We conducted a secondary scientific analysis of the NHANES database from 2011-2018, The NHANES database is an independent cross-sectional survey conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC) that assesses the health and nutritional status of a nationally representative sample of U.S. civilian noninstitutionalized individuals from the general population. Details of the survey in terms of its planning, conduct, and design are available on the official website (https://www.cdc.gov/nchs/nhanes/about_nhanes.htm). The survey included questionnaires; physical examinations; household interviews that covered demographic, dietary, and health-related questions and examinations; and laboratory tests (21, 22).

A total of 39,156 participants were included in the NHANES for the period between 2011 and 2018. After the exclusion of 10,087 participants who did not have serum selenium data, 12,937 participants younger than 20 years of age, 9,719 participants with a BMI less than 30 kg/m², 278 participants with missing serum selenium data, and 10 participants with selenium concentrations that did not fall within the mean±5 standard deviation (SD) range, the final analysis included 6,125 participants. The screening process for the study participants is illustrated in Figure 1 .

Serum samples were processed, stored, and shipped to the Division of Laboratory Sciences at the National Center for Environmental Health and the CDC for analysis. Serum selenium concentrations were quantified via inductively coupled plasma-mass spectrometry. The lower limit of detection for serum selenium was 0.311 μM for the NHANES 2011-2018. All included populations presented serum selenium values above the detection limit. Detailed instructions for sample collection, processing, and quality assessment are available on the NHANES website:(https://wwwn.cdc.gov/Nchs/Nhanes/2011-2012/PBCD_G.htm; https://wwwn.cdc.gov/Nchs/Nhanes/2013-2014/PBCD_H.htm; https://wwwn.cdc.gov/Nchs/Nhanes/2015-2016/PBCD_I.htm; https://wwwn.cdc.gov/Nchs/Nhanes/2017-2018/PBCD_J.htm).

Demographic data, including age, sex, race (non-Hispanic white, non-Hispanic black, non-Hispanic Asian, Mexican American, and other races), marital status (married and living with partner; widowed, divorced, and separated; never married), education (college and above, high school, less than high school), family size (1-3 people, 4-6 people and more than 7 people) and family income (<$45,000, $45,000-$99,999, and ≥ $1,00,000), were collected from all participants. The following physical examination data were collected: standing height (cm), weight (kg), body mass index (BMI, kg/m²), waist circumference (cm), systolic blood pressure (SBP, mmHg) and diastolic blood pressure (DBP, mmHg). Additionally, laboratory data including blood glucose (mM), glycohemoglobin (HBA1C, %), triglyceride (TG, mM), total cholesterol (TC, mM), high-density lipoprotein cholesterol (HDL-C, mM) and low-density lipoprotein cholesterol (LDL-C, mM), were obtained.

A questionnaire was used to collect data regarding the smoking history of the participants, who were classified into three categories: never, former, and current smokers. Additionally, the participants were queried regarding their alcohol consumption status (never, ever, and current), with responses from current drinkers converted to reflect their weekly alcohol intake. The final exposure categories were as follows: never, former, mild (defined as 1 time/week), moderate (1-3 times/week), and vigorous (> 3 times/week). The term “physical activity (PA)” encompasses both work and recreation activities. This definition is consistent with that presented in a previous report (23). In brief, the level of PA is determined by the PA Questionnaire in the NHANES, which is based on metabolic equivalent (MET) values, the type of activity, the frequency of engagement in the activity weekly, and the duration of the activity. The PA score was calculated via the following formula: PA=MET×weekly frequency×duration of each physical activity. Participants with a PA score of 0 were classified as exhibiting no PA, whereas those with a PA score of ≥ 1 were classified as engaging in some PAs.

Individuals meeting two or fewer of the following diagnostic criteria were classified as having MUO: BMI of 30 kg/m² or higher; triglyceride≥1.69 mmol/L; HDL-C ≤ 1.04 mmol/L for males and ≤ 1.29 mmol/L for females; blood pressure (systolic BP) ≥ 130 mmHg and/or diastolic BP ≥ 85 mmHg; and fasting glucose levels ≥ 5.6 mmol/L. MUO was diagnosed if three or more of the above diagnostic criteria were met (24). Participants were identified as having selected components of metabolic syndrome if they were taking medication to modify blood pressure or lipid or carbohydrate metabolism. This information was obtained from the participants’ questionnaire responses.

Appropriate weighting methods are employed to account for the complex sampling design, thereby ensuring the provision of nationally representative results in accordance with the guidelines set forth by the NHANES (25). The basic characteristics of categorical variables are expressed as counts and percentages (%), and the basic characteristics of continuous variables are described by weighted means and SDs. The differences between continuous variables were analyzed via weighted linear regression models, whereas the differences between categorical variables were analyzed via weighted chi-square tests.

The objective of this study was to evaluate the associations between serum selenium concentrations and MUO while controlling for confounding covariates. To ascertain the correlation between the serum selenium concentration and the incidence of MUO, a univariate logistic regression model was used for the preliminary analysis. Model 1 was adjusted for sociodemographic variables, including age, race, sex, marital status, family size, education and family income. Model 2 was adjusted for blood glucose, HBA1C, SBP, DBP, TC, TG, HDL-C, BMI, waist circumference, physical activity, smoke and alcohol use. Model 3 adjusts for both the covariates present in Model 2 and Model 1. These confounders were selected on basis of their associations with the outcomes of interest or a change in effect estimate of more than 10%.

In addition, smooth curves were fitted via a generalized additive model (GAM). A log-likelihood ratio test was conducted to compare the single-line model with the two-piecewise linear regression model to determine the existence of a saturation threshold. The inflection point (K), which connects the line segments, was identified as the maximum likelihood value calculated via the recursive algorithm method. Furthermore, the risk of serum selenium concentrations and MUO occurring on either side of the inflection point was also reassessed via weighted generalized linear modeling.

The statistical analyses were conducted using R (http://www.R-project.org) and Empower Stats (http://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA). A two-tailed P value less than 0.05 was considered to indicate statistical significance.

The NHANES study protocols used were approved by the Research Ethics Review Board of the NCHS. The methods were conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. Written informed consent was obtained from all study participants.

The final cohort consisted of 6,125 participants, selected on the basis of the pre-established inclusion and exclusion criteria, from a total of 39,156 participants in NHANES from 2011-2018. The participants were classified into two groups on the basis of their metabolic health status. The characteristics of the study population are detailed in Supplementary Table 1 . The weighted prevalence of MUO was 48.35%, with 52.24% in males and 47.76% in females. The present study indicated that participants susceptible to MUO were more likely to be male and non-Hispanic White. Additionally, these participants may have been living in households with small populations, potentially due to circumferences such as widowed, divorced, or separated. They had a history of smoking, lower alcohol consumption, and lower levels of physical activity. Furthermore, the MUO group presented elevated waist circumference, blood glucose, HBAIC, TC, TG, HDL-C, LDL-C and blood pressure, which were significantly greater than those of the MHO group (all P<0.05).

The present study demonstrated the observed discrepancies in the serum selenium concentrations observed among the participants in the MHO and MUO groups. Supplementary Table 1 shows that the weighted mean serum selenium concentration in participants with MUO was 2.48 µM (approximately 195 µg/L), which was significantly higher than that observed in the MHO group (P=0.006). The weighted prevalence of MUO was subsequently examined on the basis of tertiles of serum selenium concentrations (T1: < 2.30 µM; T2: 2.31 - 2.55 µM; T3: 2.56 - 4.09 µM). As illustrated in Table 1 , there was a positive correlation between the weighted prevalence of MUO and increasing selenium concentrations. Specifically, the weighted prevalence was 44.39% in the T1 group and increased to 51.38% in the T3 group (P=0.0094). Notably, participants in the T3 group presented an elevated risk of hyperlipidemia (P=0.001) and DM (P=0.023) than did those in the MHO group.

Four weighted multiple regression models were used to assess the correlation between the serum selenium concentration and the likelihood of MHO progressing to MUO. The results revealed a statistically significant correlation between the serum selenium concentration and the risk of MHO developing into MUO in all the models, including the crude model and the models adjusted for differences in covariates ( Table 2 ). The fully adjusted Model 3 revealed that the weighted odds ratio (OR) for MUO increased by 44% for each unit increase in serum selenium (OR=1.44, 95% CI=1.09 - 1.91; P=0.018). Furthermore, the associations between serum selenium concentrations and metabolic abnormalities in both sexes were examined. These results indicate that the influence of selenium on metabolic functions in individuals with obesity is not contingent on sex.

A log-likelihood ratio test was conducted to compare the nonsegment model to the segmented regression model, which was used to determine the threshold. The results of the log-likelihood ratio test indicate that there is a segmental relationship between the serum selenium concentration and MUO. The recursive algorithm was used to calculate the inflection point (K), which was determined to be 2.81 µM (approximately 222 µg/L) ( Figure 2A ).

Further examination was conducted to investigate the association between serum selenium concentrations and the risk of MHO developing into MUO by sex. In males, the effect size was 1.64, with a 95% CI of 1.09-2.48 and a P value of 0.018 on the left side of the K value (K=2.82). In females, the log-likelihood ratio of the two-segment linear model was 0.182, suggesting that the standard linear model provided a more accurate representation of the relationship between serum selenium and metabolic abnormalities ( Table 3 ). As illustrated in Figure 2B , the correlation between serum selenium and MUO displayed a U-shaped pattern in males and a linear correlation in females.

Metabolic abnormalities such as hypertension, hyperlipidemia, DM, stroke, and CHD are associated with dysfunctional metabolic homeostasis. The present study revealed a significantly higher prevalence of hypertension, DM, stroke, and CHD in the MUO group compared to the MHO group. In contrast, no significant difference was observed in the prevalence of hyperlipidemia between the two groups ( Supplementary Table 2 ). We subsequently examined the correlation between the selenium concentration interval and metabolism-related abnormalities on the basis of the saturation threshold. After adjusting for all potential confounding covariates, a statistically significant increased risk of developing hypertension was observed among participants with serum selenium concentrations less than 2.82 µM in males. Table 4 illustrates the impact of different serum selenium threshold intervals on blood pressure in obese males. When the selenium concentration was less than 2.82 μM, the weighted β-value was -1.69 (95% CI: -5.51 - 2.13) for SBP and 3.34 (95% CI: 0.21 - 6.44) for DBP. The corresponding P-values were 0.389 and 0.038, respectively. During the concentration interval, a positive correlation was observed between selenium and DBP, with an increase of 3.34 mmHg in DBP for every 1 μM increase in serum selenium ( Table 5 ). Conversely, when the selenium concentration exceeded 2.82 μM identified, which was negatively associated with SBP, a weighted β-value of -10.70 (95% CI: -17.83 - -3.57, P=0.005) was observed, along with a weighted β-value of -0.30 (95% CI: -6.25 - 5.66), with a P value of 0.923 for DBP. The results indicated a statistically significant inverse association between serum selenium concentrations and SBP, with a 10.70 mmHg reduction observed for each unit increase in serum selenium.

In females, a linear association was observed between the serum selenium concentration and the risk of developing metabolic abnormalities. An elevated risk of hyperlipidemia was observed in participants whose serum selenium concentrations exceeded 2.12 µM. Table 6 illustrates the effects of different selenium threshold intervals on the serum lipid profiles of females. No statistically significant differences were detected in the effects of serum selenium on TG (weighted β=-0.78, 95% CI= -2.12 - -0.55; P=0.254), TC (weighted β=0.50, 95% CI= -0.45 -1.43; P=0.314), HDL-C (weighted β= -0.03, 95% CI= -0.42 - 0.36 P=0.881), or LDL-C (weighted β= 0.39, 95% CI= -0.75 - 1.53; P=0.508) at selenium concentrations below 2.12 μM. However, when the serum selenium concentration exceeded the threshold, we observed that the elevated serum TC (weighted β=0.54, 95% CI=0.32 - 0.76; P=0.000) and TG (weighted β=0.51, 95% CI=0.27 - 0.75; P=0.000) concentrations were elevated. At a serum concentration of 1 μM, the concentrations of TC and TG increased by 0.54 μM and 0.51 μM, respectively. However, no effects on HDL-C (weighted β= -0.00, 95% CI= -0.07 - 0.06; P=0.937) or LDL-C (weighted β=0.18, 95% CI= -0.07 - 0.43; P=0.159) concentrations were detected.