Predictive nomogram for abdominal aortic calcification risk: a multi-algorithm approach integrating demographic and clinical variables Authors: Yapeng Zhu, Wei Guo, Hongpeng Zhang Published: 2026 ABSTRACT BackgroundAbdominal aortic calcification (AAC), a significant predictor of cardiovascular events and mortality, lacks a simple predictive tool for early risk stratification. This study aimed to develop and validate a clinical nomogram for predicting AAC risk using accessible demographic, comorbidity, and lifestyle factors.MethodsUtilizing data from the 2013–2014 National Health and Nutrition Examination Survey (NHANES) cohort, AAC was assessed via dual-energy X-ray absorptiometry (DXA) using the Kauppila scoring system (score > 0 defined AAC). Participants were stratified into training (n = 2,198) and validation (n = 942) cohorts. The nomogram was developed through multi-algorithm consensus, incorporating predictors selected by three robust methodologies: Least Absolute Shrinkage and Selection Operator (LASSO) regression, the best subsets regression (BSR) method and the Boruta algorithm. Model performance was evaluated through ROC, calibration, and decision curve analysis (DCA) curves.ResultsFive predictors—age, BMI, cardiovascular disease (CVD), hypertension, and smoking status—were identified. The nomogram demonstrated robust discrimination, with AUCs of 0.747 (95% CI: 0.724–0.770) and 0.700 (95% CI: 0.660–0.735) in training and validation cohorts, respectively. Calibration curves revealed strong agreement between predicted and observed probabilities. DCA curves confirmed clinical utility across risk thresholds.ConclusionsThis validated nomogram provides a practical tool for AAC risk stratification using readily obtainable clinical parameters, enabling targeted interventions to mitigate calcification progression. Its integration into routine cardiovascular assessments may enhance preventive strategies in resource-limited settings.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13019-026-04588-7. Introduction Vascular calcification (VC) is characterized by the accumulation of calcium salts within the vascular wall, leading to increased rigidity and loss of elasticity [1]. Computed tomography (CT) serves as the primary modality for assessing VC, with calcification in any vascular bed demonstrating associations with cardiovascular disease (CVD) morbidity and mortality [2]. Coronary artery calcification (CAC) has been established as a robust predictor of cardiovascular risk and mortality [3]. While calcification frequently occurs in extra-coronary vascular territories, notably, the abdominal aorta represents one of the earliest sites prone to calcification, often preceding coronary involvement [4]. Abdominal aortic calcification (AAC) is a complex pathophysiological process involving intricate cellular and molecular mechanisms, which manifests as progressive calcium deposition within the arterial wall accompanied by mineral co-aggregation, ultimately forming calcified plaques [5, 6]. Given its clinical relevance, AAC has garnered increasing scientific attention in recent decades, particularly due to its high prevalence in patients with diabetes mellitus and chronic kidney disease (CKD) [7, 8]. First, epidemiological data revealed a national prevalence of 28.8% for AAC in the U.S. population [9], with second, age-dependent incidence rates reaching 96% of individuals aged 85 years and older [10]. Longitudinal analyses further confirm significant associations between AAC progression and adverse cardiovascular outcomes, such as CVD events and myocardial infarction (MI) [11]. A decade-long cohort study found that both rapid nonlinear and gradual slow progression trajectories of AAC were associated with higher mortality risks compared to stable trajectories—indicating that progression velocity, not just severity, is a critical risk marker [12]. A 5-year randomized controlled trial (RCT) involving 1,471 healthy women demonstrated a significantly increased risk of CVD and MI in those with AAC versus those without [11], suggesting AAC may serve as an independent risk factor. These findings underscore the clinical need for healthcare providers to integrate AAC evaluation into cardiovascular risk stratification protocols [13]. Consequently, identifying high-risk populations during the early pathogenesis of AAC is critical for implementing primary prevention strategies, such as targeted lifestyle modifications or pharmacotherapies, to mitigate disease progression and improve long-term outcomes. Emerging evidence substantiates the pivotal role of oxidative stress in the pathogenesis and progression of VC [14, 15]. The disruption of redox homeostasis between pro-oxidant species and antioxidant defense mechanisms triggers oxidative modifications to nucleic acids, lipids, proteins, and carbohydrates, a process that creates a pro-inflammatory microenvironment and mechanistically drives the osteogenic transdifferentiation of vascular smooth muscle cells (VSMCs) [16]. Consequently, a growing number of studies have explored the relationship between inflammatory markers and AAC [17–20]. Despite the association between these markers and AAC, there are significant limitations, such as the inability to account for causality, the failure to take into account other confounding factors, but most importantly the fact that very few of these factors have proven effective in predicting the risk of AAC occurrence. Challenges in clinical implementation of primary AAC screening further underscore the need for alternative approaches. Established diagnostic techniques such as dual-energy X-ray absorptiometry (DXA) [21] and CT [22] require specialized imaging infrastructure. Given socioeconomic and geographic constraints that limit access to advanced medical infrastructure in certain patient populations, our research prioritizes the development of robust prediction models incorporating readily obtainable risk parameters encompassing demographic characteristics, comorbidities, lifestyle determinants, and biochemical profiles. The nomogram model, as an intuitive multivariate prognostic tool, has gained prominence in clinical risk stratification by enabling the systematic integration of diverse predictor variables [23]. Leveraging data derived from the National Health and Nutrition Examination Survey (NHANES)—a nationally representative cross-sectional epidemiological study—this investigation developed a predictive nomogram for estimating AAC risk. The aimed is to provide clinicians with a practical tool to identify early patients at high risk of AAC using factors like key demographics, comorbidities, lifestyle, and biochemical indicators, without relying on specialized tests. By providing a simple tool to predict the risk of developing AAC, this study may help to reduce the risk of developing AAC through early detection and targeted interventions. Methods Data source and study population The study population was drawn from NHANES, a nationally representative public database collected and maintained by the U.S. Centers for Disease Control and Prevention (CDC). NHANES employs a cross-sectional design to collect data on basic health status using a stratified, multistage probability sampling methodology in non-institutional demographics. The National Center for Health Statistics (NCHS) is responsible for implementing the survey. The study protocol was approved by the NCHS Ethical Review Committee, and all participants provided written informed consent. For this study, we specifically selected the NHANES dataset from 2013 to 2014, as this was the only cycle with systematic AAC assessment. Data were screened according to the content of this study, and the exclusion criteria were (1) participants under 40 years of age and (2) participants who lacked complete data on AAC. Following this comprehensive screening process, 3,140 participants were ultimately selected for analysis (Fig. 1). Notably, 947 of these participants were diagnosed with AAC, providing a robust foundation for our subsequent analyses. Fig. 1A detailed flow chart of participant recruitment A detailed flow chart of participant recruitment Measurements and definition of AAC AAC assessment was conducted through lateral DXA scanning of the lumbar spine (L1-L4 vertebrae), as previously reported [24, 25]. The severity of AAC was quantified using the Kauppila scoring system [26, 27], a validated method that correlates higher scores with more extensive aortic calcification [28]. This system segments the anterior and posterior walls of the aorta into four zones each, aligned with the L1-L4 vertebrae. Calcification within these eight zones appears visually as scattered white streaks or linear calcifications on the aortic walls. Scoring criteria are as follows: “0” indicates no calcification, “1” indicates calcification in one-third or less of the aortic wall segment, “2” indicates calcification between one-third and two-thirds, and “3” indicates calcification in more than two-thirds of the segment. Separate scoring of the anterior and posterior walls yields a per-vertebral-level score range of “0”–“6” and a total score range of “0”–“24”. According to previous studies [28], participants were classified as having AAC if their score > 0. Variables Based on prior studies and clinical insights, we identified variables potentially affecting AAC development as predictors. Our analysis focused on demographic factors, comorbidities, lifestyle factors, and key biochemical indicators to ensure model simplicity and practicality. Demographic factors included age (years), sex (male, female), race (Mexican American, other Hispanic, non-Hispanic white, non-Hispanic black, other), education level (less than high school, more than high school), body mass index (BMI), and poverty-to-income ratio (PIR). Comorbidities included hypertension (no, yes), diabetes (no, yes), CVD (no, yes) and cancer (no, yes). Hypertension was defined as systolic/diastolic blood pressure of ≥ 140/90 mmHg, self-reported hypertension, or use of antihypertensive medication. Diabetes was defined as a self-reported physician diagnosis, use of oral glucose-lowering medicines or insulin, or a fasting plasma glucose level of at least 126 mg/dL, 75 g oral glucose tolerance test (OGTT) of at least 200 mg/dL, or hemoglobin A1c (HbA1c) level of at least 6.5%. CVD was determined by asking the following five questions, “Has a doctor or other health professional ever told you that you have congestive heart failure, coronary heart disease, angina heart attack, or stroke?” If the answer to one of these questions was yes, the participant was classified as having CVD. Cancer was defined as: “Has a doctor or other health professional ever told you that you have cancer or a malignancy of any kind?” Lifestyle factors included smoking status and alcohol consumption. Smoking status was categorized as never smoker (fewer than 100 cigarettes), former smoker (at least 100 cigarettes but no longer smokes), and current smoker (at least 100 cigarettes and currently smoking). Alcohol consumption was defined as having at least 12 alcoholic drinks in the past year. Biochemical indicators included albumin, AST, ALT, creatinine, uric acid, eGFR, serum calcium, serum phosphorus, 25-hydroxyvitamin D, and TC. In this study, the eGFR was meticulously calculated utilizing standardized serum creatinine measurements, employing the well-established Modification of Diet in Renal Disease (MDRD) study equation to accurately assess renal function [29]. Detailed measurement procedures for these variables are available at https://www.cdc.gov/nchs/nhanes/. Model establishment and validation We randomly allocated 3,140 participants into a training cohort (n = 2,198) and a validation cohort (n = 942) at a 7:3 ratio. The training cohort was used for model development, while the validation cohort was reserved for internal validation. Predictor variables were selected based on the consensus results from the Least Absolute Shrinkage and Selection Operator (LASSO) regression model, the best subsets regression (BSR) method, and the Boruta algorithm. Using these variables, we developed a clinical diagnostic nomogram and assessed its discriminative capacity via receiver operating characteristic (ROC) curves [30]. The model’s performance was evaluated using the area under the curve (AUC), calibration curves, and decision curve analysis (DCA) curves. A higher AUC value signifies superior diagnostic accuracy in distinguishing patients from healthy individuals. Calibration curves visually represent the alignment between predicted and actual outcomes [31]. DCA provides a thorough evaluation of the model’s clinical utility by examining the net benefit across various probability thresholds [32]. To further ensure the reliability of our nomogram, sensitivity analyses were conducted. These included calculating the corrected C-index through 500 bootstrapped resamples of the training set and validating the results using 10-fold cross-validation, both of which confirmed the stability and robustness of the model. Statistical analysis Normally distributed continuous variables were presented as mean ± standard deviation (SD), whereas skewed continuous variables were reported as median (interquartile range [IQR]). Categorical variables were expressed as frequencies and percentages (%). For group comparisons, continuous variables were analyzed using the independent samples Student’s t-test or the Mann-Whitney U-test, depending on their distribution. Categorical data were assessed using the chi-square test as appropriate. To address missing data, we implemented a multivariate single imputation method based on an iterative imputer, utilizing a Bayesian Ridge model as the estimator during each iteration of the round-robin imputation process. All analyses were executed using R Statistical Software (Version 4.2.2, http://www.R-project.org, The R Foundation) and the Free Statistics analysis platform (Version 2.1.1, Beijing, China, http://www.clinicalscientists.cn/freestatistics). A two-sided P-value<0.05 was deemed statistically significant. Results Baseline characteristics Figure 1 provides a flowchart outlining the study selection methodology. The study comprised 3,140 participants, with a mean age of 58.6 ± 12 years; 43.8% were Non-Hispanic White, and 48.3% were male. The overall prevalence of AAC disease within this cohort was 30.2%. Table 1 details the baseline characteristics of participants according to their AAC status. Notably, significant differences were found between groups in age, race, BMI, smoking status, diabetes, hypertension, ALT, creatinine, uric acid, serum calcium, 25-hydroxyvitamin D, TC, CVD, cancer, and eGFR (all P values<0.05). Conversely, characteristics such as sex, education, PIR, alcohol consumption, albumin, AST, and serum phosphorus showed similar distributions across the AAC groups (all P values>0.05), indicating no significant associations with AAC status Table 1Baseline characteristics of participant demographicsVariablesTotal (n = 3140)Without AAC (n = 2193)With AAC (n = 947) p Age (years), Mean ± SD58.6 ± 12.055.9 ± 11.064.8 ± 11.9< 0.001Sex, n (%)0.237Male1518 (48.3)1045 (47.7)473 (49.9)Female1622 (51.7)1148 (52.3)474 (50.1)Race, n (%)< 0.001Mexican American412 (13.1)316 (14.4)96 (10.1)Other Hispanic298 (9.5)225 (10.3)73 (7.7)Non-Hispanic White1375 (43.8)869 (39.6)506 (53.4)Non-Hispanic Black620 (19.7)465 (21.2)155 (16.4)Other race435 (13.9)318 (14.5)117 (12.4)Education, n (%)0.894< High school728 (23.2)507 (23.1)221 (23.3)≥ High school2412 (76.8)1686 (76.9)726 (76.7)PIR, n (%)0.065PIR ≤ 1.3969 (30.9)675 (30.8)294 (31)1.3 < PIR ≤ 3.51084 (34.5)733 (33.4)351 (37.1)PIR > 3.51087 (34.6)785 (35.8)302 (31.9)BMI (kg/m2), Mean ± SD28.4 ± 5.628.8 ± 5.827.7 ± 4.8< 0.001Smoking status, n (%)< 0.001Never1687 (53.7)1252 (57.1)435 (45.9)Former869 (27.7)553 (25.2)316 (33.4)Current584 (18.6)388 (17.7)196 (20.7)Alcohol consumption, n (%)0.703No910 (29.0)640 (29.2)270 (28.5)Yes2230 (71.0)1553 (70.8)677 (71.5)Diabetes, n (%)< 0.001No2408 (76.7)1752 (79.9)656 (69.3)Yes732 (23.3)441 (20.1)291 (30.7)Hypertension, n (%)< 0.001No1443 (46.0)1139 (51.9)304 (32.1)Yes1697 (54.0)1054 (48.1)643 (67.9)Albumin(g/dL), Mean ± SD4.2 ± 0.34.2 ± 0.34.2 ± 0.30.291AST (U/L), Mean ± SD25.5 ± 13.825.4 ± 13.125.7 ± 15.50.614ALT (U/L), Median (IQR)21.0 (16.0, 28.0)21.0 (17.0, 28.0)20.0 (16.0, 27.0)0.04Creatinine (mg/dL), Mean ± SD0.9 ± 0.50.9 ± 0.41.0 ± 0.7< 0.001Uric acid (mg/dL), Mean ± SD5.4 ± 1.45.4 ± 1.45.6 ± 1.4< 0.001Serum calcium (mg/dL), Mean ± SD9.5 ± 0.49.4 ± 0.49.5 ± 0.40.01Serum phosphorus (mg/dL), Mean ± SD3.8 ± 0.63.8 ± 0.63.8 ± 0.60.59625-hydroxyvitamin D (nmol/L), Mean ± SD70.3 ± 29.568.4 ± 28.674.6 ± 31.1< 0.001TC (mg/dL), Mean ± SD194.9 ± 43.6196.4 ± 43.7191.5 ± 43.40.004CVD, n (%)< 0.001No2736 (87.1)1990 (90.7)746 (78.8)Yes404 (12.9)203 (9.3)201 (21.2)Cancer, n (%)< 0.001No2736 (87.1)1971 (89.9)765 (80.8)Yes404 (12.9)222 (10.1)182 (19.2)eGFR (mL/min / 1.73 m2), Mean ± SD81.8 ± 22.184.3 ± 21.576.0 ± 22.4< 0.001n, number; SD, standard deviation; IQR, interquartile range; PIR, ratio of family income to poverty; BMI, body mass index; AST: aspartate transaminase; ALT: alanine transaminase; TC, total cholesterol; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; AAC, abdominal aortic calcification Table 1Baseline characteristics of participant demographicsVariablesTotal (n = 3140)Without AAC (n = 2193)With AAC (n = 947) p Age (years), Mean ± SD58.6 ± 12.055.9 ± 11.064.8 ± 11.9< 0.001Sex, n (%)0.237Male1518 (48.3)1045 (47.7)473 (49.9)Female1622 (51.7)1148 (52.3)474 (50.1)Race, n (%)< 0.001Mexican American412 (13.1)316 (14.4)96 (10.1)Other Hispanic298 (9.5)225 (10.3)73 (7.7)Non-Hispanic White1375 (43.8)869 (39.6)506 (53.4)Non-Hispanic Black620 (19.7)465 (21.2)155 (16.4)Other race435 (13.9)318 (14.5)117 (12.4)Education, n (%)0.894< High school728 (23.2)507 (23.1)221 (23.3)≥ High school2412 (76.8)1686 (76.9)726 (76.7)PIR, n (%)0.065PIR ≤ 1.3969 (30.9)675 (30.8)294 (31)1.3 < PIR ≤ 3.51084 (34.5)733 (33.4)351 (37.1)PIR > 3.51087 (34.6)785 (35.8)302 (31.9)BMI (kg/m2), Mean ± SD28.4 ± 5.628.8 ± 5.827.7 ± 4.8< 0.001Smoking status, n (%)< 0.001Never1687 (53.7)1252 (57.1)435 (45.9)Former869 (27.7)553 (25.2)316 (33.4)Current584 (18.6)388 (17.7)196 (20.7)Alcohol consumption, n (%)0.703No910 (29.0)640 (29.2)270 (28.5)Yes2230 (71.0)1553 (70.8)677 (71.5)Diabetes, n (%)< 0.001No2408 (76.7)1752 (79.9)656 (69.3)Yes732 (23.3)441 (20.1)291 (30.7)Hypertension, n (%)< 0.001No1443 (46.0)1139 (51.9)304 (32.1)Yes1697 (54.0)1054 (48.1)643 (67.9)Albumin(g/dL), Mean ± SD4.2 ± 0.34.2 ± 0.34.2 ± 0.30.291AST (U/L), Mean ± SD25.5 ± 13.825.4 ± 13.125.7 ± 15.50.614ALT (U/L), Median (IQR)21.0 (16.0, 28.0)21.0 (17.0, 28.0)20.0 (16.0, 27.0)0.04Creatinine (mg/dL), Mean ± SD0.9 ± 0.50.9 ± 0.41.0 ± 0.7< 0.001Uric acid (mg/dL), Mean ± SD5.4 ± 1.45.4 ± 1.45.6 ± 1.4< 0.001Serum calcium (mg/dL), Mean ± SD9.5 ± 0.49.4 ± 0.49.5 ± 0.40.01Serum phosphorus (mg/dL), Mean ± SD3.8 ± 0.63.8 ± 0.63.8 ± 0.60.59625-hydroxyvitamin D (nmol/L), Mean ± SD70.3 ± 29.568.4 ± 28.674.6 ± 31.1< 0.001TC (mg/dL), Mean ± SD194.9 ± 43.6196.4 ± 43.7191.5 ± 43.40.004CVD, n (%)< 0.001No2736 (87.1)1990 (90.7)746 (78.8)Yes404 (12.9)203 (9.3)201 (21.2)Cancer, n (%)< 0.001No2736 (87.1)1971 (89.9)765 (80.8)Yes404 (12.9)222 (10.1)182 (19.2)eGFR (mL/min / 1.73 m2), Mean ± SD81.8 ± 22.184.3 ± 21.576.0 ± 22.4< 0.001n, number; SD, standard deviation; IQR, interquartile range; PIR, ratio of family income to poverty; BMI, body mass index; AST: aspartate transaminase; ALT: alanine transaminase; TC, total cholesterol; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; AAC, abdominal aortic calcification Baseline characteristics of participant demographics n, number; SD, standard deviation; IQR, interquartile range; PIR, ratio of family income to poverty; BMI, body mass index; AST: aspartate transaminase; ALT: alanine transaminase; TC, total cholesterol; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; AAC, abdominal aortic calcification Predictor screening The optimal LASSO tuning parameter (λ = 0.029) was determined via 5-fold cross-validation using the one-standard-error criterion. This identified five predictors for AAC occurrence: age, BMI, CVD, hypertension, and smoking status (Fig. 2A and B). We also applied the BSR method, which selected seven predictive factors based on the minimum Bayesian information criterion (BIC) of − 440 (Fig. 2C and D): age, BMI, CVD, hypertension, smoking status, race, and diabetes. The Boruta algorithm further identified fourteen predictors (Fig. 2E), including age, BMI, diabetes, hypertension, race, AST, ALT, CVD, sex, cancer, smoking status, 25-hydroxyvitamin D, creatinine, and eGFR. By intersecting the results of all three methods, the final five predictors were determined to be age, BMI, CVD, hypertension, and smoking status (Fig. 2F), ensuring a robust and consistent selection of variables for predicting AAC occurrence. Fig. 2Variables selection methods. (A) Lasso regression coefficient diagram. (B) Lasso regression cross-validation plot. (C, D) The selection of variables using the BSR method (BIC). (E) Feature selection based on the Boruta algorithm. (F) Five risk factors related to AAC. BSR, best subsets regression; BIC, Bayesian information criterion; PIR, ratio of family income to poverty; BMI, body mass index; DM, diabetes; HBP, hypertension; AST: aspartate transaminase; ALT: alanine transaminase; TC, total cholesterol; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate Variables selection methods. (A) Lasso regression coefficient diagram. (B) Lasso regression cross-validation plot. (C, D) The selection of variables using the BSR method (BIC). (E) Feature selection based on the Boruta algorithm. (F) Five risk factors related to AAC. BSR, best subsets regression; BIC, Bayesian information criterion; PIR, ratio of family income to poverty; BMI, body mass index; DM, diabetes; HBP, hypertension; AST: aspartate transaminase; ALT: alanine transaminase; TC, total cholesterol; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate Nomogram construction Participants were randomly divided into the training set (n = 2,198) and the validation set (n = 942) through stratified random sampling at a 7:3 ratio. The training set was used to develop the nomogram model, while the validation set evaluated the predictive model’s performance. We selected five key predictors using the LASSO regression, BIC method, and Boruta algorithm: age, BMI, CVD, hypertension, and smoking status (Fig. 2). A nomogram was then constructed based on these predictors (Fig. 3). Notably, lower BMI is associated with a higher risk of AAC in American adults. For example, an 80-year-old current smoker with a BMI of 20 kg/m² and a history of CVD and hypertension would have a total score of approximately 210 on the nomogram, indicating a 78% probability of having AAC. This demonstrates how the nomogram can be used to predict an individual’s likelihood of having AAC based on specific risk factors. Fig. 3Nomogram for predicting AAC risk Nomogram for predicting AAC risk Performance and validation of the nomogram We evaluated the nomogram model’s performance using ROC, calibration, and DCA curves across both the training and validation sets. In the training set, the nomogram achieved an AUC of 0.747 (95% CI: 0.724–0.770) for predicting AAC risk (Fig. 4A), and an AUC of 0.700 (95% CI: 0.660–0.735) in the validation set (Fig. 4B). The calibration curves showed strong correlation between the actual and predicted values across the 0–1.0 range (Fig. 4C and D), confirming the model’s reliability. Moreover, DCA curves revealed significant net gains for the nomogram model in both the training set (Fig. 4E) and validation set (Fig. 4F), demonstrating its clinical utility in identifying individuals at risk of AAC. Fig. 4Performance and validation of the nomogram model. (A) ROC curve of the nomogram in the training set. (B) ROC curve in the validation set. (C) Calibration curve of the nomogram in the training set. (D) Calibration curve in the validation set. (E) Decision curve analysis of the nomogram in the training set. (F) Decision curve analysis in the validation set Performance and validation of the nomogram model. (A) ROC curve of the nomogram in the training set. (B) ROC curve in the validation set. (C) Calibration curve of the nomogram in the training set. (D) Calibration curve in the validation set. (E) Decision curve analysis of the nomogram in the training set. (F) Decision curve analysis in the validation set Sensitivity analysis During the sensitivity analysis, we employed bootstrap and 10-fold cross-validation to assess the model’s stability. The bootstrap-corrected C-index of 0.744 indicates reliable predictive performance. The calibration curves of the bootstrap method, shown in supplementary Fig. 1, confirm the consistency of the model’s predictions. Furthermore, the 10-fold cross-validation results, provided in Fig. 2A and B, reinforce the stability and robustness of our findings, ensuring the model’s validity across different validation methods. Discussion This study developed and rigorously validated a nomogram using NHANES (2013–2014) data to predict the risk of AAC occurrence. The nomogram integrates key demographic characteristics, comorbidities, and lifestyle factors, making it a practical and computationally efficient tool for clinical application. AAC is defined by the abnormal accumulation of minerals such as calcium and phosphate within the aortic wall, positioning it as an early indicator of atherosclerotic calcification and an independent predictor of cardiovascular events and all-cause mortality [33]. Numerous studies have underscored the intricate relationships between AAC and a range of clinical outcomes. For example, a prospective cohort study indicated that individuals with severe AAC face a higher risk of developing dementia in later life [34]. Additionally, a retrospective case-control study established a link between AAC and ruptured intracranial aneurysms [35]. Nevertheless, the pathophysiological mechanisms of VC remain elusive, and therapeutic options for calcification are currently restricted [36]. While preclinical research suggests that statins and calcium channel blockers (CCBs) may mitigate VC [37], their clinical efficacy in humans has been debated [38–40]. Sodium thiosulfate has demonstrated potential in reducing vascular calcification in rodent models [41], yet its safety and effectiveness in humans require validation through large-scale RCTs. Given these challenges, there is an urgent need to identify risk factors for AAC. Although previous investigations have explored AAC-related risk factors [42, 43], no studies have focused on developing a reliable predictive model using simple and readily accessible predictors such as demographic characteristics, comorbidities, and lifestyle parameters. Therefore, this study aimed to develop and validate a novel prediction model based on these clinically practical variables. The proposed nomogram presents two significant advantages. First, it combines demographic factors, comorbid conditions, and lifestyle factors, offering a comprehensive view of AAC risk. Second, its visual format improves the understanding of risk variables, enabling clinicians to efficiently quantify individual risk profiles and supporting more informed clinical decision-making. This study utilized advanced statistical methods to select variables for predicting AAC occurrence. By combining results from LASSO regression, BSR, and the Boruta algorithm, we pinpointed five key predictors: age, BMI, CVD, hypertension, and smoking status. Unlike conventional approaches that use only univariate analysis (a practice known as “data snooping”), which can cause model overfitting and subpar performance [44], our multi-algorithm strategy reduces bias and improves model reliability. These predictors were then integrated into a nomogram. The nomogram achieved an AUC of 0.747 in the training cohort and maintained an AUC of 0.700 in the validation cohort, demonstrating strong predictive discrimination. Calibration curves showed high consistency between predicted and observed probabilities in both sets. DCA also highlighted the nomogram’s clinical value across a wide range of risk thresholds. Among these factors, the relationship between AAC and age has attracted significant scientific attention. Previous research demonstrated that 18 F-sodium fluoride (18 F-NaF) uptake in individuals aged 50–75 years was significantly higher than in those aged 25–35 years [45], likely due to age-related increases in vascular smooth muscle cell proliferation and DNA synthesis, which promote calcification [46]. Age emerged as an independent risk factor for AAC, particularly among CKD patients [47]. For example, Chen et al. reported a 72.4% prevalence of AAC in maintenance hemodialysis (MHD) patients, with a strong age-dependent correlation [47]. Similarly, age remained a critical predictor of AAC in non-CKD populations [48]. Our findings are consistent with these established observations, reinforcing the robustness of our results. Interestingly, our analysis revealed an inverse correlation between BMI and AAC risk, with higher BMI values associated with a lower likelihood of calcification. This may be explained by BMI’s limitations in accurately reflecting adipose tissue distribution [49]. Epidemiological evidence indicated a J-shaped association between BMI and all-cause mortality, wherein underweight status (low BMI) correlated with heightened mortality risk [50]. However, the ongoing debate surrounding the “obesity paradox” may be partially explained by confounding factors: individuals in lower BMI ranges often exhibit reduced lean mass rather than decreased adiposity [51]. Prior studies have similarly documented an inverse BMI-AAC relationship [9], necessitating further mechanistic investigations to elucidate this counterintuitive association. In this study, CVD was identified as a predictor of AAC. It should be emphasized that our operational definition of CVD [52]—“Has a doctor or other health professional ever diagnosed you with congestive heart failure, coronary heart disease, angina pectoris, myocardial infarction, or stroke?”—represents a narrower clinical spectrum than conventional CVD classifications. Participants reporting at least one of these five conditions were categorized as CVD-positive and demonstrated elevated AAC risk. Hypertension was identified as a significant predictor of AAC, corroborating prior findings that link elevated systolic blood pressure to increased AAC incidence [53]. The pathophysiological mechanism underlying this relationship likely involves hypertension - induced vascular remodeling. This process alters extracellular matrix deposition and activates vascular cells, thereby fostering a microenvironment conducive to arterial calcium deposition or elastocalcinosis [54, 55]. Notably, medial calcification and hypertension exhibit bidirectional pathogenic interplay [56]: hypertension serves as a calcification-promoting stressor, while medial calcification may exacerbate hypertension via reduced arterial compliance [57]. Smoking, a well-established risk factor for atherosclerosis—the primary pathological basis of AAC—demonstrated a clear dose-dependent association with AAC risk [48]. Studies have shown that smokers face a significantly higher likelihood of developing AAC [58]. The pro-calcification effects of smoking are attributed to mechanisms such as increased oxidative stress, amplified inflammatory responses, and endothelial dysfunction [59]. These five clinically accessible predictors—obtainable through routine history-taking without specialized laboratory testing—collectively enable efficient risk stratification. The model’s parsimonious design streamlines clinical workflows, facilitates early AAC detection, and allows rapid risk quantification through nomogram scoring ultimately supporting timely therapeutic interventions. The clinical utility of the proposed nomogram may be conceptualized within two distinct prevention paradigms. For primary prevention, it could serve as a screening tool for the general population without established CVD, enabling risk stratification for AAC before clinical symptoms arise and prompting lifestyle interventions—counseling, blood pressure optimization, and smoking cessation—to halt subclinical AAC progression, particularly in resource-limited settings where DXA or CT are not routinely accessible. For secondary prevention in patients with prior cardiovascular events, the nomogram offers incremental value for risk re-stratification, supporting more aggressive measures including intensive statin therapy, tighter blood pressure control, and closer surveillance. Moreover, given evidence that AAC progression trajectory independently predicts mortality [12], serial nomogram application may facilitate longitudinal risk monitoring. Future prospective studies are warranted to validate these frameworks and establish nomogram-guided intervention thresholds for both contexts. This investigation has several limitations that warrant consideration. First, the cross-sectional study design prevents definitive causal inferences between the identified predictors and AAC. Second, the analytical framework was developed using data from the U.S. adult population, which may limit its generalizability to other ethnic or geographic groups with different genetic and environmental risk factors. Third, while internal validation confirmed satisfactory discriminative performance, external validation in independent, multicenter cohorts is necessary to confirm the nomogram’s reproducibility across various healthcare settings. Fourth, CVD was defined based on self-reported medical history rather than adjudicated clinical records, which may introduce recall bias. However, this approach is consistent with established methodologies in previously published NHANES-based studies. These limitations highlight the need for future research to use longitudinal designs and conduct external validation in multiethnic cohorts. This would enhance the model’s clinical applicability and statistical robustness across diverse populations. Conclusions This study successfully developed and validated a clinically applicable nomogram for AAC risk stratification. The nomogram offers clinicians a practical tool to identify high-risk individuals by incorporating readily available clinical indicators such as demographic parameters, comorbidities, and lifestyle characteristics. By enabling early detection and facilitating precision interventions aimed at slowing calcification progression, this risk prediction model holds significant translational potential for enhancing patient management. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1.