A MACHINE learning guided approach to temporary diverting ileostomy (TDI) use in rectal cancer surgery reduces unnecessary stoma formation without increasing anastomotic leakage, according to a randomised controlled trial evaluating the Risk Guided Temporary Ileostomy Decision system (RTID).
TDI Decision Making in Rectal Cancer
TDI remains widely used in rectal cancer surgery to mitigate the consequences of anastomotic leakage, yet its use lacks standardised criteria. TDI decisions are often based on surgeon discretion, which may contribute to variation in practice and potentially unnecessary stoma formation.
The study evaluated whether a machine learning based anastomotic leakage prediction model, the RTID system, could improve decision making.
Trial Design and RTID System Evaluation
A total of 872 patients with stage I to III rectal cancer undergoing anterior resection were randomised in a 1:1 ratio to either surgeon discretion or guidance from the RTID system. The final analysis included 750 patients, with 368 in the control group and 382 in the RTID group.
The RTID system uses predicted risk of anastomotic leakage to guide TDI decisions. Its aim is to improve the appropriateness of stoma formation while maintaining patient safety following rectal cancer resection.
Reduced Stoma Use Without Safety Compromise
The RTID guided group demonstrated a significantly lower overall TDI rate compared with surgeon discretion alone, at 18.6% versus 40.5% (p<0.001). Unnecessary stoma formation was also significantly reduced, at 17.7% versus 41.3% (p<0.001), indicating improved targeting of TDI use.
Although a numerical increase in necessary TDI use was observed in the RTID group compared with control, 55.6% versus 10.0%, this difference did not reach statistical significance (p=0.057). Importantly, the incidence of anastomotic leakage was comparable between groups, at 2.4% versus 2.7% (p=0.753), suggesting no observed compromise in safety outcomes.
However, the authors noted that the study was underpowered to formally test non-inferiority for anastomotic leakage, meaning non-inferiority in anastomotic leakage could not be formally confirmed.
Overall, the findings suggest that the RTID system may support more efficient TDI decision making in rectal cancer surgery, reducing unnecessary stoma use while maintaining comparable short term safety outcomes.
Reference
Shao S et al. Machine learning model-guided selective use of TDI in rectal cancer surgery: a randomized controlled trial. Nat Commun. 2026;DOI:10.1038/s41467-026-73565-4
Featured Image: santypan on Adobe Stock
