A clinical scoring system for survival prediction in advanced gastric cancer.
436 Background: We established a scoring system using easily approachable clinical characteristics at the timing of initiating palliative chemotherapy to achieve accurate overall survival prediction to first-line treatment consisting of fluoropyrimidines in patients with advanced gastric cancer. Methods: A total of 1,733 patients were included in the study. The dataset was split into a training (n=1156, 67%) and validation set (n=577, 33%). Top-ranked variables were identified using the Random Forest for Survival algorithm and analyzed into a Cox regression model, thereby constructing the scoring system for predicting overall survival of advanced gastric cancer. Results: Five variables were finally included in the scoring system: serum neutrophil-lymphocyte ratio, alkaline phosphatase, albumin level, performance status, and histologic differentiation. The scoring system determined four distinct risk groups in validation dataset with median overall survival of 17.1 month (95% confidence interval [CI] = 14.9 to 20.5 month), 12.9 month (95% CI = 11.4 to 14.6 month), 8.1 month (95% CI = 5.3 to 12.3 month), and 3.9 month (95% CI = 1.5 to 8.2 month), respectively. AUC to estimate discrimination performance of the scoring system was 66.1 for one-year overall survival. Conclusions: We developed a simple and clinically useful predictive scoring model in a relatively homogenous population who initiate fluoropyrimidine-containing chemotherapy in advanced gastric cancer. Generalized application of the scoring model will require additional independent validation. [Table: see text]