The main objective of this study to analyze which of 31 cellular factors (resistance proteins, proliferative factors, apoptotic factors, angiogenic factors, proto-oncogenes) most accurately predict the resistance of non-small cell lung carcinomas. To this purpose, we used a short-term in vitro test that measures changes in the rate at which radioactive nucleic acid precursors are incorporated into tumor cells after the addition of doxorubicin to determine the response to doxorubicin in 94 non-small cell lung carcinomas. The results obtained by the short-term test were related to the various cellular factors which were in turn determined by immunohistochemistry and flow cytometry. A significant correlation was found between the data obtained by the short-term test and the expression of P-glycoprotein 170 (P = 0.00004), glutathione-S-transferase-pi (P = 0.0002), metallothionein (P = 0.0008), thymidylate synthase (P = 0.002), O6-methylguanine-DNA-methyltransferase (P = 0.008) and lung resistance-related protein (LRP, P = 0.03). There was only a weak correlation between heat shock proteins (HSP70) and no correlation between the expression of topoisomerase II or catalase and the short-term test results. To measure the proliferative activity, the following were determined: PCNA, cyclin A, cyclin D and cdk2. Only a weak relationship was found between the expression of cdk2 (P = 0.04) and PCNA (P = 0.05) and the doxorubicin response in vitro. Of the investigated pro-apoptotic factors (Fas/CD95, Fas ligand, caspase-3), only Fas/CD95 is significantly associated with the drug response (P = 0.007). The apoptotic index also reveals a significant correlation (P = 0.03). Angiogenesis, as measured by the microvessel density and the angiogenic factors, is inversely correlated to the resistance of non-small cell lung cancer. Platelet-derived endothelial cell growth factor (PD-ECGF) and vascular endothelial growth factor (VEGF) exhibit a significant relationship to the drug resistance (P = 0.0006 and P = 0.004, respectively). Of the investigated proto-oncogenes (Fos, Jun, ErbB-1, ErbB-2, Myc, Ras), only ErbB-2 is weakly associated with the in vitro short term test. In order to determine whether combining factors can result in improved predictive information, combinations of the factors (pairs, triplets) were analyzed. The systematic investigation of these combinations yields an improvement in the predictive information. With one factor up to 76.6% of the tumors, with two factors up to 85.4% and with three factors up to 89.5% of the tumors could be correctly diagnosed.