Predictive value of magnification chromoendoscopy for diagnosing invasive neoplasia in nonpolypoid colorectal lesions and stratifying patients for endoscopic resection or surgery

Background and Study Aims: In nonpolypoid colorectal lesions, the presence of irregular, distorted glands in the colon (a disrupted crypt pattern) on magnification chromoendoscopy (NICE) is strongly associated with submucosal invasive cancer. The aim of the present study was to evaluate the ability of MCE to differentiate between an invasive crypt pattern and a noninvasive crypt pattern, including nonneoplastic lesions, and to assess the ability of this MCE classification to predict invasiveness and allow patients to be selected for endoscopic resection or surgical resection. Patients and Methods: In a prospective study including 1560 colonoscopies, 153 flat or depressed colorectal lesions were evaluated with MCE among 534 colorectal lesions; the remainder had a polypoid appearance. The pit pattern was classified as nonneoplastic (type II) or neoplastic (types III-V), and the latter was subdivided into noninvasive (types III or IV) or submucosally invasive (type V). Lesions with a nonneoplastic and noninvasive neoplastic appearance were resected endoscopically if technically feasible, whereas those with a type V pattern were resected surgically. The resection specimens were analyzed histologically in relation to the Vienna classification. Results: Using this management strategy based on the pit pattern, 86% (n = 70) of the type II lesions were hyperplastic; the remaining II had low-grade intraepithelial neoplasia. Type III and IV lesions (n = 58) represented either low-grade or high-grade intraepithelial neoplasia in 95% of the cases. Three patients had sm1 (n = 2) or sm2/3 invasive cancers. Among the patients with type V lesions (n = 14), 11 had invasive cancers (four sml and seven sm2/3). Endoscopic differentiation based on the pit pattern thus had a positive predictive value (PPV) of 86% and a negative predictive value of 96% for distinguishing between nonneoplastic and neoplastic lesions. The pit pattern criteria for distinguishing between invasive and noninvasive neoplasia (including nonneoplastic lesions), and hence the choice between endoscopic and surgical resection, had a PPV of 79% and a NPV of 98%. Excluding nonneoplastic lesions, the NPV would be 95%. Conclusions: The endoscopic pit pattern on NICE has only a moderate predictive value for nonneoplastic lesions, so that leaving these flat hyperplastic lesions in place on the basis of the endoscopic magnification appearance alone cannot be generally recommended. However, MCE has a good predictive value for guiding management toward either endoscopic resection (if technically feasible) or surgical resection.