Predicting anatomical difficulty in robot-assisted rectal resection using a novel anatomic landmark-based grading...

This study aims to develop an anatomic landmark-based ordinal grading system for surgical difficulty in robot-assisted rectal resection and to assess its preoperative predictability, with the potential to support preoperative assessment of surgical difficulty.

We divided the procedure into two phases: mesocolon dissection and mesorectal dissection. Surgical difficulty was defined using an anatomy-based ordinal grading system (grades 1-3) for each phase. A cohort of 158 patients undergoing robot-assisted rectal resection was included. Regression analysis was performed for each phase using preoperative variables to select predictors. Pooled tenfold cross-validation was used for model validation.

For the mesocolic phase, a four-predictor model (body mass index, sex, lung disease, and age) achieved moderate performance (accuracy 0.648; AUC (≤ 1 vs > 1) 0.787; AUC (≤ 2 vs 3) 0.831). For the mesorectal phase, a three-predictor model (body mass index, sex, and neoadjuvant chemoradiotherapy) achieved good predictive performance (accuracy 0.697; AUC (≤ 1 vs > 1) 0.927; AUC (≤ 2 vs 3) 0.850). The final equations were implemented in a ready-to-use calculator to facilitate clinical application.

This study introduces a novel, phase-specific grading system that allows preoperative estimation of anatomical difficulty in robot-assisted rectal resection. The model is consistent with established preoperative risk factors. By utilizing standard preoperative data, this tool may assist surgeons in anticipating operative complexity and tailoring their surgical approach.