Accuracy of plain radiographs to exclude the diagnosis of intussusception

OBJECTIVES: To prospectively determine the test characteristics of the 3-view abdominal radiograph to decrease the likelihood of ileocolic intussusception. METHODS: We conducted a prospective cross-sectional study of children aged 3 months to 3 years suspected of having intussusception at a children's hospital emergency department. Clinicians obtained supine, prone, and left lateral decubitus radiographs. We determined the presence or absence of intussusception by air enema, ultrasound, operative report, or clinical follow-up. A masked pediatric radiologist reviewed all radiographs. The criteria evaluated were whether air was visualized in the ascending colon on each view and in the transverse colon on the supine view. RESULTS: Nineteen (14.8%) of 128 patients had intussusception. Using air in the ascending colon on all 3 views as the diagnostic criteria, the test characteristics of the 3-view radiograph were sensitivity, 100% (95% confidence interval [CI], 79.1-100); specificity, 17.4% (95% CI, 11.1-26.1); negative predictive value, 100% (95% CI, 79.1-100); and likelihood ratio of a negative test, 0. When 2 or more of 3 views had air in the ascending colon, sensitivity decreased to 89.5% (95% CI, 75.7-100) and specificity improved to 45.0% (95% CI, 35.6-54.3). Air in the transverse colon had moderate sensitivity, 84.2% (95% CI, 67.8-100), but further improved specificity, 63.3% (95% CI, 54.2-72.4). CONCLUSIONS: The presence of air in the ascending colon on the 3-view abdominal radiograph can decrease the likelihood of or exclude intussusception. When clinical suspicion is low, the presence of specific criterion on a 3-view abdominal radiograph series may obviate the need for further studies.