Differential predictive value of inflammation- and coagulation-based composite biomarkers for venous thromboembolism and mortality in patients with NSCLC receiving immune checkpoint inhibitors

BACKGROUND: Venous thromboembolism (VTE) is a clinically important complication in patients with non-small cell lung cancer (NSCLC) receiving immune checkpoint inhibitors (ICIs). Although inflammation and coagulation abnormalities may contribute to both thrombosis and poor prognosis, it remains unclear whether routinely available composite biomarkers have similar or differential predictive value for VTE occurrence and all-cause mortality in this setting. METHODS: This single-center retrospective cohort included 206 patients with advanced NSCLC who received at least two cycles of ICI therapy between January 2020 and December 2024. Baseline clinical, treatment, laboratory, and conventional thrombosis risk score data were collected. The D-dimer-to-albumin ratio (DAR), platelet-to-hemoglobin ratio (PHR), and fibrinogen-to-prealbumin ratio (FPR) were analyzed as log2-transformed continuous variables. VTE was evaluated using cumulative incidence functions and Fine-Gray competing-risk models with death without prior VTE as a competing event. All-cause mortality was assessed using time-dependent Cox models, with incident VTE modeled as a time-dependent covariate. Exploratory transition-specific Cox models evaluated baseline-to-VTE, baseline-to-death without prior VTE, and VTE-to-death pathways. E-values and proportional hazards diagnostics were used for sensitivity assessment and model checking. RESULTS: During a median follow-up of 23.0 months, 29 patients (14.1%) developed VTE, 101 died without prior VTE, and 76 were censored. The median time to VTE was 5.0 months, and the 12-, 24-, and 36-month cumulative incidences were 9.2%, 13.7%, and 14.4%, respectively. Neither the Padua nor Khorana score effectively discriminated VTE risk. In Fine-Gray models, DAR was associated with VTE risk after full adjustment, with a subdistribution hazard ratio of 1.28 per doubling (95% confidence interval, 1.00-1.64; P = 0.048), whereas PHR and FPR were not. In time-dependent Cox models, PHR and FPR remained associated with all-cause mortality, whereas DAR and incident VTE did not. Transition-specific analyses suggested that DAR was more closely related to VTE occurrence, whereas PHR and FPR were more closely related to death without prior VTE. CONCLUSIONS: In patients with advanced NSCLC receiving ICIs, DAR, PHR, and FPR showed differential predictive patterns for VTE and all-cause mortality. These findings require validation in larger multicenter cohorts.