SCAR ASSESSMENT AFTER PIECEMEAL ENDOSCOPIC MUCOSAL RESECTION - INTEROBSERVER AGREEMENT IN HISTOLOGICAL RECURRENCE PREDICTION

Aims Piecemeal endoscopic mucosal resection (pEMR) allows resection of larger non-invasive colorectal lesions. Adenoma recurrence is an important limitation and occurs in≤20%. It is unclear whether scar assessment after pEMR has to include scar biopsy in the absence of suspected recurrence. The present study aimed to evaluate interobserver agreement in adenoma recurrence based on endoscopic scar assessment, among non-experts in EMR.

Methods An image based offline analysis was performed to evaluate adenoma recurrence prediction and assess the interobserver agreement within a heterogeneous group of participants, mostly composed by non-experts in EMR. Thirty-eight high-definition pEMR scar images were selected from the cohort of patients under pEMR surveillance. The images were subsequently randomized, though simple random sampling. Participants were blinded for the randomization sequence, patient-related factors and final histology.

Results Endoscopic prediction of recurrence had high sensitivity (55.6%-88.9%) and specificity (85.0-95.0%). Negative predictive value (70.4%-94.1%) and positive predictive value (82.4-93.8%) were also calculated. Overall coefficient of agreement (k) between the participants was 0.806

(p< 0.001), indicating substantial agreement. Considering image modalities higher agreement (k) was achieved under NBI visualization (k-0.837; p < 0.001).

Conclusions Our study suggests that optical diagnosis of recurrent adenoma has high interobserver agreement between experts and non-experts in EMR.