Association of vitamins, minerals, and lead with Lipoprotein(a) in a cross-sectional cohort of US adults

Abstract

Lipoprotein(a)(Lp[a]) is a low-density lipoprotein-cholesterol (LDL-C)-like particle with potent pro-atherothrombotic properties. The association of Lp(a) with several circulating factors, including vitamins, remains unresolved. We performed an observational analysis using the National Health and Nutrition Examination Survey III cohort, a cohort used to monitor the nutrition status of US-citizens. We used multivariable linear regression to test associations of Lp(a) and LDL-C with levels of serum vitamins and minerals and whole-blood lead. Analyses controlled for factors known to associate with Lp(a) (age, sex, race/ethnicity, statin use, hemoglobin A1c, body mass index, hypertension, diabetes, glomerular filtration rate, alcohol intake, and saturated fat intake). LDL-C was corrected for Lp(a) mass. Multiple sensitivity tests were performed, including considering factors as categorical variables (deficient, normal, elevated). Among 7,662 subjects, Lp(a) correlated (β-coefficient) positively (change per 1 conventional unit increase) with carotenoids (lycopene (0.17(0.06,0.28), p=0.005), lutein (0.19(0.07,0.30), p=0.002), β-cryptoxanthin (0.21(0.05,0.37), p=0.01), β-carotene (0.05(0.02,0.09), p=0.003), and α-carotene (0.15(0.01,0.30), p=0.04)) and lead (0.54(0.03,1.05), p=0.04) levels when tested as continuous variables. LDL-C had similar associations. Lp(a) did not associate with vitamins A, B12, C, or E retinyl esters, folate, RBC-folate, selenium, ferritin, transferrin saturation, or calcium. With factors as categorical variables, Lp(a) but not LDL-C negatively associated with elevated vitamin B12 (−5.41(−9.50, −1.53), p=0.01) and folate (−2.86(−5.09, −0.63), p=0.01). In conclusion, Lp(a) associated similarly to LDL-C when vitamins, minerals, and lead were tested as continuous variables, while only Lp(a) correlated with vitamin B12 and folate when tested as categorical variables. These observations are hypotheses generating and require further studies to determine causality.

Introduction

Lipoprotein(a) (Lp[a]) is a plasma lipoprotein composed of cholesterol, cholesteryl esters, apolipoprotein B100, apolipoprotein(a) (apo[a]), and a low amount of triglycerides and carbohydrates [1, 2]. Lp(a) differs from LDL by the addition of apo(a). Apo(a) originated from a duplicated plasminogen gene and competes with plasminogen for binding of endothelial cells and altering their functions, which may account for its both atherogenic and atherothrombotic effects.

Mendelian randomization and genome wide association studies suggest that Lp(a) is causal for increasing the risk of cardiovascular disease (i.e. coronary heart disease, aortic stenosis, and stroke) independently from LDL [1]. About 1-in-7 individuals in the US have an Lp(a)≥50 mg/dL [3], the guideline-recognized threshold to consider Lp(a) a risk enhancing factor [4, 5]. Although several drugs are in the pipeline to target Lp(a), this problem will require multiple angles to limit the negative outcomes associated with elevated Lp(a).

Lp(a) levels are primarily mediated by genetic variations in LPA’s promoter region, kringle IV Type 2 domain repeats, and single nucleotide polymorphisms [1]. However, little is known about nongenetic factors that affect or associate with Lp(a) levels. Lp(a) is higher among blacks than whites, females, non-diabetics, and in end-stage renal disease or nephrotic syndrome [6, 7]. Lp(a) has intermittently associated with age, BMI, and hypertension [6–9]. C-reactive protein and N-acetyl cysteine have been shown to have no associations [10–13]. Substances associated with lower Lp (a) include niacin, androgens and estrogens [14, 15], L-carnitine [16], curcuminoids [17], co-enzyme Q [18, 19], combined omega-3 fatty acids and vitamin E [20], vitamin C [21], saturated fats [22, 23], alcohol [24, 25], and coffee [26]. A recent comprehensive review suggested that the strongest data from nutrition or nutraceuticals to lower Lp (a) were shifts in dietary fat, and higher ethanol, coffee, L-carnitine, and coenzyme Q [27]. Associations between these factors and Lp(a) may be due to influences on transcriptional activation or inhibition, alterations in mitochondrial fatty acid metabolism, hepatic lipogenesis, or alteration in inflammatory signaling cascades, mechanisms that are not fully elucidated [14–16, 18, 20, 21]. Furthermore, there remain many commonly measured metabolites and nutrients that have not been studied for their association with Lp(a).

With this in mind, we sought to determine the association between serum Lp(a) levels and serum levels of vitamins and minerals and whole-blood lead since this can generate new hypotheses and future research considerations. We used the National Health and Nutrition Examination Survey III (NHANES III), a cross-sectional US cohort designed to be nationally representative. Because both fat-soluble (e.g. vitamin E) and water-soluble factors (e.g. vitamin C) might alter Lp(a) levels, we studied all available factors. Factors identified by our study as associated with Lp(a) can be examined in future studies to test for causality and potential utility for lowering Lp(a), especially if some factors only associate with Lp(a) levels, but not LDL-C.

Methods

Study population

We performed an observational analysis using data from the NHANES III adult sample [28]. NHANES III is a nationally representative cross-sectional cohort of the US population collected in two phases between 1988–1994 from a multi-stage probability sample of about 40,000 non-institutionalized civilians aged 2 months and over. Each phase independently comprised a national probability sample [28]. Lp(a) was measured in the second phase (1991–1994). The samples were tested and recorded at the time of data collection. The plan and operations for data collection and reporting, including sampling have been previously described [28]. Samples were included if they had data on Lp(a) and excluded if they were missing any of the covariates of interest or were <18 years old. The study was approved by the Yale University institutional review board.

Assays

Lp(a), total cholesterol, high density lipoprotein cholesterol, and triglycerides were tested from serum samples. The Lp(a) assay was an enzyme-linked immunosorbent assay, which is reported in mass concentration (mg/dL) (Macra test kit, Strategic Diagnostics Inc., Newark DE) [29]. Samples were tested against standardized samples provided by the laboratory providing the test kits. Cholesterol, high density lipoprotein cholesterol, and triglycerides were measured using an enzymatic assay and reported in mass concentration (mg/dL). Additional information on all of the assays utilized for the laboratory testing of serum vitamins and minerals and whole-blood lead have been previously outlined in the laboratory procedures manual [29].

Outcome definitions

The outcomes were Lp(a) and LDL-C. Lp(a) is reported in mg/dL. LDL-C was calculated using the Martins-Hopkins method [30], corrected for Lp(a) mass (LDL-C_corr = LDL–[Lp(a)×0.3]) [31], and reported in mg/dL We excluded samples if LDL-C was calculated at <0 mg/dL (n=1).

Statistical analyses

Covariates were not normally distributed in the sample. Thus, we report descriptive statistics for covariates as the median (interquartile range [IQR]) or percent of the sample. Vitamins, minerals, and lead are reported in conventional units (Table 1).

Table 1.

Reference ranges for serum vitamins, minerals, and whole blood lead [29, 32]

Vitamin, Mineral, or Lead	Normal Range
Vitamin A (μg/dL)	25–115
Retinyl esters (μg/dL)	2–9a
Vitamin B12 (pg/mL)	165–1600
Folate (ng/mL)	2.6–12.2
RBC-folate (ng/mL)	102.6–410.9
Vitamin C (mg/dL)	0.05–1.5
Vitamin E (μg/dL)	500–2650
Lycopene (μg/dL)	3–55
Lutein (μg/dL)	5–65
β-cryptoxanthin (μg/dL)	2–4
β-carotene (μg/dL)	2–80
α-carotene (μg/dL)	1–15
Selenium (ng/mL)	<200
Ferritin (ng/mL)	10–800
Transferrin Saturation (%)	≥15
Serum Calcium (mg/dL)	9.0–11.0
Lead (μg/dL)	≤5b

^aNo reference range available, used <10^th and >90^th percentile.

^bLead reference range.

We performed 3 linear regressions (Model 1 – univariable analysis, Model 2 – multivariable analysis [primary results], and Model 3 – multivariable analysis with interaction terms) for each vitamin, mineral, and lead, including vitamin A, retinyl esters, vitamin B12, folate, red blood cell folate, vitamin C, vitamin E, lycopene, lutein, β-cryptoxanthin, β-carotene, α-carotene, selenium, iron, iron saturation, calcium, and lead. The control variables included in multivariable linear regressions were age (range: 17–90 years), sex, race/ethnicity (non-Hispanic white, non-Hispanic Black, Mexican-American, or Other), statin use, history of diabetes, history of hypertension, BMI (<18, 18–25, and >25 kg/m²), estimated glomerular filtration rate (eGFR) (via the 6-variable Modification of Diet in Renal Disease equation), active smoking, saturated fat dietary content (% of daily kcal), and daily alcohol intake (grams). Niacin was not chosen as a covariate. Niacin was not approved by the FDA at the time these data were collected, thus, no subjects were taking this at a prescriptive dose. In the multivariable model with interaction terms (Model 3) we transform covariates to their best fit (i.e. r²) linear or non-linear relationships (i.e. log-linear and polynomials). In Model 3 we also tested for any potential interactions between covariates. Interaction terms were retained if the p-value was <.05. Control variables were included in models regardless of p-value.

Sensitivity analyses

We performed four sensitivity tests. First, since a high proportion of the sample had Lp(a)=0 mg/dL (n=1,120), we repeated analyses excluding subjects with Lp(a) of 0 mg/dL. Second, analyses were repeated with the outcomes of ln(Lp[a]+1) or ln(LDL-C+1). Third, we repeated analyses with vitamin, mineral, and lead covariates as categorical variables. Categories were deficient, normal, and elevated as outlined in the NHANES III laboratory procedures manual [29]. For lead, since standards for safe lead levels have decreased since NHANES III we used the current Centers for Disease Control and Prevention definition of elevated lead value (≥5 μg/dL) rather than the laboratory manual definition (≥10 μg/dL) (Table 1) [32]. Retinyl esters had no established range for normal, hence we used <10th percentile and >90th percentile at deficient or elevated, respectively. Testing as categorical variables was chosen because some effects might only be observed at levels above or below the recognized normal range of serum values. Interactions were not included in this sensitivity test because of small counts in some categories that prevented several models from converging. Fourth, we tested associations of covariates with the outcomes (Lp(a) and LDL-C) as quartiles in multinomial logit models for both univariable (Model 4) and multivariable models (Model 5) that used the same aforementioned control variables. Again, for this sensitivity test we did not consider interaction terms since small counts prevented models from converging.

Criteria to interpret results

A 2-tailed p-value<.05 was considered statistically significant. All analyses were adjusted for complex sample design and sample weights. Outputs for multivariable models that include interactions (i.e. Model 3) were average marginal effects at mean population values. We conducted data analysis from December 2019 through February 2021. Data were analyzed using Stata 16 (StataCorp, LLC).

Results

Population characteristics

The final sample included 7,662 subjects. Unadjusted for complex sample design, median age was 42 years (IQR 29, 62), the majority were female (57.5%), and the most prevalent race/ethnicity was non-Hispanic white (38.5%). Population characteristics are available in Table 2. See Table E1 in Electronic Supplementary Material 1 for comparison of sample characteristics of included to excluded individuals.

Table 2.

Population characteristics (n=7,662)

Covariate	Median (IQR) or percent
Age (years)	42 (29, 62)
Female (%)	57.5%
Race
Non-Hispanic White	38.5%
Non-Hispanic Black	30.3%
Mexican American	26.4%
Other	4.8%
BMI
<18 kg/m2	1.4%
18–25 kg/m2	37.8%
>25 kg/m2	60.8%
Diabetes (%)	7.7%
Hypertension (%)	26.8%
eGFR (ml/min /1.73 m2)	75 (64, 87)
Taking Statin (%)	1.3%
Active smoker	26.8%
Saturated fat (% kcal)	10.5 (8.1, 13)
Daily alcohol (g per day)	0 (0, 0)

BMI: body mass index; eGFR: estimated glomerular filtration rate; IQR: interquartile range.

Univariate analysis

Lp(a) and LDL-C both positively associated (β-coefficient) with history of hypertension being on a statin, and negatively associated among Mexican-Americans (Table 3). However, Lp(a) associated positively but LDL-C associated negatively among non-Hispanic Blacks and female sex. Only Lp(a) was significantly negatively associated with saturated fat and only LDL-C was associated with diabetes, age, BMI, eGFR, and dietary alcohol.

Table 3.

Associations between control variables with Lp(a) and LDL-C in univariate, multivariate, and multivariate with interaction term models

	Model 1 – univariate			Model 2 – multivariate			Model 3 – multivariate with Interactionsa,b,c
	β	95% CI	p-value	β	95% CI	p-value	β	95% CI	p-value
Lipoprotein(a)
Age (per year)	0.05	−0.01, 0.11	0.07	0.07	−0.01, 0.15	0.10	0.08	−0.01, 0.16	0.06
Female	3.36	1.10, 5.62	0.005**	2.87	0.55, 5.18	0.02*	2.76	0.48, 5.03	0.02*
Race/ethnicity	–	–	<0.001***	–		<0.001***	–		<0.001***
Non-Hispanic White	–	–	–	–		–	–		–
Non-Hispanic Black	24.33	22.00, 26.66	–	24.87	22.34, 27.40	–	25.35	22.74, 27.96	–
Mexican-American	−4.84	−6.99, −2.68	–	−4.13	−6.17, −2.08	–	−4.40	−6.42, −2.38	–
Other	0.33	−3.92, 4.58	–	0.23	−3.97, 4.44	–	0.52	−3.50, 4.55	–
BMI	–	–	0.22	–		0.14	–		0.12
<18 kg/m2	–	–	–	–		–	–		–
18–25 kg/m2	4.98	−0.70, 10.67	–	5.34	−0.79, 11.48	–	5.62	−0.25, 11.48	–
>25 kg/m2	4.7’	−1.10, 10.52	–	4.14	−2.00, 10.28	–	4.42	−1.36, 10.20	–
Diabetes	0.62	−1.61, 2.85	0.57	−1.49	−3.77, 0.80	0.19	−1.21	−4.46, 2.05	0.45
Hypertension	2.47	1.25, 3.69	<0.001***	−0.02	−1.42, 1.39	0.98	−0.29	−1.67, 1.09	0.67
eGFR (per ml/min/1.73m2)	0.03	−0.03, 0.09	0.29	0.00	−0.08, 0.08	0.95	0.00	−0.08, 0.08	0.93
Statin	15.34	4.83, 25.85	0.006**	16.16	5.85, 26.47	0.004**	16.50	6.66, 26.35	0.002**
Smoker	−0.22	−2.07, 1.63	0.81	0.04	−1.54, 1.61	0.96	−0.24	−1.82, 1.34	0.75
Saturated fat (per % kcal)	−0.33	−0.58, −0.07	0.01*	−0.26	−0.51, −0.01	0.04*	−0.26	−0.51, −0.01	0.04*
Daily alcohol (per g)	0.01	−0.03, 0.04	0.70	0.01	−0.03, 0.05	0.54	−0.02	−0.05, 0.02	0.29
LDL-C
Age (per year)	0.69	0.60, 0.78	<0.001***	0.50	0.42, 0.58	<0.001***	0.62	0.55, 0.70	<0.001***
Female	−2.77	−5.13, −0.40	0.02*	−2.02	−4.29, 0.25	0.08	−2.13	−4.37, 0.10	0.06
Race/ethnicity	–	–	<0.001***	–	–	<0.001***	–	–	<0.001***
Non-Hispanic White	–	–	–	–	–	–	–	–	–
Non-Hispanic Black	−14.03	−16.41, −11.64	–	−12.05	−14.35, −9.75	–	−12.39	−14.88, −9.91	–
Mexican-American	−5.17	−7.62, −2.73	–	−1.30	−3.49, 0.90	–	−1.41	−3.53, 0.71	–
Other	−2.53	−10.15, 5.09	–	1.28	−4.75, 7.31	–	0.63	−4.77, 6.03	–
BMI	–	–	<0.001***		–	<0.001***	–	–	<0.001***
<18 kg/m2	–	–	–		–	–	–	–	–
18–25 kg/m2	14.72	7.37, 22.07	–	13.77	6.57, 20.98	–	14.96	7.59, 22.34	–
>25 kg/m2	34.37	28.92, 41.82	–	30.22	22.97, 37.47	–	29.57	21.58, 37.57	–
Diabetes	13.35	10.43, 16.28	<0.001***	−0.58	−4.05, 2.89	0.74	8.11	1.75, 14.47	0.02*
Hypertension	15.66	12.96, 18.36	<0.001***	4.36	1.52, 7.21	0.004**	4.51	1.39, 7.64	0.007**
eGFR (per ml/min/1.73 m2)	−0.52	−0.63, −0.40	<0.001***	−0.10	−0.19, −0.00	0.048*	−0.09	−0.18, −0.01	0.04*
Statin	13.56	3.53, 23.58	0.01*	−3.29	−13.08, 6.50	0.49	−4.12	−13.76, 5.52	0.39
Smoker	−1.59	−4.21, 1.04	0.22	2.61	0.01, 5.21	0.047*	2.19	−0.08, 4.46	0.06
Saturated fat (per % kcal)	0.37	−0.03, 0.78	0.07	0.37	0.01, 0.73	0.046*	0.38	0.03, 0.73	0.04*
Daily alcohol (per g)	−0.09	−0.12, −0.06	<0.001***	−0.06	−0.09, −0.03	<0.001***	−0.07	−0.10, −0.04	<0.001***

CI: confidence interval; Lp(a): Lipoprotein(a); LDL-C: low-density lipoprotein cholesterol; BMI: body mass index; eGFR: estimated glomerular filtration rate.

^*p<0.05;

^**p<0.01;

^***p<0.001.

^aIn the Lp(a) model, interactions included between age and race/ethnicity, sex and race/ethnicity, race/ethnic and statin, diabetes and BMI, alcohol and smoking, and hypertension and smoking.

^bIn the LDL-C model, interactions included age², between age and sex, age and diabetes, age and BMI, BMI and diabetes, eGFR and saturated fat content, eGFR and smoking, race/ethnicity and alcohol, and sex and BMI.

^cEffect sizes between Lp(a) and LDL-C in Model 3 can only be compared if variables are transformed to the same linear or non-linear association with the outcome.

Serum vitamins, minerals, and lead negatively associated with Lp(a) included red blood cell folate, ferritin, and transferrin saturation (Table 4). Covariates positively associated with Lp(a) included carotenoids (lutein and β-cryptoxanthin) and lead. These factors associated similarly with LDL-C, except for red blood cell folate, which associated positively with LDL-C.

Table 4.

Associations between Vitamins, Minerals, and Heavy Metals with Lp(a) and LDL-C in univariate, multivariate, and multivariate with interaction term models

	Model 1 – univariate			Model 2 – multivariate			Model 3 – multivariate and Interactionsa, b, c
	β	95% CI	p-value	β	95% CI	p-value	β	95% CI	p-value
Lipoprotein(a)
Vitamin A	0.00	−0.08, 0.08	0.98	0.04	−0.03, 0.10	0.22	0.05	−0.02, 0.12	0.14
Retinyl esters	−0.01	−0.21, 0.19	0.93	0.01	−0.17, 0.19	0.92	0.63	0.41, 0.84	<0.001***
Vitamin B12	0.00	−0.00, 0.00	0.14	0.00	−0.00, 0.00	0.33	1.42	−0.33, 3.16	0.11
Folate	−0.13	−0.26, 0.00	0.051	−0.12	−0.27, 0.04	0.13	−0.11	−0.27, 0.05	0.16
RBC-folate	−0.01	−0.02, −0.00	0.04*	0.00	−0.01, 0.01	0.33	−0.01	−0.02, 0.01	0.45
Vitamin C	−0.55	−3.34, 2.24	0.69	−0.20	−3.03, 2.62	0.88	0.00	−2.87, 2.87	0.99
Vitamin E	0.00	−0.00, 0.00	0.29	0.00	−0.00, 0.00	0.07	0.01	0.00, 0.01	<.0001***
Lycopene	0.11	−0.01, 0.23	0.06	0.17	0.06, 0.28	0.005**	0.18	0.07, 0.29	0.003**
Lutein	0.26	0.14, 0.38	<0.001***	0.19	0.07, 0.30	0.002**	5.43	2.98, 7.88	<0.001***
β-cryptoxanthin	0.21	0.05, 0.37	0.01	0.21	0.05, 0.37	0.01*	3.40	1.56, 5.23	0.001*
β-carotene	0.07	0.04, 0.10	<0.001***	0.05	0.02, 0.09	0.003**	2.37	1.23, 3.52	<0.001***
α-carotene	0.13	−0.01, 0.28	0.08	0.15	0.01, 0.30	0.04*	0.17	0.02, 0.32	0.03*
Selenium	−0.04	−0.12, 0.03	0.24	0.02	−0.06, 0.10	0.56	2.97	−6.53, 12.48	0.52
Ferritin	−0.04	−0.06, −0.02	0.002**	−0.01	−0.03, 0.01	0.44	−0.52	−2.38, 1.34	0.57
Transferrin saturation	−0.13	−0.22, −0.05	0.004**	−0.05	−0.13, 0.03	0.21	−0.05	−0.13, 0.04	0.25
Calcium	−0.26	−2.72, 2.19	0.83	−0.19	−2.40, 2.01	0.86	−0.13	−2.32, 2.06	0.91
Lead	0.67	0.14, 1.20	0.02*	0.54	0.03, 1.05	0.04*	0.52	0.01, 1.04	0.046*
LDL-C
Vitamin A	0.60	0.49, 0.70	<0.001***	0.42	0.32, 0.52	<0.001***	0.49	0.39, 0.58	<0.001***
Retinyl esters	2.01	1.41, 2.62	<0.001***	1.68	1.08, 2.27	<0.001***	2.84	2.43, 3.25	<0.001***
Vitamin B12	0.00	−0.00, 0.00	0.34	0	−0.00, 0.00	0.85	0.12	−2.79, 3.03	0.93
Folate	0.28	0.09, 0.48	0.006**	−0.06	−0.28, 0.15	0.55	−0.08	−0.28, 0.11	0.39
RBC-folate	0.02	0.01, 0.04	<0.001***	−0.01	−0.02, 0.00	0.13	−0.01	−0.03, 0.00	0.09
Vitamin C	−2.82	−6.66, 1.02	0.14	−4.19	−8.10, −0.28	0.04*	−5.66	−9.85, −1.47	0.01*
Vitamin E	0.03	0.03, 0.04	<0.001***	0.03	0.02, 0.03	<0.001***	0.06	0.06, 0.06	<0.001***
Lycopene	0.84	0.71, 0.98	<0.001***	1.1	1.00, 1.20	<0.001***	1.04	0.94, 1.14	<0.001***
Lutein	0.83	0.71, 0.95	<0.001***	0.75	0.64, 0.87	<0.001***	17.09	14.43, 19.75	<0.001***
β-cryptoxanthin	0.66	0.42, 0.90	<0.001***	0.79	0.56, 1.01	<0.001***	10.34	7.93, 12.75	<0.001***
β-carotene	0.22	0.14, 0.30	<0.001***	0.18	0.10, 0.25	<0.001***	7.82	5.60, 10.04	<0.001***
α-carotene	0.55	0.29, 0.81	<0.001***	0.46	0.19, 0.73	0.002**	4.79	3.13, 6.46	<0.001***
Selenium	0.27	0.16, 0.37	<0.001***	0.26	0.18, 0.34	<0.001***	0.31	0.25, 0.38	<0.001***
Ferritin	−0.03	−0.06, 0.01	0.14	0.02	−0.01, 0.05	0.21	0.02	−0.01, 0.05	0.17
Transferrin saturation	−0.15	−0.29, −0.02	0.03*	−0.07	−0.20, 0.06	0.30	0.13	0.01, 0.25	0.03*
Calcium	6.65	2.37, 10.92	0.004**	9.35	6.28, 12.43	<0.001***	9.98	6.69, 13.26	<0.001***
Lead	1.68	1.07, 2.29	<0.001***	0.48	−0.16, 1.12	0.13	2.47	0.21, 4.73	0.03*

CI: confidence interval; Lp(a): Lipoprotein(a); LDL-C: low-density lipoprotein cholesterol; RBC: red blood cell.

^*p<0.05;

^**p<0.01;

^***p<0.001.

^aIn the Lp(a) model, vitamin, mineral, and lead covariates transformed as follows (otherwise linear): Retinyl esters⁴, RBC-folate⁴, vitamin E⁴, ln(vitamin B12), ln(lutein), ln(β-cryptoxanthin, ln(β-carotene), ln(selenium), ln(ferritin). Interactions were present including between vitamin A and statins, vitamin A and saturated fat, retinyl esters and statins, retinyl esters and saturated fat, vitamin C and BMI, vitamin E and statin, vitamin E and hypertension, lycopene and saturated fat, α -carotene and diabetes, selenium and race/ethnicity, calcium and sex, and calcium and hypertension.

^bIn the LDL-C model, vitamin, mineral, and lead covariates were transformed as follow (otherwise linear): vitamin A², Retinyl esters³, ln(vitamin B12), vitamin C³, vitamin E⁴, ln(lutein), ln(β-cryptoxanthin, ln(β-carotene),ln(α-carotene), selenium², transferrin saturation⁴, and ln(lead). Interactions were present, including vitamin A and diabetes, vitamin A and race/ethnicity, retinyl esters and eGFR, retinyl esters and saturated fat, RBC-folate and statin, RBC-folate and race/ethnicity, vitamin C and race/ethnicity, vitamin E and hypertension, lycopene and BMI, lycopene and diabetes, lycopene and saturated fat, lycopene and eGFR, lutein and saturated fat, β-cryptoxanthin and race/ethnicity, β-cryptoxanthin and saturated fat, β-carotene and sex, β-carotene and diabetes, β-carotene and race/ethnicity, β-carotene and saturated fat, α-carotene and age, α-carotene and race/ethnicity, α-carotene and diabetes, α-carotene and saturated fat, selenium and hypertension, transferrin saturation and sex, calcium and age, calcium and saturated fat, lead and eGFR, and lead and sex.

^cEffect sizes between Lp(a) and LDL-C in Model 3 can only be compared if variables are transformed to the same linear or non-linear association with the outcome.

Multivariable analysis

In multivariable analysis, the primary analysis of interest, Lp(a) retained positive association with female sex, non-Hispanic Black race/ethnicity, and statin use, and negative association with Mexican-American race/ethnicity and saturated fat (Table 3). Lp(a) no longer had significant association with a history of hypertension or eGFR. LDL-C remained positively associated with age, history of hypertension, and higher BMI, and negatively associated with eGFR, Non-Hispanic Black race/ethnicity, and daily alcohol. LDL-C was no longer associated with sex or statin use and became significantly associated with BM and was now positively associated with smoking and saturated fat.

In multivariable analysis, Lp(a) was positively associated with carotenoids (lycopene, lutein, β-cryptoxanthin, β-carotene, and α-carotene) and lead (Figure 1) (Table 4). The directionality of associations between LDL-C and these covariates were consistent with those of the Lp(a) models, although the association with lead was not significant (p=0.13).

Figure 1.

Association of Lp(A) and LDL-C to serum vitamins and minerals and whole-blood lead in multivariable linear regression. *p<0.05.

Multivariable analysis with interactions

Interactions in the Lp(a) model included those between age and race/ethnicity, sex and race/ethnicity, race/ethnic and statin, diabetes and BMI, alcohol and smoking, and hypertension and smoking. Associations were unchanged. In the LDL-C model, additional terms included age², and interactions between age and sex, age and diabetes, age and BMI, BMI and diabetes, eGFR and saturated fat content, eGFR and smoking, race/ethnicity and alcohol, and sex and BMI. Associations were similar to prior analyses except that diabetes was now significantly associated with LDL-C and the association with smoking was no longer significant.

When covariates were considered as their best-fit log-linear or polynomials and all potential interactions were considered, prior associations were similar (Table 4). However, retinyl esters and vitamin E now had significant positive association with Lp(a), which was similar to the association with LDL-C. Lead was positively and significantly associated with both Lp(a) and LDL-C. See Table 4 footnotes for a list of all interactions.

Sensitivity tests

When we excluded samples with Lp(a)=0, association with control variables remained unchanged from prior analyses (Table E2 in ESM 1). Association with vitamins, minerals and lead were also similar when excluding samples with Lp(a)=0 and transforming Lp(a) to ln(Lp(a)+1) (Tables E3 and E4 in ESM 1).

Vitamins, minerals, and lead may have different associations when their levels are beyond normal values. When we tested vitamins, minerals, and lead as categorical variables (deficient, normal, or elevated), Lp(a) associated negatively with deficient vitamin A, retinyl esters, lutein, β-carotene and elevated selenium, which correlated similarly to LDL-C (Table 5). However, only for Lp(a) did vitamin B12 and folate associate negatively and lead associate positively (p=0.01, p=0.01, and p=0.04, respectively).

Table 5.

Associations between Vitamins, Minerals, and Heavy Metals as categorical variables with Lp(a) and LDL-C in univariate, multivariate, and multivariate with interaction term models.

		Model 1 – Univariate			Model 2 – Multivariate
	n	β	95% CI	p-value	β	95% CI	n
Lp(a)
Vitamin A deficient	64	−6.32	−13.15, 0.51	0.07	−9.21	−13.78, −4.64	<0.001***
Vitamin A elevated	36	6.93	−7.70, 21.57	0.34	2.19	−8.13, 12.50	0.67
Retinyl esters deficient	1451	−4.27	−6.12, −2.42	<0.001***	−4.74	−6.39, −3.09	<0.001***
Retinyl esters elevated	1000	−1.38	−4.40, 1.65	0.36	−1.25	−3.92, 1.41	0.34
Vitamin B12 deficient	114	−1.67	−7.36, 4.01	0.55	−0.05	−5.63, 5.53	0.99
Vitamin B12 elevated	76	−4.79	−10.83, 1.24	0.11	−5.41	−9.50, −1.31	0.01*
Folate deficient	801	0	−3.26, 3.26	0.99	−0.67	−3.66, 2.32	0.65
Folate elevated	951	−2.72	−5.01, −0.43	0.02*	−2.86	−5.09, −0.63	0.01*
RBC-folate deficient	1184	1.54	−0.26, 3.33	0.09	−1.52	−3.33, 0.29	0.10
RBC-folate elevated	323	−0.34	−6.10, 5.41	0.90	−0.55	−5.78, 4.67	0.83
Vitamin C deficient	125	3.49	−2.28, 9.27	0.22	2.74	−3.10, 8.59	0.34
Vitamin C elevated	263	−0.11	−5.16, 4.94	0.97	0.47	−4.51, 5.45	0.85
Vitamin E deficient	54	−4.25	−16.87, 8.36	0.49	−5.39	−16.90, 6.12	0.34
Vitamin E elevated	116	−0.63	−7.74, 6.48	0.86	0.26	−6.46, 6.99	0.94
Lycopene deficient	56	0.71	−8.24, 9.66	0.87	1.27	−7.54, 10.08	0.77
Lycopene elevated	55	2.47	−20.37, 25.30	0.83	3.08	−18.04, 24.21	0.77
Lutein deficient	10	−9.43	−19.16, 0.30	0.06	−10.7	−18.37, −3.03	0.008**
Lutein elevated	53	−1.61	−9.55, 6.33	0.68	−6.14	−14.46, 2.18	0.14
β-cryptoxanthin deficient	53	−8.07	−17.79, 1.65	0.10	−6.73	−15.80, 2.35	0.14
β-cryptoxanthin elevated	84	1.28	−7.51, 10.06	0.77	1.58	−6.61, 9.77	0.69
β-carotene deficient	47	−6.06	−17.51, 5.39	0.29	−9.42	−16.13, −2.71	0.008**
β-carotene elevated	141	3.86	−0.82, 8.53	0.10	2.84	−2.41, 8.09	0.28
α-carotene deficient	94	−2.63	−9.46, 4.19	0.43	−2.57	−8.76, 3.62	0.40
α-carotene elevated	231	3.6	−2.22, 9.41	0.21	2.78	−3.50, 9.06	0.37
Selenium elevated	15	−12.49	−19.36, −5.63	0.001**	−9.18	−15.44, −2.91	0.006**
Iron deficient	245	3.52	−1.56, 8.60	0.17	−0.04	−4.22, 4.14	0.98
Transferrin saturation low	1405	3.87	0.92, 6.82	0.01**	1.4	−1.19, 3.99	0.28
Calcium low	2054	1.06	−1.39, 3.52	0.38	0.83	−1.26, 2.91	0.42
Calcium elevated	10	−1.33	−18.83, 16.17	0.88	1.68	−14.01, 17.38	0.83
Lead elevated	1457	4.19	0.73, 7.65	0.02*	3.17	0.14, 6.20	0.04*
LDL-C
Vitamin A deficient	64	−38.55	−48.54, −28.56	<0.001***	−39.89	−53.90, −25.88	<0.001***
Vitamin A elevated	36	18.33	−1.55, 38.22	0.07	4.37	−12.37, 21.10	0.59
Retinyl esters deficient	1451	−10.86	−14.26, −7.46	<0.001***	−11.6	−14.65, −8.55	<0.001***
Retinyl esters elevated	1000	15.62	9.96, 21.29	<0.001***	12.21	6.38, 18.04	<0.001***
Vitamin B12 deficient	114	3.81	−3.58, 11.21	0.30	−1.66	−8.51, 5.19	0.62
Vitamin B12 elevated	76	−7.09	−21.54, 7.36	0.32	−7.76	−20.51, 4.99	0.22
Folate deficient	801	−3.21	−6.14, −0.29	0.03*	−0.43	−3.53, 2.67	0.78
Folate elevated	951	4.35	0.02, 8.68	0.049*	−0.27	−4.80, 4.26	0.90
RBC-folate deficient	1184	−2.46	−6.14, 1.22	0.18	2.21	−1.71, 6.14	0.26
RBC-folate elevated	323	9.21	2.42, 16.01	0.01*	−1.25	−8.80, 6.30	0.74
Vitamin C deficient	125	−4.31	−15.02, 6.40	0.41	−4.68	−14.14, 4.78	0.32
Vitamin C elevated	263	2.11	−6.06, 10.27	0.60	−0.03	−6.04, 5.97	0.99
Vitamin E deficient	54	−64.7	−80.22, −49.18	<0.001***	−56.52	−73.40, −39.64	<0.001***
Vitamin E elevated	116	29.4	16.21, 42.59	<0.001***	18.97	4.65, 33.29	0.01*
Lycopene deficient	56	−8.38	−25.37, 8.61	0.32	−19.99	−36.99, −2.99	0.02*
Lycopene elevated	55	51.25	40.46, 62.04	<0.001***	52.22	39.16, 65.27	<0.001***
Lutein deficient	10	−38.72	−48.74, −28.71	<0.001***	−38.01	−49.28, −26.73	<0.001***
Lutein elevated	53	38.73	23.35, 54.11	<0.001***	30.78	14.68, 46.89	0.001**
β-cryptoxanthin deficient	53	−12.98	−19.99, −5.98	0.001**	−15.42	−24.68, −6.15	<0.001***
β-cryptoxanthin elevated	84	12.96	0.04, 25.89	0.049*	19.23	8.27, 30.18	0.002**
β-carotene deficient	47	−37.05	−47.13, −26.98	<0.001***	−33.72	−46.24, −21.20	<0.001***
β-carotene elevated	141	5.95	−8.41, 20.32	0.40	3.35	−8.28, 14.97	0.56
α-carotene deficient	94	−16.53	−25.56, −7.49	0.001**	−11.83	−22.28, −1.37	0.03*
α-carotene elevated	231	3.18	−6.47, 12.83	0.50	3.49	−5.10, 12.07	0.41
Selenium elevated	15	−22.74	−38.56, −6.92	0.007**	−20.83	−33.57, −8.09	0.003**
Iron deficient	245	−13.9	−20.94, −6.87	<0.001***	−10.4	−16.46, −4.35	0.002**
Transferrin saturation low	1405	−6.32	−10.24, −2.40	0.003**	−5.55	−10.05, −1.05	0.02*
Calcium low	2054	−5.14	−9.51, −0.77	0.02*	−6.65	−10.21, −3.09	0.001**
Calcium elevated	10	−7.12	−31.75, 17.51	0.56	−3.22	−8.44, 1.99	0.21
Lead elevated	1457	6.24	2.60, 9.89	0.002**	−0.39	−4.72, 3.94	0.85

CI: confidence interval; Lp(a): Lipoprotein(a); LDL-C: low-density lipoprotein cholesterol; RBC: red blood cell.

^*p<0.05;

^**p<0.01;

^***p<0.001.

With Lp(a) as quartiles, a one-unit increase in lycopene, β-carotene, and α-carotene was significantly associated with lower odds of being in the first quartile rather than the second quartile of Lp(a) (Table E5 in ESM 1). In contrast, a one-unit increase in vitamin E, lycopene, β-cryptoxanthin, and β-carotene were significantly associated with higher odds of being in the fourth-quartile rather than the second-quartile of Lp(a). LDL-C had similar associations.

Discussion

We found that Lp(a) positively associated with several serum factors, including carotenoids and lead. Several sensitivity analyses confirmed these associations. However, there were important differences between LDL-C and Lp(a). In analyses that tested associations of covariates as categorical variables, (deficient, normal, or elevated), only Lp(a) but not LDL-C inversely correlated with elevated serum vitamin B12 and folate, and positively with elevated serum lead levels. Our analyses do not inform on the causality of these factors for higher or lower Lp(a) levels. Instead, these findings raise new questions about the relationship between Lp(a) levels and vitamins, minerals, and lead. Additional studies are needed to test the hypotheses driven by our findings.

Our data support several prior associations of Lp(a), including higher levels with among females, non-Hispanic Blacks, and those on statins and lower levels with higher dietary saturated fat [7, 27, 33, 34]. Although prior smaller studies had been inconclusive to whether Lp(a) associated with sex, more recent studies have confirmed that Lp(a) tends to be slightly higher among females [7, 34]. Lp(a) levels are well-known to vary by race/ethnicity. Several prior studies have found non-Hispanic Blacks to have the highest levels of Lp(a), followed by non-Hispanic whites and Mexican-Americans/Hispanics, respectively [35–37]. Among our control variables, we found significant interactions between sex, race/ethnicity, statin use, diabetes, BMI, alcohol, and smoking in their relationship to Lp(a). The influence of physiologic states (e.g. diabetes, menopause, and obesity) on Lp(a) is incompletely understood. Taken together, our findings lend credence to the concept that there are likely nongenetic factors that influence Lp(a), but require further study to better delineate these relationships.

In our study, we tested associations between both fat- and water-soluble vitamins, minerals, and lead to Lp(a). Lp(a) differs from LDL-C only by the addition of apo(a). Thus, a priori one could predict that, because Lp(a) is similar to LDL-C, associations would be present for primarily fat-soluble molecules known to be carried by LDL-C. Carotenoids and lead are also fat-soluble and have been previously observed to be carried or associated with LDL-C [38–41]. Notably, carotenoids may be exclusively carried by lipoproteins [42]. However, this is the first time carotenoids and lead were documented to associate with Lp(a). In one study of 50 subjects randomized to 20 mg of β-carotene, Lp(a) levels did not change despite a 15-fold increase in serum β-carotene levels [43]. Although there may not be a direct effect on serum Lp(a) levels, in vitro studies suggest that β-carotene may prevent Lp(a) and LDL-C oxidation and macrophage uptake [44, 45]. Dietary carotenoids are associated with reduced cardiovascular mortality [46–48], thus additional research understanding whether carotenoids influence the risk between Lp(a) and CVD is warranted.

We did not find evidence to confirm prior associations of vitamin C with Lp(a) levels, although did associate with LDL-C. In only a few models did vitamin E associate with Lp(a) levels, although with a very small effect size and positively instead of previous research finding negative association [20].

Differences from the primary analyses were observed for levels of vitamins, minerals, and lead were above the normal range when considered as categorical variables. Elevated vitamin B12 and folate only associated negatively, while lead associated positively with Lp(a) but not LDL-C. Elevated levels of vitamin B12, folate, and lead may have unique regulatory actions on Lp(a) compared to LDL-C that are not captured in linear regression models. However, considering groups with deficient or elevated levels were often small, potentials for unmeasured confounders should be considered. These observations require further validation by analysis of other datasets and other methodologies and do not suggest causality. Notably, supplementation with folate, vitamin B6, and vitamin B12 among a cohort of dialysis patients decreased Lp(a) levels while LDL-C remained unchanged [49].

We also found Lp(a) to associate more often than LDL-C to lead in this analysis. Lead has been associated with higher LDL-C, but never studied in association with Lp(a) [41]. In our study, Lp(a) and LDL-C associated with lead in most cases. In the cases where LDL-C did not associate, the p-values were only slightly above the .05 threshold for statistical significance. Taken together we interpret our findings to suggest that both Lp(a) and LDL-C associate positively with lead. However, the connection is more consistent with Lp(a). Notably, there is a possible explanatory link between lead to Lp(a) and LDL-C. Lead was found to reduce apolipoprotein E in children [50], a protein that influences Lp(a) and LDL-C levels [51].

Our data does not inform on the mechanism of interaction between Lp(a) and different plasma factors. Although further studies are needed to test this relationship, our findings raise awareness for potential nongenetic factors. Lead is atherosclerotic and acts via multiple mechanisms, including endothelial dysfunction and increased free radicals [41]. It is associated with negative health outcomes, including all-cause mortality, cardiovascular mortality, and cardiovascular disease [41]. Our finding that lead associates with elevated Lp(a) offers an additional mechanism that lead might exert negative health consequences. However, our findings require confirmation with additional studies.

Although, Lp(a) levels are genetically determined, they are likely influenced by gene-environmental interactions. It is noteworthy, that little is known about environmental influence on Lp(a) levels. This is particularly important since there is currently no therapy available to reduce concentration of Lp(a) levels and improves outcomes. Niacin, which is a vitamin, has been shown to reduce Lp(a), but with no improvement in outcome even amongst those with highest Lp(a) at baseline [52]. PCSK9 inhibitors have been shown to modestly reduce Lp(a) and may improve clinical outcome. Major hope lies in the application of antisense oligonucleotides [53]. Although phase 2 trials have shown safety of these drugs [54], there will likely be individuals who will not tolerate, cannot afford, or show inadequate response to the drug. Therefore, identifying additional plasma or environmental factors that can be altered to reduce Lp(a) levels are of great importance. Our findings act as a first phase for these findings and suggest that some factors, particularly carotenoids, could be considered as factors to further test for causality beyond the observational associations in our study.

Strengths and limitations

Strengths to our study include that LDL-C was calculated using the Hopkins-Martin formula and corrected for Lp (a). Furthermore, we repeated analyses in several sensitivity tests that confirm findings from the primary analysis. We did a broad search for associating factors and considered all fat- or water-soluble factors since prior data had suggested that vitamin C, a water-soluble molecule, might alter apo(a) expression [21].

Our study has several limitations. First, these data are from 1991–1994, the only iteration of NHANES that includes Lp(a). Estimations could differ at the present time due to shifts in population demographics. However, it is unlikely that associations of Lp(a) with variables have changed over time. Rather, what could have changed in the interim are the prevalence of covariates (e.g. obesity, diabetes, hypertension, and lipid level). In addition, more modern assays for Lp(a) and other factors may provide higher accuracy. These data lack information on lipoprotein particle size and Lp(a) molar concentration, which might be important in analyzing associations. However, no more current data were available from a nationally representative sample. These limitations emphasize the need to test Lp (a) in future iterations of NHANES, which should be performed with an isoform independent assay. There were also significant missing datapoints and some covariates that represented a small proportion of the sample which may have limited the potential to detect some differences. Nevertheless, these data are the only cohort of the general US population to measure Lp(a).

Conclusions

Our results suggest that carotenoids and lead are positively associated with both serum Lp(a) and LDL-C levels. This is the first time that lead has been tested for associated with Lp (a). When considered as categorical variables, vitamin B12, folate, and lead only associated with Lp(a) but not LDL-C. These observations require further investigation of other cohorts and with other methodologies to confirm our findings and test for causality of these associations.