Beyond Pathology: A Procedure-Based Approach to Planning and Predicting Outcomes in Robotic Gynaecological Oncology Surgery
Abstract
Background
Current surgical planning in robotic gynaecology relies heavily on pathological diagnosis, yet operating theatre utilisation may depend more on procedural requirements. Recent advances in machine learning-based surgical prediction have highlighted the need for more accurate planning models whilst challenging fundamental assumptions about surgical complexity.
Objectives
To compare the impact of procedural requirements versus traditional complexity markers on operating time and complications in robotic gynaecological surgery, and to develop a practical framework for procedure-based surgical planning that could complement contemporary machine learning approaches.
Methods
Retrospective analysis of 80 consecutive robotic gynaecological surgeries (2021–2024) at a single tertiary centre. We examined relationships between procedural requirements (lymphadenectomy, adhesiolysis), traditional complexity markers (BMI > 35, pathology type, previous surgery), and outcomes (operating time, complications). A Preoperative Procedural Demand Score (PPDS) was developed and validated against pathology-based predictions. Multivariable regression analysis included β-coefficients, confidence intervals, and comprehensive model diagnostics.
Results
Procedural requirements explained operating time variation better than pathology type. Standard procedures averaged 135.4 ± 28.6 min. Adding lymphadenectomy increased time by 33.1 min (95% CI: 24.8–41.4), adhesiolysis by 19.8 min (95% CI: 11.5–28.1). Traditional complexity markers showed minimal impact: BMI > 35 added 8.1 min (95% CI: -7.2 to 23.4, p = 0.42), previous surgery 7.6 min (95% CI: -8.4 to 23.6, p = 0.48). All complications (6.3%) were minor (Clavien-Dindo Grade I-II). Operating time decreased from 178.0 ± 47.2 to 135.9 ± 36.8 min between study halves ( p < 0.001), whilst intraoperative complications decreased from 12.5 to 0% ( p = 0.02), despite similar case complexity.
Conclusions
This study demonstrates that procedural requirements better predict operating time than traditional markers in robotic gynaecological surgery, representing a novel conceptual framework for surgical planning. However, the single-centre design and modest predictive accuracy compared to contemporary machine learning approaches indicate that comprehensive multi-centre prospective validation is essential before recommending widespread adoption. The excellent safety profile across all complexity levels supports the feasibility of robotic surgery in traditionally “complex” cases, though larger studies are needed to establish definitive safety parameters.