak from damaged he- patocytes first and mitochondria1 enzymes leak later or with more severe damage.6.20 Another explanation may be that inappropriate enzymes were studied. A more successful effort has been made to discern a biliary or intralobular location for hepatic necrosis us- ing serum enzymes. 14~18 The heterogeneous distribution of hepatic enzymes within the liver lobule and biliary tract has been recognized for The clinical importance of the intralobular or biliary distribution of enzymes was demonstrated by the increased diag- nostic sensitivity which can be obtained by selecting assays to detect a specific hepatic le~ion.~'.~~,~~ Statistical models can be developed to determine which clinical chemistry tests provide the most useful information.28 Multivariate discriminant analysis is an effective method for determining which tests provide the most diagnostic information in man.23 This statis- tical method is particularly applicable to situations where large amounts of data must be analyzed and experimental subjects classified into a few well-char- acterized groups (e.g., affected or ~naffected).~~ Multi- variate discriminant analysis has proven to be robust against deviations caused by data not normally dis- tributed or which contain unequal population vari- ance~.'~,~~ Both problems are common in clinical chem- istry data. Our study was designed to determine the extent to which different groups of rats are separable based on the results of serum biochemical assays and how lab- oratory test results can be efficiently used to classify rats into various treatment (or control) groups using multivariate discriminant analysis.