Incremental prognostic value of a preoperative immune-metabolic-inflammatory score for disease-free survival in endometrial cancer

Vollständiger Abstract

Worum geht es in dieser Arbeit?

The systemic immune-inflammation index, fibrinogen, triglyceride/high-density lipoprotein cholesterol ratio, and prognostic nutritional index have each been linked to endometrial cancer outcomes, but their optimal combination is uncertain. We compared alternative formulations and evaluated incremental prognostic value for disease-free survival (DFS). This single-center cohort included 853 surgically treated patients. Four biomarkers and 3 risk-aligned, equally weighted scores were compared: the 4-component immune-metabolic-inflammatory score (IMIS), 3-component simplified IMIS (sIMIS), and a 2-component score. Evaluation included Cox association strength, discrimination, prediction error, out-of-fold (OOF) calibration, and exploratory decision curves. The selected score was assessed alone and after addition to a 4-variable pathology reference model using 1000 patient-level bootstrap samples and stratified 10-fold OOF prediction. Alternative formulations were retained as sensitivity analyses. Median follow-up was 4.93 years, with 94 DFS events. All 4 biomarkers were associated with DFS. sIMIS had the largest likelihood-ratio statistic and numerically highest optimism-corrected and OOF C-indices, while the alternative formulations yielded broadly concordant performance. Each 1 – standard deviation increase in sIMIS was associated with a higher hazard of a DFS event (multivariable hazard ratio, 1.65; 95% confidence interval [CI], 1.45–1.88; P < .001). The pathology + sIMIS model had optimism-corrected and OOF C-indices of 0.820 and 0.819. Relative to pathology alone, the paired C-index difference was 0.0664 (95% CI, 0.0330–0.1025). The 5-year Brier-score gain was 0.0104 (95% CI, 0.0044–0.0165); the 3-year CI included zero. Treatment-by-risk interaction estimates were imprecise and were not statistically significant. In this single-center development cohort, sIMIS provided the most favorable overall balance of prognostic performance, component count, and construct coverage under internal resampling, although the alternative formulations performed broadly similarly. sIMIS remains a candidate risk marker rather than an established clinical or treatment-selection tool; external validation and comparison with contemporary molecularly integrated risk models are required.