Association between the perivascular fat attenuation index and CCTA-defined anatomically significant left anterior descending coronary artery stenosis: relationship with serum inflammatory biomarkers

Objective To investigate whether lesion-specific perivascular fat attenuation index (FAI) is associated with CCTA-defined anatomically significant (≥50%) left anterior descending coronary artery (LAD) stenosis and to evaluate the diagnostic performance of a composite model integrating serum inflammatory biomarkers. Methods We retrospectively analyzed 160 patients with suspected coronary artery disease (CAD) who underwent coronary computed tomography angiography (CCTA). Patients receiving anti-inflammatory medications were excluded. Lesion-specific perivascular FAI was measured around the LAD. The prespecified primary outcome was CCTA-defined anatomically significant LAD stenosis (≥50%), analysed as a binary variable. Multivariable logistic regression identified independent predictors, and the final model was internally evaluated with stratified 5-fold cross-validation, 10-fold cross-validation, and 1,000-bootstrap resampling. Model building was repeated within each resample. Calibration (slope, calibration-in-the-large, Brier score) and incremental discrimination (DeLong test) were assessed. Results Patients with CCTA-defined anatomically significant stenosis (≥50%) had higher, less negative FAI values, higher hs-CRP and LDH levels, and a lower ApoA1/ApoB ratio than those with stenosis <50% (all P < 0.05). FAI correlated with stenosis both as a continuous percentage ( ρ = 0.54, P < 0.001) and as a binary ≥50% variable ( ρ = 0.64, P < 0.001), with less negative FAI values accompanying more severe stenosis. The final model identified FAI (OR = 1.055, 95% CI: 1.014–1.098), hs-CRP (OR = 1.063, 95% CI: 1.005–1.125), LDH (OR = 1.014, 95% CI: 1.005–1.023), and the ApoA1/ApoB ratio (OR = 0.209, 95% CI: 0.070–0.624) as independent predictors. The composite model achieved an AUC of 0.915 (95% CI: 0.865–0.965); internal validation yielded AUCs of 0.872 (5-fold), 0.881 (95% CI: 0.826–0.936; 10-fold), and 0.895 (95% CI: 0.848–0.942; bootstrap). On formal DeLong testing for correlated ROC curves, the composite model showed significantly better discrimination than hs-CRP alone ( P = 0.020) and than FAI alone ( P < 0.001). Conclusions Higher, less negative perivascular FAI values were associated with CCTA-defined anatomically significant LAD stenosis. A composite model integrating FAI, hs-CRP, LDH, and the ApoA1/ApoB ratio showed high apparent diagnostic accuracy and remained stable across multiple internal validation strategies. Because this was a single-centre retrospective study without external validation or outcome data, these findings should be regarded as preliminary and hypothesis-generating; the model is not ready for clinical use.