Abdominal aortic calcification (AAC) is a common subtype of vascular calcification acknowledged not only as a marker of atherosclerosis [1, 2] but also as a predictor of poor cardiovascular prognosis. Studies have linked AAC to increased risks of all-cause mortality [3, 4], cardiovascular mortality [4], as well as both fatal and nonfatal cardiovascular events [3, 5]. Furthermore, in dialysis patients, AAC correlates with an increase in left ventricular mass [6]. Given its association with these clinically significant outcomes, analyzing risk factors for AAC is crucial for its prevention and the mitigation of adverse cardiovascular events.

Smoking is an acknowledged risk factor of atherosclerosis. Pro-inflammatory and endothelial-damaging components in cigarette smoke have been demonstrated to promote the development of this endothelial injury and atherosclerosis [7]. Despite the well-established link between atherosclerosis and vascular calcification [8], research into the association between smoking and AAC has been predominantly limited to populations with certain characteristics, such as males or individuals afflicted by chronic kidney disease (CKD). For instance, Jung et al. [9] found smoking-related indices like cumulative smoking duration and cumulative smoking amount correlated with AAC in a cohort of 218 middle-aged and elderly men. Similarly, Lioufas et al. [10] also found smoking to be associated with AAC in CKD patients. Furthermore, Pham et al. [11] found that, after adjusting for age and other cardiovascular risk factors, current smokers exhibited an increased risk of AAC progression compared to non-smokers. Zhang et al. [12] found that populations with a smoking history faced an increase in risk of AAC exacerbation. These evidence underscores the need for further investigation into smoking’s role in AAC development across broader demographic groups.

Research on the relationship between alcohol consumption and AAC has yielded inconsistent results. Forbang et al. [13] found that alcohol consumption was correlated with an increase in the natural logarithm of AAC volume. In contrast, Wu [14] found no significant association between alcohol consumption history and AAC in a population of patients with CKD Stage 3–5. In a more recent study examining a cross-sectional sample of the general United States (U.S.) population no significant differences in AAC prevalence among never, former, mild, moderate and heavy alcohol consumers were found [15]. These divergent outcomes highlight the complexity of the relationship between alcohol consumption and AAC development, suggesting that additional research is needed to delineate the underlying mechanisms and potential moderating factors.

Despite the well-documented link between smoking and atherosclerosis, the specific association between smoking and AAC remains underexplored, particularly in the general middle-aged and elderly population. Studies utilizing nationwide survey data for the examination of this association are notably scarce. Similarly, existent research on the link between alcohol consumption and AAC is limited and yields inconsistent results. Consequently, there is a significant need to undertake comprehensive studies using nationwide survey data to investigate these associations. Such research could provide crucial insights into the prevention of AAC and its associated adverse events.

Vascular calcification is a heterogeneous disease comprised of various subtypes with distinctly different risk factors. Vascular calcification can be classified as (1) intimal calcification, which is closely associated with atherosclerosis; (2) medial calcification, which is closely related to declined renal function and diabetes, and (3) genetic calcification [8]. As the nomenclature suggests, genetic calcification is associated with genetic disorders such as pseudoxanthoma elasticum (PXE) and Marfan syndrome [8], whose incidence are generally minuscule. For instance, the incidence of PXE between January, 2011 to December, 2020 stood at a mere 0.08 per 100,000 person-years [29]. Given the low prevalence of genetic disorders associated with genetic calcification, risk factors for intimal (atherosclerosis) and medial (renal function impairment and diabetes) are the major risk factors of AAC in the general middle-aged and elderly population. This is supported by findings from our study which observed an increase in AAC scores associated with advancing age and a decrease in eGFR. Similar conclusions concerning the prevalence of other AAC risk factors are also drawn, including smoking history as well as hypertension and diabetes diagnosis. These results reaffirmed the close association of AAC and atherosclerosis as well as the close relationship of traditional cardiovascular risk factors and AAC. However, the proportion of current smokers did not vary significantly across the groups. This, as is verified in the present study, is associated with the heterogeneity of risks of AAC within the subpopulation of current non-smokers, which is discussed later in this section.

The biological pathways through which smoking contributes to atherosclerosis and other cardiovascular diseases involve a highly complex mixture of compounds found in cigarette fumes and their interaction with the smoker’s personal traits, including environmental and genetic factors [30]. Endothelial dysfunction, a widely and long-lastingly studied topic, has been recognized as a pivotal mechanism associated with smoking-induced atherosclerosis. In as early as 1992, Celermajer et al. [31] found clinical evidence linking smoking to endothelial dysfunction as measured by decreased flow-mediated dilation. Subsequent biological studies have provided further insight into this phenomenon. Notably, nicotine has been shown to suppress endothelial nitric oxide synthase activation, leading to reduced nitric oxide production and increased nitric oxide depletion, culminating in endothelial and vascular dysfunction along with associated vascular calcification [32]. In addition, nicotine induces the production of inflammatory cytokines including interleukin (IL)-1$\beta$ and IL-18 in human aortic endothelial cells, prompting vascular inflammation [33], another key mechanism of atherosclerosis. A recent study also revealed the role of oxidative stress and the release of extracellular vesicles in the development of nicotine-induced vascular calcification [34].

Clinical studies have also revealed the link between smoking and AAC. Jung et al. [9] analyzed risk factors for AAC in a cohort of 218 middle-aged and elderly men, identifying links between AAC and smoking-related indices, including cumulative smoking time and cumulative smoking amount. Similarly, Lioufas et al. [10] found a correlation between smoking and AAC in a cohort of 278 CKD patients. Extending these observations, this study utilized a larger and more representative sample drawn from a nationwide survey that was not restricted to sex or disease, aligning with previous research that underscored the association between smoking and AAC.

Results of this study indicate that the risks to which current smokers were exposed did not differ significantly from those of current non-smokers in terms of both the presence of AAC and severe AAC. From this, a hypothesis that those former smokers still faced a higher risk of AAC arose. If current non-smokers (both former smokers and never smokers) truly benefited from a lower risk of AAC, they should have been more prevalent in the group with an AAC score of 0 or the one with an AAC score between 0 and 6. In other words, the proportion of current non-smokers in the two groups should had been higher, while the proportion of current non-smokers in the group with an AAC score not less than 6 should had been lower. However, findings from the data contradicted this scenario, suggesting that the hypothesis of a homogeneous benefit among all current non-smokers might be incorrect. Further analysis comparing the risk of AAC among never, former, and current smokers confirmed this corollary. This finding is consistent with the research of Lv et al. [21], who analyzed the risk factors of severe AAC in the NHANES population using an identical definition of severe AAC as the one adopted by the current study. Their results confirmed a significantly increased risk of severe AAC for both former and current smokers compared to never smokers. Although our results parallel these findings in the sense of also revealing the association of AAC and smoking, a common atherosclerotic risk factor, as well as the cardiovascular harm of smoking, Lv et al. [21] found no significant difference in risk of severe AAC between former and current smokers while it was found in the present study that former smokers were benefited from a decrease in risk of presence of AAC compared to the currently smoking population. Taken together, these results suggest that the cardiovascular benefits associated with smoking cessation primarily manifest as reduction in the risk of AAC presence rather than severe AAC. Given the consistently elevated risks of AAC presence in elderly ever and current smokers revealed by subgroup analyses, promoting smoking cessation is of significant clinical and epidemiological importance, particularly for elderly individuals. This emphasizes the critical need for tobacco control measures targeting this population to mitigate their elevated risk of AAC and their associated cardiovascular complications.

The potential roles of alcoholic drinks, including ethanol and its coexistent compounds, in cardiovascular health are complex and not fully understood. From a biological perspective, alcohol affects the development of atherosclerosis in many respects. Alterations of blood lipid profile is one them. Alcohol has been shown to elevate high-density lipoprotein (HDL) levels except in individuals with severe hepatic dysfunction [35] and affect low-density lipoprotein (LDL) levels. However, the exact impact of alcohol consumption on LDL varies among different populations and is influenced by genetic variations [36, 37] and consumption patterns, such as binge versus regular drinking [35]. This contributes to the inconsistencies in the observed associations between alcohol consumption and atherosclerotic lesions like AAC.

Alcohol has also been found to modulate inflammatory mediators. Particularly, ethanol and non-alcoholic components of alcoholic drinks, such as polyphenols, downregulate the expression of intercellular adhesion molecule-1 (ICAM-1), vascular cell adhesion molecule-1 (VCAM-1) and E-selectin while also reducing monocyte adhesion to the vascular endothelium [38]. Despite these findings, evidence is sparse regarding the mechanisms via which excessive alcohol consumption leads to atherosclerotic lesions like AAC. This gap was noted by the finding in the subgroup of our study’s alcohol consumers who consumed more than two drinks per day over the past 12 months.

It has been postulated that the duo of ethanol and its metabolite, acetaldehyde, is crucial in determining the shift from atheroprotective to atherogenic effects of alcohol consumption [39]. More specifically, acetaldehyde has been shown to enhance monocyte adhesion to the endothelium, contributing to atherogenesis [40]. Once the level of acetaldehyde surpasses a certain threshold where the atherogenic effect of acetaldehyde balances the atheroprotective effect of ethanol, the effect of the former becomes dominant [39]. The interplay between alcohol and the elevation of blood pressure it induces [41] is another proposed conjecture of this transition where the detrimental effects of increased blood pressure may outweigh the cardiovascular benefits alcohol might provide [39].

Given the complexity of the biological actions of ethanol and other chemicals in alcoholic beverages, it is no surprise that clinical studies conducted thus far have reached inconsistent results. For example, Forbang et al. [13] observed a significant association between alcohol consumption and an increase in the natural logarithm of AAC volume among 1413 community-dwelling adult individuals with an AAC score above 0. However, Wu [14] found no association between alcohol consumption history and AAC in a Chinese cohort of patients with CKD stages 3–5. Similarly, this study found no association between alcohol consumption history and either the presence of AAC or severe AAC after adjusting for confounders in the entire cohort, which is in line with Wu’s results.

However, discrepancy between our results and those reported by Forbang et al. [13] exists. This may be attributed to several factors. First is measurement variability. AAC score assessment was conducted by different raters (radiologists or clinical doctors inspecting the radiographs and calculating AAC scores), which may introduce inter-observer variability. Second, AAC scores gauged the severity of AAC in terms of the extent of abdominal aorta afflicted. However, AAC may also be concentrated in one or more specific arterial segments and hence may not be perfectly correlated with overall extension. Third is the possibility of population differences. For example, the study by Forbang et al. [13] comprised mainly U.S. residents that included some individuals with Chinese ancestry, contrasting with Wu’s study [14] which focused on hospitalized patients in central China. These demographic and clinical differences can significantly impact the outcomes of studies examining the associations of alcohol and AAC.

In a more detailed analyses in the present study where the entire study population was grouped into never, former and current alcohol consumers to investigate the association of alcohol consumption with the presence of AAC, results indicated that neither former nor current alcohol consumers experienced significantly different risk of AAC compared to never alcohol consumers after adjusting for confounders. This finding aligns with the one obtained by McClelland et al. [42], who investigated the association of alcohol consumption with coronary artery calcification in an American population and also found no significant differences among never, former, and current alcohol consumers. This consistency may be attributed to two factors. First, vascular calcification in different arterial segments share a common taxonomy (i.e., intimal, medial and genetic calcification), each associated with specific risk factors and potentially considered as manifestations of a single disease in different locations. Second, both the current study and the study by McClelland et al. [42] involved American, middle-aged and elderly populations (investigated subjects were $\ge$40 years old in the present study while the ones in McClelland et al. [42] aged between 45 and 84 years old), which may suggest a geographic or demographic consistency in these findings. However, the study by McClelland et al. [42] did not involve age subgroup analyses while it was found in this study that elderly ever and current alcohol consumers were both exposed to a significantly elevated risk of presence of AAC.

The results discussed above were obtained from both univariable and multivariable analyses, with confounders fully adjusted for in the latter. A confounder is defined by three essential characteristics: (1) it is associated with the exposure, (2) it is related to the outcome, and (3) it is not an intermediary in the causal pathway between the exposure and the outcome [43]. Based on these criteria, this study carefully selected several variables as confounders for adjustment: sex, age, diabetes, hypertension, CHD, LDL-C, and usage of lipid-lowering agents. The following paragraphs outline evidence in establishing the association of the confounder and the outcome (i.e., AAC) as well as the exposures (i.e., smoking and alcohol consumption), justifying their role as confounders adjusted in the analyses.

(1) Sex plays ubiquitous roles in various physiological and pathophysiological conditions, with vascular calcification being no exception. Males typically exhibit an increased prevalence of vascular calcification compared to females [44]. Additionally, significant difference in the proportions of smokers between males and females have been observed, highlighting the association between sex and the risk factor [45]. Furthermore, there is also a notable sex difference in alcohol consumption patterns, specifically in the mean number of standard drinks consumed, which has been found to vary significantly between males and females, providing evidence for the association between sex and alcohol consumption [46].

(2) Age is a well-recognized risk factor for atherosclerosis and vascular calcification. Age has been shown to be significantly associated with calcification in both the coronary artery and the aorta, underscoring its influence on vascular health across different body regions [47]. Additionally, age has been shown to be statistically significantly different among never, former and current smokers [11]. Moreover, an age-associated increase in the mean number of standard drinks consumed has been observed [46].

(3) Diabetes is intricately linked with vascular calcification, a relationship that has been emphasized early in this discussion. Moreover, there is a documented correlation between smoking and diabetes [48]. Additionally, the consumption of alcohol, specifically at low and moderate levels, has also been found to be associated with diabetes [49].

(4) Hypertension is significantly correlated with vascular calcification [50]. Furthermore, smoking is recognized as a risk factor for hypertension, establishing a direct link between the risk factor and the disease [51]. Additionally, the proportion of hypertensive individuals increases with the amount of alcohol consumed. This observation underscores the association between hypertension and alcohol consumption [42].

(5) Atherosclerosis, as mentioned above, has close connections to vascular calcification. Coronary artery calcification has been shown to be associated with AAC [52], underscoring the shared pathological processes between these conditions and the link between coronary atherosclerosis and AAC. Moreover, smoking is a widely recognized risk factor for CHD [53]. Additionally, there is a documented association between alcohol consumption and CHD [54].

(6) The accumulation of LDL-C is one of the first steps in atherosclerosis [1]. Smoking has been found to be associated with alterations in LDL-C concentrations [55]. Research has indicated a lack of association between LDL-C and alcohol consumption [42]. However, there is evidence suggesting an association between the ratio of LDL-C and high-density lipoprotein cholesterol (HDL-C) and alcohol consumption [56].

(7) Usage of lipid-lowering agents, particularly statins, has been associated with vascular calcification [57]. When examining the connection between statin use and smoking, the available evidence presents a mixed picture. A research project reported no direct association between statin use and smoking [58], while another one observed a noticeable difference in smoking prevalence—approximately 10%—between statin users and non-users [59]. The relationship of alcohol consumption and statin use is similarly complex. While one study found little difference in the proportions of various categories of daily alcohol consumption between statin users and non-users [60], another documented a statistically significant association between alcohol consumption and usage of lipid-lowering agents [61]. Taken together, it is decided to accept lipid-lowering agents as a confounder to adjust for in the analyses.

Despite rigorous attempts, the current study is subject to several limitations that may impact the validity of its findings. The first is the risk of recall bias. Histories of smoking and alcohol consumption relied on self- or proxy-reported information. This method is prone to recall bias, where respondents may not accurately remember or may choose to selectively report their consumption habits. The second is the problem of inter-observer variability. The AAC score was based on radiograph inspection. Different observers may interpret radiographic images differently, leading to inconsistencies in the AAC scores calculated. A third limitation regards the AAC score itself, which quantifies disease severity by examining its extent. However, when AAC is localized and not widely distributed, the AAC score may not accurately reflect the true severity of the condition. Fourth is the possibility of residual confounding variables. While efforts were made to control for confounders, the possibility of residual confounding variables remain. For instance, the NHANES dataset does not include information on certain comorbidities like systemic lupus erythematosus, which could influence the results. These limitations necessitate cautious interpretation of the study’s results and suggest that future research should consider more robust methods for data collection and analysis to mitigate these issues.

Given the above limitations, we propose a series of proactive steps that can be used to mitigate similar limitations in future studies. The implementation of standardized training and operating procedures for AAC raters should be employed to limit measurement errors and inter-observer variability. Consistent systematic training ensures that all raters apply the same criteria when evaluating radiographs, leading to more reliable and comparable results across various studies. Modern medicine is increasingly integrating artificial intelligence (AI) technology. This includes utilizing AI tools which can significantly enhance the accuracy and consistency of measurements. AI algorithms can analyze imaging data with high precision and less bias, providing a standardized assessment of calcification that is not influenced by human error. To address recall bias, the use medical insurance reimbursement data and medical records can serve as an effective method for cross-validating responses to the survey questions. By cross-referencing participants’ responses with their medical records, researchers can confirm the accuracy of reported smoking and alcohol consumption habits and other relevant health information. This approach helps enhance the reliability of the data and strengthens the study’s findings.

From a statistical and mathematical perspective, diagnosis and treatment in clinical care, two essentials of everyday clinical practice, can be conceptualized as classification and decision-making tasks, respectively. These tasks align closely with areas extensively explored within mathematics and statistics. For instance, as research continues to reveal the heterogeneity in the associations of alcohol consumption on AAC across different populations, such as those with varying genotypes, there is an emerging opportunity to apply statistical decision-theoretic methods. These methods could quantitatively assess the benefits and risks of alcohol consumption, supporting the development of decision support systems incorporated as modules within hospital information systems (HIS). Such systems or modules would enable health care professionals to readily formulate personalized responses to the question of “Is alcohol good for my health?” in the outpatient clinic and formulate practical guidance on alcohol consumption (e.g., amount, frequency or type of alcohol intake that ensures enhancement of health) or even “alcohol prescriptions” tailored for each patient. Additionally, mathematical models can be valuable in simulating the microenvironment of cigarette fume- or ethanol-exposed endothelium. These models offer not only a dynamic description of pathophysiological processes but also a quantitative framework for understanding these biological interactions more deeply than qualitative descriptions alone. Given the practical challenges in replicating prolonged exposure to cigarette or alcohol in humans through animal, cell, and other biological studies, mathematical modeling presents a promising alternative for quantifying these long-term biological effects. The integration of interdisciplinary scientific approaches promises to enhance our understanding of human health and disease comprehensively. In line with the principles of precision and evidence-based medicine, this interdisciplinary approach could herald the advent of the epoch of algorithm-supported healthcare.

Using nationwide survey data, the current study identifies a significant association between smoking history and both the presence of AAC and severe AAC, whereas current smoking shows no such association. Notably, compared to never smokers, former smokers are at a significantly higher risk for AAC, yet their risk remain lower than that of current smokers. Subgroup analyses identify elderly ever and current smokers as victims of increased AAC risk, as do middle-aged ever smokers. These results indicate that smoking is an independent risk factor for AAC, with stronger risk in the elderly population. Results of this study and the ones obtained by previously conducted studies collectively indicate that the cardiovascular benefits associated with smoking cessation primarily manifest as reduction in the risk of AAC presence rather than severe AAC.

By contrast, the study found no significant associations between either ever or current alcohol consumption and AAC or severe AAC in the entire cohort after adjusting for confounders. Compared with never alcohol consumers, AAC risk do not significantly differ among former or current alcohol consumers. However, subjects consuming more than 2 drinks of alcohol per day as well as the elderly ever and current alcohol consumers suffer from increased AAC risk. These results indicate that alcohol consumption is not an independent risk factor of AAC except in the heavy drinking and elderly subpopulations.

This study utilizes data of NHANES 2013-2014, a publicly available dataset that can be downloaded from its website: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2013. All data this study used can be directly downloaded there.

YZJ: Conceptualization, Methodology (original design of study), Data curation and Investigation (data acquisition and data cleaning), Formal analysis (data analyses and mathematical modeling via SAS), Software (compiling SAS codes), Writing (drafting of original manuscript and edited manuscript submitted for publication). AMD: Conceptualization, Methodology (original design of study), Supervision, Project administration, Funding acquisition, Writing-review& editing. NQL: Funding acquisition, Project administration, Conceptualization, Methodology (original design of study), Writing-review& editing. All authors have full and direct access to and verified the underlying data in this study, and were responsible for the decision to submit the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

The study was carried out in accordance with the guidelines of the Declaration of Helsinki. This study was approved by Ethics Committee of Fuwai Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (Approval No.: 2021-1461). Informed consent was waived since data from the electronic database of medical records do not involve any personally identifiable information.

This study was funded by National Key Research and Development Program of China (2022YFC3602400, 2022YFC3602405); Key Project of the National Health Commission of The People’s Republic of China (2020-ZD13). Funders of the study had no role in study design, data collection, data analysis, data interpretation, or drafting of the manuscript.

The authors declare no conflict of interest.