Many patients undergoing surgery for gastric carcinoma will develop peritoneal metastases. A method to identify those patients at risk of peritoneal recurrence would help in the selection of patients for adjuvant therapy. Peritoneal cytology has received little attention in the West, but may prove a useful additional means of evaluating patients with gastric cancer. The aims of this study were to evaluate sampling techniques for peritoneal cytology in patients with gastric cancer, to assess the prognostic significance of free peritoneal malignant cells and to discover the effect of the operative procedure on dissemination of malignant cells. The study is based on 85 consecutive patients undergoing surgical treatment of gastric cancer and followed up for 2 years or until death. Peritoneal cytology samples were collected at laparoscopy, and at operation prior to resection by intraperitoneal lavage and serosal brushings. After resection, samples were taken by peritoneal lavage, imprint cytology of the resected specimen and post-operatively by peritoneal irrigation via a percutaneous catheter. Malignant cells were diagnosed by two independent microscopists. Preoperative peritoneal lavage yielded malignant cells in 16 out of 85 cases (19%). The yield of free malignant cells was increased by using serosal brushings (by four cases) and imprint cytology (by two cases); all of the cases had evidence of serosal penetration. One serosa-negative case exhibited positive cytology in the post-resection peritoneal specimen in which the preresection cytology specimen was negative. Survival was worse in the cytology-positive group (chi2 = 25.1; P< 0.0001). Among serosa-positive patients, survival was significantly reduced if cytology was positive, if cases yielded by brushings and imprint cytology were included (log-rank test = 8.44; 1 df, P = 0.004). In conclusion, free peritoneal malignant cells can be identified in patients with gastric cancer who have a poor prognosis; the yield can be increased with brushings and imprint cytology in addition to conventional peritoneal lavage. Evaluation of peritoneal cytology by these methods may have a role in the selection of patients with the poorest prognosis who may benefit most from adjuvant therapy.