The presence of peritoneal micrometastasis in the abdomen is a poor prognostic factor in advanced gastric cancer. However, there is no standardized method for detection of peritoneal micrometastases. The aim of this study was to establish an animal model mimicking early phase peritoneal dissemination of advanced gastric cancer and then to use this model to compare the sensitivity and specificity of reverse transcription-polymerase chain reaction (RT-PCR) with peritoneal lavage fluid, randomly collected omentum with or without milky spots stained with activated carbon particles for detection of disseminated cancer cells. MKN-45-EGFP gastric cancer cells (1 x 10(3), 5 x 10(3) and 1 x 10(4)) were injected intraperitoneally into 4-week-old female BALBc nu/nu mice on day 0. Mice were sacrificed at 7 or 14 days postinjection. Peritoneal seeding was assessed by fluorescence stereomicroscopy. The sensitivities of RT-PCR with peritoneal lavage fluid, omentum with or without milky spots, were compared. Peritoneal seeding was confirmed in 8 of 10 mice injected with 1 x 10(4) MKN-45-EGFP cells by fluorescence stereomicroscopy. On days 7 and 14, the rate of the detection of CEA mRNA was 0, 50.0, and 87.5% in the omentum, omentum with milky spots and peritoneal lavage fluid, respectively. On days 7 and 14, the average level of CEA mRNA was 0, 51113+/-28225, and 3556+/-2842 in the omentum, omentum with milky spots and peritoneal lavage fluid, respectively. Molecular diagnosis using peritoneal lavage fluid is more sensitive than that using the omentum.