Higher serum 25(OH)D level is associated with decreased risk of impairment of glucose homeostasis: data from Southwest China

We used data from an ongoing population-based prospective study conducted in Southwest China initiated in September 2013, which aimed to investigate the health impact of nutritional and lifestyle factors on the development of several chronic diseases, as described elsewhere [19]. Using a cluster random sampling design stratified by urban and rural locations, a representative sample of civilian aged 25–65 years was recruited from the general population in Chengdu, Southwest China. The participants were invited to the study center for interviews. Generally, each visit included anthropometric measurements, medical examinations, questionnaires and face-to-face interviews by trained investigators about nutrition-related behaviors, lifestyles and social status. However, the following participants were excluded from the study: a) if they had major organ diseases, including heart, liver or kidney disease; b) if they had mental diseases; c) if they were taking hormone-based drugs and other medicines that affect blood glucose and lipids; or d) if they were pregnant or lactating women. The study was approved by the Ethics Committee of Sichuan University, and all participants provided written informed consent.

For the reason that serum 25(OH)D concentration was not measured in 2013 and 2014, eligible data in the present analysis were identified from the baseline survey conducted from March to October, 2015. Participants in survey 2013-2014 did not differ in gender, age, location and educational status from those who were included in our study.

All participants were requested to have an overnight fast of at least 10 h. Peripheral venous blood samples were centrifuged, aliquoted and stored at − 80 °C until measurement. 25(OH)D, calcium²⁺ and insulin were assayed from serum samples, while plasma samples were used to measure the concentrations of FPG. Finally, HbA_1c was quantified from resolved erythrocytes. Serum 25(OH)D was measured using high-performance liquid chromatography (Agilent 1260 HPLC, Shanghai, China) and the intra-assay Coefficient of Variation (CV) was less than 5%. Vitamin D nutritional status was assessed as “deficiency” (< 20 ng/ml), “insufficiency” (20-30 ng/ml) or “sufficiency” (> 30 ng/ml) [20]. Serum calcium²⁺, which is closely related to serum 25(OH)D, was measured by automatic biochemistry analyzer. Serum insulin was assayed with chemiluminescence enzyme immunoassay within 4 h, and the intra-assay CV was 2.4%. Plasma glucose was measured by hexokinase assay on blood collected into fluoridated EDTA tubes within 2 h with an intra-assay CV of 2.5%. HbA_1c was quantified with high-performance liquid chromatography (Bio-Rad D10 automatic analyzer, Shanghai, China) (intra-assay CV: 1.1%) at the clinical laboratory center in Chengdu, which was certified by the National Glycohemoglobin Standardization Program. Finally, the insulin resistance index (HOMA2-IR) was calculated using updated homeostasis model assessment methods (http://​www.​dtu.​ox.​ac.​uk/​homacalculator/​) according to the Wallace formula [21].

Anthropometric measurements were performed by trained medical workers according to the standard procedures [22], with the participants dressed in underwear only, barefoot, and women’s hair uncovered. Waist circumference (WC) was measured without clothes midway between the lower rib margin and iliac crest, to the nearest 0.1 cm, after inhalation and exhalation, using inelasticity tape. Height and weight were measured to the nearest 0.1 kg and 0.1 cm, respectively, with an ultrasonic meter (Weight and Height Instrument DHM-30, China). Weight, height and WC were each averaged based on two measurements. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m²) and was categorized as underweight (BMI < 18.5 kg/m²), normal weight (18.5 kg/m² ≤ BMI < 24 kg/m²), overweight (24 kg/m² ≤ BMI < 28 kg/m²), or obese (BMI ≥ 28 kg/m²) using the standard of Working Group on Obesity in China [23].

Pre-diabetes, based on glycaemic parameters above normal but below diabetes thresholds, is a high risk state for diabetes [24]. In our study, it was defined using the updated classification and diagnosis of diabetes of American Diabetes Association [3] as presentation of one or more of the following results: a) HbA_1c of 5.7-6.4%; b) Fasting blood glucose of 100-125 mg/dl (5.6-6.9 mmol/L).

Information on socio-demographic characteristics, lifestyle, dietary intake, and other potential confounders were collected by interviewer-administered questionnaires in a face-to-face interview.

For the present analysis, we assessed socio-demographic factors potentially associated with serum 25(OH)D level and glucose homeostasis, which included gender, age (years), education level (≤ 6, 6~ 12, or > 12 years of schooling), occupation (mental worker, physical worker, retired or unemployed) [25] and monthly personal income (≤ 1800 Yuan, 1800~ 3200 Yuan, or > 3200 Yuan) [26]. We also collected data on lifestyle, including smoking status (current smokers, ex-smokers and non-smokers), sleeping and stress, and physical activities. To quantify the intensity of physical activities, the energy cost of moderate-to-vigorous physical activity (MVPA) was measured in metabolic equivalents-hours per week (MET-hours/week) [27].

Dietary data were collected on two random days within a 10-day period by trained investigators using a validated 24-h dietary recall [19]. Participants were asked to recall all foods and beverages they consumed and the corresponding timing. Dietary intake data from 24-h dietary recall were converted into energy using the continuously updated in-house nutrient database based on China Food Composition 2009 [28]. And in this analysis, total energy intake for each participant was calculated as individual means of two-day 24-h dietary recall in kcal/day. Alcohol beverage consumption (cups/d) and coffee consumption (cups/wk) were accessed by food frequency questionnaire.

In addition, season of blood drawn (spring: March to May, summer: June to August; autumn: September to November; winter: December to February) was recorded.

All statistical analyses were performed with SAS software (SAS, version 9.3, 2011, SAS Institute Inc., Cary, NC, USA.). A P value < 0.05 was considered statistically significant, except for interaction tests, where P < 0.1 was considered significant. Normality of all continuous variables was examined using normal probability plots and the Kolmogorov-Smirnov test. Given their non-normality, all continuous variables were presented as median (25th percentile, 75th percentile). As the initial analysis indicated no interaction between gender and relations of serum 25(OH)D with FPG, HbA_1c, insulin levels, HOMA2-IR (range of P-value: 0.4–0.9), we pooled the sample in the follow-up analyses.

We cross-classified the study sample into categories of tertiles (T1-T3) of serum 25(OH)D to examine the distribution of baseline parameters. We tested differences in proportions using Student t-tests for normally-distributed continuous variables, the Wilcoxon rank-sum for non-normally distributed continuous variables and the Chi-square test for categorical variables, respectively.

To investigate the associations of serum 25(OH)D with glucose homeostasis, multivariable linear generalized regression models (PROC GLM in SAS) were performed. Serum 25(OH)D was defined as the independent variable in separate models. Glucose homeostasis including fasting insulin, FPG, HbA_1c and HOMA2-IR were dependent variables in separate models. The independent and dependent variables that enter the linear regression models were non-normally distributed continuous variables. To improve the fitting effect of the models, log-transformed values of insulin, FPG, HbA_1c and HOMA2-IR were used in the models.

In the basic models, the correlation analyses between serum 25(OH)D and glucose homeostasis (insulin, FPG, HbA_1c, HOMA2-IR) were carried out first. In a further analysis, the following variables potentially affecting these associations were added: gender, age (years), educational level (≤ 6 years, 6~ 12 years, or > 12 years of theoretical education), monthly personal income (≤ 1800 Yuan, 1800~ 3200 Yuan, or > 3200 Yuan), occupation (mental worker, physical worker, retired or unemployed), smoking status (current smokers, ex-smokers and non-smokers), MVPA (MET-hour/week), total energy intake (kcal/d), alcohol beverage consumption (cups/d), coffee consumption (cups/wk), serum calcium²⁺ (mmol/L), season of blood drawn (spring, summer, autumn and winter) and WC (cm) or BMI (kg/m²). Each variable was initially considered separately, and variables that had their own independent significant effect in the basic models or that substantially modified the association of serum 25(OH)D with each variable of glucose homeostasis were included in the multivariate analyses. Thus, age, gender, monthly personal income, smoking status and season of blood drawn were retained in model A. In a further step, we additionally adjusted for MVPA and energy intake (model B). WC was checked as a potential confounder in model C. The adjusted means were the least-squares means predicted by the model when the other variables were held at their mean values. Then the least-squares means and 95% confidence interval (95%CI) computed by the linear models were back transformed and then presented in the results.

Finally, multivariate logistic regression analyses were used to determine the association of serum 25(OH)D with the odds of pre-diabetes. To enhance comparability, models were constructed in analogy to the multivariable linear regression analyses. To explore interaction of weight status on this association, we divided the participants into two groups according to their weight status (overweight or not) and performed stratified analyses. Estimates are presented as odds ratios (ORs) with 95% CI.