Consistent liver metastases in animal models is generally observed only with certain cancer cell lines. With the aim of improving on existing animal models of liver metastases, we hypothesized that cancer cells encased in 300 microm microcapsules, mimicking micrometastatic foci, might be effective seeds of liver metastases. A total of 3,000 microcapsules, containing 700 to 1,500 viable cells/capsule in logarithmic growth phase of three human pancreatic cancer cell lines (SUIT-2, AsPC-1, and BxPC-3), were transplanted in nude rats by portal injection. The rate of liver metastases was 100% (12 of 12), 100% (6 of 6), and 83% (5 of 6) for SUIT-2, AsPC-1, and BxPC-3 microcapsules, respectively. In contrast, the administration of an identical number of single cancer cells (2.1-4.5 x 10(6)) did not lead to liver metastases. Metastases was strictly limited to the liver, was quite stable, and could be proportionately tailored by varying the number of cancer microcapsules administered. Microscopic observation showed that two-thirds of the cancer microcapsules were lodged in the peripheral small (20-50 microm) portal veins, although one-third of the cancer microcapsules were trapped in the central wide (200-400 microm) portal vein. Capsules began to burst at day 3, with recognizable metastases produced at day 7, resulting in overt metastases production at days 28 to 42. The present cancer microcapsule method may be useful for obtaining liver metastases in animal models, especially for cell lines that will not form liver metastases with conventional single cell injection methods and/or for experiments requiring the consistent formation of liver metastases.