The National Health and Nutrition Examination Survey is an ongoing, health-related survey that assesses the nutritional and health status of the American population. Survey participants were recruited by a stratified multistage probability sampling method to ensure the sample was nationally representative (23).

The original study protocol was available on the website of the ethics review board of the national center for health statistics research (https://www.cdc.gov/nchs/nhanes/irba98.htm), which was further approved by the ethical review committee (protocol # 2005–06; protocol # 2011–17). The current study was based on the existing data retrieved from NHANES, and the details were extracted from the official website (24).

This study used public data retrieved from four cycles of NHANES (2007–2008, 2009–2010, 2011–2012, and 2013–2014). Adult patients aged 20 or older with available data for HOMA-IR and vitamin D were included. The exclusion criteria were the presence of Type 1 diabetes mellitus (T1DM) and T2DM (since in patients with diabetes, HOMA-IR may not be a representative indicator of insulin resistance due to diminished insulin secretion) (25), CKD, and the use of drugs that can influence insulin sensitivity, including antidiabetic drugs, glucose elevating agents, antineoplastics and anti-retroviral agents, adrenal cortical steroids, selective estrogen receptor modulators, parathyroid hormone and analogs, antiandrogens, aromatase inhibitors, calcimimetics, antipsychotics, other metabolic agents, bone resorption inhibitors (bisphosphonates, etc.), and niacin. Although some anti-hypertensive drugs, sex hormones (including contraceptives), and statins can influence insulin sensitivity, the subjects who used those medications were not excluded from the study, because a substantial number of the subjects were using these agents (N = 1,081, N = 269, and N = 561, respectively). Nevertheless, to account for their potential influence on insulin sensitivity, the usage of these drugs was included in covariates in our regression analyses. Participants with any covariates missing were excluded.

Type 2 diabetes mellitus is diagnosed based on plasma glucose levels, including either the fasting plasma glucose value or the 2-h plasma glucose value during a 75 g oral glucose tolerance test or the glycosylated hemoglobin A1c criteria (26). However, either doctor-diagnosed or self-reported diabetes is included for certain. The participants with impaired glucose tolerance or impaired fasting glucose, in case they were not using antidiabetic drugs, were included. CKD was diagnosed based on an increased albumin/creatinine ratio (≥30 mg/g) and a decreased estimated glomerular filtration rate (<60 ml/min/1.73m²) (27). The data on the prescription medications were inquired and collected by trained interviewers.

Plasma and serum samples for fasting plasma glucose, serum insulin, 25(OH)D, bicarbonates, total calcium, phosphorus, and triglycerides were obtained and stored in the Mobile Examination Center until shipped to the Centers for Disease Control and Prevention Environmental Health Laboratory (Atlanta, Georgia). The HOMA-IR model was used to evaluate insulin resistance, calculated using the following formula: fasting serum insulin (μU/L) × fasting plasma glucose (mmol/L)/22.5 (28). Concentrations of 25(OH)D3 and 25(OH)D2 in the serum samples were analyzed using super high-ultra performance liquid chromatography-tandem mass spectrometry. Total 25(OH)D (or vitamin D) was defined as the sum of 25(OH)D3 and 25(OH)D2. In terms of the serum total vitamin D levels, the participants were classified as deficient (<50 nmol/L), suboptimal (50–75 nmol/L), and sufficient (>75 nmol/L), as recommended by the American Endocrine Society (29).

We tested all covariates if they were associated with HOMA-IR or vitamin D levels, and the significantly associated covariates were included in the adjusted linear regression models. The eligible covariates included age, gender, race/ethnicity, education, body mass index (BMI), physical activity level (PAL), the season of examination, current smoking, hypertension, the usage of antihypertensive drugs, sex hormones (30, 31), statins (32, 33), serum bicarbonates (34, 35), triglycerides (36, 37), and calcium and phosphorus levels (38–40). The race/ethnicity was divided into five groups: Mexican Americans, Other Hispanics, Non-Hispanic Whites, Non-Hispanic Blacks, and Other races/ethnicities (including Asians and mixes). Education levels were categorized as < 9th grade, 9th−11th grade (including 12th grade with no diploma), high school graduate/general educational development (GED) or equivalent, college/associate of arts (AA) degree, college graduate or above, refused, and unknown. The season of examinations was classified into November to April and May to October. The current smokers were separated from the former and never smokers. Participants who reported smoking either some days or every day at the time of the interview were considered current smokers. Participants who smoked more than 100 cigarettes during their lifetime but did not smoke currently were former smokers. Body mass index (BMI, kg/m²) was defined as body weight in kilograms divided by squared body height in meters. Physical activity level (PAL) scores were calculated to assess physical activity based on the different levels of activity, including vigorous (2 points) or moderate (1 point) work-related activity, vigorous (2 points) or moderate (1 point) leisure-time physical activity, and walking or bicycling for transportation (1 point). The minimum PAL score was 0, and the maximum PAL score was 5. Hypertension was defined as having systolic blood pressure ≥ 130 mmHg or diastolic blood pressure ≥ 80 mmHg, which were measured on more than or equal to two occasions to acquire an average (41).

Median and interquartile range (IQR) were used to describe a non-normal distribution. The mean and standard deviation (SD) were used to describe a normal distribution. To compare the differences between various vitamin D status categories, the χ² test (for nominal data), the one-way analysis of variance (ANOVA) (for continuous variables with normal distribution), and the Kruskal-Wallis's test (for continuous variables with non-normal distribution) were used. In linear correlation analyses, any continuous variable that was not normally distributed underwent log 10 transformation to ensure its normal distribution (HOMA-IR, triglycerides). Pearson correlation coefficient (r) was used for normally distributed continuous variables, while Spearman correlation coefficient (r_s) was used for non-normally distributed continuous variables or ordered categorical variables. The Point-biserial correlation coefficient (r_pb) was used for dichotomous variables.

The association between total vitamin D and HOMA-IR was evaluated by employing the enter-type linear regression models. Standardized beta was utilized to compare the relative predictive strength of different covariates in the regression models. The variance inflation factor (VIF) was used to assess the multicollinearity of all covariates in the regression model. In linear regression analyses, HOMA-IR and triglycerides underwent log 10 transformation. Stratified regression analyses were used to account for differences between races. Two-tail p < 0.05 were considered statistically significant. All analyses were performed using Empower stats (http://www.empowerstats.net/cn/) and SPSS software Version 21.0.

Following the exclusions, this study included a total of 6,079 participants aged 20 years or older (Figure 1). Baseline characteristics of the selected participants were classified according to varying serum vitamin D status categories as provided in Table 1.

The proportion of vitamin D deficiency did not differ between men and women; however, a higher proportion of vitamin D suboptimal and a lower proportion of vitamin D sufficient was found among men. The highest proportion of vitamin D deficiency was among the Non-Hispanic Blacks (63.25%), followed by Mexican Americans, other races-ethnicities (including Asians and mixes), other Hispanics, and finally, the Non-Hispanic Whites (15.25%). Meanwhile, the highest proportion of vitamin D sufficiency was observed among the non-Hispanic Whites (45.56%), while the lowest was among the non-Hispanic Blacks (11.62%).

As expected, a higher prevalence of vitamin D deficiency was observed in samples collected during the winter period (from November to April). With the exception of those who refused to answer or were left unknown about their education levels, the group with the most vitamin D deficiency belonged to the education groups of 9th−11th grade, whereas the group with the most vitamin D sufficiency was the group of college graduates or above. In the present study population, patients with hypertension and current smokers were more vitamin D deficient. Compared with the vitamin D sufficient subgroup, participants in the deficient subgroup were at the highest HOMA-IR.

Serum vitamin D levels were related to age, sex, race, educational level, BMI, PAL score, the season of examination, smoking status, usage of antihypertensive drugs, sex hormones, and statins, serum fasting insulin, triglycerides, bicarbonates, and calcium levels, as well as HOMA-IR. There was no association found between serum vitamin D levels and plasma fasting glucose, serum phosphorus, and hypertension (Table 1).

Table 2 shows that all covariates, except age and serum calcium, were linearly related to HOMA-IR. Regarding the stratified racial analysis, insulin resistance was found to be different among various races: Mexican Americans and other Hispanics were more prone to higher HOMA-IR, while the Non-Hispanic Whites and other races/ethnicities (including Asiatic) were less susceptible, while the Non-Hispanic Blacks were in the middle (Table 2).

Linear regression analysis (Table 3) revealed that HOMA-IR was inversely associated with vitamin D levels prior to the adjustments for covariates (Model 1). The unadjusted model described only a small variance in HOMA-IR by using only vitamin D levels (2.8%). Following the adjustments of covariates that included age, gender, specific race/ethnicity, education, BMI, physical activity, the season of examination, current smoking, hypertension, the usage of antihypertensive drugs, sex hormones, and statins, as well as serum bicarbonates, calcium, and phosphorus levels (Model 2), this inverse association between vitamin D and HOMA-IR remained significant, although it decreased. This model explained a much higher variance in HOMA-IR (36.1%). The association of vitamin D with HOMA-IR in the fully adjusted model with added log-transformed triglycerides was even more significant since the standardized regression coefficient for vitamin D increased (Model 3). This model explained the highest percentage of variance in HOMA-IR (41.3%).

In stratified regression analyses (Table 4), only in the Non-Hispanic Blacks, there was no significant inverse association between vitamin D and insulin resistance in the fully adjusted model with serum triglycerides included (Model 3).

In the general population or ethnic subgroups, BMI contributed the most to HOMA-IR, as shown in Supplementary Tables S1, S2. The influence of vitamin D on HOMA-IR was the strongest among other races/ethnicities (including Asiatic) compared to Mexicans and other Hispanics and the Non-Hispanic Whites, while the association was not observed in the Non-Hispanic Blacks (Supplementary Table S2).