Nonlinear Dose-Response Relationship between D-Dimer and Venous Thromboembolism in Patients with Heart Failure: A Restricted Cubic Spline Analysis

<title>Abstract</title> <p>Objective To investigate the dose--response relationship between D-dimer levels and venous thromboembolism (VTE) risk in heart failure (HF) patients, and to assess effect modification by HF subtype and sex. Methods We included 276 HF patients (148 VTE events). Multivariable restricted cubic spline (RCS) logistic regression with four knots (5th, 35th, 65th, 95th percentiles) was used to analyze nonlinear associations, adjusted for age, fibrinogen, LVEF, LVEDV, PT, and NYHA class. Subgroup RCS analyses were performed by HF subtype (HFmrEF, HFpEF, or HFrEF) and sex, with interactions tested by the likelihood ratio. Bootstrap (200 resamples) was used to estimate the 95% CI for the nonlinear P value. Model performance was assessed by the C-index, Hosmer-Lemeshow test, and calibration curves. Results In the overall cohort, the association between D-dimer levels and VTE risk was significantly nonlinear (P &lt; sub&gt;nonlinear&lt;/sub &gt; = 0.014), the predicted probability rose steeply at low levels and then plateaued. The HFrEF subgroup (n = 162, events = 54) showed significant nonlinearity (P = 0.001), whereas the HFpEF subgroup (n = 77, events = 72) did not (P = 0.540); HFmrEF (n = 37, events = 22) was not analyzed because of insufficient events. Both sexes showed significant nonlinearity (males P = 0.012; females P = 0.049). Interactions were not significant for the HF subtype (P = 0.567) or sex (P = 0.606). The bootstrap 95% CI for the nonlinear P value was 4.41×10⁻⁷–0.094. The model showed good discrimination (C-index = 0.879, 95% CI: 0.839–0.919) and calibration (Hosmer-Lemeshow χ²=7.07, df = 8, P = 0.529). Conclusion D-dimer levels exhibit a significant nonlinear dose--response relationship with VTE risk for HF patients. This association was significant in HFrEF and both sexes, but not for HFpEF. Neither the HF subtype nor sex significantly modified the association. D-dimer may be a useful biomarker for VTE risk stratification in HF, especially in HFrEF where nonlinear risk from elevated levels should be monitored.</p>