Critical appraisal of predictive tools to assess the difficulty of laparoscopic liver resection: a systematic review

BACKGROUND: Objective assessment of the difficulty of laparoscopic liver resection (LLR) preoperatively is key in improving its uptake. Difficulty scores are proposed but are not used routinely in practice. We identified and appraised predictive models to estimate LLR difficulty. METHODS: We systematically searched the literature for tools predicting LLR difficulty. Two independent reviewers selected studies, abstracted data and assessed methodology. We evaluated tools' quality and clinical relevance using the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) guidelines. RESULTS: From 1037 citations, we included 8 studies reporting on 4 predictive tools using data from 1995 to 2016 in Asia and Europe. In 4 development studies, tools were designed to predict difficulty as assigned by experts using a 10-level difficulty index, operative time, post-operative morbidity or intra-operative complications. Internal validation and performance metrics were reported in one development study. One tool was subjected to external validations in 4 studies (1 independent and geographic). Validations compared post-operative outcomes (operative time, blood loss, transfusion, major morbidity and conversion) between the risk categories. One study validated discrimination (AUROC 0.53). Calibration was not assessed. CONCLUSION: Existing tools cannot be used confidently to predict LLR difficulty. Consistent objective clinical outcomes to predict to define LLR difficulty should be established, and better-quality tools developed and validated in a wide array of populations and clinical settings, following best practices for predictive tools development and validation. This will improve risk stratification for future trials and uptake of LLR