OBJECTIVE Recently, attention has been drawn to the significant occurrence of respiratory and other complications during upper intestinal endoscopy. This prospective study was designed to compare the incidence and severity of oxygen desaturation when two different methods of oxygen delivery were used during elective oesophagogastroduodenoscopy. METHODS After local medical ethics committee approval and written informed consent, one hundred patients undergoing elective oesophagogastroduodenoscopy were randomly allocated to receive supplemental oxygen at either four litres per minute via nasal specular or ten litres per minute through a specially modified non-rebreathing mask. Oxygen saturations during endoscopy facilitated by midazolam sedation were recorded, and non-parametric tests were used to compare the oxygen saturations in the two groups. RESULTS The mean oxygen saturations were significantly better during the procedure for American Society of Anesthesiology (ASA) grade 2 and 3 patients who received oxygen with the modified mask than for those who received oxygen via nasal specular (98.6% vs. 97.0%, P = 0.004 for ASA grade 2 and 98.4% vs. 95.5, P = 0.006 for ASA grade 3). CONCLUSIONS For ASA grade 2 and 3 patients, the modified non-rebreathing mask significantly improves oxygen saturations during upper intestinal endoscopy. This technique should be more widely used for patients at moderate and high risk.