Selection for oesophagectomy and postoperative outcome in a defined population.

OBJECTIVE To measure the extent of use of, and perioperative mortality from, oesophagectomy for carcinoma of the oesophagus, and to examine the association between oesophagectomy and long term survival. DESIGN Retrospective cohort study of cases of oesophageal carcinoma notified to the Thames Cancer Registry. SETTING South East Thames and South West Thames health regions. PATIENTS 3273 patients first registered with carcinoma of the oesophagus during 1985-9, 789 of whom were excluded because of incomplete data, leaving 2484 (75.9%) for further analysis. MAIN MEASURES Treatment of oesophagectomy, mortality within 30 days of oesophagectomy, and duration of survival from date of diagnosis to death, according to patient and tumour characteristics. RESULTS Oesophagectomy was performed in 571(23.0%) patients. Its use decreased with increasing age (odds ratio (95% confidence interval) 0.935(0.925 to 0.944) per year) and was less common for tumours of the middle or upper third of the oesophagus than the lower third (0.56(0.42 to 0.75)). The proportion of patients undergoing oesophagectomy varied threefold among the 28 districts of residence. The perioperative mortality rate was 15.1(86/571) (12% to 18%); it increased with age (odds ratio 1.05(1.02 to 1.08) per year) and for tumours of the middle or upper third of the oesophagus compared with the lower third (2.52(1.31 to 4.84)). Long term survival was slightly higher for patients undergoing oesophagectomy (0.5% v 0.2%). CONCLUSIONS Despite a high perioperative mortality rate patients selected for oesophagectomy showed better long term survival than those who were not, suggesting that clinical judgements used in selection were independent markers of a better prognosis. The nature of this selection needs to be more completely characterised to permit a valid evaluation of outcome of oesophagectomy.