Blood vessel invasion in gastric carcinoma.

In an effort to identify factors in primary tumors that would predict liver metastases, we retrospectively reviewed 102 patients with gastric cancer, and their tissue blocks were restained. New staining methods for elastin and endothelium were used to identify intratumoral vessels. Blood vessel invasion, thus detected, was analyzed quantitatively, as well as qualitatively, according to the location of invasion, the size of the involved vessel, and the mode of invasion. The invasion was then compared with the presence of liver metastasis by the chi 2 test, the Mann-Whitney U test, and the Student t test. Discrimination analysis of factors significantly correlated with liver metastasis was performed with linear discrimination function to identify a predictive model for liver metastasis. Significant differences in qualitative frequency of blood vessel invasion (p less than 0.01), the number of lymph node metastases (p less than 0.05), and the depth of tumor invasion (p less than 0.05) were found in those patients in whom liver metastasis developed, as compared with 5-year disease-free survivors. Quantitative analysis of blood vessel invasion revealed eight other factors correlated with liver metastasis; frequency of blood vessel invasion in the 0.01 to 0.1 mm and 0.1 to 1 mm diameter vessels, in the forms of complete, partial tumor thrombi, and vessel wall invasion, in the submucosa and the subserosa, and the number of anatomic stomach layers involved. Application of the discrimination coefficient to these factors allows prediction of liver metastasis with 81.8% sensitivity, 85.3% specificity, and 83.6% accuracy. Liver metastasis can be predicted from the qualitative and quantitative examination of blood vessel invasion within the primary tumor by the use of an elastic fiber stain.