This was a cross-sectional study using data from the four consecutive cycles of NHANES from 2011 to 2018, which was conducted in the US by the National Center for Health Statistics of the Centers for Disease Control and Prevention. NHANES is a cross-sectional survey program conducted every 2 years, aimed at including individuals who are representative of the general, noninstitutionalized population of all age groups. A stratified, multistage, clustered probability sampling design is applied during the survey. The survey consists of a structured interview conducted in the home, followed by a standardized health examination that includes a physical examination and laboratory tests. Full methodology of the data collection process is available elsewhere (12). Initially, we included participants enrolled in the four consecutive study cycles from 2011 to 2018 (n = 39,157). Participants whose age < 20 years (n = 16,540), had missing values on components of metabolic syndrome (n = 13,930), did not undergone DXA scan examination (n = 4,273), were underweight (n = 9), had missing values on kidney stones status (n = 7), or did not fast for 8–24 h before blood sample collection (n = 147) were excluded from the study. Finally, 4,287 participants were included in the analyses (Figure 1).

The original survey was approved by the Centers for Disease Control and Prevention Research Ethics Review Board and written informed consent was obtained from all adult participants. The present analyses were approved by the Institutional Review Board of Peking university People’s Hospital (approval ID: 2020PHB 125–01).

Data on demographic characteristics, educational level, smoking status, alcohol consumption, physical activities, and dietary intake were collected using standardized questionnaires. Height, weight, waist circumference, and systolic and diastolic blood pressure were measured using calibrated instruments with standard protocols during the mobile examination center visit. Fasting venous blood samples were collected in a subsample of the total participants in the NHANES. Laboratory methods of measuring fasting blood glucose, insulin, total cholesterol, triglycerides, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, serum creatinine, uric acid, blood urea nitrogen, alkaline phosphatase, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and albumin are reported in detail elsewhere (13).

The whole-body DXA scans were conducted in the four study cycles using a Hologic QDR 4500A fan-beam densitometer (Hologic, Inc., Bedford, MA, USA) following the manufacturer’s protocol (14). The principle of whole-body DXA scan relies on the different X-ray absorptions of tissues with different densities (e.g., bone vs. soft tissue) to determine the type and characteristics of the tissues. The accuracy of using DXA to measure body fat has been verified as good or very good by previous studies (15, 16). Each DXA scan was reviewed and analyzed by the Department of Radiology of University of California, San Francisco, using standard radiologic techniques and study-specific protocols developed for the NHANES. Hologic Discovery software version 12.1 was used to analyze DXA examinations and provide body composition data. %BF was determined as a ratio of fat mass over total body mass (including bone mineral mass). In the present study, %BF were extracted from the DXA scan data for each participant to conduct the analyses.

Participants were classified as meeting criteria for normal weight, overweight, or obesity according to %BF based on the criteria recommended by Bray (Supplementary Table S1) (17). Metabolic health was defined based on the 2009 harmonized criteria of metabolic syndrome and homeostasis model assessment of insulin resistance (HOMA-IR) (18, 19). Participants who had none of the following five components of metabolic syndrome were considered metabolically healthy: (1) systolic blood pressure ≥ 130 mmHg or diastolic blood pressure ≥ 85 mmHg or self-reported hypertension or taking antihypertensive medication; (2) fasting blood glucose ≥ 5.6 mmol/L or self-reported diabetes or using antidiabetic drugs; (3) high-density lipoprotein cholesterol <1.03 mmol/L for men and < 1.29 mmol/L for women or using lipid-lowering medication; (4) triglycerides ≥1.7 mmol/L or using lipid-lowering medication; and (5) HOMA-IR ≥ 2.5 (calculated by insulin*glucose/22.5). Those who had ≥1 component were categorized as metabolically unhealthy. Finally, participants were categorized into six phenotypes: MHN, metabolically healthy overweight (MHOW), MHO, metabolically unhealthy normal weight (MUN), metabolically unhealthy overweight (MUOW), and MUO. To comprehensively explore the association of obesity with kidney stones, we further categorized participants into different groups using BMI and waist circumference (WC) criteria. BMI was calculated as body weight in kilograms divided by the square of height in meters. For BMI criteria, normal weight (BMI 18.5–24.9 kg/m²), overweight (BMI 25.0–29.9 kg/m²), and obese (BMI ≥ 30.0 kg/m²) were categorized according to the cut-offs suggested by the World Health Organization (20). For WC criteria, participants were also categorized into three WC groups corresponding to the normal weight, overweight, and obese groups in BMI category. According to the recommendation of the World Health Organization, WC ≥ 80 cm for females and WC ≥ 94 cm for males (WC ≥ 90 for Asian American males) were the cut-off points for central obesity (18); moreover, the median values of WC among females (75.3 cm) and males (85.8 cm for others or 83 cm for Asian American males) without central obesity in this study were used as the cut-offs between normal and medium WC.

Questions on kidney conditions, including kidney stones status, were asked in the home, by trained interviewers, using the Computer-Assisted Personal Interview (CAPI) system. Participants reported their status of kidney stones by responding to the question: “Have you ever had kidney stones?” Participants who answered “Yes” were defined as having kidney stones (21).

Sex was categorized as male or female. Age was analyzed as continuous variable. Race and ethnicity were categorized as non-Hispanic white, non-Hispanic black, Hispanic, Asian, and others. Educational level was categorized as less than 12 grade, high school, college, or above. Smoking status was categorized into never, former, and current smokers. Alcohol consumption was categorized into not excess, and excess (≥ 30 g/d for male, ≥ 20 g/d for female). Physical activity was classified as inactive (no moderate-to-vigorous physical activity), and active (any moderate-to-vigorous physical activity during a typical week) groups. Daily water intake was categorized into three groups according to the 33.33 and 66.67% percentiles. Estimated glomerular filtration rate (eGFR) was calculated according to the Chronic Kidney Disease Epidemiology Collaboration equation (22), and chronic kidney disease (CKD) stage 3–5 was defined as an estimated glomerular filtration rate < 60 mL/min/1.73 m² (23). Hyperuricemia was defined as uric acid ≥ 420 μmol/L. Serum analytes, including lipid profile, serum creatine, uric acid, etc., were analyzed as continuous variables. Dietary factors, including protein, carbohydrate, fiber, and total fat were analyzed as continuous variables.

Given the data collected in the NHANES were obtained through a stratified, multistage probability sample design, we used survey analyses procedures to account for sample weights, stratification, and clustering in our study. Sample size and characteristics were described according to different metabolic health-obesity phenotypes. Data were presented as weighted proportions (SE) for categorical variables and as weighted means ± SE for continuous variables. Logistic regression models were used to calculate the odds ratios (ORs) and 95% confidence intervals (CIs). Model 1 was not adjusted. Model 2 was adjusted for age, sex, and race and ethnicity. Model 3 was adjusted for age, sex, race and ethics, education level, smoking status, alcohol consumption, physical activity, daily water intake, CKD stage 3–5, and hyperuricemia. The covariates adjusted in the model were risk factors of kidney stones that were well-established in previous studies (1, 2, 24–27). Trend tests were conducted by treating the metabolically healthy or unhealthy phenotypes as continuous variables, which were generated by assigning a median %BF/BMI/WC value to each phenotype. Restricted cubic splines with knots at the 5^th, 27.5^th, 50th, 72.5^th, and 95^th percentiles were used to further explore the shape of the dose–response relationship between %BF and kidney stones in metabolically healthy and unhealthy participants, respectively. The multivariable-adjusted model was stratified by demographic and lifestyle characteristics as potential modifiers: age (<40 or ≥40 years), sex (male or female), smoking status (current or none-current), and physical activity (inactive or active). The multiplicative interaction terms between these subgroups and metabolic health-obesity phenotypes were added to the fully adjusted model, and models with and without multiplicative interaction terms were compared using the likelihood ratio test. The following sensitivity analyses were conducted to examine the robustness of our findings: (1) additionally adjusting systolic blood pressure, diastolic blood pressure, glucose, triglycerides, high-density lipoprotein-cholesterol, and HOMA-IR (all were analyzed as continuous variables) in the multivariate logistic regression model and (2) additionally adjusting dietary protein intake, dietary carbohydrate intake, dietary fiber intake, and dietary intake of total fat (all were analyzed as continuous variables) in the multivariate logistic regression model.

All statistical analyses were performed using R version 3.6.2. The “survey” package in R was used to account for the survey design analyses. The “rms” package in R was used to conduct logistic regression analyses. The “rms” and “ggplot2” packages were used to perform restricted cubic spline analyses. All p-values were two-sided, and statistical significance was defined as p < 0.05.

The characteristics of the study participants across metabolic health-obesity phenotypes defined by %BF are presented in Table 1. Among the 4,287 study participants, the mean (SE) age was 39.16 (0.27) years, and 50.53% were male. Metabolically healthy phenotypes accounted for 26.45% of the total study population. Among metabolically healthy participants, 277 (24.43%) participants were defined as having the MHO phenotype. Compared with the MUO phenotype, individuals with MHO had a favorable metabolic profile. On the contrary, compared with individuals with the MHN phenotype, MHO individuals were more likely to have fewer healthy metabolic parameters, CKD stage 3–5, and hyperuricemia.

A total of 358 participants had kidney stones [weighted prevalence (SE): 8.61% (0.56%)]. The weighted prevalence (SE) of kidney stones in MHN, MHOW, and MHO groups was 3.13% (1.10%), 4.97% (1.36%), and 8.55% (2.09%), while the figures were relatively higher in metabolically unhealthy groups, with the weighted prevalence (SE) increased to 6.81% (1.47%), 9.55% (1.14%), and 10.98% (0.82%) in MUN, MUOW, and MUO groups (Table 1).

The association between kidney stones and metabolic health-obesity phenotypes (defined by %BF) is presented in Table 2. After adjusting for age, sex, race, education level, smoking status, alcohol consumption, physical activity, daily water intake, CKD stage 3–5, and hyperuricemia, individuals with MHO (OR: 2.90, 95% CI: 1.18, 7.08) had a significantly higher risk of kidney stones than those with MHN. The corresponding OR (95% CI) for MHOW was 1.62 (0.56–4.05). Individuals with MUO had the highest OR (95% CI) among all phenotypes (OR: 3.82, 95% CI: 1.79, 8.14). In metabolically healthy participants, a 5% increment in %BF was associated with a significantly higher risk of kidney stones (OR: 1.60, 95% CI: 1.20, 2.14), while the corresponding OR (95%CI) for metabolically unhealthy participants was 1.17 (1.00–1.37). Significant trend associations were observed for %BF and kidney stones in metabolically healthy and unhealthy participants (All P for trend <0.05).

Figure 2 shows the results of restricted cubic spline analyses. A non-linear dose–response relationship between %BF and the kidney stones was observed in metabolically healthy participants (P for non-linearity = 0.046) (Figure 2A). However, the dose–response relationship was approximately linear throughout the range of its levels in metabolically unhealthy participants (P for non-linearity = 0.161) (Figure 2B).

The results were largely unchanged when using the BMI or WC criteria to categorize various obesity phenotypes (Supplementary Table S2). When using BMI to define obesity, the multivariate adjusted ORs (95% CIs) comparing individuals with MHOW, MHO, MUN, MUOW, and MUO to MHN were 2.20 (1.13–4.32), 2.89 (1.01–8.24), 1.72 (0.86–3.44), 2.72 (1.54–4.82), and 3.70 (2.06–6.64), respectively. The corresponding ORs (95% CIs) when using WC categories exhibited a similar pattern.

Subgroup analyses are presented in Table 3. No significant interactions of metabolic health-obesity phenotypes with sex, age, smoking status, and physical activity on the risk of kidney stones were detected (all P for interaction >0.05).

In the sensitivity analyses (Supplementary Table S3), the associations between metabolic health-obesity phenotypes and kidney stones were largely unchanged when additionally adjusting blood pressures, glucose, lipid profiles, and HOMA-IR in the multivariate logistic regression model, or additionally adjusting important dietary factors in the model.