The colony formation of human tumor xenografts from nude mice, of murine tumors, and of human bone marrow (CFU-C) has been investigated in vitro using a modification of the double-layer agar assay described by Hamburger and Salmon. Systematic modification of growth conditions and careful selection of viable tumor tissue enhanced the growth rate (at least 30 colonies per dish) of human tumor xenografts to 86% (98/114). The median plating efficiency was 0.07% which is comparable to the results observed by others using fresh human tumors. The growth of human bone marrow was stimulated with a placenta-conditioned medium, which allowed growth of granulocytic stem cell colonies (CFU-C). The median plating efficiency of the bone marrow was 0.08%. The murine tumors P388, L1210, B16 melanoma, Lewis lung carcinoma and colon carcinoma 38 grew very well in vitro. Excluding the Lewis lung carcinoma, the plating efficiency of these tumors was markedly higher than that of the human tumor xenografts and human bone marrow. The colony assay may have potential as a secondary screening system for identifying new active structures and also for indicating which tumor types are most responsive to a new antitumor agent. We test new structures in 20 well-selected human tumor xenografts and in the P388 mouse leukemia in dose-response relationships. The two most responsive xenograft tumors are subsequently studied in vivo in nude mice in order to determine if a new compound presents antitumor activity in an in vivo organism at a dose around the LD10 level. If a remission or at least no change is observed in the subcutaneously growing tumor, the new compound undergoes large disease-oriented testing usually in 60 xenografts. The in vivo studies are necessary in determining whether a compound has a more specific effect on tumor cells than on the dose-limiting normal tissue. The comparison of in vitro/in vivo activity allows an assessment of the relevant in vitro dose based on in vivo pharmacological behavior of a drug. It seems justifiable to apply the conclusions of this approach to the clinical setting because mouse toxicity data, e.g. the LD10, correspond well to the maximal tolerable doses in man. Moreover, for compounds whose dose-limiting toxicity is bone marrow suppression, the comparison of drug dosages effective in vitro on human bone marrow and tumor xenografts may prove helpful. The proposed testing strategy has been applied to TGU and Tiazofurin. At the relevant dosages TGU exhibited very limited activity in 67 human tumor xenografts studied.(ABSTRACT TRUNCATED AT 400 WORDS)