Abstract
BACKGROUND: The aim of the present study was to establish a predictive model to predict the peritoneal cancer index (PCI) preoperatively in patients with pseudomyxoma peritonei (PMP). METHODS: A total of 372 PMP patients were consecutively included from a prospective follow-up database between 1 June 2013 and 1 June 2023. Nine potential variables, namely, gender, age, Barthel Index (BAI), hemoglobin (Hb), albumin (Alb), D-dimer, carcinoembryonic antigen (CEA), carbohydrate antigen 125 (CA 125), and CA 19-9, were estimated using multiple linear regression (MLR) analysis with a stepwise selection procedure. The established MLR model was internally validated using K-fold cross-validation. The agreement between the predicted and surgical PCI was assessed using Bland-Altman plots and intraclass correlation (ICC). A p-value of less than 0.05 was considered statistically significant. RESULTS: Six independent predictors were confirmed by the stepwise MLR analysis with an R (2) value of 0.570. The predicted PCI formula was represented as follows: PCI = 19.567 + 2.091 * Gender (male = 1, female = 0) - 0.643 * Alb +4.201 * Lg (D-dimer) + 2.938 * Lg (CEA) + 5.441 * Lg (CA 125) + 1.802 * Lg (CA 19-9). The agreement between predicted and surgical PCI was assessed using Bland-Altman plots, showing a limit of agreement (LoA) between -15.847 (95%CI: -17.2646 to -14.4292) and +15.847 (95%CI: 14.4292-17.2646). CONCLUSION: This study represents the first attempt to use an MLR model for the preoperative prediction of PCI in PMP patients. Nevertheless, the MLR model did not perform well enough in predicting preoperative PCI. In the future, more advanced statistical techniques and a radiomics-based CT-PCI-participated MLR model will be developed, which may enhance the predictive ability of PCI.