Performance of Multidetector Computed Tomography and Negative Versus Positive Enteric Contrast for Evaluation of Gastrointestinal Neuroendocrine Neoplasms

Background Routine computed tomography (CT) scans are thought to have poor performance for detection of gastrointestinal (GI) neuroendocrine neoplasms (NENs), which leads to delayed workup. Detection of even 1 bowel tumor can guide diagnostic workup and management. The purposes of this study were to assess the accuracy of multidetector computed tomography (MDCT) and to compare negative versus positive enteric contrast in detecting at least 1 GI tumor per patient with suspected or confirmed diagnosis of a NEN. Methods This retrospective study included 107 patients with intravenous and oral contrast (65 positive, 40 negative, and 2 no oral contrast) abdominopelvic MDCT. Two abdominal radiologists independently analyzed the CTs for detection and localization of bowel NENs. Surgical pathology was considered the reference standard. Analyses included kappa and summary statistics, McNemar test, Pearson chi(2) test, and Fisher exact test. Results Among the 107 CT scans, there were 30 pathology negative studies and 77 studies with positive pathology for GI NEN. Interreader agreement for CT evaluation was substantial (kappa = 0.61). At least 1 GI NEN per patient was detected with 51% to 53% sensitivity, 87% to 93% specificity, 91% to 95% positive predictive value (PPV), 42% negative predictive value, and 63% accuracy for each reader, and 57% accuracy when only the concordant (ie, matching) results of the 2 readers were considered. Computed tomography scans with negative enteric contrast had significantly higher sensitivity for concordant results than CTs with positive enteric contrast (58% vs 30%, P = 0.01). Specificity (100% vs 95%, P = 0.5), PPV (100% vs 93%, P = 0.49), negative predictive value (39% vs 39%, P = 0.99), and accuracy (67% vs 51%, P = 0.10) were not significantly different for negative versus positive enteric contrast for the concordant results. There was no significant difference in GI NEN localization between the readers. Conclusions Routine MDCT with either positive or negative enteric contrast can detect at least 1 GI tumor per patient with more than 90% PPV and more than 50% accuracy in patients suspected of GI NEN. Using negative enteric contrast improves sensitivity for GI NEN versus positive enteric contrast. In addition, there is high accuracy in localizing the bowel tumor with positive or negative enteric contrast, which may guide surgery. Radiologists should have heightened awareness that evaluating such scans closely may lead to detection of primary bowel NENs at a higher rate than previously reported.