Objective To evaluate the respiration-induced dosimetric variance in 3DCRT for midthoracic esophageal carcinoma, in order to guide the radiation oncologist to choose the expansion margin. Methods Ten patients with mid-thoracic esophageal carcinoma were scanned by multi-spiral CT simulator respectively in free breathing ( FB), breath-hold after normal inspiration and expiration ( IBH and EBH )with the same scanning range. Then the CT images of three series were transferred to the treatment planning system. The target volume was outlined following the same standard. Plan1 was designed in the images of FB and transported completely to the images of IBH and EBH as Plan2 and Plan3 respectively to observe the dosimetric variance in target volume. Results For GTV, there was a statistical difference only in V100 of the three plans ( H = 6.423, P = 0.040 ) and no significant difference was found in other indexes. For CTV, the V100 and V95 were better in Plan1 (F=3.992, P=0.030; H=9.920, P=0.007) and no significant difference was found in other indexes. While ()TV, the Dmin, V100 and V95 was better in Plan1 ( F = 3.677, P = 0.039; F = 4.539, P = 0.020; H = 6.846, P = 0.033 ) and no significant difference was found in other indexes. There were no significant differences in all the indexes for the spinal cord and lung in the three plans. Conclusions The change in dose distribution was not so much with the standard expansion. It can meet the needs of clinical treatment.