In an attempt to identify anticancer agents more active in the major groups of solid cancers, the National Cancer Institute has replaced the primary murine tumor systems with an in vitro system using human cancer cell lines. A broader approach is being followed in Europe, where considerable reliance will still be placed on data from animal tumor systems. We argue the case for broader preclinical evaluation using animal model systems that are resistant to standard anticancer agents and thus reflect the clinical disease. Selection of compounds for clinical trials should be based on a therapeutic index that will give a more realistic representation of the margin between antitumor activity and toxicity to normal cells. Furthermore, preclinical evaluation can establish understanding of the factors responsible for treatment efficacy, including pharmacokinetics, metabolism, and drug bioavailability.