Frailty and an Inflammation-Nutrition Composite for Predicting Postoperative Outcomes in Older Patients Undergoing Endometrial Cancer Surgery

Abstract

Background Frailty and systemic inflammation/immune–protein reserve may drive adverse outcomes after endometrial cancer (EC) surgery in older patients, but their incremental predictive value is unclear. Methods We analyzed a retrospective cohort of women ≥65 years undergoing primary EC surgery between January 2016 and December 2024. Frailty was assessed using mFI‑5. Inflammation–nutrition status (INS) was derived from routine preoperative laboratory tests as exploratory composite. The primary endpoint was 30‑day severe complications (Clavien–Dindo III–V), and the secondary endpoint was 90‑day unplanned readmission to the index hospital. One‑year non–endometrial cancer (non‑EC) death was exploratory with EC death as a competing event. Logistic regression and Fine–Gray models were fitted. Incremental prediction versus a prespecified baseline perioperative clinical model was evaluated with bootstrap optimism‑corrected discrimination, calibration, and decision‑curve analysis. Results Severe complications occurred in 62 (7.1%) and readmission in 71 (8.1%). One‑year non‑EC and EC deaths were 28 (3.2%) and 19 (2.2%), respectively. mFI‑5 and INS predicted severe complications (OR per 1‑point mFI‑5 1.43 [95% CI 1.22–1.67]; OR per 1‑SD INS 1.34 [1.15–1.55]) and readmission (OR 1.28 [1.10–1.50]; OR 1.22 [1.05–1.41]). For non‑EC death, mFI‑5 (SHR 1.52 [1.23–1.88]) and INS (SHR 1.37 [1.11–1.69]) were significant. Adding mFI‑5+INS improved discrimination (AUC 0.69 to 0.77 for complications; 0.63 to 0.70 for readmission) and the time-dependent C-index for the exploratory competing-risk endpoint (0.66 to 0.73), with acceptable calibration. Decision-curve findings were descriptive rather than definitive. Conclusion Combining frailty with an objective inflammation–nutrition risk composite modestly improved internal model performance for 30-day severe complications and 90-day readmission in older EC patients, and 1-year non-EC death analysis remains exploratory. These findings should be interpreted as perioperative rather than purely preoperative prediction, and decision thresholds require external validation before protocolized use.