Palliation is necessary in over 50% of patients with oesophageal cancer, and the most effective means of achieving this is still debated. Plastic or stainless steel reinforced endoprostheses have been available for some 20 years, but have the disadvantages of bulky introducing systems, a significant incidence of perforation and, frequently, sub-optimal palliation. The introduction of self-expanding metallic stents (SEMS) in 1990 was received enthusiastically on account of their relative ease of insertion, with low perforation risk and greater internal diameter of 20-25 cm, resulting in better relief of dysphagia. However, disadvantages of SEMS include the high cost of the stents and disposable delivery systems, the difficulty in removing or repositioning these stents, and the high rate of re-intervention because of tumour ingrowth with uncovered stents, stent displacement with their covered counterparts, and obstruction owing to stent compression or tumour overgrowth at either end of the stent. Published studies include a randomized study between conventional plastic prostheses and uncovered Wallstents, a non-randomized study comparing uncovered Wallstents and Ultraflex stents, and the study published in this issue comparing uncovered Ultraflex stents with covered Wallstents. Somewhat surprisingly, 30 day mortality and relief of dysphagia were similar between conventional prostheses and uncovered Wallstents, and despite a 10-fold increase in cost of the SEMS over plastic prostheses, the overall cost of palliation was less because of a mean hospital stay of 5.4 days compared with 12.5 days for plastic prostheses, which is higher than many reported series and may relate to their insertion under general anaesthesia in this study. From the comparative studies of different SEMS, uncovered stents are associated with a higher incidence of tumour ingrowth and covered stents with a higher incidence of stent migration, particularly when they traverse the cardia. Thirty day mortality is relatively high (16-27%), although one study found no procedure-related mortality using the uncovered Ultraflex stent, but the reintervention rate was uniformly higher with those stents as compared with covered or uncovered Wallstents. Improvements in SEMS design are likely to overcome many of the technical problems, at which point it would be necessary to conduct prospective randomized studies against conventional prostheses inserted under sedation, with quality of life and economic assessment and sufficient numbers to enable sub-group analysis for variables such as tumour site and morphology. In the meantime, specialized centres should have facilities for each of the current palliative modalities so as to be able to deploy those most suited to individual circumstances.