T1 colorectal polyps with at least 1 risk factor for metastasis to lymph node should be treated surgically and are considered endoscopically unresectable. Optical analysis, based on the Narrow-Band Imaging International Colorectal Endoscopic (NICE) classification system, is used to identify neoplasias with invasion of the submucosa that require endoscopic treatment. We assessed the accuracy of the NICE classification, along with other morphologic characteristics, in identifying invasive polyps that are endoscopically unresectable (have at least 1 risk factor for metastasis to lymph node).
We performed a multicenter, prospective study of data collected by 58 endoscopists, from 1634 consecutive patients (examining 2123 lesions) at 17 university and community hospitals in Spain from July 2014 through June 2016. All consecutive lesions >10 mm assessed with narrow-band imaging were included. The primary end point was the accuracy of the NICE classification for identifying lesions with deep invasion, using findings from histology analysis as the reference standard. Conditional inference trees were fitted for the analysis of diagnostic accuracy.
Of the 2123 lesions analyzed, 89 (4.2%) had features of deep invasion and 91 (4.3%) were endoscopically unresectable. The NICE classification system identified lesions with deep invasion with 58.4% sensitivity (95% CI, 47.5-68.8), 96.4% specificity (95% CI, 95.5-97.2), a positive-predictive value of 41.6% (95% CI, 32.9-50.8), and a negative-predictive value of 98.1% (95% CI, 97.5-98.7). A conditional inference tree that included all variables found the NICE classification to most accurately identify lesions with deep invasion (P < .001). However, pedunculated morphology (P < .007), ulceration (P = .026), depressed areas (P < .001), or nodular mixed type (P < .001) affected accuracy of identification. Results were comparable for identifying lesions that were endoscopically unresectable.
In an analysis of 2123 colon lesions >10 mm, we found the NICE classification and morphologic features identify those with deep lesions with >96% specificity-even in non-expert hands and without magnification. ClinicalTrials.gov number NCT02328066.