Sciarrone SS, Di Leo M, Fini L, Losco A, De Luca L. Letter to the Editor: Splitting the risk: Interpretable machine learning enters the era of split liver transplantation. World J Gastroenterol 2026; 32(22): 116533 [DOI: 10.3748/wjg.v32.i22.116533]
Corresponding Author of This Article
Luca De Luca, MD, Director, Department of Surgery, Gastroenterology and Digestive Endoscopy Unit, ASST Santi Paolo e Carlo, University of Milan, Via Antonio Rudini’ 2, Milan 20142, Lombardy, Italy. luca.deluca@asst-santipaolocarlo.it
Research Domain of This Article
Gastroenterology & Hepatology
Article-Type of This Article
letter
Open-Access Policy of This Article
This article is an open-access article which was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Baishideng Publishing Group Inc, 7041 Koll Center Parkway, Suite 160, Pleasanton, CA 94566, USA
Share the Article
Sciarrone SS, Di Leo M, Fini L, Losco A, De Luca L. Letter to the Editor: Splitting the risk: Interpretable machine learning enters the era of split liver transplantation. World J Gastroenterol 2026; 32(22): 116533 [DOI: 10.3748/wjg.v32.i22.116533]
World J Gastroenterol. Jun 14, 2026; 32(22): 116533 Published online Jun 14, 2026. doi: 10.3748/wjg.v32.i22.116533
Letter to the Editor: Splitting the risk: Interpretable machine learning enters the era of split liver transplantation
Salvatore S Sciarrone, Milena Di Leo, Lucia Fini, Alessandra Losco, Luca De Luca
Salvatore S Sciarrone, Milena Di Leo, Lucia Fini, Alessandra Losco, Luca De Luca, Department of Surgery, Gastroenterology and Digestive Endoscopy Unit, ASST Santi Paolo e Carlo, University of Milan, Milan 20142, Lombardy, Italy
Author contributions: Sciarrone SS, Di Leo M, Fini L, Losco A, De Luca L contributed to writing the manuscript.
AI contribution statement: AI with chat GPT was only used for a more accurate bibliography and to check the DOI and rights comments for some corrections for some sentences. None AI tools were used in the article writing, making completely letter to editor and making corrections of English.
Conflict-of-interest statement: The authors declare that they have no conflict of interest.
Corresponding author: Luca De Luca, MD, Director, Department of Surgery, Gastroenterology and Digestive Endoscopy Unit, ASST Santi Paolo e Carlo, University of Milan, Via Antonio Rudini’ 2, Milan 20142, Lombardy, Italy. luca.deluca@asst-santipaolocarlo.it
Received: November 17, 2025 Revised: January 4, 2026 Accepted: February 3, 2026 Published online: June 14, 2026 Processing time: 196 Days and 21 Hours
Abstract
Split liver transplantation (SLT) has always existed in a delicate tension between expanding access and accepting complexity. By converting one deceased donor liver into two grafts, SLT directly addresses the persistent organ shortage, but at the cost of extended transection planes, altered hemodynamics, and a consistently higher rate of early postoperative complications (EPC) compared with whole liver transplantation, especially biliary and vasculobiliary events. In this context, the work by Wang et al, published in the recent issue of World Journal of Gastroenterology, on right tri-segment SLT and EPC represents an important conceptual step forward. The authors do not simply refine the list of risk factors, but deliberately place interpretable machine learning at the center of perioperative risk stratification in this highly vulnerable population.
Core Tip: In a single-center retrospective cohort of 109 adult recipients of right tri-segment split liver transplantation, Wang et al report an early postoperative complications (EPC) rate of 33.9%, including hemorrhage, vascular complications, biliary events, ascites and infections. They evaluated four modeling strategies, random forest, support vector machine, extreme gradient boosting and logistic regression, identifying random forest as the best-performing algorithm. Critically, they did not stop at prediction performance. Using SHapley Additive exPlanations, they quantified the contribution of each variable to the model and highlighted six major drivers of EPC risk.
