The actual benefit of extended lymphadenectomy in terms of survival in the surgical treatment of gastric cancer is still a debated issue. The aim of this non-randomized prospective multicentre study was to evaluate long-term survival in a group of patients with involvement of the second level lymph nodes, which would not have been removed with a limited lymphadenectomy. From 1991 to 1997, 451 patients with primary gastric cancer underwent curative resection with extended lymphadenectomy in three italian surgical departments. Lymph node stations were removed and classified according to the rules of the Japanese Research Society for Gastric Cancer; in all cases, retrieval of the lymph nodes was performed by the surgeon on the fresh specimen. Metastases to lymph node stations 7-12 were found in 126 patients out of 451 (27.9%). A mean number of 13 +/- 9 positive lymph nodes (range: 1-42) was found in these cases. Lymph node stations 7 and 8 showed the highest incidence of metastases (61.1% and 44.4%, respectively). Morbidity and mortality rates were 17.1% and 2% in 451 cases treated by extended lymphadenectomy, and 21.4% and 3.2%, respectively, in 126 cases with involvement of second level lymph nodes. In this group of patients, the five-year survival rate was 32 +/- 4%. Multivariate analysis, identified depth of invasion (P < 0.0001, relative risk (RR) 2.4) and the number of positive lymph nodes (P < 0.001, RR 1.6) as significant predictors of a poor prognosis. Japanese-type extended lymphadenectomy is associated with low morbidity and mortality rates if performed in specialised centres. The incidence of metastases in lymph node stations removed with this technique is by no means negligible. This procedure could be beneficial as regards long-term survival even in patients with involvement of regional lymph nodes.