Contribution of biostatistics to the standardized evaluation of tumor drug resistance tests

flag

Klin Onkol 2000; 13(Speciál2 2000): 22-29.

Summary: Specific role of drug resistance tests in the concept of predictive oncology determines the necessity of objective parametric outputs with biologically clear interpretation. In the case of in vitro tests, the problem implies quantitative model description of dose response curves and standardization of descriptors. Comparative evaluation of the calibration data set (46 cell lines tested factorially for the effect of 28 different chemoterapeutic agents in MTT test) revealed 3 basic types of dose response relationships: sigmoidal curves, hormesis model with response stimulation in low concentrations and complex curves consisting of two or more sigmoidal (or hyperbolic) components that might be related to different mechanisms of influence (Carboplatin, Cisplatin, 5-Fluorouracil). Final spectrum of obtained curves was however more diverse as a consequence of numerous types of interaction between drug mechanism and properties of tumor cells. The dose-effect diversity could be only partially attributed to different mechanisms of influence, it however appeared to be significantly associated with the heterogeneity of tumor cells, namely with the degree of resistance. Model analyses performed on the calibration data set proved the median effect equation as an powerful diagnostic technique for uniform analysis of all dose response patterns that occurred in drug resistance tests. A minimum, but clinically relevant set of parameters was proposed to be routinely estimated from each curve: EC50 estimates, threshold concentration (or proper alternative as ECx, e.g. EC10) and fraction of cells surviving the highest applied concentration of examined compound.