CT of gastric volvulus: Interobserver reliability, radiologists' accuracy, and imaging findings

OBJECTIVE. The objective of this study was to identify CT findings and determine interobserver reliability of surgically proven gastric volvulus. MATERIALS AND METHODS. This single-center retrospective study included 30 patients (21 women, nine men; mean age, 73 years old) with surgically proven gastric volvulus who underwent preoperative CT and 31 age- and sex-matched control subjects (21 women, nine men; mean age, 74 years old) with large hiatal hernias who were imaged for reasons other than abdominal pain. Two blinded radiologists reviewed the CT images and recorded findings of organoaxial and mesenteroaxial gastric volvulus and ischemia. Interobserver reliability, reader accuracy, sensitivity, specificity, and likelihood ratios of each CT finding were calculated. RESULTS. The radiologists were overall 90% accurate (55/61; six false-negatives per reader) in identifying gastric volvulus. Interobserver agreement was substantial (κ = 0.71) for identifying the presence or absence of gastric volvulus. Agreement for most CT findings of gastric volvulus (11/14, 79%) was excellent (5/14, 36%) or substantial (6/14, 43%); the remaining findings showed moderate agreement (3/14, 21%). The most frequent and sensitive CT findings of volvulus with high positive likelihood ratios were stenosis at the hernia neck (reader 1, sensitivity = 80%, positive likelihood ratio = 26.66; reader 2, sensitivity = 77%, positive likelihood ratio = 12.83) and transition point at the pylorus (reader 1, sensitivity = 80%, positive likelihood ratio = 17; reader 2, sensitivity = 70%, positive likelihood ratio = 15). The presence of perigastric fluid or a pleural effusion were significantly more frequent in patients with ischemia at surgical pathology (p < 0.05 in all comparisons, both radiologists). CONCLUSION. In our series, CT showed substantial interobserver agreement and fair accuracy in identifying the presence of gastric volvulus.