British Journal of Surgery 1998, 85, 255–260 Prognostic factors in a series of 297 patients with gastric adenocarcinoma undergoing surgical resection F. S ´ ANCHEZ-BUENO, J. A. GARCIA-MARCILLA, D. PEREZ-FLORES*, J. M. P ´ EREZ-ABAD, R. VICENTE, F. ARANDA, P. RAMIREZ and P. PARRILLA Department of General Surgery, ‘Virgen de la Arrixaca’ University Hospital and *Department of Medical Statistics, University of Murcia, Murcia, Spain Correspondence to: Professor F. S´anchez-Bueno, Servicio de Cirug´ıa, 3 a Planta, Hospital Universitario ‘Virgen de la Arrixaca’, 30 120, Murcia, Spain Background Gastric cancer has a poor prognosis. The aim of this study was to determine the influence of several clinicopathological variables on outcome in a series of 297 Western patients undergoing surgical resection for gastric adenocarcinoma. Methods The results were analysed retrospectively and prognostic factors were identified in a univariate and Cox proportional hazards regression model. Mean patient age at the time of operation was 61·9 years; 65·7 per cent were men. Mean follow-up was 7·8 (range 1–15) years. Of the 297 patients undergoing surgery, 70 per cent had subtotal gastrectomy, 26·3 per cent underwent total gastrectomy and 3·7 per cent had proximal gastrectomy. Results The overall survival rate was 38·9 per cent at 5 years. In the univariate analysis, survival- related factors were weight loss (P  0·05), abdominal mass (P  0·01), dysphagia (P  0·001), type of gastrectomy (subtotal gastrectomy versus total gastrectomy, P  0·001), intention of resection (curative versus palliative resection, P  0·001), tumour site (P  0·001), histopathological grade (low versus high grade, P  0·05), tumour diameter less than 3 cm (P  0·001), degree of gastric wall invasion (P  0·001), degree of lymph node invasion (P  0·001) and stage of the neoplasia (P  0·001). Other variables had no significant influence. In the multivariate analysis, degree of gastric wall invasion, lymph node invasion, tumour size and dysphagia at presentation were the only independent prognostic variables. Conclusion From these data it was possible to derive a prognostic index with which patients could be classified as at low, intermediate or high risk. Gastric cancer is one of the major causes of cancer- related death in the world, even though its incidence has decreased over the past decade 1 . The prognosis is generally poor, especially in Western countries 2,3 where the overall survival rate at 5 years has not changed, oscil- lating between 8 and 26 per cent, even though the resect- ability rate has increased (currently 60–80 per cent) and the postoperative mortality rate has dropped from 14 to 6 per cent 4–8 . The poor survival is related to delayed diag- nosis and frequent local and regional recurrence. Several variables representing pathological, clinical and thera- peutic characteristics have already been studied in numerous retrospective reports 9–15 in an attempt to identify prognostic indicators in patients with gastric cancer and, recently 16 , to help define high-risk patients who may benefit from adjuvant therapy. The aim of this paper was to determine the influence of several clinicopathological variables on outcome in a series of 297 patients undergoing surgical resection for gastric adenocarcinoma. The results were analysed retro- spectively and prognostic factors were identified in a univariate and Cox proportional hazards regression model. Patients and methods Between January 1979 and December 1994, 321 patients with gastric adenocarcinoma (excluding patients with cancer of the gastric cardia) underwent surgical resection. Excluding 24 operative deaths (defined as those occurring within 30 days of operation), the remaining 297 patients were discharged from hospital and are the subject of study. Mean follow-up was 7·8 (range 1–15) years. Mean patient age at the time of operation was 61·9 (range 30–88) years; 195 patients were men (65·7 per cent) and 102 women (34·3 per cent). Of the 297 patients under- going surgery, 208 (70·0 per cent) had subtotal gastrectomy, 78 (26·3 per cent) underwent total gastrectomy and the remaining 11 (3·7 per cent) had proximal gastrectomy. Three types of variable were analysed in the survival analysis: (1) clinical variables including sex, age, evolution time (time elapsed from the appearance of the first symptoms to the time of treatment) and the presence or absence of more meaningful clinical data (epigastric pain, pyloric obstruction, gastrointestinal haemorrhage, weight loss, dysphagia and a palpable abdominal mass); (2) treatment-related variables including the type of operation performed and the intention of the resection – curative (all gross disease is removed at operation) or palliative (gross disease remaining at the conclusion of the operation); and (3) anatomicopathological variables including tumour location, size, tumour structure (exophytic versus non-exophytic), degree of tumour differentiation, extent of the disease (tumour confined to the gastric wall versus invading adjacent organs or distant metastasis) and tumour node metastasis (TNM) stage (Union Internacional Contra la Cancrum classification 17 ). Deaths from causes other than gastric cancer were treated as censored data at the time of death. The survival data are expressed as 5-year survival rates with 95 per cent confidence intervals (c.i.). Survival rates in the different groups were calculated with the Kaplan–Meier estimation method, and survival curves were compared with the log rank test or with its version for testing trends if the categories of the grouping variable were ordered 18 . The possible prognostic factors were evaluated with the Cox proportional hazards regression model 19 by means of a backward Paper accepted 9 June 1997 © 1998 Blackwell Science Ltd 255 Downloaded from https://academic.oup.com/bjs/article/85/2/255/6269845 by guest on 02 January 2024