BACKGROUND: Small, single-institution studies have suggested risk factors for bowel ischemia/necrosis (I/N) in patients with computed tomography (CT) findings of pneumatosis (PN) and portal venous gas (PVG). Here, analysis has been expanded in a large, multicenter study. MATERIALS & METHODS: Logistic regression models and receiver operating characteristic curves were used to construct a scoring system for I/N in cases of PN/PVG. RESULTS: Of 265 patients with PN/PVG identified, 209 had adequate data. In unadjusted analyses the following variables were significantly associated with I/N: age, peritoneal signs, ascites, the presence of both PVG and PN, blood urea nitrogen (BUN), CO2, albumin, and a history of hypertension, myocardial infarction, or stroke. In contrast, the CT findings of mesenteric stranding, bowel-wall thickening, and type of PN were not associated with I/N. In adjusted analyses, three variables were significantly associated with I/N: age ≥60 y (odds ratio = 2.51, 95% confidence interval: 1.26-4.97), peritoneal signs (10.58, 4.23-26.4), and BUN >25 mg/dL (3.08, 1.54-6.17), whereas presence of both PN and PVG (versus only one) was associated with an increase (but not statistically significant increase) in odds (2.01, 0.94-4.36). Although several ad hoc models were used to maximize diagnostic ability, with maximal odds ratio = 174, the areas of receiver operating characteristic curves were all below 0.80, revealing suboptimal accuracy to diagnose I/N. CONCLUSIONS: Older age, peritoneal signs, and high BUN are associated with I/N, suggesting an ability to predict which patients need operation. CT findings traditionally suggestive of ischemic PN/PVG, however, do not diagnose I/N accurately enough to reliably identify patients needing operation.
BACKGROUND: Small, single-institution studies have suggested risk factors for bowel ischemia/necrosis (I/N) in patients with computed tomography (CT) findings of pneumatosis (PN) and portal venous gas (PVG). Here, analysis has been expanded in a large, multicenter study. MATERIALS & METHODS: Logistic regression models and receiver operating characteristic curves were used to construct a scoring system for I/N in cases of PN/PVG. RESULTS: Of 265 patients with PN/PVG identified, 209 had adequate data. In unadjusted analyses the following variables were significantly associated with I/N: age, peritoneal signs, ascites, the presence of both PVG and PN, blood ureanitrogen (BUN), CO2, albumin, and a history of hypertension, myocardial infarction, or stroke. In contrast, the CT findings of mesenteric stranding, bowel-wall thickening, and type of PN were not associated with I/N. In adjusted analyses, three variables were significantly associated with I/N: age ≥60 y (odds ratio = 2.51, 95% confidence interval: 1.26-4.97), peritoneal signs (10.58, 4.23-26.4), and BUN >25 mg/dL (3.08, 1.54-6.17), whereas presence of both PN and PVG (versus only one) was associated with an increase (but not statistically significant increase) in odds (2.01, 0.94-4.36). Although several ad hoc models were used to maximize diagnostic ability, with maximal odds ratio = 174, the areas of receiver operating characteristic curves were all below 0.80, revealing suboptimal accuracy to diagnose I/N. CONCLUSIONS: Older age, peritoneal signs, and high BUN are associated with I/N, suggesting an ability to predict which patients need operation. CT findings traditionally suggestive of ischemic PN/PVG, however, do not diagnose I/N accurately enough to reliably identify patients needing operation.