Introduction
Endoscopic characterization aims to predict the histology of a lesion and thus the risk of cancer. This characterization step is essential, especially in the case of large superficial colorectal lesions, as the choice of resection technique directly depends on it. The JNET (Japan NBI Expert Team) classification is the classification proposed by European and Japanese guidelines to select the optimal therapeutic approach between EMR (endoscopic mucosal resection), ESD (endoscopic submucosal dissection), or surgery, but has not been evaluated in large laterally spreading lesions (LSTs) (>20 mm). We therefore evaluated the diagnostic performance of the JNET classification for predicting the risk of submucosal invasive cancer (SMIC) in LSTs >20 mm.
Aims & Methods
Single-center, prospective, observational study including all patients referred for endoscopic resection of lesions >20 mm, between 2017 and 2022. All LST samples with SMIC underwent a secondary pathological examination to establish the correlation between "covert" carcinoma (SMICs categorized as JNET 2A) and "buried" carcinoma (existence of SMICs more than 0.5 mm from the surface). The primary outcome was the diagnostic accuracy of the JNET classification for histological prediction of LSTs >20 mm. The secondary objective was to assess the combined performance of the JNET classification and pejorative macroscopic (macronodule >1 cm, Paris 0-IIc area, JNET 2B area, Non-Granular LST) criteria also called the CONECCT classification. We also attempted to determine the characteristics and risk factors of LSTs with covert or buried carcinoma, and to identify potential associations between the two.
Results
972 lesions were removed by en-bloc ESD, of which 110 (11.3%) of these lesions contained SMIC. The median lesion size was 57 mm. Of the 110 submucosal cancers, 33 had covert carcinoma and 35 were classified as buried carcinoma. The sensitivity and specificity for predicting high-grade dysplasia (HGD) to SMIC were 46% and 75% for JNET 2B and 87% and 38% for CONECCT IIC, respectively; and for predicting superficial SMIC were 67% and 67% for JNET 2B and 94% and 27% for CONECCT IIC, respectively. The false negative rates (FNR) (probability of missing SMIC) of JNET 2B and CONECCT IIC were 31% (31/100) and 3% (3/100) for predicting SMIC, respectively; and 33% (16/48) and 6% (3/48) for predicting superficial SMIC, respectively. JNET 2B exhibited significantly higher specificity than CONECCT IIC for predicting HGD to superficial SMIC (75% vs. 38%; p< 0.001) and for predicting isolated superficial cancer (67% vs. 27%; p< 0.0001). And CONECCT IIC was significantly more sensitive than JNET 2B for predicting HGD to superficial SMIC (46% vs. 87%; p< 0.001) and for predicting isolated superficial cancer (67% vs. 97%; p< 0.0001). No association was found between covert carcinoma and buried carcinoma (p 0.889).
Conclusion
The JNET classification has low diagnostic accuracy for predicting the risk of submucosal cancer for large superficial colorectal lesions and for making decisions on en bloc endoscopic treatment using ESD. If a selective ESD and piecemeal EMR strategy is used based on the submucosal cancer risk, the CONECCT classification is preferred to prevent under-treatment of SMIC.
Disclosure
Nothing to declare.
