In an attempt to increase the predictability of preclinical antitumor testing, the value of human tumor lines in immune-deficient nude mice is assessed by reviewing the relevant literature. This test model is rather elaborate due to the nature of the animal as well as test and evaluation procedure. However, it represents a realistic simulation of clinical drug treatment. This is demonstrated by (a) a good correlation of drug effects in the nude mouse with clinical results in the donating patient's tumor and (b) by a good predictability of a panel of human tumor lines for clinically effective drugs. In order to avoid clinical trials with inactive drugs and no therapeutic benefit for a large number of patients, the application of human tumor xenografts in anticancer drug development is warranted.