Inter‐rater reliability of frozen donor liver macrovesicular steatosis assessment applying the Banff consensus method: a real‐world, multi‐specialty validation study

Introduction Reliability of intraoperative assessment of donor liver allografts for steatosis, an important risk factor for future graft dysfunction, is limited by poor inter‐observer agreement. To address this issue, consensus guidelines by the Banff working group in 2022 proposed a standardized steatosis algorithm to improve agreement; however, the practical impact of the proposal has not been well‐validated across subspecialty practices and experience levels. Methods This single‐institution inter‐observer agreement study entailed two rounds of evaluation of 27 frozen donor liver biopsies by 15 pathologists with a range of subspecialty expertise and experience levels. Raters estimated the amount of macrovesicular steatosis (MVS) by individual methods and, subsequently, using the proposed Banff algorithm. Results Moderate agreement was observed among pooled raters, irrespective of method used (ICC 0.55 vs. 0.47 for individual and Banff methods, respectively) for precise estimation of MVS. Importantly, though, when precise estimates were reduced to clinically relevant grades, the Banff method improved agreement from ‘fair’ to ‘substantial’ compared with individual methods, with the most marked improvement among gastrointestinal subspecialists and pathologists with > 10 years of experience. The Banff method also resulted in overall lower grade estimates, with no ‘grade 3’ estimates provided, including from biopsies with abundant and azonal MVS, which are notoriously challenging. Conclusion The findings support utilization of the Banff method as a means of improving inter‐rater agreement among diverse pathologists, resulting in more reliable estimation of MVS at clinically significant thresholds.