30-Second Takeaway
- Perioperative probiotics may speed GI recovery and reduce some postoperative infections in colorectal cancer.
- Neoadjuvant chemoimmunotherapy improves pathological complete response but not survival for resectable NSCLC with PD-L1 <1%.
Week ending June 6, 2026
Concise evidence briefs for surgical oncologists: probiotics, RWD emulation, AI education, ML in sarcoma, and neoadjuvant chemoimmunotherapy in PD-L1 <1% NSCLC
Probiotics in cancer: signals for reduced GI toxicity and faster postoperative GI recovery
Systematic review of trials and meta-analyses found probiotics associated with reduced clinically relevant chemotherapy diarrhea and improvements in other GI symptoms. In surgical oncology, perioperative probiotics/synbiotics linked to faster GI functional recovery and fewer postoperative infections in pooled colorectal surgery trials. Effects on anastomotic leak were inconsistent and serious probiotic-related infections were not reported in controlled studies. Heterogeneity in strains, formulations, endpoints, and small trial sizes limit guideline-ready adoption.
Feasibility framework for oncology RCT emulation using commercial EHR-derived RWD
CARE systematically screened six EHR-based RWD sources across 23 oncology RCTs to assess fitness for emulation. Of 54 RCT–RWD combinations, nine passed initial screening and three advanced to protocol development. Key limitations for emulation included complex eligibility, biomarker capture, performance status, and outcome ascertainment. Authors recommend pre-emulation feasibility assessment and linking oncology-specific data to claims or custom abstraction.
Protocol: AI-assisted education for older postoperative head and neck cancer patients
Single-center randomized protocol will test a 12-month AI-driven personalized education program versus standard SMS education in 100 patients aged ≥60 after HNC surgery. Primary outcomes target mental health, social support, and quality of life measured at five timepoints up to 12 months. Analysis will use intention-to-treat linear mixed models to handle repeated measures and missing data. Single-center design and reliance on self-reported outcomes may limit generalizability; multicenter validation is needed.
References
Numbered in order of appearance. Click any reference to view details.
Additional Reads
Optional additional studies from this edition.