An Evaluation of Multidetector Computed Tomography in Detecting Pancreatic Injury: Results of a Multicenter AAST Study

Abstract

Efforts to determine the suitability of low-grade pancreatic injuries for nonoperative management have been hindered by the inaccuracy of older computed tomography (CT) technology for detecting pancreatic injury (PI). This retrospective, multicenter American Association for the Surgery of Trauma-sponsored trial examined the sensitivity of newer 16- and 64-multidetector CT (MDCT) for detecting PI, and sensitivity/specificity for the identification of pancreatic ductal injury (PDI). Patients who received a preoperative 16- or 64-MDCT followed by laparotomy with a documented PI were enrolled. Preoperative MDCT scans were classified as indicating the presence (+) or absence (-) of PI and PDI. Operative notes were reviewed and all patients were confirmed as PI (+), and then classified as PDI (+) or (-). As all patients had PI, an analysis of PI specificity was not possible. PI patients formed the pool for further PDI analysis. As sensitivity and specificity data were available for PDI, multivariate logistic regression was performed for PDI patients using the presence or absence of agreement between CT and operative note findings as an independent variable. Covariates were age, gender, Injury Severity Score, mechanism of injury, presence of oral contrast, presence of other abdominal injuries, performance of the scan as part of a dedicated pancreas protocol, and image thickness < or =3 mm or > or =5 mm. Twenty centers enrolled 206 PI patients, including 71 PDI (+) patients. Intravenous contrast was used in 203 studies; 69 studies used presence of oral contrast. Eight-nine percent were blunt mechanisms, and 96% were able to have their duct status operatively classified as PDI (+) or (-). The sensitivity of 16-MDCT for all PI was 60.1%, whereas 64-MDCT was 47.2%. For PDI, the sensitivities of 16- and 64-MDCT were 54.0% and 52.4%, respectively, with specificities of 94.8% for 16-MDCT scanners and 90.3% for 64-MDCT scanners. Logistic regression showed that no covariates were associated with an increased likelihood of detecting PDI for either 16- or 64-MDCT scanners. The area under the curve was 0.66 for the 16-MDCT PDI analysis and 0.77 for the 64-MDCT PDI analysis. Sixteen and 64-MDCT have low sensitivity for detecting PI and PDI, while exhibiting a high specificity for PDI. Their use as decision-making tools for the nonoperative management of PI are, therefore, limited.