The National Center for Health Statistics (NCHS) administers the NHANES program, which collects comprehensive health data from a representative sample of adults and children through interviews, physical examinations, and laboratory analyses. The NCHS Institutional Review Board approved the study protocol, and informed consent was obtained from all participants. Detailed information and access to the data are available on the NHANES website.¹

The study population was limited to females aged 20–49 years for several reasons. On one hand, AGP measurements in the NHANES dataset were only available for females aged 12–49 years and children aged 1–5 years during the 2017–2020 and 2021–2023 cycles, while males were not included in the AGP measurements, representing a limitation of the dataset. On the other hand, the gallstone questionnaire, which was used to identify participants’ gallstone history, was only administered to individuals aged 20 years and older.

Of the initial 27,493 enrollees, 10,452 were excluded because they were under 20 years of age and therefore did not complete the gallstone questionnaire. Further exclusions included 128 participants who were pregnant, 42 participants who either refused to answer or were unsure about their gallstone history, 14,100 who lacked serum AGP test results, and 868 participants with incomplete data on key covariates. After these exclusions, the final analytic sample comprised 1,903 participants. The selection process is illustrated in Figure 1, and Supplementary Table 1 provides details on missing data for key covariates.

Participants were assessed for the presence of gallstones by asking, “Has a doctor ever told you that you have gallstones?” Individuals who answered “Yes” were categorized as having gallstones, while those who answered “No” were categorized as not having gallstones.

AGP levels were measured in serum samples as part of the NHANES 2017–2020 and 2021–2023 cycles using a high-sensitivity turbidimetric immunoassay. This method involves the reaction of AGP in the sample with specific antibodies, leading to the formation of antigen–antibody complexes that cause turbidity, which is then quantitatively measured. The assay was performed using a clinical chemistry analyzer, and results were standardized against known controls to ensure accuracy. This robust method allows for the precise quantification of AGP, which is crucial for assessing its association with various health outcomes, including gallstone formation.

The statistical model incorporated the following variables as covariates based on previous research (16–18): age (years), race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Other), level of education (less than high school, high school, college or above), ratio of family income to poverty (<1.3, 1.3–3.5, and >3.5), obesity, alcohol intake, sedentary activity, total cholesterol, hs-CRP, total energy intake, total fiber intake, total fat intake, total cholesterol intake, total moisture intake, and history of smoking, diabetes, and hypertension.

Age was divided into two categories: 20–39 years and 40–49 years.

Diabetes was identified based on an HbA1c level of ≥6.5%, a self-reported diabetes diagnosis, a history of using antidiabetic medications, or fasting blood glucose levels of ≥7 mmol/L.

Hypertension was defined by a mean systolic blood pressure of ≥140 mmHg, a mean diastolic blood pressure of ≥90 mmHg, or a self-reported diagnosis of hypertension.

Obesity was determined as a body mass index (BMI) of ≥30 kg/m².

Alcohol intake was assessed using responses to two specific questionnaire items: ALQ121, which asked, “In the past 12 months, how often did you drink alcoholic beverages?” and ALQ131, which queried, “During the past 12 months, on days when you consumed alcoholic beverages, how many drinks did you typically have?” Participants who reported consuming 12 or more alcoholic drinks per year were classified as alcohol consumers.

Individuals classified as smokers if they had smoked at least 100 cigarettes in their lifetime.

Total cholesterol levels and hs-CRP were derived from laboratory test results.

Sedentary activity was assessed using the NHANES questionnaire, which recorded the self-reported average daily time spent in sedentary behaviors, including activities such as sitting and watching television or using a computer, measured in minutes per day.

Total energy intake, total fiber intake, total fat intake, total cholesterol intake, and total moisture intake were derived from the first 24-h dietary recall interview conducted by NHANES. Participants reported all foods and beverages consumed in the previous 24 h, and these data were subsequently processed to quantify the total intake of each nutrient for the first day.

Two-sided statistical tests were conducted, with significance set at p < 0.05. All analyses were performed using R version 4.4.0. The NHANES official website recommends the use of appropriate sampling weights for statistical analysis and provides detailed guidance on how to apply these weights. In this study, the specific sample weights were used, as AGP data were measured in participants aged 3–5 years and females aged 12–49 years. The “survey” package in R was utilized to account for the NHANES complex sampling design, including weights, strata, and primary sampling units (PSUs), ensuring accurate variance estimation and nationally representative results.

Baseline characteristics were analyzed using the “svyCreateTableOne” function, which automatically applies survey-weighted statistical methods. Unweighted data were generated using the “CreateTableOne” function. Continuous variables are presented as weighted means with standard deviations, while categorical variables are presented as unweighted counts with weighted percentages. All statistical tests, including normality tests and non-parametric tests for continuous variables that did not meet the normality assumption, were performed using survey-weighted methods. An overall summary across all groups was also provided.

In accordance with STROBE guidelines (19), three weighted multivariable regression models were constructed. Model 1 was unadjusted for any covariates. Model 2 was adjusted for age, race/ethnicity, level of education, and the ratio of family income to poverty. Model 3 was adjusted for all potential covariates. Potential collinearity among variables included in the complete model (Model 3) was assessed using generalized variance inflation factors (GVIFs) with adjustments for degrees of freedom. The complete GVIF results for all variables are provided in Supplementary Table 2. While some variables, such as ratio of family income to poverty (2.64), total energy intake (4.36), and total fat intake (4.93), exceeded the commonly accepted threshold of √5 (approximately 2.236), these were not central to the primary research objective. Furthermore, even when total energy intake or total fat intake was excluded from the model, the odds ratio (OR), 95% confidence interval (CI), and p-values for the association between AGP and gallstone risk showed only slightly changes. Thus, overall multicollinearity was not considered a significant concern for evaluating the association between AGP and gallstone risk.

Subgroup analyses were performed using weighted multivariate regression analysis for all categorical variables. Interaction terms were added to the models, and the log-likelihood ratio test was used to examine heterogeneity in associations across subgroups. A smoothed curve was generated to identify the dose–response relationship between AGP and gallstone risk using generalized additive models (GAM) via the “mgcv” package in R. GAM was chosen for its ability to model non-linear relationships flexibly using penalized regression splines. Non-linearity was assessed by comparing the Akaike information criterion (AIC) values of the GAM and linear models. Additionally, the smoothed curves were visually inspected for deviations from linearity to further confirm the presence of non-linear relationships.

To mitigate bias due to missing data, MI was performed using the “MICE” package in R software (20). The imputation model included variables listed in Supplementary Table 1, selected based on the variables included in model 3 described earlier. MI was performed using predictive mean matching (PMM) with five imputed datasets (m = 5) and 20 iterations to ensure algorithm convergence. Diagnostics, including density plots and trace plots, were performed to validate the stability and convergence of imputed values. The results from the imputed datasets were combined using Rubin’s rules to account for within- and between-imputation variability.

Table 1 summarizes the baseline characteristics of the study population across tertiles of AGP, adjusted for survey weights. Among the 1,903 participants, gallstones were identified in 185 individuals, corresponding to a sample-weighted prevalence of 9.9%. The weighted tertile ranges for AGP were T1 [0.261–0.674], T2 [0.674–0.880], and T3 [0.880–2.020]. Significant differences were observed among the weighted AGP tertile groups in terms of education level, ratio of family income to poverty, smoking history, obesity, total cholesterol, total fiber intake, hs-CRP, history of diabetes, history of hypertension, and prevalence of gallstones (p < 0.05).

The association between AGP and the risk of gallstones was examined using three different models, adjusting for various confounders. The results were presented in Table 2.

After adjusting for all covariates (Model 3), AGP, treated as a continuous variable, was still positively associated with gallstone risk (OR: 3.07; 95% CI: 1.16–8.11; p = 0.036). Participants in the highest AGP tertile (T3) remained significantly associated with gallstone risk, with an OR of 1.87 (95% CI: 1.11–3.14; p = 0.030), compared to those in the lowest tertile (T1).

Smoothed curve fitting by the GAM displayed positive relationship between AGP and gallstone risk (Figure 2).

Subgroup analysis revealed a positive association between higher levels of AGP and the risk of gallstone in several clinical categories (Table 3). Significant associations were observed in older adults aged 40–59 years, Non-Hispanic White individuals. Additionally, those with college or above, ratio of family income to poverty >3.5, no hypertension, no diabetes, alcohol consumers and smokers showed a significant positive correlation between higher AGP and gallstone risk (p < 0.05).

Potential subgroup interactions were assessed using the likelihood ratio test. Among all subgroups, only the ratio of family income to poverty showed a significant interaction (p for interaction < 0.05). No other significant interactions were observed across the subgroups. Multiple comparisons were not adjusted, as these analyses were exploratory and hypothesis-generating in nature.

A sensitivity analysis was conducted on the multiply imputed data to assess the robustness of the findings. The analysis divided the data into weighted tertiles based on AGP levels: T1 [0.261–0.672], T2 [0.672–0.880], and T3 [0.880–2.760]. The results demonstrated that the estimates derived from the imputed data were consistent with those obtained from the original data (Supplementary Tables 3, 4).

Firstly, utilizing data from NHANES 2017–2020 and 2021–2023, this study benefits from its national representativeness, making the findings generalizable to the broader U.S. adult population. Secondly, the study thoroughly adjusted for key demographic and lifestyle confounders, enhancing the reliability of the observed association between AGP and gallstone risk. Additionally, the study employed robust statistical methods, including weighted logistic regression and MI to validate the findings. Finally, this study systematically explores the link between AGP and gallstone risk in women, contributing to the understanding of this association and suggesting new avenues for future research.

Firstly, the cross-sectional design limits causal inference, making it unclear whether elevated AGP levels contribute to gallstone formation or are a consequence of the disease. Secondly, gallstone disease was defined based on self-reported interview data rather than medical imaging or clinical diagnoses, which may introduce recall bias or misclassification. This limitation is inherent in large-scale population surveys like NHANES, and future studies should aim to use objective diagnostic measures, such as abdominal ultrasound, to confirm gallstone presence. Thirdly, while we adjusted for several key demographic and lifestyle confounders, other potential risk factors, such as fasting status and a history of biliary tract disease, were not comprehensively accounted for due to limitations in the dataset. These factors have been reported as important contributors to gallbladder-related conditions (27) and may influence the observed association. Finally, the study’s focus on American women aged 20–49 restricts generalizability to men and other age groups. Expanding the study population in future research would provide a more comprehensive understanding of AGP’s association with gallstone disease. Despite these limitations, identifying interventions to reduce AGP levels could have significant clinical implications, potentially aiding in gallstone prevention and management.

Future longitudinal studies are needed to clarify causality and investigate AGP’s biological role in gallstone formation. Specifically, prospective cohort studies incorporating objective outcome measures, such as imaging techniques (e.g., abdominal ultrasound), are essential to confirm the presence of gallstones. The inclusion of ultrasound data would address potential misclassification bias due to self-reported gallstone diagnoses. Furthermore, future studies should incorporate a broader panel of inflammatory and metabolic markers, such as interleukins, tumor necrosis factor-alpha (TNF-α), adiponectin, and lipid profiles, to disentangle the causal pathways underlying the AGP-gallstone association. These markers would provide a more comprehensive understanding of the interplay between systemic inflammation, bile metabolism, and gallstone pathogenesis. Finally, mechanistic studies using animal models or human bile samples could elucidate AGP’s specific role in bile supersaturation, mucin secretion, and cholesterol crystallization. Identifying interventions to reduce AGP levels could have significant clinical implications, particularly for individuals at high risk for gallstone formation. Such interventions might include lifestyle modifications, anti-inflammatory therapies, or metabolic-targeted treatments, which warrant further exploration.