[Evaluation by multiple regression analysis of factors influencing the chemosensitivity of human tumors xenografted into nude mice].

The chemosensitivity of human cancer lines is thought to be expressed as a result of contributions by various interacting factors. Multiple regression analyses were performed in order to clarify the weighting of factors responsible for the chemosensitivity of 15 human cancers xenografted into nude mice. Inhibition rates of 11 anticancer agents predetermined for each line of human cancer were used as the criterion variables. As the explanatory variables, 9 parameters characteristic of each cancer or cancer-bearing mouse were selected as follows; grade of differentiation, vascularity, percentage necrosis, volume doubling time, labeling index, LDH activity, tissue/serum LDH ratio, thymidine phosphorylase activity and serum CEA. By applying this analysis with stepwise deletion, the estimated multiple regression equations for drug sensitivity were clarified for each drug. Although all equations were composed of different factors and their partial repression coefficients varied from drug to drug, those among analogous drugs such as FT-207 and UFT, or MMC and M-83 had similar factors. The equations for M-83, ACNU and ADR consisted of a number of parameters with a sufficiently high coefficient of determination of over 80%. Even in cases of MXT that showed no significant factor upon simple correlation analysis, an equation with 7 factors revealed a coefficient of determination of 0.83. The estimated values of effectiveness for these drugs showed remarkable coincidence with each actual value. For some drugs, the in vivo mode of action was inferred through this analysis.