Scoring systems for differentiating gastrointestinal stromal tumors and schwannomas from leiomyomas in the stomach

There is no practical predictive model for the diagnosis of gastrointestinal stromal tumors (GISTs). To establish a practical predictive model for the diagnosis of subepithelial lesions in the stomach, we reviewed patients with GISTs (n = 89), schwannomas (n = 7), and leiomyomas (n = 28). The tumor was more frequently found along the gastric cardia in the leiomyoma group (57.1%) than in the GIST/schwannoma group (2.1%, P < .01). Contrast enhancement (57.3% vs 0%, P < .01) and intra-tumoral necrosis (34.4% vs 0.0%, P < .01) were more frequently observed in the GIST/schwannoma group than in the leiomyoma group. On endoscopic ultrasonography, 58.3% of GISTs/schwannomas showed uneven echogenicity, whereas the echogenicity was uneven in 21.4% of leiomyomas (P < .01). There were no differences between the tumor color and the presence or absence of ulcer formation, tumor bleeding, irregularity of the tumor margin, cystic spaces, and hyperechoic spots between the 2 groups. Based on these results, we developed a 2-step diagnostic algorithm for GISTs/schwannomas. The first step comprises 1 endoscopic feature: a cardiac or non-cardiac location. Tumors with a cardiac location were judged as leiomyomas and those with a non-cardiac location were judged as GISTs/schwannomas, with 96.9% sensitivity and 57.1% specificity for GIST/schwannoma diagnosis. The second step comprises a combination of endoscopic (non-cardiac location), radiologic (positive contrast enhancement and intra-tumoral necrosis), and endosonographic (uneven echogenicity) features for a total of 4 points. We assigned 1 point to each feature. Tumors with scores of 2 to 4 were judged as GISTs/schwannomas, with 81.3% sensitivity and 92.9% specificity for GIST/schwannoma diagnosis. Our predictive model will be a practical guide for the management of gastric subepithelial lesions.