Validation of the BEST-J score, a prediction model for bleeding after endoscopic submucosal dissection for early gastric cancer: a multicenter retrospective observational study

Yasuyuki Kagawa,Masakatsu Fukuzawa,Mitsushige Sugimoto,Daiki Nemoto,Takahiro Muramatsu,Hirokazu Shinohara,Taisuke Matsumoto,Akira Madarame,Hayato Yamaguchi,Kumiko Uchida,Takashi Morise,Yohei Koyama,Akihiko Sugimoto,Yoshiya Yamauchi,Shin Kono,Sakiko Naito,Kei Yamamoto,Yoshiko Kishimoto,Mitsuko Inuyama,Takashi Kawai,Takao Itoi

DOI: https://doi.org/10.1007/s00464-022-09096-y

Abstract:Background: A new scoring system, the BEST-J score, using ten risk factors to assign cases to different post-endoscopic submucosal dissection (ESD) risk groups for bleeding, has been shown to be accurate for risk stratification. We first aimed to validate the BEST-J score at four hospitals not specialized in performing ESD and then aimed to identify other risk factors for post-ESD bleeding. Methods: We evaluated the incidence of post-ESD bleeding in 791 cases of early gastric cancer (EGC) between October 2013 and December 2020 as a retrospective, multi-center observational study conducted at four hospitals. Multivariate logistic regression models to examine the effect of independent variables on post-ESD bleeding firstly included ten possible factors raised by the BEST-J score and secondly included statistically significant (p < 0.01) in univariate analysis. The prediction accuracy of the model was evaluated by receiver-operating characteristic analysis and the areas under the curve (AUC). Results: The incidence of post-ESD bleeding was 4.8% (38/791, 95% confidence interval [CI] 3.4-6.5%). On multivariate analysis, the risk factors were P2Y12 receptor antagonist (odds ratio [OR]: 5.870, 95% CI 1.624-21.219), warfarin (8.382, 1.658-42.322), direct oral anticoagulant (DOAC) (8.980, 1.603-50.322), and tumor location in lower third of stomach (2.151, 1.012-4.571), respectively. When we categorized cases into low-risk by BEST-J score, intermediate-risk, high-risk, and very high-risk groups, the bleeding rates were 2.8%, 7.3%, 12.8%, and 19.0%, respectively. The AUC for our cohort was 0.713 (95% CI 0.625-0.802) for the BEST-J score. In the multivariate analysis in our cohort, the risks were age, body mass index, P2Y12 receptor antagonist, warfarin, DOAC, respectively. Discussion: The BEST-J score is equally accurate in risk stratification of patients with EGC for post-ESD bleeding at non-specialized facilities for ESD as in specialized hospitals. BMI and age may be helpful additional risk factors at hospitals not specialized.