Reconsidering predictive models in perihilar cholangiocarcinoma

We sincerely appreciate the thoughtful and constructive comments regarding our recently published article, “Predictive model for very early recurrence of patients with perihilar cholangiocarcinoma: A machine learning approach” (1). We are grateful for the opportunity to address the methodological considerations raised, which indeed highlight important aspects for improving generalizability and robustness in clinical prediction modeling.

We fully agree with the importance of external validation to assess a model’s generalizability, particularly when dealing with multi-center datasets collected over an extended period. In our study, we employed bootstrapping resampling (n=5,000) for internal validation to mitigate overfitting; however, we acknowledge that truly independent external validation is the gold standard (2). As mentioned in our limitations section, we emphasized that future studies should seek to validate the proposed model in independent external cohorts to fully evaluate its transportability across institutions and practice settings (1). We appreciate the suggestion and are currently planning a collaborative effort to apply the model to separate external datasets.

The candidate variables included in the XGBoost model were pre-specified based on a combination of clinical relevance and prior literature, rather than derived through automated feature selection methods such as least absolute shrinkage and selection operator (LASSO) or Boruta (1). Given the relatively limited number of events (n=65), we prioritized domain-informed selection to avoid overfitting and maintain model interpretability. We agree that explicitly stating this selection approach would provide additional transparency, and we will ensure to highlight this more clearly in future publications.

XGBoost was selected for its strong track record in structured tabular data and prior demonstrated performance in similar clinical prediction settings (3-6). However, we agree that no single machine learning algorithm is universally superior. Due to the relatively small sample size, we initially opted not to pursue extensive algorithmic comparisons to avoid unnecessary model complexity. Nevertheless, we concur that comparative analyses with other techniques such as random forests, support vector machines, or ensemble approaches would provide valuable insights and could further strengthen the robustness of prediction. This will be an important consideration for future model refinement and external validation phases.

In conclusion, we appreciate insightful feedback, which will undoubtedly help guide future efforts to optimize predictive modeling for perihilar cholangiocarcinoma and beyond. We are grateful for the academic dialogue and look forward to continuing to improve risk stratification tools for this challenging malignancy.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, HepatoBiliary Surgery and Nutrition. The article did not undergo external peer review.

Funding: None.

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://hbsn.amegroups.com/article/view/10.21037/hbsn-2025-272/coif). T.M.P. serves as an unpaid Deputy Editor-in-Chief of HepatoBiliary Surgery and Nutrition. The other author has no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Kawashima J, Endo Y, Rashid Z, et al. Predictive model for very early recurrence of patients with perihilar cholangiocarcinoma: a machine learning approach. Hepatobiliary Surg Nutr 2025;14:3-15. [Crossref] [PubMed]
2. Collins GS, Dhiman P, Ma J, et al. Evaluation of clinical prediction models (part 1): from development to external validation. BMJ 2024;384:e074819. [Crossref] [PubMed]
3. Kawashima J, Endo Y, Woldesenbet S, et al. Preoperative identification of early extrahepatic recurrence after hepatectomy for colorectal liver metastases: A machine learning approach. World J Surg 2024;48:2760-71. [Crossref] [PubMed]
4. Jabeur SB, Mefteh-Wali S, Viviani JL. Forecasting gold price with the XGBoost algorithm and SHAP interaction values. Ann Oper Res 2024;334:679-99.
5. Altaf A, Kawashima J, Khalil M, et al. Identification of a gene signature and prediction of overall survival of patients with stage IV colorectal cancer using a novel machine learning approach. Eur J Surg Oncol 2025;51:109718. [Crossref] [PubMed]
6. Endo Y, Tsilimigras DI, Munir MM, et al. Machine learning models including preoperative and postoperative albumin-bilirubin score: short-term outcomes among patients with hepatocellular carcinoma. HPB (Oxford) 2024;26:1369-78. [Crossref] [PubMed]