Background and purpose: Extended-field intensity-modulated radiotherapy (IMRT) and concurrent chemotherapy may lead to severe toxicity. This study aims to investigate the incidence and dosimetric predictors of acute gastrointestinal toxicity in cervical cancer patients. Methods and materials: Dosimetric data of the duodenum and small bowel loop from 273 eligible patients were analyzed. Logistic regression and Pearson's pairwise correlation were used to identify the best predictors. Receiver operating characteristic curves were used to determine the optimal cutoff values. Logistic models were established to predict the normal tissue complication probability. Results: Twenty-six patients experienced grades >= 3 (G3(+)) vomiting, while seven experienced G3(+) diarrhea. Either the duodenal volume receiving >= 29 Gy (V-29; p = 0.002) or V-49 (p = 0.014) could predict the incidence of G3(+) vomiting. The incidence was 1.3 % vs. 12.8 % in patients with V-29 <= 43.5 % vs. > 43.5 % (p = 0.008). For patients with V-49 <= 9.2 % vs. > 9.2 %, the incidence was 7.6 % vs. 16.1 % (p = 0.077). The plans for 22 patients had to be modified due to vomiting. It was found that duodenal V-29 (p = 0.012) was a predictor for replanning, with an incidence of 3.3 % vs. 17.6 % in patients with V-29 <= 51.3 % vs. > 51.3 % (p < 0.001). The V-22 of the small bowel was selected to predict the incidence of G3(+) diarrhea, which were 0.6 % and 5.6 % in patients with V-22 <= 62.3 % and > 62.3 %, respectively (p = 0.031). Conclusion: Limiting V-29 and V-49 of the duodenum, as well as V-22 of the small bowel loop, may help reduce the risk of acute G3(+) gastrointestinal toxicity when treating cervical cancer with extended-field concurrent IMRT.