Association between use of vitamin and mineral supplement and non-alcoholic fatty liver disease in hypertensive adults

Non-alcoholic fatty liver disease (NAFLD) is the most common hepatic metabolic disorder in hypertensive adults. Impaired metabolism of micronutrients may increase NAFLD risk by exacerbating oxidative stress, insulin resistance, and inflammation among hypertensive adults. In this first cross-sectional analysis of 7,376 hypertensive adults with 2,015 NAFLD cases in the Korea National Health and Nutrition Examination Survey, vitamin and mineral supplements (VMS) use was identified via questionnaire. NAFLD was defined by a hepatic steatosis index > 36. Multivariable-adjusted odds ratios (MVOR) and 95% confidence intervals (CIs) were calculated using logistic regression models. In our study, 18.6% were current users of VMS; of these, 76.7% used multi-vitamin/mineral supplements. Current VMS users had significantly lower odds of NAFLD, compared with non-users (MVOR [95% CI]: 0.73 [0.58–0.92]). The inverse association became attenuated and non-significant among those consuming VMS at higher frequency (≥ 2 times/day), for longer duration (> 16 months), and taking ≥ 2 VMS products. The inverse association with current use of VMS was only evident in those aged < 56 years (MVOR [95% CI]: 0.54 [0.40–0.72]) and men (MVOR [95% CI]: 0.56 [0.40–0.80])(Pinteraction ≤ 0.04). Our results suggest that VMS use may lower NAFLD risk, particularly among younger or male hypertensive adults, if taken in moderation.

NAFLD ascertainment.NAFLD was defined using the hepatic steatosis index (HSI), one of the mostwidely used, validated, noninvasive method to diagnose NAFLD, developed in Korean adults 32 .The constituents of HSI include the ALT, AST, BMI, sex, and diabetes status, and the index is calculated as follows: 8 × (ALT/ AST ratio) + BMI (+ 2 if female; + 2 if had diabetes) 32 .The presence of NAFLD was defined as having an HSI value > 36 32 .The diagnostic accuracy of HSI in detecting NAFLD is known to be high, particularly in Asian populations [32][33][34][35] , with an area under receiver-operating curve of 0.81 (95% CI:0.82-0.83)and > 90% sensitivity and specificity 32 .Consequently, 2,015 NAFLD cases (27.3% of the study population) were ascertained.Statistical analyses.All analyses applied sampling weight to account for the stratified, multistage, clustered sampling design of the KNHANES 36 and sampling-weighted estimates were reported in all tables, except for the number of study participants.Participants' characteristics according to their current use of VMS were summarized using the SAS SURVEYMEANS procedure for continuous variables and the SAS SURVEYFREQ procedure for categorical variables.
The associations between VMS use (including frequency, duration, and multiple product use) and NAFLD risk were estimated by calculating odds ratios and 95% confidence intervals (CIs) using the SAS SURVEYLO-GISTIC procedure.The multivariable model included potential confounding factors by adding, a priori selected, well-known risk factors for NAFLD 4 : age, sex, obesity status, waist circumference, household income, education, alcohol drinking status, smoking status, regular exercise, status of diabetes, status of hypercholesterolemia, serum triglyceride concentration, a comorbidity measure, hypertension stage, current use of hypertension medication, and intakes of total energy and fruits and vegetables (see Table 2 for the categorization of confounding variables).A missing indicator for missing responses of each covariate (< 4.6%) was created, if applicable.The P-trend for frequency and duration of VMS use was performed by modelling the median of each VMS frequency and duration category as a continuous term.Stratified analyses were conducted to assess if the association between VMS use and NAFLD differed by participants' demographic, lifestyle, and clinical characteristics.Statistical significance for potential effect modification was tested using the Wald test of the product term between VMS use and the stratification factor.
All statistical analyses were conducted using SAS statistical software (version 9.4; SAS Institute Inc., Cary, NC) and were considered to be statistically significant if p value < 0.05.

Results
In our study, 18.6% hypertensive adults were current VMS users.Compared with nonusers, current VMS users were more likely to be female, earn a higher household income, have at least college degree, not currently smoke, drink less alcohol, and consume more fruits and vegetables.They were also less likely to be obese, had fewer comorbidities, and had healthier levels of biochemical marker profiles (Table 1).Among current VMS users, the major types of VMS consumed were M_VMS (76.7%), followed by S_VIT (20.3%) and S_MIN (3.0%) (Table S1).
Among hypertensive adults, current users of VMS had significantly lower odds of NAFLD compared to nonusers; the multivariable-adjusted odds ratio (MVOR) (95% CI) was 0.73 (0.58-0.92) (Table 2).In analyses with specific types of VMS use (Table S2), significant inverse associations was observed with the use of M_VMS or S_MIN, though not with S_VIT, showing MVOR ranging from 0.34 to 0.76.The inverse association among M_VMS users, who were the majority among those who used multiple VMS products, was only evident among those who used one product, but not among those who took additionally other supplements (data not shown).
In sensitivity analyses, additional adjustment for dietary intakes of vitamins and minerals, macronutrient intakes, and eating frequency or excluding severe NAFLD cases with ≥3 BARDs score 37 did not change our results materially (data not shown).When the comprehensive NAFLD score 34 was used to defined NAFLD cases, the results were weakened and nonsignificant (data not shown).
There was no significant effect modification by the other factors including obesity, smoking status, alcohol consumption status, fruit and vegetable intake, stage and types of hypertension, use of hypertension medication, and blood pressure control status (Tables 4 and 5).Also, there was no significant effect modification by carbohydrate, fat, or protein intake, or eating frequency (Table S3).However, age and sex significantly modified the association between current use of VMS and odds of NAFLD (P interaction ≤ 0.04) with inverse association being observed only in those aged <56 years (MVOR [95% CIs]:0.54[0.40-0.72])and in men (MVOR [95% CIs]: 0.56 [0.40-0.80]).

Discussion
In this large nationally representative sample of hypertensive adults in Korea, nearly 20% reported using VMS, and current use of VMS was significantly inversely associated with odds of NAFLD.When the association of VMS use with frequency, duration, and use of multiple products was explored, the inverse association was attenuated and non-significant among those who took VMS at high frequency, for long duration, or taking multiple products.The inverse association of current VMS use did not significantly vary across most of demographics, lifestyles, and clinical characteristics, but was observed evidently only in the younger population aged <56 years or in men, but not their respective counterparts.
The suggested hepatoprotective potential of VMS use against NAFLD risk, observed in our study, can be further explained by the involvement of vitamins and minerals in VMS in the pathogenesis of NAFLD 11 .For example, anti-oxidant vitamins and minerals, such as vitamins A, C, and E, and selenium, may reduce fibrosis in the liver by scavenging free radicals associated with lipid peroxidation, protein damage, and the induction of inflammatory cytokines such as TGF-beta, IL-6 and TNF-alpha.Vitamin D were found to reduce the secretion of fibrogenic growth factors, decrease proinflammatory cytokines, and exert an insulin-sensitizing effect favoring glucose uptake in the liver or release of adiponectin.Vitamin B3 acts as a precursor in lipid metabolisms.Folate may lower NAFLD risk via lowering homocysteine concentrations, which promote lipid accumulation.Vitamins B 2 , B 6 , B 9 and B 12 participate in one-carbon metabolisms that helps maintain DNA integrity and epigenetic signature.Deficiency of zinc and copper is found to augment oxidative stress.NAFLD is also considered as a hepatic component of metabolic syndrome having obesity and diabetes as strong risk factors 4 .VMS might reduce NAFLD risk by altering these metabolic conditions.Table 3. Age-adjusted and multivariable-adjusted a,b odds ratio (OR) and 95% confidence intervals (95% CIs) of NAFLD according to the use of VMS defined by frequency, duration, and use of multiple products among hypertensive adults.CI confidence intervals, NAFLD non-alcoholic fatty liver disease, OR odds ratio, VMS vitamin and mineral supplements.a All results presented in this table are sampling-weighted estimates except the number of study participants.b Multivariable model was adjusted for age (quartiles), sex (male, female), household income level (quartiles, missing), education level (elementary school, middle school, high school, college or higher), smoking status (never, past, current), alcohol drinking status (never, past, ≤ 1 time/month, 2-≤ 4 times/month, ≥ 2 times/week), regular exercise (no, yes), body mass index (< 18.5 kg/m 2 , 18.5-< 23 kg/m 2 , 23-< 25 kg/m 2 , ≥ 25 kg/m 2 ), waist circumference (cm, continuous), diabetes mellitus (no, yes), hypercholesterolemia (no, yes), number of comorbidities (0, 1, 2, 3, ≥ 4), serum triglyceride concentration (mg/ dL, continuous), hypertension stage (stage1, ≥ stage2, other), current use of hypertension medications (no, yes), and intakes of total energy (quartiles) and fruits and vegetables (quartiles).c P-trend was tested from model including the median value of specific VMS usage categories as a continuous term and using the Wald test of it.d Duration of VMS use was categorized according to the quartiles of duration among current users of VMS.www.nature.com/scientificreports/However, when we further explored associations with frequencies, durations, and multiple use of VMS to understand the optimal use of VMS, the inverse association observed with VMS use was, interestingly, attenuated and became no longer significant at high frequency (≥ 2 times/day) or long-term uses (> 16 months).Similarly, there was a significant inverse association with a single use of VMS product, but the association became null when ≥ 2 VMS products were taken.The explanation for these findings remains unclear.However, along with the widespread use of VMS, the possible risk of adverse events due to overdose and interactions with prescribed medications raises the safety concerns especially among those using VMS prevalently 47 or with clinical disease conditions 48 .
Our results may caution the overuses of VMS against NAFLD risk among hypertensive adults.Indeed, although no studies have examined specific usages of VMS with NAFLD risk, prior studies of metabolic diseases, which tend to coexist with NAFLD 6 , have shown the harmful effects of excessive use of VMS.For instance, a previous randomized controlled trial (RCT) of 31,210 Swedish men revealed a 21-59% increased risk of agerelated cataract with high doses of vitamin C (1000 mg) and E (100 mg) supplementation, which was not seen with a multivitamin supplementation containing low doses of vitamins and minerals 49 .In that study, the positive association with vitamin C supplementation was also even stronger among long-term users 49 .Similarly, a RCT of selenium supplementation (200 µg/d) showed increased risk of diabetes among those in the highest tertile (> 121.6 ng/ml), but not in the lower tertiles, of plasma selenium concentration (P-interaction = 0.038) 50 .This result was further confirmed in subsequent studies 51,52 .A large cross-sectional study of 200,000 US military service members reported 2.3 to 17.9 fold risk of adverse events among those consuming ≥ 5 dietary supplements than those consuming 1-2 dietary supplements 47 .The US Preventive Services Task Force also concluded  www.nature.com/scientificreports/insufficient evidence to use ≥ 2 VMS products for the prevention of cardiovascular disease 53 .However, sample size in the highest categories of specific use of VMS in our study were small; individuals tended to be older and use more hypertensive medications.Our results could be due to chance or unmeasured vulnerabilities of those groups.
The inverse association between VMS use and NAFLD was only observed in adults aged < 56 year.Advancing aging is associated with diminishing nutrient absorption and progressive loss of physiological integrity.In particular, the aging was found to be associated with annual decline of hepatic clearance by 0.80% from 40 years of age, which suggests the regenerative capacity of the liver to metabolize VMS 54 .Our lack of association observed in older population might be due to loss of liver function 55 and dysregulation of lipid and glucose metabolisms with aging, which may render body becoming less responsive to the anti-oxidants, anti-inflammatory and insulin sensitizing activities of VMS.Although no previous study on VMS use and NAFLD has examined an interaction with age, our results are consistent with those observed in the UK Biobank cohort study for cardiovascular disease outcomes, which share common risk factors and pathogenesis with NAFLD 6 .In that study of 465,268 adults, multivitamin use was inversely associated with cardiovascular disease outcomes, only among adults aged < 60 years 56 .
Furthermore, in our study VMS use was associated with a significantly lower odds of NAFLD in men, but not in women.Though the underlying mechanisms for this observation are not clear, sex-specific expression of hepatic drug metabolizing enzymes has been found 57 .Women also generally have a slower gastric emptying time and lower blood volume, liver size, and hepatic clearance than men, which might affect the absorption, distribution, bioavailability, and excretion of VMS 57,58 .Our results might suggest greater metabolizing capacity of VMS among men than women, aligning with the well-known sex difference in pharmacokinetics 58,59 .In addition, men have greater hepatic insulin insensitivity and visceral adiposity than women 60 .These sex-specific biological Table 4. Multivariable-adjusted a,b odds ratio (OR) and 95% confidence intervals (95% CIs) of NAFLD according to use of VMS according to demographic and lifestyle factors.BMI body mass index, CI confidence intervals, NAFLD non-alcoholic fatty liver disease, OR odds ratio, VMS vitamin and mineral supplements.a All results presented in this table are sampling-weighted estimates except the number of study participants.b Multivariable model was adjusted for age (quartiles), sex (male, female), household income level (quartiles), education level (elementary school, middle school, high school, college or higher), smoking status (never, past, current), alcohol drinking status (never, past, ≤ 1 time/month, 2-≤ 4 times/month, ≥ 2 times/week), regular exercise (no, yes), body mass index (< 18.5 kg/m 2 , 18.5-< 23 kg/m 2 , ≥ 23-< 25 kg/m 2 , ≥ 25 kg/m 2 ), waist circumference (cm, continuous), diabetes mellitus (no, yes), hypercholesterolemia (no, yes), number of comorbidities (0, 1, 2, 3, ≥ 4), serum triglyceride concentration (mg/dL, continuous), hypertension stage (stage1, ≥ stage2, other), current use of hypertension medications (no, yes), and intakes of total energy (quartiles) and fruits and vegetables (quartiles).c P interaction was tested by including the product term between current use of VMS and each of stratification factors.d Age groups were defined using the median value of our study population as a cut-off value.e BMI of 23 kg/m 2 is the cut-off criteria to define overweight status in the Asian population.f Fruit and vegetable consumption groups were categorized using the median value of our study population as a cut-off value.www.nature.com/scientificreports/variations associated with the pathogenesis of NAFLD might also predispose men to greater metabolic stress and lipoapoptosis in hepatocytes 60 , resulting in a greater benefit from VMS for men than women.
Our study has several strengths.To the best of our knowledge, ours is the first study that comprehensively evaluated the association between VMS use and NAFLD among hypertensive adults, and thus has highly significant clinical implication.We had detailed data on VMS usages and were able to explore associations, beyond the current status of VMS, including frequencies, duration, and use of multiple products.We comprehensively adjusted for potential confounding factors for NAFLD and conducted stratified analyses by demographics, lifestyle, and clinical conditions.However, our cross-sectional study design limits to infer causality.Nonetheless, NAFLD is asymptomatic.Individuals are less likely to change VMS use recognizing NAFLD status, though we cannot confirm this due to lack of data on change of VMS use.Second, NAFLD status was not ascertained by ultrasound or biopsy, the gold standard for diagnosing NAFLD.However, HSI has high sensitivity and specificity 32 , and is considered the most appropriate index for Asian populations 32 .Third, the distribution of VMS frequency, duration, and use of multiple VMS in our study was skewed limiting our power on examining appropriate use of VMS.Fourth, we did not have details on prescribed drugs, which may have an interaction with VMS, although we observed no significant effect modification by hypertension stage or hypertension medication use.Fifth, our study population are hypertensive adults, limiting generalizability of our findings.Finally, we cannot rule out unmeasured or residual confounding.
In conclusion, in this first large nationally representative study of hypertensive adults, current use of VMS was significantly associated with lower risk of NAFLD.The inverse association was stronger in younger than older individuals and in men than women.The attenuation of inverse association with current use of VMS at higher frequency, with longer duration of use, and with use of multiple products corroborates the positional statements of health authorities, which cautions overuses of VMS despite its potential benefit 48,53 .Large-scale cohort studies are warranted to replicate our finding.Table 5. Multivariable-adjusted a,b odds ratio (OR) and 95% confidence intervals (95% CIs) of NAFLD according to use of VMS according to population's hypertension-related factors.CI confidence intervals, IDH Isolated diastolic hypertension, ISH Isolated systolic hypertension, NAFLD non-alcoholic fatty liver disease, OR odds ratio, SDH systolic and diastolic hypertension, VMS vitamins and minerals supplements.a All results presented in this table are sampling-weighted estimates except the number of study participants.b Multivariable model was adjusted for age (quartiles), sex (male, female), household income level (quartiles), education level (elementary school, middle school, high school, college or higher), smoking status (never, past, current), alcohol drinking status (never, past, ≤ 1 time/month, 2-≤ 4 times/month, ≥ 2 times/week), regular exercise (no, yes), body mass index (< 18.5 kg/m 2 , 18.5-< 23 kg/m 2 , 23-< 25 kg/m 2 , ≥ 25 kg/m 2 ), waist circumference (cm, continuous), diabetes mellitus (no, yes), hypercholesterolemia (no, yes), number of comorbidities (0, 1, 2, 3, ≥ 4), serum triglyceride concentration (mg/dL, continuous), hypertension stage (stage1, ≥ stage2, other), current use of hypertension medications (no, yes), and intakes of total energy (quartiles) and fruits and vegetables (quartiles).c P interaction was tested by including the product term between current use of VMS and each of stratification factors.d Hypertension stage 1 was defined as 130 mmHg ≤ SBP < 140 mmHg or 80 mmHg ≤ DBP < 90 mmHg; and hypertension stage ≥ 2 was defined as 140 mmHg ≤ SBP or 90 mmHg ≤ DBP. e ISH was defined as 130 mmHg ≤ SBP and DBP < 80 mmHg; IDH was defined as SBP < 130 mmHg and 80 mmHg ≤ DBP; SDH was defined as 130 mmHg ≤ SBP and 80 mmHg ≤ DBP.f Controlled group were defined as those taking hypertension medication and having normal blood pressure levels (SBP < 130 mmHg or DBP < 80 mmHg; and uncontrolled group were defined those with hypertensive blood pressure levels despite the use of hypertension medication.

Figure 1 .
Figure 1.Flow diagram of subject inclusion and exclusion.

Figure 2 .
Figure 2. Association of NAFLD risk according to uses of vitamin and mineral supplements (VMS) defined by frequency, duration, and use of multiple products among hypertensive adults.

Table 1 .
Population characteristics a according to the current use of VMS among hypertensive adults in the Korea National Health and Nutrition Examination Survey, 2005 and 2007-2009 (N = 7376).ALT alanine aminotransferase, AST aspartate aminotransferase, BMI body mass index, HDL-C high density lipoprotein cholesterol, SES socio economic status, TG triglyceride, VMS vitamin and mineral supplements.a Values are presented as mean ± SE or N (percentages); all results presented in this table are sampling-weighted estimates except the number of study participants.b Household income level was grouped based on quartiles in our study population.c BMI cut-off is based on Asian obesity criteria.d Hypercholesterolemia was defined as participants having fasting total serum cholesterol level ≥ 240 mg/mL or currently taking lipid-lowering agents.e Comorbidity status score was defined as 0 point to 5 points by counting 1 point each for comorbid status of NAFLD such as cancer, chronic obstructive pulmonary disease, diabetes mellitus, cardiovascular disease, and chronic kidney disease.f Hypertension stage 1 was defined as 130 mmHg ≤ SBP < 140 mmHg or 80 mmHg ≤ DBP < 90 mmHg; hypertension stage ≥ 2 was defined as 140 mmHg ≤ SBP or 90 mmHg ≤ DBP; others were those who had blood pressure within the normal range, but used hypertensive medications.

Table 2 .
Age-adjusted and multivariable-adjusted a,b odds ratio (OR) and 95% confidence intervals (95% CIs) of NAFLD according to current use of VMS use among hypertensive adults.CI confidence intervals, NAFLD non-alcoholic fatty liver disease, OR odds ratio, VMS vitamin and mineral supplements.a All results presented in this table are sampling-weighted estimates except the number of study participants.