Utility of Prognostic Prediction Models in the Terminal Stage of Gastrointestinal Cancer

Purpose For patients receiving palliative care, information about prognosis is important to help them set priorities and expectations for care and to assist clinicians in decision-making. The purpose of this study was to investigate prognostic models applicable to the terminal stage of gastrointestinal cancer, especially in terms of accuracy of prediction regarding 3-week survival. Methods We validated retrospectively the accuracy of a prognosis prediction model for 354 end-stage gastrointestinal cancer patients who underwent palliative care at our hospital. Using receiver operating characteristic analysis and the area under the curve (AUC), we selected the cut-off value for 3-week survival and evaluated the predictive ability using sensitivity, specificity, positive predictive value, negative predictive value, and accurate diagnosis rate. Results In our analysis of various models, Palliative Prognostic Index (PPI) and Biological Prognostic Score (BPS) version 3 showed excellent predictive performance with AUCs of 0.85 and 0.83, respectively, and accurate diagnosis rates of 80.0 and 79.0, respectively. BPS version 2 showed fair predictive performance with an AUC of 0.76 and an accurate diagnosis rate of 72.0. Using these models, stratification of prognostic prediction was possible. Conclusions PPI and BPS were found to be accurate prediction models for short-term survival of terminal gastrointestinal cancer patients.