Prognostic models for intrahepatic cholangiocarcinoma after resection: A systematic review and meta-analysis
Yizhao Liu1, Xia Lei1, Jinyong Hao2
1The second clinical medical school, Lanzhou University, Lanzhou, Gansu, China.
Objective:
To systematically characterize existing prognostic models and to evaluate and compare their discriminative performance in patients with intrahepatic cholangiocarcinoma (iCCA) after curative resection.
Methods:
We systematically searched Web of Science, Embase, Cochrane Library, PubMed, and MEDLINE (January 2010 to January 2026). Two reviewers independently screened studies, extracted data, and assessed risk of bias using PROBAST+AI. Tested and selected variables were summarized using frequencies. Discriminative performance estimates (C-index and area under the curve [AUC]) with 95% CI were pooled via random-effects meta-analysis after logit transformation. All analyses were performed using R, version 4.4.2.
Results:
A total of 60 studies involving 82 prediction models were included. Frequencies revealed that traditional variables were consistently considered, including markers of tumor burden and invasiveness, inflammation-related and nutrition-related markers, and individual characteristics. Meta-analysis of 36 studies showed a pooled training set C-index of 0.723 (95% CI: 0.705-0.740) and AUC of 0.767 (95% CI: 0.700-0.823), while the validation set showed a pooled C-index of 0.696 (95% CI: 0.673-0.718) and AUC of 0.787 (95% CI: 0.690-0.860). These performance metrics were consistently higher than those of the AJCC staging system (7th edition), the AJCC staging system (8th edition), and the LCSGJ staging system. Subgroup analyses suggested that incorporating treatment-related variables, distinguishing resection status, and using recurrence as an endpoint tended to improve model discrimination. Considerable heterogeneity was observed.
Conclusions:
In this systematic review and meta-analysis, existing prognostic models for iCCA demonstrated good discriminative performance. These models outperformed conventional staging systems and provided predictive value for both survival and recurrence. Nevertheless, methodological shortcomings remain in most models and warrant further refinement.
