From Predicting Difficulty to Choosing the Safest Operative Strategy: A Systematic Review of Decision Support in...

BackgroundDifficult laparoscopic cholecystectomy (LC) remains a major contributor to bile duct injury, bailout procedures, and conversion to open surgery. Although numerous preoperative prediction models have been developed, their ability to meaningfully support operative decision-making and safe bailout strategies remains uncertain. This systematic review aimed to evaluate current preoperative risk prediction models for difficult LC, with particular emphasis on their clinical applicability, integration into surgical decision-making, and relevance to contemporary safe cholecystectomy concepts.MethodsA systematic review was conducted in accordance with PRISMA guidelines. PubMed, Scopus, and Web of Science were searched from inception to March 2026 for studies evaluating preoperative prediction models for difficult LC, including both traditional clinical scoring systems and machine learning-based approaches. Data regarding study design, predictors, model performance, validation strategies, and clinical applicability were extracted. Artificial intelligence applications related to intraoperative guidance and operative workflow were analyzed separately. Because variance measures were inconsistently reported across studies, a formal meta-analysis was not feasible.ResultsFourteen studies were included in the qualitative synthesis, comprising nine studies on preoperative prediction models and five studies on artificial intelligence applications in LC. Most studies were retrospective, single-center, and methodologically heterogeneous. Traditional clinical and radiological models demonstrated moderate discriminatory performance, whereas external validation was uncommon. Across five studies with extractable AUC data involving 1738 patients, reported AUC values ranged from 0.735 to 0.960, with substantial variability and limited generalizability. Most artificial intelligence studies focused on intraoperative image analysis and critical view of safety assessment rather than actionable preoperative risk stratification. Importantly, current prediction systems remain insufficiently integrated into real-world operative decision-making and safe bailout strategy selection.ConclusionCurrent evidence suggests that preoperative prediction models may assist in risk stratification and operative planning; however, their real-world clinical applicability remains limited. At present, no model has demonstrated sufficient external validation or integration into surgical workflow to support routine standalone use. Future research should move beyond predictive accuracy alone toward clinically integrated decision-support systems that combine preoperative risk stratification, intraoperative grading frameworks, and AI-assisted guidance to support safe operative strategy selection, bailout planning, and intraoperative risk management.