Background Training simulators have been used for decades with success; however, a standardized educational strategy for diagnostic EGD is still lacking. Objective Development of a training strategy for diagnostic upper endoscopy. Study design Prospective, randomized trial. Settings A total of 28 medical and surgical residents without endoscopic experience were enrolled. Basic skills evaluations were performed following a structured program involving theoretical lectures and a hands-on course in diagnostic EGD. Subsequently, stratified randomization to clinical plus simulator training (group 1, n = 10), clinical training only (group 2, n = 9), or simulator training only (group 3, n = 9) was performed. Ten sessions of simulator training were conducted for groups 1 and 3 during the 4-month program. Group 2 underwent standard training in endoscopy without supplemental simulator training. The final evaluation was performed on the simulator and by observation of 3 clinical cases. Skills and procedural times were recorded by blinded and unblinded evaluators. Main Outcome Measurements Time to reach the duodenum, pylorus, or esophagus. Results All trainees demonstrated a significant reduction in procedure time during a simple manual skills test ( P < .05) and significantly better skills scores ( P = .006, P = .042 and P = .017) in the simulator independent of the training strategy. Group 1 showed shorter times to intubate the esophagus (61 ± 26 seconds vs 85 ± 30 seconds and 95 ± 36 seconds) and the pylorus (183 ± 65 seconds vs 207 ± 61 seconds and 247 ± 66 seconds) during the clinical evaluation. Blinded assessment of EGD skills showed significantly better results for group 1 compared with group 3. Blinded and unblinded evaluations were not statistically different. Limitations Small sample size. Conclusions Structured simulator training supplementing clinical training in upper endoscopy appears to be superior to clinical training alone. Simulator training alone does not seem to be sufficient to improve endoscopic skills.