Bicuspid Aortic Valve (BAV) is the most common congenital heart disease, with a prevalence of 1%–2% in the general population (1, 2). The most serious complications are ascending aortic dilatation (AAD) and dissection, which is associated with proximal aortic dilation (called bicuspid aortopathy) in about 50% of patients with BAV, resulting in a significantly higher aorta-related mortality rate and an eight-fold increase in the risk of aortic dissection than in the general population (3, 4). Traditionally, the diameter of the aorta has been the mainstay of surgical intervention (recommended threshold 5.0–5.5 cm) (5). However, approximately 40% of BAV-associated dissections occur in patients with an aorta diameter <5.5 cm (4, 6), and anatomical markers alone are not effective in distinguishing the risk of disease progression between different phenotypes (e.g., root or tubular) (7, 8). This highlights the inadequacy of the current risk stratification strategy.

In addition, the pathological mechanism of BAV-related aortic lesions involves multifactorial interactions. First, in terms of genetic factors, mutations in genes such as NOTCH1 and GATA5 were significantly associated with root expansion, especially in young patients (9). Family studies have shown a 6.9-fold increased risk of aortic disease in first-degree relatives (10). Second, there is an abnormal hemodynamic, and the abnormal leaflet opening of BAV leads to a disordered distribution of wall shear stress (WSS) in the ascending aorta wall, especially in right-left fusion BAV, and 4D Flow Magnetic Resonance Imaging (MRI) confirms that the high WSS region is spatially consistent with elastic fiber degradation and differential expression of microRNA (such as miR-128-3p) (11–13). And the molecular pathway dysregulation of BAV-related diseases. Downregulation of endothelial nitric oxide synthase (eNOS) promotes the phenotypic switch of vascular smooth muscle cells toward a synthetic phenotype (14). The imbalance between matrix metalloproteinase (MMP-2/9) and tissue inhibitor (Tissue Inhibitor of Metalloproteinases-1,TIMP-1) leads to the degradation of the extracellular matrix, and MMP-2 is regionally heterogeneous in the aortic wall of BAV patients (15, 16). Oxidative stress is exacerbated: the plasma antioxidant markers (α-tocopherol, HDL-C) levels and inflammatory factors (C-Reactive Protein,CRP) are increased in BAV patients (17).

Although cyclic markers (e.g., MMP-2, TIMP-1, endotrophin) and imaging parameters (e.g., flow displacement index on 4D Flow MRI, WSS) have been proposed for risk stratification, there are the following limitations (12, 18). First, the predictive value of a single biomarker is limited (e.g., TIMP-1 is only negatively correlated with the Z-score of some aortic segments) (16); Second, the predictive performance of the traditional model for Tricuspid Aortic Valve (TAV)-related aortic aneurysm (AUC>0.8) was significantly better than that of BAV (AUC = 0.5–0.55). Third, there is a lack of individualized prediction tools that integrate clinical, imaging, and molecular multi-dimensional parameters (19).

In summary, this study is the first to combine clinical features (age), ultrasound hemodynamic parameters [ascending aortic maximum velocity (AAoV) and ascending aortic mean pressure gradient (AAoMPG)], and serum lipid markers (HDL-C and ApoB). By combining machine learning algorithms (such as LASSO) and Nomogram to construct a prediction model for ascending aortic dilation in BAV patients, it aims to break through the limitation of a single parameter and provide accurate decision-making tools for early surgical intervention.

In this study, a novel integrated method was established to combine hemodynamic indicators of 2D-TTE with serum lipid biomarkers for predicting ascending aortic dilation in patients with BAC. The nomogram based on LASSO algorithm showed good discrimination efficiency, highlighting the synergistic effect of abnormal hemodynamic stress and lipid metabolism disorder in BAV aortic lesions. Five key predictors: AAoV, HDL-C, ApoB, LVMI, and age were screened, which together translate complex pathophysiological interactions into clinically applicable risk assessment tools.

First, this study confirmed that AAoV and AAoMPG were significantly increased in the BAV plus ascending aortic dilation group, which was directly related to the unique blood flow pattern of BAV. Abnormal leaflet morphology of BAV (e.g., right-left coronary fusion) can produce eccentric jets that increase WSS in the ascending aorta wall, promoting endothelial injury and stromal remodeling (11, 20, 21). At the same time, the 4D Flow MRI study further showed that the abnormal flow displacement and WSS distribution of ascending aorta in BAV patients were positively correlated with the progression of aortic elasticity and dilation (22–24). This suggests that the AAoV and AAoMPG in our study have certain predictive value and also confirm the central role of hemodynamics in aortopathy.

Notably, HDL-C levels were significantly higher in the BAV-D group than in the non-dilated group, contrary to the traditional perception of cardiovascular risk. The possible mechanism of this phenomenon is that HDL-C may play a dual role by regulating oxidative stress and inflammatory response in aorta dilation in BAV patients. Studies have shown that patients with BAV have a systemic decrease in antioxidant capacity (e.g., decreased α-tocopherol) and inflammatory activation (elevated CRP), while HDL-C may be upregulated as a compensatory protective factor (1, 17). In addition, Martinez-Micaelo et al. found that the plasma antioxidant capacity of aorta dilation in BAV patients was significantly reduced, which was associated with HDL dysfunction (17). Thus, elevated HDL-C may reflect a compensatory response to oxidative damage in the blood vessels, rather than a mere protective marker.

The Lasso regression model constructed in this study integrated five parameters: age, HDL-C, ApoB, LVMI and AAoV, and its AUC reached 0.825, which was significantly better than that of a single anatomical index (such as aortic diameter). The Lasso model has three advantages to working with nomogram. First, the joint model covered multi-dimensional influencing factors, including hemodynamics (AAoV), metabolism (HDL-C and ApoB), myocardial remodeling (LVMI) and age factors, which fit the multifactorial pathogenesis of aorta dilation in BAV patients (8, 25, 26); Secondly, the joint model quantifies and concretizes the individualized risk assessment, and the Nomogram visualization tool quantifies the individual risk to assist clinical decision-making. For example, a 50-year-old patient with a score of 22 points for age, HDL-C of 1.2 mmol/L (30 points), ApoB of 1 g/L (28 points), LVMI of 80 g/m² (20 points), and AAoV of 1.8 m/s (44 points) has a total score of 154 points, corresponding to a predicted risk probability between 0.4 and 0.8, approaching 0.8. This indicates a substantially elevated risk, warranting modifications in both the frequency and modalities of follow-up. Thirdly, the combined model can predict the occurrence and development of the disease at an early stage, and AAoV and LVMI may be abnormal before the aortic structure changes, providing a window for subclinical intervention (27).

In addition, our findings are consistent with those of Lopez-Sainz et al. quantifying abnormal blood flow in BAV by 4D Flow MRI compared to previous studies (12), but this study is the first to incorporate simple ultrasound parameters (AAoV and AAoMPG) into a prediction model, which is more applicable to grassroots outreach. At the same time, traditional studies have suggested that LDL-C promotes aortic atherosclerosis, but this study suggests that HDL-C and ApoB may be involved in aortic remodeling through non-atherosclerotic pathways (such as modulation of matrix metalloproteinase activity) in aorta dilation in BAV patients (28, 29). And machine learning predicted aortic dilation in TAV/BAV patients and found that the BAV group had lower predictive power (AUC 0.5–0.55), attributed to its stronger genetic dominance (19). However, the AUC of 0.825 in this study may be due to the fact that the study focused on BAV subsets and included hemodynamic markers. Compared with our model, their parameter selection relied solely on lipid biomarkers and failed to effectively integrate ultrasound-derived parameters—particularly the hemodynamic index AAoV, which was not selected as a key feature, thereby reducing the predictive performance of their model. This study is of great significance in clinical translation. First, the visual nomogram serves as a practical risk stratification tool. For instance, a hypothetical patient with a nomogram-derived total score > 150 (indicating high risk) might be recommended for more intensive surveillance, such as annual cardiac magnetic resonance imaging (MRI) to closely monitor aortic dimensions and wall stress. In contrast, a patient with a score < 100 (low risk) could potentially adhere to the standard biennial follow-up with transthoracic echocardiography. This individualized approach may optimize resource allocation and patient safety; Second, this study provides a clinical reference for the timing of clinical intervention. For example, early pharmacological intervention (e.g., statins) may be considered for patients whose aortic diameter is <4.5 cm predicted to be high-risk by the model (30); Third, complementary surgical decision-making, by combining the US guidelines (diameter threshold) and this model, can be individualized for patients with moderate dilation (4.0–5.0 cm) but high risk of the model (31).

This study has several limitations that should be acknowledged. First, the relatively small sample size constitutes a major limitation of this study; future external validation of the model in larger, multicenter cohorts is warranted. Furthermore, the resampling analysis based on bootstrapping, which yielded a reduced mean accuracy, underscores the limited generalizability of our findings and further diminishes the statistical power to detect subtle associations. Similarly, due to the limited sample size, the association between valve morphology and hemodynamics was not analyzed. Second, the analysis did not incorporate established genetic markers (e.g., GATA5 mutations) or novel circulating biomarkers such as endotrophin (9, 18), which may provide deeper mechanistic insights and enhance risk stratification. Third, the absence of longitudinal data precludes an assessment of the model's predictive efficacy for tracking the progression of aortic dilation over time. Addressing these limitations in subsequent studies will be crucial for confirming the robustness and clinical utility of our results.