Neurosurgery

PROTECT: a 61,289‑patient geriatric perioperative cohort for risk modeling

PROTECT is an ambispective, multicenter perioperative cohort of patients aged ≥65 with 61,289 participants and median age 71. Participants were followed to 30 days with high early follow-up and a reported 30‑day mortality of 0.7%. Common postoperative complications included pulmonary infection (13.0%), nausea (11.9%), hepatic dysfunction (9.9%), and ICU admission (8.6%). The dataset already supports deployed prediction models and can inform geriatric risk profiling and decision‑support tool development. 1

Many RCTs labeled 'personalized' are heterogeneous and poorly transparent

Survey of 262 RCTs using 'personalized/individualized/precision' labels found heterogenous, mostly non‑genomic interventions. Most trials compared a labeled personalized intervention versus non‑personalized control and reported favorable conclusions in 70.6% of comparisons. Transparency indicators were rare (data sharing 5.0%, code sharing 0.4%), and 68.6% of trials had high overall risk of bias. Clinical claims of personalization often lack empirical grounding and should be interpreted cautiously. 2

EAST‑BMS: an East Asian ML model for survival after nonspinal bone metastasis surgery

EAST‑BMS is a multinational machine‑learning survival model developed on 1,045 East Asian patients undergoing nonspinal bone metastasis surgery. Model discrimination was favorable with a 3‑month tdAUC of 0.81 and 12‑month tdAUC of 0.78; 3‑month Brier score was 0.12. Important predictors included albumin, Karnofsky performance status, lymphocyte percentage, and CRP. The tool may improve surgical counseling in East Asian populations but requires external validation before broad adoption. 3