Predictive Factors Of Gastrointestinal Stromal Tumors In A Consecutive Series Of Patients Evaluated By Endoscopic Ultrasound

Background: GISTs has been increasingly recognized since the development of immunohistochemical techniques allows for determination of CD117 or c-kit, an specific marker for GISTs. On the other hand, the diagnosis of GIST has important prognostic and therapeutic implications. Endosonographic predicting factors of GIST have never been investigated. Aim: To identify predictive factors of GIST in a consecutive series of patients with subepithelial tumors evaluated by EUS. Patients and Methods: We included all consecutive patients referred for EUS evaluation of a subepithelial tumor from September 1997 to December 2003 and who had histological or cytological definitive diagnosis. EUS was performed under conscious sedation with a radial echoendoscop (GF UM20 or GF UM160, Olympus). Variables analysed were: age, gender and US findings such as tumour size, location, echogenicity, echo pattern , margins, layer of origin, presence of calcifications, intratumoral nodules, cystic areas and lymph nodes. Gold standard used was histology of resected specimens (n = 58), endoscopic biopsies (n = 10), endoscopic resection (n = 5), EUS-guided trucut biopsy (n = 2) and EUS-guided FNA cytology (n = 5). Immunohistochemical determination of C-kit was positive in all cases. Results: Eighty six patients (44M/42F, mean age 59 + 16) with 88 lesions (44 GISTs/44no-GISTs) were evaluated. Final diagnosis was as follows: GISTs (n = 44), carcinoid tumour (n = 9), lipoma (n = 7), heterotopic pancreas (n = 7), leiomyoma (n = 4), esophageal cancer (n = 4), endometriosis (n = 2), granulous cell tumour (n = 2), metastases (n = 2), and miscellanea (n = 7). Malignancy was significantly higher in GISTs (18 vs 9, p = 0.04). Univariate analysis showed that GISTs were larger (46 ± 17 mm vs 30 ± 20 mm, p = 0.001), more heterogeneous (24 vs 11, p = 0.001), more often located in the stomach (39 vs 28, p = 0.006) originating in the muscularis propria (27 vs 11, p = 0.001) and with cystic areas (28 vs 6, p = 0.001) than non-GISTs tumors. The gastric location, origin in muscularis propria and the presence of cystic areas were the only variables that independently predicted GIST in the multivariate analysis. The ROC curve made with these three variables showed an AUC of 0.810. Conclusions: Localization in the stomach, origin in the muscularis propria and the presence of cystic areas are good predictors for GIST. This has important impact in terms of prognosis and therapeutic management. Background: GISTs has been increasingly recognized since the development of immunohistochemical techniques allows for determination of CD117 or c-kit, an specific marker for GISTs. On the other hand, the diagnosis of GIST has important prognostic and therapeutic implications. Endosonographic predicting factors of GIST have never been investigated. Aim: To identify predictive factors of GIST in a consecutive series of patients with subepithelial tumors evaluated by EUS. Patients and Methods: We included all consecutive patients referred for EUS evaluation of a subepithelial tumor from September 1997 to December 2003 and who had histological or cytological definitive diagnosis. EUS was performed under conscious sedation with a radial echoendoscop (GF UM20 or GF UM160, Olympus). Variables analysed were: age, gender and US findings such as tumour size, location, echogenicity, echo pattern , margins, layer of origin, presence of calcifications, intratumoral nodules, cystic areas and lymph nodes. Gold standard used was histology of resected specimens (n = 58), endoscopic biopsies (n = 10), endoscopic resection (n = 5), EUS-guided trucut biopsy (n = 2) and EUS-guided FNA cytology (n = 5). Immunohistochemical determination of C-kit was positive in all cases. Results: Eighty six patients (44M/42F, mean age 59 + 16) with 88 lesions (44 GISTs/44no-GISTs) were evaluated. Final diagnosis was as follows: GISTs (n = 44), carcinoid tumour (n = 9), lipoma (n = 7), heterotopic pancreas (n = 7), leiomyoma (n = 4), esophageal cancer (n = 4), endometriosis (n = 2), granulous cell tumour (n = 2), metastases (n = 2), and miscellanea (n = 7). Malignancy was significantly higher in GISTs (18 vs 9, p = 0.04). Univariate analysis showed that GISTs were larger (46 ± 17 mm vs 30 ± 20 mm, p = 0.001), more heterogeneous (24 vs 11, p = 0.001), more often located in the stomach (39 vs 28, p = 0.006) originating in the muscularis propria (27 vs 11, p = 0.001) and with cystic areas (28 vs 6, p = 0.001) than non-GISTs tumors. The gastric location, origin in muscularis propria and the presence of cystic areas were the only variables that independently predicted GIST in the multivariate analysis. The ROC curve made with these three variables showed an AUC of 0.810. Conclusions: Localization in the stomach, origin in the muscularis propria and the presence of cystic areas are good predictors for GIST. This has important impact in terms of prognosis and therapeutic management.