BACKGROUND Curative oesophageal resection for carcinoma may be carried out by either the transhiatal or the Ivor-Lewis transthoracic technique. The aims of this study were to compare the morbidity, 30-day mortality and long-term survival of the two techniques in the treatment of oesophageal carcinoma and to provide data to calculate the sample sizes for a prospective randomized trial. METHODS Results from 44 series published between January 1986 and December 1996 were reviewed. Thirty-three papers reported results on 2675 patients having transhiatal (THO) and 29 papers reported results on 2808 patients having Ivor-Lewis oesophagectomy (ILO). RESULTS The two groups were comparable in terms of age, sex and stage of the disease. There was no apparent difference in postoperative morbidity between the two groups with respect to respiratory complications (24% for THO, 25% for ILO), cardiovascular complications (12.4% for THO, 10.5% for ILO), wound infection (8.8% for THO, 6.2% for ILO) and chylothorax (2.1% for THO, 3.4% for ILO). The transhiatal group appeared to have a higher incidence of anastomotic leaks (16% for THO, 10% for ILO), anastomotic strictures (28% for THO, 16% for ILO) and recurrent laryngeal nerve injuries (11.2% for THO, 4.8% for ILO). The 30-day mortality was 6.3% for transhiatal and 9.5% for Ivor-Lewis oesophagectomy. Overall long-term survival at 5 years was similar (24% for THO, 26% for ILO). CONCLUSIONS The surgical approach to oesophagectomy was not an important determinant of morbidity and long-term survival in patients with oesophageal carcinoma. Transhiatal oesophagectomy was associated with a higher incidence of anastomotic complications and recurrent laryngeal nerve injury. Ivor-Lewis oesophagectomy had a higher mortality. In order to demonstrate a significant difference in morbidity or long-term survival between the two techniques 3100 patients would be required in each arm of a prospective randomized trial.