External validation of COMPASS and BIOSCOPE prognostic scores in colorectal peritoneal metastases treated with cytoreductive surgery and hyperthermic intraperitoneal chemotherapy (HIPEC)

Introduction: The selection of patients undergoing cytoreductive- surgery (CRS) followed by hyperthermic intraperitoneal chemotherapy (HIPEC) is crucial. BIOSCOPE and COMPASS are prognostic scores designed to stratify survival into four classes according to clinical and pathological features. The purpose of this study is to analyze the prognostic role of these scores using a large cohort of patients as an external reference. Methods: Overall survival analysis was performed using Log-Rank and Kaplan-Meier curves for each score. The probability of survival at 12, 36, and 60 months was tested using receiver operating characteristic (ROC) curves to determine sensitivity and specificity. Results: From the validation cohort of 437 patients, the analysis included 410 patients in the COMPASS group and 364 patients in the BIOSCOPE group (100% data completeness). We observed a different patient distribution between classes (high-risk for BIOSCOPE compared to COMPASS, p = 0.0001). Nevertheless, both COMPASS and BIOSCOPE effectively stratified overall survival (Log-Rank, p = 0.0001 in both cases), with a lack of discrimination between COMPASS classes II and III (p = n.s.). COMPASS at 12 m and BIOSCOPE at 60 m showed the best performance in terms of survival prediction (AUC of 0.82 and 0.81). The specificity of the two tests is good (median 81.3%), whereas sensibility is quite low (median 64.2%). Conclusion: Following external validation in a large population of patients with CRC-PM who are eligible for surgery, the COMPASS and BIOSCOPE scores exhibit high inter-test variability but effectively stratify cancer-related mortality risk. While the quality of the scores is similar, BIOSCOPE shows better inter-tier differentiation, suggesting that tumor molecular classification could improve test discrimination capability. More powerful stratification scores with the inclusion of novel predictors are needed.