Integrated biomarker model based on cTnI, BNP, and D-dimer for early risk stratification in acute aortic dissection

Background Early identification of patients at high risk of death following acute aortic dissection (AAD) remains a major clinical challenge because mortality is driven by multiple interacting pathophysiological mechanisms, including myocardial injury, hemodynamic stress, and activation of coagulation pathways. We aimed to develop and internally validate an integrated biomarker-based model for predicting 30-day mortality in patients with AAD. Methods We conducted a retrospective cohort study to develop and internally validate a multivariable prediction model for 30-day mortality in 572 consecutive patients with AAD. Admission levels of cardiac troponin I (cTnI), creatine kinase-MB (CK-MB), B-type natriuretic peptide (BNP), D-dimer, and fibrin/fibrinogen degradation products (FDP) were assessed. Multivariable logistic regression was performed after adjustment for age, Stanford classification, symptom-onset-to-admission time, systolic blood pressure, serum creatinine, and treatment strategy. Model performance was evaluated using discrimination, calibration, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Internal validation was conducted using bootstrap resampling with 1,000 iterations. Results Elevated cTnI (adjusted OR = 3.76, 95% CI: 1.87–7.56; P < 0.001), BNP (adjusted OR = 6.59, 95% CI: 2.76–15.76; P < 0.001), and D-dimer (adjusted OR = 2.70, 95% CI: 1.17–6.24; P = 0.020) were independently associated with 30-day mortality, whereas CK-MB and FDP were not independently associated with mortality after multivariable adjustment. The integrated prediction model incorporating cTnI, BNP, D-dimer, and clinical covariates demonstrated good discrimination, with an optimism-corrected AUC of 0.89 compared with 0.84 for the clinical model alone, representing a modest but statistically significant improvement (ΔAUC = 0.05). At the optimal probability threshold, the integrated model achieved a sensitivity of 76.2% and a specificity of 96.8%. It also significantly improved risk classification over the clinical model alone, with an NRI of 0.32 ( P < 0.01) and an IDI of 0.11 ( P < 0.01). A nomogram derived from the final multivariable model was developed to facilitate individualized estimation of 30-day mortality risk using routinely available clinical and laboratory variables. Conclusions An integrated model incorporating cTnI, BNP, D-dimer, and key clinical variables demonstrated good discrimination (optimism-corrected AUC 0.89) and statistically significant incremental value over clinical variables alone (NRI = 0.32, IDI = 0.11). However, the modest AUC improvement (ΔAUC = 0.05) and lack of head-to-head comparison with established risk scores (IRAD, Penn, GERAADA) warrant cautious interpretation and external validation before clinical implementation.