Colorectal Surgery

Meta-analysis: ML predicts postoperative complications and early recurrence with moderate-to-good accuracy.

This meta-analysis of 31 studies reports pooled AUCs of 0.83 for postoperative complications and 0.80 for early recurrence. Pooled sensitivity and specificity were roughly 0.75 and 0.78 for complications, and 0.74 and 0.73 for early recurrence. Performance varied meaningfully by ML model type, tumor type, sample size, publication year, and country. Authors found no evidence of publication bias but call for larger, standardized studies to improve reliability. 1

Population study: colorectal cancer incidence and mortality rising fastest in Taiwanese adults aged 30–49.

From 1994–2018 Taiwan demonstrated marked increases in CRC incidence and mortality driven by the 30–49 age cohort. Age-period-cohort analysis reported cohort annual increases of 3580 (incidence) and 1000 (mortality) per 100,000 for ages 30–49. Patients aged 15–29 more often presented at stage 3–4 and had 7.8% lower survival than the 30–49 group. Authors highlight urgent need for tailored prevention and policy interventions for younger adults. 2

RCT: ML-guided ileostomy decision reduces diverting stomas without apparent increase in leakage.

In 872 randomized rectal cancer patients (750 analyzed), the RTID system lowered temporary diverting ileostomy (TDI) use to 18.6% versus 40.5% with usual care. Unnecessary stoma formation fell from 41.3% to 17.7% with RTID guidance. Anastomotic leakage rates were similar (2.4% RTID vs 2.7% control), although the trial was underpowered for non-inferiority on leakage. RTID provides an objective adjunct for TDI decisions but requires local validation before adoption. 3