Predicting incomplete cytoreduction and aborted hyperthermic intraperitoneal chemotherapy procedures in patients with peritoneal carcinomatosis

Background: Selection of patients with peritoneal carcinomatosis (PC) for cytoreductive surgery and hyperthermic intraperitoneal chemotherapy (CRS/HIPEC) is challenging. Even in peritoneal surface malignancy centers, the rates of incomplete cytoreduction (IC) and aborted hyperthermic intratperitoneal chemotherapy (AHIPEC) procedures varied from 17 to 30%. This study aims to develop and validate a risk predictive model (RPM) to identify patients who would be at the greatest risk for IC/AHIPEC. Material and Methods: Data from 450 attempts of CRS/HIPEC procedures (including aborted HIPEC and incomplete cytoreduction) was used to apply a combination of univariate analysis and multivariate binary logistic regressions. Preoperative risk factors for IC/AHIPEC were selected to develop RPM. Data on all CRS/HIPEC attempts was randomly separated in two subsets, 225 attempts each. The first subset defined the model and the second was used to validate it. Results: From 450 attempts, 305 was from PC of appendiceal, 61 ovarian, and 56 of colon origin. Thirty attempts were due to peritoneal mesothelioma and the remaining 38 due to PC of other origin. On univariate analyses, followed by exclusion of multicollinearity and multivariate logistic regression, 5 variables remained: high grade, extensive surgical history prior to CRS/HIPEC (ESH), three times the upper limit of normal (ULN) for CA-125 or CA-19-9 level, neutrophil-lymphocyte ratio (NLR) >2 and C-reactive protein (CRP) >1mg/L. Internal validation of the RPM on subset 1 produced an area under ROC curve (AUROC) of 0.84, a sensitivity of 83% and specificity of 69%. During external validation on subset 2, the AUROC was 0.77 with sensitivity of 75% and specificity of 70%. The weights attributed to each variable in the score were obtained from the logistic regression b-coefficients of the whole population logistic regression analysis (450 attempts). The weights were high grade (5 points), ESH (10 points), three times the ULN for CA-125 or CA-19-9 level (14 points), NLR >2 (10 points) and CRP >1mg/L (14 points). During internal validation of the scoring model on the entire sample of patients, AUROC was 0.81, sensitivity was 80%, and the specificity was 68% for the cutoff of 24 points (cutoff determined based on Youden's Index). The correlation between the score and percentage of IC/AHIPEC is shown in the table. Conclusion: We developed and validated a model predicting IC/AHIPEC procedures in patients with PC. Clinical implications of having high cumulative score may include laparoscopy to determine resectability prior to CRS/HIPEC, avoiding surgery in patients with comorbidities, and preoperative systemic chemotherapy to decrease tumor burden. External validation of the model by other peritoneal surface malignancy centers is needed. (Table Presented).