The hollow fiber test has been developed for the preliminary in vivo assessment of cancer chemotherapeutic efficacy of selected natural products. Using this model, we have established growth conditions for HL-60, HUVEC, Ishikawa, KB, KB-V1, LNCaP, Lu1, MCF-7, Mel2, P-388, and SW626 cells implanted at the intraperitoneal (i.p.) and subcutaneous (s.c.) compartments of athymic mice. Five cytotoxic natural product isolates (2-6) were tested in this model, along with paclitaxel (taxol) (1). Among the compounds tested, dioscin (2) and 13-methoxy-15-oxozoapatlin (3) were found to be active, indicating their potential to function as cancer chemotherapeutic agents. On the other hand, ochraceolide A (4), alpha-lapachone (5), and 2-(1-hydroxyethyl)naphtha[2,3-b]furan-4,9-quinone (6), all of which were significantly cytotoxic to cultured mammalian cells, did not mediate significant responses with the hollow fiber model. In further xenograft studies using KB cells implanted at the subcutaneous site, compound 3 mediated a statistically significant response which was consistent with the response observed at the subcutaneous compartment in the hollow fiber tests. In sum, these studies illustrate the usefulness of the hollow fiber model in natural product drug discovery programs. Preliminary indications of potential therapeutic efficacy can be provided quickly at relatively low expense. Agents capable of mediating a response at the subcutaneous site would appear to warrant greatest attention.