The significance of preoperative serum fibrinogen in patients with gallbladder carcinoma after curative-intent resection: A Lasso predictive model with an extremely high predictive accuracy and a meta-analysis

Abstract

Objective To systematically evaluate the prognostic value of preoperative fibrinogen in surgically-treated patients with gallbladder carcinoma (GBC). Various prognostic models were established to evaluate their predictive accuracy versus the latest 8th American Joint Committee on Cancer (AJCC) staging system. Methods Post-surgical patients with GBC between 2010 and 2020 were identified. Time-dependent receiver operating characteristic (ROC) curve was used evaluate the diagnostic performance of fibrinogen. Lasso regression-based predictive model and Cox regression-based predictive model were developed. Model performance was evaluated via area under curve. Results Based on our cohort and the following meta-analysis, an elevated preoperative fibrinogen level in GBC indicated a more advanced stage and a worse prognosis. Even after propensity score matching, the recurrence rate was still significantly higher in patients with hyper-fibrinogen. The diagnostic performance of fibrinogen on overall survival was inferior to CA199 while satisfactory performance was acquired when they were combined. Lasso regression-based predictive model with eight predictors incorporated (postoperative chemotherapy, age, preoperative fibrinogen, surgical margin, node metastasis, lymph-vascular invasion, liver invasion, and T stage) showed the highest predictive accuracy (AUC: 0.9) and the latest 8th AJCC staging system has the worst predictive accuracy (AUC: 0.85). Conclusion An elevated preoperative fibrinogen in resected GBC indicated a more advanced stage and a worse prognosis. An established Lasso model with eight predictors included, especially the incorporation of fibrinogen, showed the most superior predictive performance than Cox-based model or the latest 8th AJCC staging system.