New Model to Predict Recurrence After Endoscopic Mucosal Resection of Non-pedunculated Colonic Polyps ≥ 20 mm.

BACKGROUND: Polyp recurrence is common after endoscopic mucosal resection (EMR) of non-pedunculated colonic polyps ≥ 20 mm. Two models haven been published for polyp recurrence prediction: Sydney EMR recurrence tool (SERT) and the size, morphology, colonic site, and access to target (SMSA) score. None of these models have been evaluated in a real-world United States (U.S.) cohort. We aimed to evaluate the external validity of these two models and develop a new model. METHODS: Retrospective cohort study of patients with non-pedunculated polyps ≥ 20 mm that underwent EMR between 1/1/2012 and 6/30/2020. Univariate and multivariate analysis were performed to identify predictors of polyp recurrence to build a new model. Receiver Operating Characteristic (ROC) curves for the new model, SERT and a modified version of SMSA were derived and compared. RESULTS: A total of 461 polyps from 461 unique patients were included for analysis. The average polyp size was 29.1 ± 12.4 mm. Recurrence rate at first or second surveillance colonoscopy was 29.0% at a 15.6 months median follow up (IQR 12.3-17.4). A model was created with 4 variables from index colonoscopy: size > 40 mm, tubulovillous adenoma histology, right colon location and piecemeal resection. ROC curves showed that the Area Under the ROC (AUC) for the new model was 0.618, for SERT 0.538 and for mSMSA 0.550. CONCLUSION: SERT score and mSMSA have poor external validity to predict polyp recurrence after EMR of non-pedunculated polyps > 20 mm. Our new model is simpler and performs better in this multiethnic, non-referral cohort from the U.S.

Evolution of endoscopic mucosal resection (EMR) technique and the reduced recurrence of large colonic polyps from 2012 to 2020.

BACKGROUND: Endoscopic mucosal resection (EMR) is an effective method for removing non-pedunculated polyps ≥ 20 mm. We aimed to examine changes in EMR techniques over a 9-year period and evaluate frequency of histologic-confirmed recurrence. METHODS: We identified patients who underwent EMR of non-pedunculated polyps ≥ 20 mm at a safety net and the Veteran's Affairs (VA) hospital in Houston, Texas between 2012 and 2020. Odds ratios (ORs) and 95% confidence intervals (CI) for associations with recurrence risk were estimated using multivariable logistic regression. RESULTS: 461 unique patients were included. The histologic-confirmed recurrence was 29.0% at 15.6 months median follow up (IQR 12.3 - 17.4). Polyps removed between 2018 and 2020 had a 0.43 decreased odds of recurrence vs. polyps removed between 2012 and 2014. The use of viscous lifting agents increased over time (from 0 to 54%), and the use of saline was associated with increased risk of recurrence (OR 2.28 [CI 1.33 - 3.31]). CONCLUSIONS: Histologic-confirmed recurrence after EMR for non-pedunculated polyps ≥ 20 mm decreased over the seven year-period. Saline was associated with a higher risk of recurrence and the use of more viscous agents increased over time.

Using Natural Language Processing to Automatically Identify Dysplasia in Pathology Reports for Patients With Barrett's Esophagus.

BACKGROUND & AIMS: Identifying dysplasia of Barrett's esophagus (BE) in the electronic medical record (EMR) requires manual abstraction of unstructured data. Natural language processing (NLP) creates structure to unstructured free text. We aimed to develop and validate an NLP algorithm to identify dysplasia in BE patients on histopathology reports with varying report formats in a large integrated EMR system. METHODS: We randomly selected 600 pathology reports for NLP development and 400 reports for validation from patients with suspected BE in the national Veterans Affairs databases. BE and dysplasia were verified by manual review of the pathology reports. We used NLP software (Clinical Language Annotation, Modeling, and Processing Toolkit; Melax Tech, Houston, TX) to develop an algorithm to identify dysplasia using findings. The algorithm performance characteristics were calculated as recall, precision, accuracy, and F-measure. RESULTS: In the development set of 600 patients, 457 patients had confirmed BE (60 with dysplasia). The NLP identified dysplasia with 98.0% accuracy, 91.7% recall, and 93.2% precision, with an F-measure of 92.4%. All 7 patients with confirmed high-grade dysplasia were classified by the algorithm as having dysplasia. Among the 400 patients in the validation cohort, 230 had confirmed BE (39 with dysplasia). Compared with manual review, the NLP algorithm identified dysplasia with 98.7% accuracy, 92.3% recall, and 100.0% precision, with an F-measure of 96.0%. CONCLUSIONS: NLP yielded a high degree of sensitivity and accuracy for identifying dysplasia from diverse types of pathology reports for patients with BE. The application of this algorithm would facilitate research and clinical care in an EMR system with text reports in large data repositories.