Cardiothoracic Surgery

Protocol: AI-assisted education versus SMS in older postoperative head-and-neck cancer patients

This single-center RCT protocol will randomize 100 postoperative head-and-neck cancer patients aged ≥60 to AI-assisted education (n=50) versus standardized SMS (n=50). Intervention delivers 12 months of personalized, phased content and real-time AI Q&A; outcomes include stress, QoL, nutrition, sleep, and mood at five timepoints to 12 months. Analysis plans intent-to-treat linear mixed models to estimate mental health and QoL effects and manage missing data. Generalisability is limited by single-center design and reliance on self-reported outcomes. 1

ESPB modestly reduces early pain and opioid use after median sternotomy

Meta-analysis of 19 RCTs (N=1344) found ESPB reduced 24‑hour pain after median sternotomy (MD -0.65; 95% CI -1.14 to -0.16). ESPB also lowered intraoperative and 24‑hour opioid consumption (MD about -80.8 mg and -45.8 mg OME), time to extubation (MD -1.28 h), and ICU stay (MD -13.1 h) in the sternotomy subgroup. Heterogeneity was very high and wide prediction intervals crossed the null, implying effects vary across centers. No consistent benefit was seen in minimally invasive cardiac surgery; certainty is low to very low for many outcomes. 2

Explainable XGBoost predicts textbook outcome after free flap oral cancer reconstruction

Single-center retrospective cohort (n=752) developed an XGBoost model predicting textbook outcome with AUC 0.843, accuracy 0.764, and sensitivity 0.733 in validation. Overall textbook outcome rate was 38.2%, and key predictors included albumin, age, operative time, T stage, and NLR. SHAP analyses suggested threshold effects at age >50, albumin >45 g/L, and operative time >400 minutes that may guide perioperative risk discussions. Model is internally validated only and requires external validation before clinical deployment. 3