Prediction of outcome of Patients Undergoing Whipple’s Pancreatoduodenectomy before Operation: A Prospective Validation of a Novel Risk Scoring System

Abstract

Background: Despite the high morbidity following Whipple’s pancreatoduodenectomy operations, there is still a lack of an objective pre-operative tool, based only on clinical and biochemical parameters to predict the outcome following the procedure that might be implemented. Materials and Methods: Using a multivariate regression model, the significant predictors of post-operative outcome were identified in a set of retrospective database of patients (2006-2017), and a risk score developed by binary logistic regression method. This was validated in a set of prospective patients (2017-2020). The model’s predictive accuracy and discriminative ability were assessed using the receiver operating characteristics (ROC) analysis. Results: On multivariate analysis in the retrospective cohort (n=442), the significant predictors of post-operative outcome were identified as peak bilirubin levels, pre-operative stenting and nature of the disease (Benign/Malignant). A risk score was derived and validated on the prospective cohort (n=182). The mean risk for an unfavourable outcome was 24% for a score of /=15. This was further tested on the validation cohort for individual risk scores (AUC=0.793). There was no significant difference between observed and expected risk of major complications (p=0.31). Conclusion: The risk score showed a fair accuracy in predicting post-operative morbidity in the prospective cohort. Therefore, we propose this to be used as a quick aid to predict the operative outcome in patients posted for pancreatoduodenectomy on an outpatient basis using simple pre-operative clinical and laboratory variables.