Background and study aims: Endoscopic submucosal dissection (ESD) is a new method for the curative treatment of early gastrointestinal neoplasms, which was developed in order to increase the en bloc and R0 resection rate, especially for lesions larger than 20 mm in diameter. Drawbacks of ESD include the fact that it is technically a substantially more difficult procedure and that it is associated with a higher perforation rate. A retrospective study was therefore carried out to analyze cases in relation to the procedure time and resection success, and these factors were correlated with the characteristics of the lesions. Patients and methods: From January 2002 to November 2005, 196 lesions in 185 patients with early gastric cancer were treated using ESD in our hospital. The rates of curative en bloc resection, the incidence of perforation, and the procedure times were analyzed in relation to lesion size (small, 20 mm or less in diameter; large, over 20 mm), location (upper, middle, or lower third of the stomach) and the presence or absence of ulceration. Results: The rate of curative en bloc resection was 84 % (93 % of the lesions overall were resected in one piece), with a perforation rate of 6.1 % (all perforations were managed endoscopically) and a mean procedure time of 68 min. The rate of curative en bloc resection differed significantly depending on the location of the lesion (upper vs. middle vs. lower, 74 % vs. 77 % vs. 91 %; P < 0.05), as well as on the size of the lesion (> 20 mm vs. 20 mm or less, 59 % vs. 89 %; P < 0.0001). There were also significant differences in the mean procedure times in relation to the location of the lesion (upper vs. middle vs. lower, 105 min vs. 81 min vs. 45 min; P < 0.0001) and the size of the lesion (> 20 mm vs. 20 mm or less, 124 min vs. 55 min; P < 0.0001), as well as the presence of ulceration (positive vs. negative, 97 min vs. 65 min; P < 0.05). With regard to perforation rates, significant differences were also observed in relation to the location of the lesion (upper vs. middle vs. lower, 22.6 % vs. 2.8 % vs. 3.2 %; P < 0.0005) and size of the lesion (> 20 mm vs. 20 mm or less, 16.2 % vs. 3.8 %; P < 0.005). No local recurrences of curatively resected lesions (n = 119) were observed after a follow-up period of 1 year. Conclusions: The difficulty of ESD depends on the location and size of the lesion, as well as on the presence of ulceration. We would recommend that trainees should begin by carrying out ESD on lesions with a diameter of less than 20 mm without ulceration that are located in the lower third of the stomach.