Accuracy of histopathology evaluation in diminutive colonic polyps diagnosed as neoplastic by computer-aided characterisation

Abstract

The removal of colorectal polyps followed by histological evaluation is the current standard of care. Intracolonoscopy computer-aided characterisation (computer-assisted diagnosis (CADx)) has emerged as an alternative strategy to possibly avoid histopathology after removal. However, each of these approaches can be limited by cases in which a diagnosis is not possible or inaccurate. We conducted a post hoc analysis of a prospective clinical study involving 249 diminutive polyps diagnosed by CADx in 164 consecutive patients. Of those polyps, 29 were diagnosed as ‘normal mucosa’ on histopathology (11.6%). Polyps diagnosed as ‘normal mucosa’ were reviewed by two independent blinded pathologists, and 24.1% were reassessed as adenomas after additional sections were performed. Besides superficial sectioning (24.1%), fragmentation (44.8%) and inadequate specimen preparation and handling (31.0%) were considered as reasons for misdiagnosis. Reassessment decreased the overall accuracy of histopathology to 88.4% (95% CI 83.7% to 92.1%). On the other hand, in these 29 cases, CADx diagnosed all polyps as neoplastic, accounting for an overall accuracy of 85.5% (95% CI 80.2% to 89.8%). Therefore, a combined approach may allow for improved diagnostic accuracy. The current standard of care for colorectal polyps involves analysis in a pathology laboratory. However, this process includes several steps including resection and retrieval, during which specimens may fracture or be lost, as well as processing, embedding, sectioning and final analysis, all of which can introduce errors and misdiagnoses. Recent publications have indicated that the final pathology report can indeed be incorrect after polyp resection, with cases discordant with optical diagnosis undergoing a series of additional reviews, which continued until concordance was achieved.1 This approach …