Abstract: Objective To investigate the extent of inter-observer variation in delineation of the organs at risk before and after studying the recommendation and its impact on estimating doses in blind mode. Methods Eight patients with nasopharyngeal carcinoma were asked to contour bilateral inner ear， middle ear and temporal lobe on computed tomography transverse section in blind mode by four observers. Interobserver variability were compared in the delineation of the above organs and estimated doses for these before and after studying the recommendation.
Results No difference of absolute volumes of inner ear， middle ear were observed（P＞0.05）， that of bilateral temporal lobe decreased significantly（P＜0.05）；The agreement index （AI） of three organs at risk volume contoured by four observers increased significantly（P＜0.05）. Based on analysis of dosage， there was significant difference for Dmean of right inner ear and V50 of bilateral inner ear except Dmean of left inner ear（P＜0.05）； There was no significant difference for Dmean of right middle ear， V50of left middle ear except Dmean of left middle ear（P＞0.05）； Significant difference were observed for D2 of right temporal lobe and Dmax， D1 and D2 of left temporal lobe （P＜0.05）；The recommendation significantly improved Kappa value of Dmean and V50 of bilateral inner ear， V50 of left middle ear， D1 of left temporal lobe and D2 of right temporal lobe（P＜0.05）.
Conclusion Different volumetric and dosimetric parameters may hinder the different researchers. The unified and standardized training can improve the accuracy and concordance of delineation, which caused evaluation of treatment plannings more objectively.