In an attempt to increase the predictability and to extend the differential capacity of the anticancer drug development program the American National Cancer Institute has recently proposed the introduction of a screening system consisting of human tumor cell lines to select drugs in a disease-oriented fashion rather than by the previously applied drug-oriented strategy. Although this new approach offers great advantages, assay limitations can be identified in testing unknown compounds for antitumor activity in vitro. Human tumor xenografts grown in nude mice may play an additional role in the prediction of clinical activity and the assessment of the spectrum of activity of potential anticancer drugs, because they have a better relationship with the clinical situation of cancer treatment. In a European multicenter collaboration it has been proposed to use panels of human tumor lines from solid tumor types to study: the antitumor activity of three different drugs per tumor type; the reliability of 'preclinical' phase II studies by comparison of the obtained data with results of phase II clinical trials; the feasibility of this joint project, such as the methodology, the reproducibility of experimental data and the introduction of uniform activity criteria. If preclinical phase II studies in human tumor lines generate reliable results, this in vivo screening system will create a unique possibility to better identify promising clinical candidate compounds or analogs of conventional cytostatic agents as well as those tumor types likely to respond to the selected investigational drugs.