Poorly differentiated component in gastric pinch biopsies predicts submucosal invasion.

BACKGROUND: Endoscopic resection has become standard therapy for selected patients with early gastric carcinoma (EGC). However, the preoperative diagnostic accuracy for excluding submucosal (SM) invasion is not precise. Moreover, histologic features predicting SM invasion in gastric carcinomas (SMiGC) have not been studied extensively. METHODS: Pre-treatment gastric biopsies from 60 patients with SM invasion who underwent endoscopic resection were reviewed and compared to 58 biopsies of lesions confirmed to be intramucosal carcinomas (IMC). For validation of the results, an independent cohort consisting of 616 gastric biopsies confirmed as EGC were analyzed. For statistical analyses, χ-square test, Fisher's exact test and multiple logistic progression tests were used. RESULTS: In the biopsy specimens of patients with SMiGCs, differentiated histology, poorly differentiated component, wisps of muscularis mucosa, tumor cribriforming, papillary architecture, desmoplasia and intraglandular eosinophilic necrotic debris (IEND) were observed in 96.7%, 36.7%, 16.7%, 16.7%, 23.3%, 40%, and 46.7% of cases, respectively, while the same features were observed in 100%, 5.2%, 0%, 1.7%, 5.2%, 19%, and 22.4% of biopsies with IMC. In multivariate analyses, poorly differentiated component [odds ratio (OR), 9.59, p = 0.002], IEND [OR, 6.23, p = 0.012], tumor cribriforming [OR, 4.66, p = 0.03] and papillary architecture [OR, 5.52, p = 0.018] were significantly associated with the detection of SM invasion. In the validation cohort, poorly differentiated component (p = 0.003) and papillary architecture (p = 0.008) remained significant. CONCLUSION: Poorly differentiated component and papillary architecture are significant histopathologic predictors of SM invasion in pretreatment gastric biopsies of lesions considered for endoscopic therapy. Additional prospective studies are warranted to confirm our findings. VIRTUAL SLIDE: The virtual slide(s) for this article can be found here: http://www.diagnosticpathology.diagnomx.eu/vs/1588557731103084.