Anticancer agents so far available and their mechanisms of action suffer from the problem of their relatively low selectivity. Their insufficient clinical efficacy against the common, slowly growing solid tumors of the lung, gastrointestinal system, kidneys, urinary bladder, and brain remains disappointing. Recently the possibility has been discussed that the limited clinical activity of current anticancer drugs could result from the screening models and methods used in their selection. The initial approach to drug discovery used by the National Cancer Institute, Bethesda, USA (NCI), the greatest oncological research unit in the world, has been empirical large-scale screening in transplantable rodent tumor models. In the past, these preclinical models have been changed periodically in line with retrospecitve analyses of preclinical predictivity for clinical efficacy. Recently, as a new strategy, a “disease-orientated” concept has been developed to screen agents against particular types of human cancer on the basis of the human tumor colony-forming assay in vitro. Each compound should now be tested directly against a spectrum of human tumor lines without passing through a rodent prescreen. Additional assays in vivo may be performed later on. This new screening concept seems to be suitable for identifying the cytotoxicity of new chemical structures and for an evaluation of sensitivity or resistance of the different tumor types. The contrasting concept of “rational drug design” is exemplified by the development of the oxazaphosphorinecytostatics. The basis of this concept was the application of the transport form/active form principle to the antiproliferative nitrogen mustard. Cyclophosphamide, the first representative of this group, had already largely reached the given objective. Generalizing conclusions from the different concepts are as follows. 1. Methods and perceptions of general pharmacotherapy must be the principal basis for the development of antitumor compounds. 2. Progress and essential new developments in cancer chemotherapy are based on experiments in intact animals. 3. An important feature of rational drug design is the stepwise or sequential procedure: the design of new drugs is based on the screening results of former drugs to achieve an optimal progress. 4. For the analysis of the activity of alkylating agents on rats and mice, the panel of test tumors for screening and pharmacological evaluation must be selected according to a different degree of chemoresistance or chemosensitivity, respectively. It should be aimed at a complete dose/activity curve with cyclophosphamide as standard at least for the most sensitive tumors of that panel. 5. The therapeutic index, i.e. LD5/CD95, has proved to be a valuable tool for chemotherapeutic usefulness, as has the danger coefficient for the quantification of organotoxic side-effects. These values provide a measure of the therapeutic range and, consequently, of the selectivity of the antitumor activity. Results from a given tumor have proved to be predictive for other tumors and turned out to be relevant also for clinical trials. 6. The different sensitivities of experimental tumors against alkylating agents is not a fundamental property but a quantitative feature. With sub- or even supra-lethal doses it is possible to overcome vitality and transplantability of even the very most resistant tumors. A new product with impressively increased selectivity is consequently expected to achieve remissions in more resistant tumors also. This evaluation system also remains applicable and useful in the context of a “disease-oriented” concept. 7. A profound knowledge of the mechanism of action of cytostatic agents and a deep insight into the metabolic patterns in the host and in tumor cells are the essential basis for a rational augmentation of cytostatic activity. 8. Promising leads, identified initially by empirical large-scale screening programs, mostly need forther optimization through the rational approach. Thus there is most often an essential and intimate interplay between the rational and large-scale screening strategies.