Associations of dietary PUFA with dyslipidaemia among the US adults: the findings from National Health and Nutrition Examination Survey (NHANES) 2009–2016

Abstract Dyslipidaemia, a significant risk factor of CVD, is threatening human health worldwide. PUFA are crucial long-chain fatty acids for TAG synthesis and removal, potentially decreasing dyslipidaemia risk. We examined dyslipidaemia prevalence among 15 244 adults aged ≥ 20 years from National Health and Nutrition Examination Survey 2009–2016. Dyslipidaemia was defined as total cholesterol ≥ 240 mg/dl, or HDL-cholesterol < 40 mg/dl/50 mg/dl for males/females, respectively, or LDL-cholesterol ≥ 160 mg/dl, or TAG ≥ 200 mg/dl, or taking lipid-modifying medications. We measured the daily PUFA intake using a 24-h dietary recall. Demographics, social economics, and lifestyle factors were collected using questionnaires/interviews. Additionally, we measured Se and Hg levels in the whole blood. Logistic regression models were used to examine the association between PUFA and dyslipidaemia. The unweighted and weighted dyslipidaemia prevalences were 72·4% and 71·0 %, respectively. When grouped into tertiles, PUFA intake above 19·524 g/d was associated with an independent 19 % decrease in dyslipidaemia risk (OR = 0·81 (95 % CI 0·71, 0·94)) compared with the lowest tertile (PUFA intake ≤ 12·349 g/d). A threshold inverse association was further determined by the restricted cubic spline analysis. When PUFA intake was increased to its turning point, that is, 19 g/d, the lower nadir risk for dyslipidaemia was obtained (OR = 0·72 (95 % CI 0·56, 0·89)). When the exposure was the sum of α-linolenic acid and octadecatetraenoic acid, the inverse linear association remained. Dietary PUFA intake is a beneficial factor for dyslipidaemia among American adults, independent of many potential confounders, including Hg and Se.


Introduction
Dyslipidemia is one of the manifestations of metabolic syndrome characterized by a change in the blood lipid profiles (1) .The accumulation of low-density lipoprotein cholesterol (LDL-C), total cholesterol (T.C.), and triglycerides (T.G.), as well as the decrease of high-density lipoprotein cholesterol (HDL-C), are important risk factors of atherosclerotic plaques, which originates in the medium and large arteries, principally leading to cardiovascular disease (CVD) and coronary heart disease (CHD) (1)(2)(3) .The increased prevalence and destructive power of dyslipidemia are threatening human health worldwide (4,5) .From the epidemiological survey data, there were more than 100 million adults aged 20 years or older having T.C. levels of 200 mg/dL(5.17mmol/L) or greater, and almost 31 million have levels of 240 mg/dL(6.20 mmol/L) or greater from 2009 to 2012 in the U.S. (2) .In 2013, the age-standardized mortality attributable to all CVD and CHD were 223.9 and 102.6 per 100,000 persons, respectively, in the U.S. (6) .Traditionally recognized risk factors for CVD included age, gender, obesity, hypertension, smoking status, type 2 diabetes, familial predisposition, and high levels of low-density lipoprotein cholesterol (LDL-C) (4) .Lowering LDL-C level with statin therapy can reduce CVD mortality (7,8) .
Additionally, dyslipidemia is also an essential consideration for controlling CVD risk (2,4) .Polyunsaturated fatty acids (PUFAs) are fatty acids that contain more than one double bond in their backbone, generally including n-3 PUFAs and n-6 PUFAs (9) .Docosahexaenoic acid (DHA), eicosapentaenoic acid (EPA), docosapentaenoic acid (DPA), α-linolenic acid (ALA), and octadecatetraenoic acid (ODTA), are crucial long-chain n-3 PUFAs for glucose and lipid metabolism as well as metabolic inflammation (10) .The mechanism of n-3 PUFAs action, especially for DHA and EPA, is mediated by surface or intracellular fatty acid receptors and sensors.Intake of PUFAs is associated with increased expression of adiponectin.As an anti-inflammatory cytokine, adiponectin promotes the hepatic metabolic enhancement and reduces atherosclerosis risk by increasing HDL-C and reducing T.G.ALA and ODTA, often referred to as plant n-3 PUFAs, are precursors of DHA Accepted manuscript and EPA.The physiological and functional attributes of plant n-3 PUFAs appear to derive from their conversion to EPA or DHA via desaturase (11,12) .Dietary intake and supplements are the primary sources of n-3 PUFAs in our daily life.Supplements are usually the derivatives with high EPA and DHA concentrations from fish, such as salmon, mackerel, or other oily fishes (13) .
Dietary habit changes with more unsaturated fatty acids are proposed to benefit metabolic disorders (12) .The beneficial effects of n-3 PUFA intake have been established widely in recent years (11,14,15) .Increased oily fish consumption was associated with a lower risk of hypertriglyceridemia in the general Korean population (1) .In another randomized controlled trial (RCT) study (16) , six-month use of DHA and EPA 900 mg/day was associated with a reduction in LDL-C and T.G. levels.Furthermore, they observed an increase in HDL-C level.A recent meta-analysis of studies totaling 693 CHD, type 2 diabetes, or nonalcoholic fatty liver diseases patients with an average age from 50 to 70 years old declared that high-dose n-3 fatty acids (purity >90%) slowed the atherosclerosis progression significantly, which is a potential mechanism in reducing CVD risk (17) .An authoritative advisory from the American Heart Association (AHA) claimed that consuming non-fried seafood 1 to 2 times per week promotes a positive effect on the cardiovascular system, resulting in a reduced risk of cardiac death, CHD, and ischemic stroke (15) .In addition, PUFAs promote effects on glucose and metabolic inflammation and as a good source of protein and vitamin D (12,15) .However, direct evidence on the association between dietary PUFAs and dyslipidemia risk among a nationally representative large sample-sized adults across various age groups is sparse if not lacking.The linear or non-linear association between dietary PUFAs intake and dyslipidemia is warranted to research.
Of note, fish consumption is a major source of long-chain PUFA intake.In addition to having PUFAs, fish, as a package, also contains other trace minerals, including both beneficial elements (e.g., selenium [Se] within a reasonable range) and toxic metals (e.g., mercury [Hg]).Se is an antioxidant that elicits its beneficial effects and may modify the Accepted manuscript association between n-3 PUFAs and dyslipidemia (18,19) .However, another study claimed that comparing the highest with lowest quartiles, the prevalence ratios of the selenium-zinc pattern is 1.36 (95%C.I.: 1.13-1.63)for metabolic syndrome with National Health and Nutrition Examination Survey (NHANES) 2011-2014 (20) .This result corroborated with a study, which found that high Se level is associated with increased T.C., LDL-C, HDL-C, and T.G. (21) .Hg, which has heavy metal toxicity and accumulates along the ocean's food chain (19) , may enter the body along with PUFAs.Some scholars suggested that Hg be evaluated in any participant with CHD or other vascular diseases (22) .Another research indicated that Hg influences include thrombosis, immune and mitochondrial dysfunction, and dyslipidemia (23) .Thus, it deserves an investigation whether Se and Hg will confound and modify the association between dietary PUFAs and dyslipidemia risk.
Therefore, in this study, we aim to examine the associations between the consumption of PUFAs and the risk of dyslipidemia, and whether the associations will be confounded and

Study population
The NHANES is a nationally representative measurement of the civilian non-institutionalized U.S. population with a stratified multistage probability cluster design

Measurement of dyslipidemia
Blood dyslipidemia was defined as T.C. ≥240 mg/dL, or HDL-C <40 mg/dL for males, HDL-C <50 mg/dL for females, or LDL-C ≥160 mg/dL, or T.G. ≥200 mg/dL, or self-reported usage of prescribed lipid-modifying medication.The T.C., HDL-C, LDL-C, and T.G. information all derived from the laboratory measurements.The biospecimens were collected at the Mobile Examination Center (MEC), including the collecting, processing, storing, and shipping of blood specimens.The MEC's controlled environment allowed laboratory measurements to be done under identical conditions at each survey location (24) .

Measurement of PUFAs
In the NHANES study, the dietary intake information was collected by the trained interviewers for all NHANES examinees and used to estimate the types and amounts of foods and beverages (including all types of water) consumed during the 24-hour period prior to the interview (midnight to midnight) and to estimate intakes of energy, nutrients, and other components from those foods and beverages.Daily intake of PUFAs was reported from the average of two 24-h dietary recall interviews.The first dietary recall interview was collected in-person at the MEC, and the second interview was conducted by telephone after 3 to 10 days, but not on the same day of the week as the first MEC interview.If a participant did not finish the second dietary call interview, only the first dietary interview was used as average.Daily intakes of the common long-chain n-3 fatty acids, DHA, EPA, and DPA, and their sum were calculated for each participant from the average of two 24-h dietary recall interviews.Other two n-3 unsaturated fatty acids, α-linolenic acid (ALA) and octadecatetraenoic acid (ODTA), and their sum were also obtained from the interviews.

Measurement of Hg and Se
The measured Hg and Se were derived from blood specimens.Whole blood specimens were collected at the MEC center, processed, stored, and shipped to the Division to Laboratory Sciences.Some measurements below the detection limit (DL) were imputed with the value of "DL/√2" (25) .In the wave of the year 2009-2010, all specimens were not detected Se in the serum.Hg was measured for all four waves.The Hg and Se in µg/L were converted to nmol/L by multiplying 4.99 nmol/µg and 0.0127 nmol/µg, respectively.

Potential confounders
In the NHANES study, demographic, behavior, social-economic information, and clinical status could be obtained from questionnaires and examinations.In our research, we considered the following variables as potential confounders: age group (groups for every 15 years starting from 20 years old), ethnicity (Non-Hispanic white, Non-Hispanic black, Hispanic, and others), sex (male and female), body mass index (BMI), education (under senior high, senior high, and college or university and above), smoking status (never, past, and current), alcohol beverages (total alcoholic drinks per year classified by median), diabetes status(non-diabetes, and diabetes), and hypertension (yes vs. no).We adjusted these confounders in multiple models sequentially, which would be detailed in the statistical analysis part.

Statistical analysis
Accounting for the complex, stratified, multistage cluster sampling design structure, we used survey procedures in SAS statistical software version 9.4 (SAS Institute Inc., Cary, NC, U.S.) for all statistical analyses.Four waves of continuous survey data were combined, and an eight-year sampling weight was calculated for the analyses.
Descriptive statistics, including counts and percentages for categorical variables, weighted means, and standard errors for continuous variables, were calculated for dyslipidemia and regular lipid group, respectively.Due to the highly skewed distribution of

Accepted manuscript
Se and Hg, we computed weighted geometric means and 95%C.I.s for these variables (26) .For all PUFAs, the weighted median with inter-quartile range was calculated in two groups, respectively, based on their sampling distributions.
We constructed multivariable logistic regression models with adjusting confounders to investigate the odds ratios (O.R.s) as well as 95% confidence intervals (C.I.s) for the risk of dyslipidemia in association with PUFAs.In model 1, we adjusted for age, sex, and race; model 2 further adjusted for education, BMI, smoking status, and alcoholic beverages; model 3 additionally adjusted for hypertension and diabetes.Also, the intakes of protein and cholesterol were involved in model 4. In model 5, we mainly considered the effects of PUFAs after adjusting the impact of blood Hg and Se.
We also fitted PUFAs with restricted cubic spline functions to explore the non-linear association with the original data.Both first-and second-order interactions among PUFA intake, blood Se (≥ vs. < median), and Hg (≥ vs. < median) were considered in the full model.The stratification analyses were done for each Hg and Se level to explore the effect modification.
Considering the potential impact of missing data on the analysis results, we performed multiple imputation, mainly for missing values of Se and Hg.We performed logistic regression models to consider the odds ratios for different intake levels of PUFAs and interactions with imputed data for the main analysis.All logistic regression results of imputed data were obtained by using PROC MIANALYZE in SAS for considering the imputation effect.

Results
Study population characteristics were presented in Table 1.The unweighted prevalence of dyslipidemia was 72.4%, with 11,041 interviewees classified as dyslipidemia among a total of 15,244 participants.Participants with dyslipidemia were more likely to be older, smokers, diabetic, non-Hispanic whites, and had higher BMI and blood pressure and lower education than those without dyslipidemia.Lower intakes of PUFAs, protein, DPA, ALA, ODTA, the sum of ALA and ODTA, and Hg were also observed in dyslipidemia participants.
As for sex, consumption of alcoholic beverages, intakes of cholesterol, DHA, EPA, the sum of DHA, EPA and DPA, Se, there was no difference between participants with or without dyslipidemia.
An inverse association was found between PUFA intake and dyslipidemia (Table 2).
Comparing with those in the lowest tertile of PUFA intake, those in the middle and upper tertiles had a 17% [ O.R. (95% C.I.) of 0.83 (0.75, 0.93)] and 19% [0.81 (0.71, 0.94)] lower risk, respectively, after controlling for multiple confounders (P for linear trend = 0.008).The inverse association was further confirmed when PUFA intake was fitted using restricted cubic spline functions (Figure 2a).Compared with the lowest PUFA intake, a threshold inverse association of PUFAs with dyslipidemia was observed.A nadir risk for dyslipidemia was reached [O.R. = 0.72 (95% C.I.: 0.59-0.86)]when PUFA intake arrived at its tuning point of 19 g/day.
When the exposure of interest was the sum of intakes of ALA and ODTA, the inverse association remained [T3 vs. T1: O.R. as 0.87 (0.76, 0.99); P for linear trend = 0.04] (Table 3).However, we did not observe a threshold phenomenon while fitting the exposure with restricted cubic spline functions (Figure 2b).We found that neither Se (≥ vs. < median) nor mercury (≥ vs. < median) modified the associations of interest in both the categorical analyses and the restricted cubic spline analyses (results not shown).There was no interaction between Se, Hg, and PUFA intake, Downloaded from https://www.cambridge.org/core.IP address: 54.70.40.11, on 04 Jul 2021 at 02:39:47, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.https://doi.org/10.1017/S0007114521002300Accepted manuscript either.Stratification analysis by gender was also performed for effect modification exploration.The associations were consistent between males and females.

Discussion
In this cross-sectional study of a nationally representative sample of the U. S. adults using data from the NHEANES 2009-16, we found inverse associations of both intakes of PUFAs, and the sum of ALA and ODTA with the risk of dyslipidemia, after controlling for some potential cofounders including blood Se and Hg.
The inverse relation between PUFAs and dyslipidemia was consistent with the findings from previous studies (11,(14)(15)(16)(17) .The n-3 PUFA intake operated on the decrease of plasma T.G., very-low-density lipoproteins (VLDL), and apolipoprotein B-100 (APOB-100), as well as the increase of high-density lipoprotein (HDL) (12,(27)(28)(29) .The potential mechanisms of n-3 PUFAs exerting effects are as follows.N-3 PUFAs may decrease the expression of sterol regulatory element-binding protein-1c(SREBP-1C), contributing to reduced expression of cholesterol-, fatty acid-, and T.G.-synthesizing enzymes.They could also increase the mitochondrial oxidation rates, or peroxisome, resulting in a reduction in available substrate required for T.G. and VLDL synthesis.In addition, they have been shown to inhibit key enzymes involved in hepatic T.G. synthesis and increase the expression of lipoprotein lipase, leading to decreased T.G. synthesis and increased T.G. removal from circulating VLDL and chylomicron particles (12,27,28,30) .Some studies reported that n-3 PUFAs reduced serum T.G. in a population with hypertriglyceridemia and increased LDL-C and HDL-C; however, the increase in LDL-C was less than the reduction in VLDL-C resulting in a decrease of non-HDL-C (VLDL-C and LDL-C) (31,32) .However, the large-scale RCT studies ─ VITAL (NCT01169259) (33) and ASCEND (NCT00135226) (34) revealed that there was no significant difference in the incidence of major cardiovascular events between the n-3 fatty acid supplementation group and the Downloaded from https://www.cambridge.org/core.IP address: 54.70.40.11, on 04 Jul 2021 at 02:39:47, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.https://doi.org/10.1017/S0007114521002300Accepted manuscript placebo group.In these two studies, the main exposure included n-3 fatty acids 1g/day.But in the VITAL research, a lower incidence of the primary cardiovascular endpoint in the n-3 supplementation group could be observed in the low fish consumption strata (33) .In fact, the resources of daily PUFAs are not only derived from supplementation but dietary seafood.Thus, the biologically plausible effect of PUFAs raised the question of the potential difference between cardiovascular events and intermediate cardiovascular endpoints, such as dyslipidemia, diabetes, inflammatory reaction, and hypertension.
The other two large-scale RCTs ─ REDUCE-IT (NCT01492361) (35) and JELIS (NCT00231738) (36) ─ showed that highly purified EPA would reduce the incidence of cardiovascular events in patients who had been under a statin therapy for treating hypertriglyceridemia or hypercholesterolemia.Both trials claimed that EPA was efficacious for participants who had established CVD for secondary prevention.What is noteworthy was that the imbalanced gender ratio and high fish consumption diet failed to detect a significant effect on primary prevention with an underpowered analysis in a Japanese study (36) .Based on a high PUFA diet, purified EPA was also useful as a prescription medicine for secondary prevention.Pharmacological interventions and nutritional observations are difficult to inter-extrapolate because PUFAs contain many fatty acids other than EPA.
Further food-based or nutrient trials are warranted to explore the dietary PUFAs effect on primary or secondary prevention of different types of CVD.
To our knowledge, this study was the first large-scale observational research on the associations between dietary PUFAs and dyslipidemia risk using the NHANES data.We clearly and newly illustrated the linear relation and a threshold phenomenon between PUFA intake and dyslipidemia risk.The benefits of dietary PUFAs on dyslipidemia have a potential translative value to the treatment and prevention of hyperlipidemia in clinical practice.This study further considered Se and Hg's potential confounding effect due to PUFA intake often deriving from deep ocean oily fish.In participants with higher Hg levels Accepted manuscript in the blood, possible stronger associations were observed.When participants had lower blood Se levels, there was a significantly decreased trend of O.R. for dyslipidemia along with the increased PUFA intake.The toxicity effect of Hg included increased oxidative stress, inflammation, and dyslipidemia (23) .Participants with a higher level of blood Hg had an increased risk of dyslipidemia.N-3 fatty acid intake could antagonize Hg's toxicity (23,37) .However, the interrelationship among PUFA intake, blood Se, and Hg has not been well elucidated so far.
In Figure 2a, the spline's slope changed substantially before and after the turning point of PUFA intake (i.e., about 19 g/day).This slope alteration appeared to be related to differences in the size of relation estimates or different dose-response relationship rather than a lack of association (1,17,38) .After the turning point, there was a threshold phenomenon appeared.The main source of PUFAs is seafood, especially deep-sea oily fish.As the intake of fish increases, the PUFAs are elevated with an initially protective relation but gradually offset with accumulated organic pollutants (39,40) .However, the threshold was not observed in the spline result of the sum of ALA and ODTA with dyslipidemia risk.One possible explanation was that the ALA source was vegetable oil in the American population, especially soybean oil and canola (41,42) , which are less likely to be contaminated by organic pollutants than seafood.Another possible reason for the steady decline for ALA and ODTA was the insufficient dose with a maximum 5.5g/day compared with daily PUFA intake.The continuous declination could also be observed in the spline result of daily PUFA intake in Figure 2a with a dose from minimum to 5.5g/day.On the other hand, plant n-3 PUFAs could be converted to DHA/EPA for functional attributes in metabolism, but this limited conversion resulted in lower biological potency (11) .This research also had some limitations.First, residual confounding from the unmeasured covariates may affect the results from observational data.In the NHANES dataset, there was no information about familial predisposition to dyslipidemia, which may have potential effect on the risk of dyslipidemia.Additionally, data reporting in this research Downloaded from https://www.cambridge.org/core.IP address: 54.70.40.11, on 04 Jul 2021 at 02:39:47, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.https://doi.org/10.1017/S0007114521002300 modified by Se and Hg levels in the U. S. population based on National Health and Nutrition Examination Survey (NHANES) 2009-2016 data.
conducted by the National Center for Health Statistics of the Centers for Disease Control and Prevention.For this study, four NHANES survey waves (2009-2010, 2011-2012, 2013-2014, and 2015-2016) were combined using adjusted sampling weights.A total of 40,439 individuals was selected from these four waves.Eligible participants were 23,234 adults aged 20 or older, with no missing information about sex, ethnicity, education, and weight.Seventeen participants were excluded due to missing smoking and diabetes information.We excluded 2,624 missings on the polyunsaturated fatty acids (PUFAs) who Downloaded from https://www.cambridge.org/core.IP address: 54.70.40.11, on 04 Jul 2021 at 02:39:47, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.https://doi.org/10.1017/S0007114521002300Accepted manuscript did not complete the two 24-hour dietary recall interviews.These exclusions resulted in the final sample size of 15,244 adults after 5,349 participants were eliminated due to missing measurements on the laboratory examinations, including T.C., HDL-C, LDL-C, and T.G., or dyslipidemia medication.The detailed sample elimination process was shown in Figure 1.

Figure 1 .
Figure 1.Flow chart of data exclusion process of this study.