Clinical Validation Of Acer, GSSED-RE, GSSED-RE2 Scores For Prediction Of Delayed Bleeding After Endoscopic Mucosal Resection Of Large Colorectal Lesions

Aims Several risk factors have been proposed to predict delayed bleeding (DB) after EMR of large colon polyps. Three prediction models (ACER, GSSED-RE, GSSED-RE2) evaluated risk factors, of which GSSED-RE2 reported the highest prediction scores (AUC 0.71). We aimed to validate these three scoring systems for their accuracy to predict DB.

Methods Patients with a ≥20mm non-pedunculated colon polyp evaluated in a multicenter, randomized trial conducted at 18 centers in North America and Spain (4/2013-10/2017) were included. Scoring systems identified age >75yrs, ASA ≥3, polyp size, proximal polyp location, antithrombotic use, epinephrine use, and clip closure as predicting factors. Risk of DB was classified as low, medium, and high. A high risk score was considered accurate if a patient classified as high risk for DB had a severe bleeding event.

Results A total of 919 patients (mean age 65.13 ± 9.62 years, male 59.5 %) were included. DB was observed in 49 (5.3 %) patients. 610 (66.3 %) patients had a large polyp in the proximal. 195 (21.2 %), 58 (6.3 %) and 103 (11.2 %) patients were classified as high risk for DB by ACER, GSSED-RE and GSSED-RE2, respectively. DB, however, occurred in less than 10 % among those patients classified as high risk, resulting in high specificity but low sensitivity across the three scoring systems (table). The overall predictive performance of the three scoring systems among patients classified as high risk was similarly poor with AUROC’s of 0.57 (95 % CI, 0.49-0.66) for ACER, 0.52 (95 % CI, 0.43-0.61) for GSSED-RE, 0.53 (95 % CI, 0.44-0.67) for GSSED-RE2, respectively.

Tab. 1

Scoring system	Patients classified as high risk, n (%)	Patients with true DB event, n (%)	Sensitivity/Specificity/Accuracy	AUROC (95 % CI)
ACER	195 (21.2)	17 (8.7)	35.4/79.3/77.0	0.57 (0.49-0.66]
GSEED-RE	58 (6.3)	5 (8.6)	10.4/93.9/89.5	0.52 (0.43-0.61)
GSEED-RE2	103 (11.2)	8 (7.8)	16.7/89.1/85.3	0.53 (0.44-0.67)

Conclusions Current models are suboptimal in predicting DB after EMR of large colon polyps when validated in a large external cohort of a randomized trial. Improved scoring systems are needed, likely with the use of novel statistical techniques.

Citation: Kamal A, Akshintala VS, Rex DK et al. OP203 CLINICAL VALIDATION OF ACER, GSSED-RE, GSSED-RE2 SCORES FOR PREDICTION OF DELAYED BLEEDING AFTER ENDOSCOPIC MUCOSAL RESECTION OF LARGE COLORECTAL LESIONS. Endoscopy 2021; 53: S84.

Publikationsverlauf

Artikel online veröffentlicht:
19. März 2021