Gastrointestinal stromal tumors: computed tomographic features and prediction of malignant risk from computed...

Gastrointestinal stromal tumors (GISTs) are specific, generally Kit (CD117)-positive, mesenchymal tumors of the gastrointestinal tract encompassing a majority of tumors previously considered gastrointestinal smooth muscle tumors. Our aim was to characterize the computed tomographic findings and predict malignant risk from computed tomography for the evaluation of GISTs.

The computed tomographic images of 39 patients with pathologically and immunohistochemically proven GISTs were reviewed by 2 radiologists, and the final interpretations were reached by consensus. Images were assessed for the size, contour, growth pattern, boundary, degree of enhancement, and necrosis of the tumors. The presence of calcification within the lesions, abdominal lymphadenopathy, ascites, and bowel obstruction were also recorded. Categorical variables were compared using Fishers exact test. Univariate and multivariate logistic regression analyses were used for selection of significant predictors of high-risk malignancy. In addition, the relationships between computed tomographic features and tumor size were assessed by means of nonparametric univariate analysis with the MannWhitney U test and KruskalWallis test.

Both old age and larger tumor size (>or= 5 cm) were statistically significant in the univariate logistic analysis for high-risk malignant tumors (p < 0.25). However, in multivariate logistic regression, only larger tumor size (>or= 5 cm) was found to have final statistical significance for high-risk malignant GISTs (p < 0.05). In addition, more exophytic growth pattern (p < 0.01), more lobulated appearance (p < 0.01), good enhancement (p < 0.05),and more necrosis (p < 0.01) of masses were more often observed in larger GISTs than small ones on computed tomography.

Larger tumor size (>or= 5 cm) was found to have a predictive value with respect to high-risk malignant GISTs.