Improvement of the present mortality from cancer depends upon a means of controlling those cancers that have spread from the original site to distant organs. Killing of tumor cells by means of drugs, whether used by themselves or in combination with surgery or irradiation, provides the strategy most likely to achieve this goal in the near future. There are now 10 disseminated cancers in man wherein drug treatment has resulted in normal life expectancy. The biological pattern of growth of these tumors is different from that of the tumors that have not been so susceptible to drugs. The new knowledge concerning the interdependence of growth pattern and drug susceptibility has led to rational animal models for selection of active drugs, in which more confidence can be placed than upon empiric models. Similarly, the prediction of drug toxicity and pharmacology from animal models has been placed on a firm scientific base. Recent developments in molecular biochemistry hold forth the hope that some day the selection of safe and effective antitumor drugs may be based on a knowledge of in vitro biochemistry.