为了方便选择惠及腹膜转移性结直肠癌（CRPM）患者接受巨大切除手术（CRS）±腹腔内化疗（IPC）治疗，制定了预后评分。其中三种突出的预后评分为腹膜表面疾病严重程度评分（PSDSS）、结直肠腹膜转移预后外科评分（COMPASS）和改良型结直肠-腹膜预后评分（mCOREP）。我们对这些评分进行了外部验证，并比较它们的预测准确性。使用1996年至2018年接受CRS/IPC治疗的连续CRPM患者的数据对COMPASS、PSDSS和mCOREP进行外部验证。分析评估了每个评分在预测以下方面的有效性：（1）开腹-关腹腹腔镜检查，在此过程中被发现不适合治愈性CRS/IPC的患者；（2）手术无效性-接受开腹-关腹腹腔镜检查、姑息性减灭切除术或总存活期不足12个月的患者；（3）总生存期和无复发生存期（OS、RFS）。预测评分计算了外部验证队列的174名患者。COMPASS在预测开腹-关腹腹腔镜检查、失败手术和生存（OS和RFS）方面最准确。开腹-关腹腹腔镜检查的曲线下面积（AUC）为0.78（95%置信区间CI: 0.68-0.87），代表有用的区分度。然而，失败性预测的AUC为0.62（95%CI: 0.52-0.71），OS的C统计量为0.65，仅表示可能有帮助的区分度。RFS的C统计量为0.59，表示区分度较差。尽管COMPASS表现出最佳的统计特性，但对于一些临床相关的结果，准确性仍然较低，因此在临床实践中的适用性有限。© 2023 部分作者。由Wiley Periodicals LLC出版的《外科肿瘤学杂志》发布。

Prognostic scores are developed to facilitate the selection of patients with colorectal cancer peritoneal metastases (CRPM) for treatment with cytoreductive surgery (CRS) ± intraperitoneal chemotherapy (IPC). Three prominent prognostic scores are the Peritoneal Surface Disease Severity Score (PSDSS), the Colorectal Peritoneal Metastases Prognostic Surgical Score (COMPASS), and the modified COloREctal-Pc (mCOREP). We externally validate these scores and compare their predictive accuracy.Data from consecutive CRPM patients who underwent CRS/IPC from 1996 to 2018 was used to externally validate COMPASS, PSDSS, and mCOREP. Analysis evaluated the efficacy of each score in predicting (1) open-close laparotomy-those found at laparotomy to not be eligible for curative intent CRS/IPC, (2) surgical futility-those who underwent open-close laparotomy, palliative debulking surgery, or had an overall survival of less than 12 months, and (3) overall and recurrence-free survival (OS, RFS).Prognostic scores were calculated for the 174-patient external validation cohort. COMPASS was most accurate in predicting open-close laparotomy, futile surgery, and survival (OS and RFS). Area under the curve (AUC) for open-close prediction was 0.78 (95% confidence interval, CI: 0.68-0.87), representing useful discrimination. However, AUC for futility prediction was 0.62 (95% CI: 0.52-0.71), and C-statistic for OS was 0.65 indicating only possibly helpful discrimination. C-statistic for RFS was 0.59 indicating poor discrimination.While COMPASS showed the best statistical behavior, accuracy for several clinically relevant outcomes remains low, and thus applicability to clinical practice limited.© 2023 The Authors. Journal of Surgical Oncology published by Wiley Periodicals LLC.