Abstract / Summary
Preoperative staging of upper tract urothelial carcinoma (UTUC) relies on cross-sectional imaging and ureteroscopic biopsy, both with recognised limitations for predicting muscle invasion. We compared the fibrinogen-to-albumin ratio (FAR) with four other inflammation–nutrition indices (NLR, PLR, LMR, SII) for predicting muscle-invasive (pT ≥ 2) and locally advanced (pT3–4) UTUC, and explored whether it adds information to a model of routinely available clinical variables. We retrospectively analysed a single-centre cohort of 127 UTUC patients who underwent robot-assisted radical nephroureterectomy (RA-RNU) between 2018 and 2025. The nomogram was developed by multivariable logistic regression using pre-specified variables, with a Firth penalised-likelihood sensitivity analysis. Performance was evaluated by discrimination (AUC), calibration, reclassification (NRI, IDI), decision-curve analysis, bootstrap optimism correction, internal temporal validation (2018–2022 development, n = 83; 2023–2025 validation, n = 44) and case-mix re-weighting. Prevalence of pT ≥ 2 and pT3–4 disease was 76.4% and 40.2%, respectively. FAR showed the numerically highest discrimination among the five indices (AUC 0.631 and 0.611) but was not statistically superior to any of the others. FAR independently predicted pT ≥ 2 (adjusted OR per 0.01 increase 1.15, 95% CI 1.00–1.32) and pT3–4 (aOR 1.13, 1.02–1.26). The combined nomogram (FAR, sex, tumour location, eGFR, diabetes) raised AUCs from 0.699/0.697 to 0.745/0.728; the change was not statistically significant for either endpoint (DeLong P = 0.154 and 0.287), whereas reclassification, calibration and decision-curve analysis were directionally supportive. On internal temporal validation, the combined model had AUCs of 0.703 (pT ≥ 2; 0.53–0.85) and 0.741 (pT3–4; 0.59–0.87), with wide confidence intervals. Preoperative FAR was independently associated with pT ≥ 2 and pT3–4 disease. Adding FAR to a model based on sex, tumour location, eGFR and diabetes improved reclassification, calibration and estimated net benefit without a statistically significant gain in discrimination; this added contribution should be regarded as exploratory and preliminary. Because the reference model did not include biopsy grade, hydronephrosis or radiological invasion, these data cannot establish whether FAR adds information beyond contemporary UTUC risk-assessment tools. The model requires only routinely available preoperative variables, and its plausible role is as a complementary, rule-out-oriented adjunct rather than a replacement for established approaches. Multicentre external validation is required before any clinical use. In 127 UTUC patients undergoing RA-RNU, preoperative fibrinogen-to-albumin ratio (FAR) was independently associated with both muscle-invasive (pT≥2) and locally advanced (pT3–4) disease. Among five inflammation–nutrition indices (FAR, NLR, PLR, LMR, SII), FAR showed the numerically highest discrimination, although it was not statistically superior to any of the other four. Adding FAR to a clinical-only model (sex, tumour location, eGFR and diabetes) yielded improvements in reclassification, calibration and estimated net benefit, without a statistically significant gain in discrimination; this added contribution should be regarded as exploratory and preliminary. Performance was directionally consistent on internal temporal validation (development 2018–2022, n=83; validation 2023–2025, n=44), although confidence intervals were wide; multicentre external validation is required before any clinical use. The model requires only routinely available preoperative variables but was not evaluated specifically in patients with failed or non-diagnostic ureteroscopic biopsy; its plausible role is as a supplementary, rule-out-oriented input to preoperative multidisciplinary assessment alongside biopsy and imaging, not as a substitute for them.