BACKGROUND AND AIMS
The identification of a demarcated area (DA), where a regular microvascular or pit pattern appears disordered, is a fundamental principle of optical evaluation and can predict the presence of submucosal invasive cancer (SMIC) in large (≥20 mm) nonpedunculated colorectal polyps (LNPCP).1-3 While virtual chromoendoscopy (VCE) is the primary method for performing optical evaluation, it has shown modest performance for LNPCP. Dye-based chromoendoscopy (DBC) is an alternative which has shown excellent performance characteristics with traditional magnification.4 The authors therefore sought to evaluate the incremental benefit of DBC in addition to high-definition white light (HDWL) and VCE for DA identification and the prediction of SMIC in LNPCP.
A prospective observational study of consecutive LNPCP at a single tertiary referral centre was performed.5Prior to resection, all LNPCP were initially assessed for a DA with HDWL plus VCE (Narrow Band Imaging [Olympus Corporation, Tokyo, Japan]) and then by DBC, by two trained independent observers. DA diagnostic performance (sensitivity, specificity, positive predictive value, and negative predictive value) and interobserver agreement (k statistic) were calculated.
Over 22 months to September 2019, 205 consecutive LNPCP (median size: 38mm; interquartile range: 30-50 mm; 46.8% right colon) were enrolled. The overall frequency of SMIC was 9.3%. The absence of a DA had a negative predictive value of 95.6% (95% confidence interval: 92.2–97.6%) for SMIC, independent of the use of DBC. A high rate of interobserver agreement was recorded for the identification of a DA with HDWL plus VCE (99.5%; k=0.98) and with HDWL plus VCE plus DBC (99%; k=0.95).
Lesion assessment is a critical component in determining the suitability of endoscopic resection for LNPCP.6-8 In this study, the authors demonstrated that the use of HDWL combined with VCE had a high rate of interobserver agreement for DA identification, independent of the use of DBC. More importantly, they showed that the absence of a DA on the surface of LNPCP is a very strong predictor for the absence of SMIC, also independent of the use of DBC. Taken together, there is no role for the universal application of DBC in addition to HDWL plus VCE for LNPCP. Moreover, the results supported that LNPCP not demonstrating a DA, and in absence of lesion characteristics associated covert SMIC, can be safely resected by piecemeal endoscopic mucosal resection. These study findings do require validation outside of an expert setting and provide an avenue for future research.
In conclusion, the absence of a DA within LNPCP is strongly predictive for the absence of SMIC. It can be determined without the need for DBC with a high rate of interobserver agreement among experts.