FRIES score: predicting conversion in robotic liver surgery

Résumé fourni par la source

Conversion to open surgery during robotic liver surgery is a critical quality indicator1. Unlike laparoscopy, robotic conversion requires a specific undocking phase, which may complicate the management of intraoperative emergencies2. A predictive score was developed and validated3 for conversion using the FRIES-ACHBPT-2024 multicentre cohort4. Data from 650 patients across 10 French centres (2010–2024) were analysed. The primary endpoint was unplanned conversion to laparotomy2. The cohort was randomly split (1:1) into training and test sets. Multivariable logistic regression identified independent predictors to develop a 0–100 risk score (supplementary methods). The open conversion rate was 10.8% (70 patients). Conversion to conventional laparoscopy occurred in eight patients, two of whom were subsequently converted to laparotomy; only six procedures (< 1% of the cohort) remained purely laparoscopic. Primary indications were oncological concerns (18, 25.7%), exposure issues (13, 18.6%), and bleeding (13, 18.6%). Conversions for oncological reasons often represented proactive decisions to ensure R0 margins rather than technical failures. Converted patients had significantly higher median blood loss (928 versus 304 ml; P < 0.001) and severe complications (Clavien–Dindo ≥ III: 14 (20.0%) versus 46 (8.1%); P < 0.001; Table S1). Five independent predictors were identified: previous abdominal surgery (odds ratio (OR) 2.88, 95% confidence interval (c.i.) 1.15 to 7.42, P = 0.025), Institut Mutualiste Montsouris (IMM) difficulty classification5 class II (OR 3.62, 95% c.i 1.11 to 12.3, P = 0.034) or III (OR 3.08, 95% c.i. 1.05 to 9.73, P = 0.045), associated procedures (OR 5.19, 95% c.i. 1.30 to 19.5, P = 0.015), multiple nodules (OR 4.24, 95% c.i. 1.50 to 12.3, P = 0.007), and tumour size > 7 cm (OR 2.80, 95% c.i. 0.98 to 7.78, P = 0.049). The FRIES (French Robotic lIvEr Surgery) score achieved an area under the curve of 0.802 (95% c.i. 0.721 to 0.883) in the training set and 0.708 (95% c.i. 0.617 to 0.800) in the test set (supplementary results and Figs S1–S6). To improve transparency regarding procedural volumes, the number of procedures performed per centre together with the conversion rate in each centre is provided in Table S2. The FRIES score identified key anatomical and technical determinants of conversion. Although surgical experience is a primary driver of success, this model, derived from a national implementation phase, provides a pragmatic tool for teams navigating their learning curve. Centre volume and learning curve effects may also influence conversion risk, but were not incorporated into the model given the limited number of centres, the heterogeneity of robotic exposure, and the difficulty of precisely defining and integrating such variables without risking model overfitting. The higher morbidity observed in converted patients underscores the need for careful patient selection, particularly when complex resections (IMM class III)5 or large/multiple tumours are involved. The distinction between emergency conversions for bleeding (18.6%), which carry immediate safety implications, and oncological conversions (25.7%), which may serve to protect long-term oncological outcomes, is important when interpreting conversion events. Oncological conversions may result from suspected tumour perforation, inadequate tumour visualization, or difficulty achieving safe margins. However, in some patients conversion may reflect suboptimal preoperative planning, with the robotic approach also increasingly used to attempt minimally invasive management of technically demanding tumours. The reason for conversion was not specified in a proportion of conversions (16, 22.8%), which may limit detailed interpretation of conversion causes. Although limited by its retrospective design and the need for external validation, the FRIES score, available as a web-based application https://brustia.shinyapps.io/shiny_web_application/ (Fig. 1), offers a standardized framework for risk stratification and shared decision-making during the critical phase of robotic programme initiation. In addition to the importance of outcomes such as safety and oncological adequacy, providing patients with quantitative estimates of conversion risk, rather than approximate statements, may help them become more involved in their care. As robotic liver surgery continues to mature and experience becomes more widespread, recalibration of the model will likely be required to maintain its predictive accuracy over time. Online calculator predicting conversion to open surgery AP-HP Hôpitaux Universitaires Henri-Mondor founded the publication fees for this article. The authors have no funding to declare. Raphael Venezia, MD (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Guillaume Millet, MD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Stylianos Tzedakis, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Heithem Jeddou, MD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Bastien Le Floch, MD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Alain Valverde, MD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Nicolas Peru, MD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Patrick Pessaux, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Fabio Giannone, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Jean Yves Mabrut, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Kayvan Mohkam, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Antonio Sa Cunha, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Chady Salloum, MD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Pierre Yves Blanc, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing) Bertrand Le Roy, MD, PhD (Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.