Methods 802 previously untreated patients with resectable oesophageal cancer of any cell type were randomly allocated either two 4-day cycles, 3 weeks apart, of cisplatin 80 mg/m(2) by infusion over 4 h plus fluorouracil 1000 mg/m(2) daily by continuous infusion for 4 days followed by surgical resection (CS group, n=400), or resection alone (S group, 402). Clinicians could choose to give preoperative radiotherapy to all their patients irrespective of randomisation. Primary outcome measure was survival time, Analysis was by intention to treat,

Findings No patients dropped out of the study. Resection was microscopically complete in 233 (60%) of 390 assessable CS patients and 215 (54%) of 397 S patients (p<0.0001). Postoperative complications were reported in 146 (41%) CS and 161 (42%) S patients. Overall survival was better in the CS group (hazard ratio 0.79; 95% Cl 0.67-0.93: p=0.004). Median survival was 512 days (16.8 months) in the CS group compared with 405 days (13.3 months) in the S group (difference 107 days; 95% Cl 30-196), and 2-year survival rates were 43% and 34% (difference 9%: 3-14).

Interpretation Two cycles of preoperative cisplatin and fluorouracil improve survival without additional serious adverse events in the treatment of patients with resectable oesophageal cancer.