Supplementary Material for: Comparison of pre-endoscopic C-WATCH score with established risk assessment tools in patients with upper gastrointestinal bleeding

Background: Use of risk scores for early assessment of patients with upper gastrointestinal bleeding (UGIB) is recommended by various guidelines. We compared C-WATCH score with Glasgow Blatchford Score (GBS), Rockall score (RS) and pre-endoscopic RS (p-RS). Methods: Patients with UGIB between January to December 2017 were retrospectively analyzed regarding 30 day-mortality and composite endpoints risk of complications and need for intervention using areas under the receiver operating characteristics curve (AUROC). Subgroup analysis was conducted for patients with UGIB on admission and in-hospital UGIB. Results: 252 patients were identified (67.5% men, mean age 63.8±14.9 years). In-hospital UGIB occurred in 49.6%. AUROCs for 30 day-mortality, risk of complications and need for intervention (not applicable to RS) were 0.684 (95%-CI 0.606-0.763), 0.665 (95%-CI 0.594-0.735) and 0.694 (95%-CI 0.612-0.775) for C-WATCH score, 0.724 (95%-CI 0.653-0.796) and 0.751 (95%-CI 0.687-0.815) for RS, 0.652 (95%-CI 0.57-0.735), 0.653 (95%-CI 0.579-0.727) and 0.673 (95%-CI 0.602-0.745) for p-RS and 0.652 (95%-CI 0.572-0.732), 0.663 (95%-CI 0.592-0.734) and 0.752 (95%-CI 0.683-0.821) for GBS. RS outperformed pre-endoscopic scores in predicting risk of complications, while there were no significant differences between pre-endoscopic scores except GBS outperforming p-RS in predicting need for intervention. The subgroup analysis obtained similar results. Positive predictive values for patients with estimated low risk for all three endpoints (C-WATCH score ≤1, RS ≤2, p-RS <1 and GBS ≤1) were 89%, 69%, 78% and 92%. Conclusion: C-WATCH score performed similar to established pre-endoscopic risk scores in patients with UGIB regarding relevant patient related endpoints with no significant differences between both subgroups.

背景：各类临床指南均推荐采用风险评分对上消化道出血（Upper Gastrointestinal Bleeding, UGIB）患者开展早期评估。本研究将C-WATCH评分与格拉斯哥布拉奇福德评分（Glasgow Blatchford Score, GBS）、罗卡尔评分（Rockall Score, RS）及内镜前罗卡尔评分（pre-endoscopic RS, p-RS）进行了对比分析。
方法：本研究对2017年1月至12月收治的上消化道出血患者进行回顾性分析，以受试者工作特征曲线下面积（Receiver Operating Characteristic Curve, AUROC）评估患者的30天死亡率、并发症复合终点风险及干预需求，并针对入院时即发生上消化道出血与住院期间新发上消化道出血的患者开展亚组分析。
结果：本研究共纳入252例患者，其中男性占比67.5%，平均年龄为63.8±14.9岁；住院期间新发上消化道出血者占49.6%。针对30天死亡率、并发症风险及干预需求（罗卡尔评分不适用于干预需求预测），C-WATCH评分的受试者工作特征曲线下面积分别为0.684（95%置信区间0.606～0.763）、0.665（95%置信区间0.594～0.735）及0.694（95%置信区间0.612～0.775）；罗卡尔评分的对应评估值为0.724（95%置信区间0.653～0.796）与0.751（95%置信区间0.687～0.815）；内镜前罗卡尔评分的对应评估值分别为0.652（95%置信区间0.570～0.735）、0.653（95%置信区间0.579～0.727）及0.673（95%置信区间0.602～0.745）；格拉斯哥布拉奇福德评分的对应评估值分别为0.652（95%置信区间0.572～0.732）、0.663（95%置信区间0.592～0.734）及0.752（95%置信区间0.683～0.821）。罗卡尔评分在并发症风险预测方面优于内镜前各类评分；除格拉斯哥布拉奇福德评分在干预需求预测方面优于内镜前罗卡尔评分外，其余内镜前评分之间无显著统计学差异。亚组分析得到了一致的研究结果。三类临床终点均评估为低风险的患者（C-WATCH评分≤1、罗卡尔评分≤2、内镜前罗卡尔评分<1且格拉斯哥布拉奇福德评分≤1）的阳性预测值分别为89%、69%、78%及92%。
结论：在上消化道出血患者中，C-WATCH评分在相关临床终点的预测效能与已确立的内镜前风险评分相当，且两个亚组之间无显著统计学差异。