Noninvasive prediction model for diagnosing gastrointestinal stromal tumors using contrast-enhanced harmonic endoscopic ultrasound

Abstract

BACKGROUND & AIMS: Subepithelial tumors (SETs) are difficult to diagnose accurately without invasive pathological confirmation. We created a noninvasive prediction model for diagnosing gastrointestinal stromal tumors (GISTs) using contrast-enhanced harmonic endoscopic ultrasound (CEH-EUS).

METHODS: We retrospectively reviewed 176 patients who underwent CEH-EUS from October 2011 to August 2017. Seventy patients with a diagnosis of GIST (n = 37) or leiomyoma (n = 33) were included. The long-to-short axis ratio (LSR) and enhancement patterns (vascularity, diffuse enhancement) on CEH-EUS were assessed. Logistic regression and classification and regression tree (CART) analyses were performed.

RESULTS: The mean age of all patients was 54.9 ± 13.68 years. The GIST group showed significantly higher rates of positive vascularity (81.1% vs. 15.2%, p < 0.001) and diffuse enhancement (51.4% vs. 15.2%, p = 0.001), and lower LSR (1.30 vs. 1.76, p < 0.001). In multivariate logistic regression, positive vascularity (odds ratio [OR] 27.765, 95% confidence interval [CI] 5.336-144.458) and low LSR (OR 18.940, 95% CI 3.623-99.007) were independent predictors of GIST. A noninvasive prediction model for GISTs was developed using the CART model, by allocating patients according to statistically significant variables.

CONCLUSIONS: The LSR and vascularity of SETs on CEH-EUS can be used as parameters for a noninvasive prediction model of GISTs. This model may be helpful in the early identification and treatment of GISTs.