Surgical Oncology

Preoperative ML model predicts cancer-specific mortality after nephrectomy

An interpretable preoperative survival-tree model using eight clinical features estimated cancer-specific mortality for nonmetastatic RCC. In external validation (n=580) the model outperformed GRANT with a C-index 0.88 and Brier score 0.02. The tool showed high accuracy in the first postoperative year and is available as a web application for individualized risk counseling. Apply this for preoperative counseling and risk stratification, but combine model output with clinical judgment and multidisciplinary discussion. 1

5x-multiplier reduces discharge opioids after open intra-abdominal cancer surgery

In 150 randomized patients, the 5x-multiplier algorithm reduced median discharge OME to 25 mg versus 75 mg with a 3-tier model. Forty-four percent of 5x-multiplier patients were discharged opioid-free versus 1% in the 3-tier arm. There were no significant differences in 14-day opioid consumption, refill rates, satisfaction, or symptom scores. Consider implementing a 5x-multiplier approach with a nonopioid discharge bundle and clear refill pathways. 2

Tumor-size dynamics reveal mutations linked to treatment response in NSCLC

Using ~17,000 radiology reports and an Andersen-Gill tumor-dynamics framework, investigators identified 27 mutations modifying treatment response. Thirteen mutations associated with progression/response were found, of which 9 were not detected by OS or PFS analyses. Deletions in MYBL1 and GATA3 validated in an external EGFR-TKI sensitivity screen. These genomic–dynamics associations generate hypotheses but require clinical validation before changing treatment selection. 3