The VDI is associated with several components of MetS [10]. The inverse correlation between Vitamin D concentration and the development of MetS could be explained by the dependence of insulin production in pancreatic β cells on Vitamin D and its receptor interaction [11], obesity resulting in more sequestration of vitamin D in adipose tissue, and greater volumetric dilution of serum 25(OH)D level, less sun exposure, poor dietary habits, and varied genetic expression of enzymes involving vitamin D metabolism [17, 42]. Although the mean 25(OH)D concentration was 29.59 ± 10.80 ng/mL, 54.9% of the participants with VDI and 43.9% of them with MetS in the present cohort, there was no statistical difference in the incidences of MetS and its related components between the VDI and the sufficiency group. We also found that the serum 25 (OH)D level of participants with MetS is not statistically different from that of those with no MetS (29.10 ± 10.52 vs. 29.97 ± 11.01, P = 0.204). Notably, the multivariate analysis revealed an association between the risk of VDI and several sociodemographic variables, including sex, age, educational attainment, and alcohol consumption. Therefore, our results reinforce the essential issue that population characteristics, study designs, and targeted clinical scenarios could partially contribute to the discrepancies among studies and determine the association between vitamin D status and MetS.
The present study identified factors associated with the development of VDI, which were classified into risk factors (female sex, age < 65 years, college education and above, and high serum LDL levels) and protective factors (alcohol use and better emotional role of HRQoL). These factors are consistent with those reported in previous publications. Risk factors such as female sex, young people, and higher education attainment were reported in one US adult -based report and two Taiwanese adult-based studies, and their associations with VDI were comprehensively discussed from several aspects including lifestyle, occupation, physical activity, residency location, sun exposure, and metabolic difference [7, 30, 43]. Furthermore, the correlation between high serum LDL levels and VDI was elucidated by the regulation of lipid metabolism, including the inhibition of lipogenesis and increased lipolysis via vitamin D-mediated calcium spillage into fat cells and pancreatic islet function in the previous studies [44–47]. Although a meta-analysis reported a controversial relationship between vitamin D status and alcohol use, recent research has shown a positive association between serum vitamin D levels and alcohol use [48] suggesting a protective role in reducing the development of VDI. Another possible reason for the protective function of alcohol consumption could be that Taiwanese people often eat protein-rich food and wine pairing when drinking alcohol and gain resources for vitamin D supplementation. Nonetheless, the possible effect of the interaction between these factors, the HBM, and HRQoL on VDI remains unknown and deserves further discussion.
We analyzed health belief behavior classified by factors identified for risk of VDI accordingly, and found the following: women had lower scores of the perceived barrier but higher scores of self-efficacy than men (Supplementary Table 1); participants with age < 65 years or with college and above education attainment had higher scores of perceived susceptivity, perceived severity, perceived benefit, and cu-to-action than those with age ≥ 65 years or with high school education (supplementary Table 2 and Table 3); participants with higher LDL level presented lower score of self-efficacy than those with lower LDL level (supplementary Table 4); participants exposed alcohol drinking had higher score of perceived barrier but lower score of self-efficacy than those with no alcohol (supplementary Table 5), whereas participants with higher RE scores had lower score of the perceived barrier but a higher score of self-efficacy than those with lower RE scores (Supplementary Table 6). Taken together, these results indicate that even though participants had these risk factors, they still had higher scores on the health belief behavior assessment, strengthening the weak-to-none association between health belief behavior and the risk of VDI in our multivariate analysis. Regarding the association between factors related to VDI and HRQoL, we found that women had worse quality scores in both physical and mental health than men (supplementary Table 1), whereas participants aged < 65 years or with college and above education attainment had better quality scores in physical health perceive but worse in mental health than those with age ≥ 65 years or with high school education (supplementary Table 2 and Table 3); participants with higher LDL level presented better physical functioning than those with lower LDL level (supplementary Table 4); participants exposed alcohol drinking had similar HRQoL to those with no alcohol exposure (supplementary Table 5), whereas participants with higher RE scores had better HRQoL including all domains of physical health, mental well-being, PCS and MCS than those with lower RE scores (Supplementary Table 6). Hence, except for female sex and alcohol exposure, participants with factors such as age < 65 years, college or higher education attainment, and higher LDL and RE scores had better HRQoL in terms of either physical or mental health. Finally, there was an intimate association between HBM and HRQoL (Supplementary Table 7). This link could partially explain why certain items of the HBM or HRQoL that were significantly detected in the univariate analysis were no longer statistically meaningful in the multivariate analysis.
Table 3
Univariate and multivariate analyses to assess variables associated with the physical component summary (PCS) in 558 participants with vitamin D insufficiency
| Univariate | Multivariate |
Variable | P* | Coefficient | 95% CI | P* |
Age (ref: <65years) | 0.004 | -1.991 | -3.614 to -0.367 | 0.016 |
Sex (ref: female) | 0.037 | 1.493 | -0.209 to 3.195 | 0.085 |
Educational attainment (ref: College) | 0.004 | 1.029 | -0.434 to 2.492 | 0.168 |
Abnormal waist (ref: no) | 0.001 | -0.122 | -1.675 to 1.431 | 0.877 |
Diet (ref: non-vegetarian) | 0.439 | | | |
Smoking (ref: no) | 0.002 | 2.169 | 0.129 to 4.209 | 0.037 |
Drinking (ref: no) | 0.912 | | | |
Betel nut usage (ref: no) | 0.701 | | | |
BMI | 0.001 | -0.129 | -0.282 to 0.024 | 0.098 |
Body weight | 0.29 | | | |
Systolic blood pressure | 0.402 | | | |
Diastolic blood pressure | 0.009 | 0.103 | 0.046 to 0.159 | < 0.001 |
Glucose_AC | 0.001 | -0.005 | -0.044 to 0.034 | 0.804 |
HbAlc (mmol/mol) | < 0.001 | -0.6 | -1.712 to 0.513 | 0.29 |
Insulin, mU/L | 0.001 | -0.005 | -0.25 to 0.24 | 0.968 |
HOMA_IR | < 0.001 | -0.177 | -0.876 to 0.523 | 0.62 |
Total cholesterol (mg/dl) | 0.916 | | | |
LDL-C (mg/dl) | 0.411 | | | |
HDL-C (mg/dl) | 0.281 | | | |
Triglyceride (mg/dl) | 0.397 | | | |
Health belief | | | | |
Perceived susceptibility | 0.053 | | | |
Perceived severity | 0.407 | | | |
Perceived benefits | 0.133 | | | |
Perceived barriers | 0.025 | -0.053 | -0.095 to -0.012 | 0.012 |
Self-efficacy | 0.028 | 0.03 | -0.038 to 0.099 | 0.385 |
Cue to action | 0.047 | 0.063 | -0.072 to 0.198 | 0.358 |
*P < 0.05, statistically significant. |
The other salient feature of the present study was the identification of factors that were associated with HRQoL (PCS and MCS) in participants with VDI, including age, smoking, blood pressure, perceived barriers of the HBM for PCS, educational attainment, perceived barriers, and self-efficacy of the HBM for MCS (Tables 3 and 4). Hence, public health promotion, such as smoking cessation, blood pressure monitoring, and health knowledge improvement via adopting health belief behaviors, is a practical strategy for improving the quality of life of people with hypovitaminosis D.
Table 4
Univariate and multivariate analyses to assess variables associated with the mental component summary (MCS) among the 558 participants with vitamin D insufficiency
| Univariate | Multivariate |
Variable | P* | Coefficient | 95% CI | P* |
Age (ref: <65years) | 0.097 | | | |
Sex (ref: female) | 0.107 | | | |
Educational attainment (ref: College) | 0.004 | -2.006 | -3.951 to -0.06 | 0.043 |
Abnormal waist (ref: no) | 0.246 | | | |
Diet (ref: non-vegetarian) | 0.911 | | | |
Smoking (ref: no) | 0.074 | | | |
Drinking (ref: no) | 0.075 | | | |
Betel nut usage (ref: no) | 0.230 | | | |
BMI | 0.237 | | | |
Body weight | 0.942 | | | |
Systolic blood pressure | 0.151 | | | |
Diastolic blood pressure | 0.172 | | | |
Glucose_AC | 0.05 | | | |
HbAlc (mmol/mol) | 0.104 | | | |
Insulin, mU/L | 0.578 | | | |
HOMA_IR | 0.071 | | | |
Total cholesterol (mg/dl) | 0.531 | | | |
LDL-C (mg/dl) | 0.486 | | | |
HDL-C (mg/dl) | 0.194 | | | |
Triglyceride (mg/dl) | 0.684 | | | |
Health belief | | | | |
Perceived susceptibility | 0.382 | | | |
Perceived severity | 0.631 | | | |
Perceived benefits | 0.512 | | | |
Perceived barriers | < 0.001 | -0.124 | -0.18 to -0.068 | < 0.001 |
Self-efficacy | 0.002 | 0.081 | 0.001 to 0.16 | 0.048 |
Cue to action | 0.819 | | | |
*P < 0.05, statistically significant |
The present study had some limitations. First, selection bias was inevitable in this cross-sectional study. Our results were obtained from a community population who actively and voluntarily sought medical counseling and complied with health guidance; thus, their external validity in the general population and different social conditions, such as staying at nursing homes, living alone with no assistance, and unwillingness to go out, requires determination. Second, causal inferences underlying the observed associations could not be drawn in a cross-sectional study, and reverse causation may be present in our results. As there is a lack of disease-specific measurements for vitamin D insufficiency, our study applied generic measurements for HBM and HRQoL assessments and offered the only viable option at present. Different measurements used to assess the HBM and HRQoL may have varied outcomes. Third, we were unable to include medication, physical activity, and comorbidity status as covariables because these variables influence vitamin D metabolism [48–50].