Race-ethnic, family income, and education differentials in nutritional and lipid biomarkers in US children and adolescents: NHANES 2003–2006^,

Abstract

Background: Children from ethnic minority and low-income families in the United States have higher rates of poor health and higher mortality rates. Diet, an acknowledged correlate of health, may mediate the known race-ethnic and socioeconomic differentials in the health of US children.

Objective: The objective was to examine the independent association of race-ethnicity, family income, and education with nutritional and lipid biomarkers in US children.

Design: We used data from the NHANES 2003–2006 to examine serum concentrations of vitamins A, D, E, C, B-6, and B-12; serum concentrations of folate, carotenoids, and lipids; and dietary intakes of corresponding nutrients for 2–19-y-old children (n = ∼2700–7500). Multiple covariate–adjusted regression methods were used to examine the independent and joint associations of race-ethnicity, family income, and education with biomarker status.

Results: Non-Hispanic blacks had lower mean serum concentrations of vitamins A, B-6, and E and α-carotene than did non-Hispanic whites. Both non-Hispanic blacks and Mexican Americans had higher mean serum vitamin C, β-cryptoxanthin, and lutein + zeaxanthin but lower folate and vitamin D concentrations compared with non-Hispanic whites. In comparison with non-Hispanic whites, non-Hispanic blacks were less likely to have low serum HDL cholesterol or high triglycerides. Family income and education predicted few biomarker or dietary outcomes, and the observed associations were weak. Moreover, modification of race-ethnic differentials by income or education (or vice versa) was noted for very few biomarkers.

Conclusion: Race-ethnicity, but not family income or education, was a strong independent predictor of serum nutrient concentrations and dietary micronutrient intakes in US children and adolescents.

INTRODUCTION

Children from ethnic minority and low-income families in the United States have higher rates of poor health and higher mortality rates (1–4). The prevalence of acknowledged risk factors for disease (higher adiposity, total cholesterol, and glycated hemoglobin concentrations) also shows variation by ethnicity and socioeconomic position in US children (5–10). Ethnic and socioeconomic disparities in the health of US children and adolescents reflect a confluence of cultural norms, family resource constraints, risk environments, and gaps in knowledge of desirable health behaviors. Diet, an intervenable risk factor for a number of chronic diseases, may be one of the mediators of the ethnic and socioeconomic differences in health outcomes. Arguably, dietary intakes of key protective and risk-increasing nutrients among children may also be expected to vary with ethnicity and family socioeconomic position. Yet, few studies have systematically examined the independent association of ethnicity, income, and education with nutrient profiles in US children. An additional problem in examination of these associations is the acknowledged bias in assessment of dietary intakes of children and adolescents (11, 12). Nutritional biomarkers, however, are not subject to respondent errors and can reflect longer-term dietary exposures, micronutrient bioavailability, and in vivo utilization (13, 14); therefore, biomarkers may be particularly valuable in examining possible ethnic and socioeconomic disparities in dietary exposures in children.

A sizable body of evidence has contributed to our understanding of race-ethnic differences in micronutrient status in adults (15–27). For children, race-ethnic differences in prevalence of marginal vitamins A and D and iron nutritional status have received the greatest attention (28–36). However, there is a surprising paucity of information on whether other nutritional biomarkers in children also show ethnic differentials. Moreover, in most studies that examined these associations in children, few have examined the independent associations of race-ethnicity, family income, and education. Because race-ethnic group membership, family income, and education can potentially relate to health (and diet) by several interrelated and unique pathways, it is important to understand their independent contributions for design of interventions. To fill these gaps, we used nationally representative data to examine the independent associations of race-ethnicity, family income, and family education with available nutritional and lipid biomarkers in US children and to compare the associations of race-ethnicity, income, and education with biomarkers to those observed for dietary intakes.

SUBJECTS AND METHODS

We used public domain data from the continuous NHANES 2003–2004 and 2005–2006 for this study (37–39). The study was approved by the Queens College Institutional Review Board with an exempt review. The continuous NHANES is conducted annually by the National Center for Health Statistics (NCHS)⁵, CDC, and data are released for 2 y at a time (40). Each NHANES is a stratified, multistage, cluster-probability sample of the US population and includes an at-home interview and an examination in a mobile examination center (MEC). The MEC examination includes a complete medical examination, anthropometric measurements, an interview to collect dietary information, and collection of blood and urine samples with the use of standardized procedures. The 2003–2004 and 2005–2006 surveys oversampled low-income persons, adolescents, blacks, and Mexican Americans. In each survey, the unweighted response rates for the MEC-examined sample of 2–19-y-olds were >80% (40).

Biomarkers

We examined serum concentrations of water-soluble vitamins [vitamin C, vitamin B-6 (as pyridoxal phosphate), serum and red blood cell (RBC) folate, and vitamin B-12], fat-soluble vitamins and carotenoids [vitamin A (as retinol), vitamin D (as 25-hydroxycholecalciferol), vitamin E (as α-tocopherol), α-carotene, β-carotene, β-cryptoxanthin, lutein + zeaxanthin, and trans lycopene], and serum lipids (total cholesterol, HDL cholesterol, LDL cholesterol, and triglycerides) for this study. In both survey cycles (2003–2004 and 2005–2006), serum folate and vitamins D and B-12 were available for 2–19-y-olds; carotenoids, total or HDL cholesterol, and vitamins A, C, and E were available for 6–19-y-olds; LDL cholesterol and triglycerides were available for 12–19-y-olds. For vitamin B-6, we followed the NCHS advisory and limited our analysis to data from the 2005–2006 survey for 2–19-y-olds (the NCHS advisory cautions against combining the data for 2003–2004 with 2005–2006 because of methodologic differences) (39). Although serum lipids and serum and RBC folate were available for the NHANES 2007–2008, because of changes in laboratory method, site, or equipment from the earlier surveys, we limited our analysis to the 2003–2004 and 2005–2006 surveys.

All procedures for phlebotomy, subsequent handling of blood samples, analysis of samples, and reporting procedures followed a standardized protocol specified by the NHANES (38, 39). The assay methods used for measurement of serum analytes are detailed in the NHANES documentation and included the following: radioimmunoassay for vitamin D, radioassay for serum/RBC folate and serum vitamin B-12, direct immunoassay for HDL cholesterol, and enzymatic methods for serum total cholesterol and triglycerides, reversed-phase HPLC with fluorometric detection for vitamin B-6, HPLC with electrochemical detection for vitamin C, 2-step Diasorin procedure (Diasorin Corp) for vitamin D, and HPLC with photodiode array detection for vitamins A and E and the carotenoids. Total LDL cholesterol was calculated by using the Friedewald equation (38, 39).

Race-ethnicity and measures of socioeconomic position

In the US population, race-ethnicity tends to be related with socioeconomic position (SEP) (41–44). However, each SEP indicator (eg, income and education) also relates to health independent of race-ethnicity, and it has been suggested that investigation of race-ethnic health disparities should include ≥2 different measures of SEP (42). Thus, we included family income and education of head of household as indicators of family SEP. In the continuous NHANES, race-ethnicity is self-reported. The NCHS combines this information and provides adjudicated race-ethnicity categories as non-Hispanic white, non-Hispanic black, Mexican American, other Hispanic, and all others. For the 1999–2002 and 2003–2006 survey periods, the NHANES analytic guidelines do not recommend estimates for all Hispanics or non–Mexican American Hispanics because of inadequate sample size for non–Mexican American Hispanics and oversampling of Mexican Americans (45). For our analysis, race-ethnicity was categorized as non-Hispanic white, non-Hispanic black, and Mexican American. We considered family income assessed by family poverty-income ratio (PIR) and educational level of the head of household as indicators of family SEP. PIR is computed by the NCHS and is a ratio of family income to the poverty threshold established by the US Census Bureau. The poverty threshold varies with family size and is inflation adjusted; a PIR <1 is considered below poverty. We operationalized the PIR as <130%, 130–349%, and ≥350% for 2 reasons: 1) a PIR of <130% is used as the cutoff for several federal food assistance programs and 2) these cutoffs are often used by the NCHS for reporting of NHANES dietary and health data. The educational level of the family head was operationalized as <12 y, 12 y, some college, and college degree or higher.

Analytic sample

All MEC-examined respondents aged 2–19 y in the NHANES 2003–2004 and 2005–2006, with at least one measured biomarker of interest, were eligible for inclusion in the study (n = 7527). We excluded pregnant and lactating respondents (n = 115), those who were missing family PIR (n = 402) or family education information (n = 253), those with unknown supplement use status in the past month (n = 8) or who reported supplement use since fasting time (n = 12), those with other race-ethnicity (n = 658), and those who were missing information on chronic disease status (n = 16) or minutes of fasting before phlebotomy (n = 1) from the eligible sample for a final analytic sample of 6062. Some biomarkers were measured for ages ≥6 y; the final analytic sample for 6–19 y olds was 4908. Serum LDL cholesterol and triglycerides were available for 1502 respondents aged ≥12 y.

Dietary nutrient intake

We also examined 24-h dietary intake of energy, fat, and the micronutrient corresponding to the respective biomarker. The NHANES collected a 24-h dietary recall by using the USDA's Automated Multiple-Pass Method, administered by a trained dietary interviewer, during the MEC interview (38, 39). Unlike the NHANES 2007–2008, estimates of 24-h total nutrient intake (diet + supplements) are not available for the 2003–2006 survey years. The supplement use information in the NHANES 2003–2004 and 2005–2006 was obtained in the household interview, and the NCHS does not provide estimates of 24-h supplemental nutrient intake for these survey years (38, 39) because of issues related to the format of supplement data and where and how they were collected.

Covariates

To determine whether race-ethnicity, family income, and education were independently associated with serum biomarker concentrations, we accounted for the effect of other potential confounders of these associations. These included sex, vitamin/mineral supplement use over the previous 30 d, self-reported disease condition (asthma, diabetes, or high blood pressure), nicotine exposure (assessed as serum cotinine concentration), month of MEC examination, BMI percentile for age and sex, and inflammatory condition (assessed as serum C-reactive protein concentration). Additional covariates were as follows: for vitamin D and serum lipids, hours of television and computer use; for vitamin E and carotenoids, serum total cholesterol. For dietary outcomes, the covariates included age, sex, month of MEC examination, day of recalled intake, self-reported history of chronic disease, and BMI percentile for age and sex.

Statistical methods

The independent associations of race-ethnicity, family income, and education with biomarker and dietary outcomes were examined by using multiple regression methods with biomarker concentration or dietary intake as a continuous outcome. Because the age interval of 2–19 y or even 6–19 y represents a range of developmental stages and differences in dietary autonomy, and possible interactions of race-ethnicity with family income and education, we began with models that included all covariates and 5 interaction terms (age by race, age by PIR, age by education, race by PIR, and race by education) to examine whether the associations of race-ethnicity, income, and education differed by age and SEP categories. In a backward stepwise multiple linear regression, we subsequently deleted the interaction term with the least significance and reexamined the significance of the remaining interaction term or terms. We used P < 0.01 as the criterion for determining the significance of the interaction terms to adjust for the number of steps in the elimination procedure (0.05/5 interaction terms for each biomarker = 0.01). If no significant interactions were observed, we present the main effects of race-ethnicity, PIR, and education; otherwise, we present findings that reflect the relevant interactions. All examined P values are based on a Satterthwaite adjusted F test statistic that takes account of the complex sample design and sample weighting and makes a correction for the df (46).

We also examined the independent association of race-ethnicity, income, and education with prevalence of serum concentration of selected micronutrients and lipids relative to a threshold indicating at-risk status by using multiple logistic regression models. The cutoffs used were based on previous NCHS publications (20, 47), and recommendations of expert committees (48–50) and were as follows: vitamin C (<0.2 mg/dL), vitamin B-6 (<20 nmol/L), serum folate (<2 ng/mL), RBC folate (<95 ng/mL), vitamin B-12 (<200 pg/mL), retinol (<20 μg/dL), vitamin D (<20 ng/mL), vitamin E (<500 μg/dL), total cholesterol (≥200 mg/dL), HDL cholesterol (≤40 mg/dL), LDL cholesterol (≥130 mg/dL), and triglycerides (≥130 mg/dL).

We log-transformed each reported biomarker concentration for multiple linear regression analyses. The estimates presented in the tables and figures were adjusted for multiple covariates and back-transformed to obtain geometric means and their 95% CIs. All analyses were performed by using SAS version 9.2 (SAS Institute) and SAS callable SUDAAN release 10.0.1 (Research Triangle Institute) software for analysis of data from multistage stratified cluster survey designs. The analyses included the appropriate sample weights to account for unequal probability of sampling and sample nonresponse (51). For serum LDL cholesterol and triglycerides, we followed the NCHS recommendations and limited our analysis to the morning-examined subsample with >8.5 h of fasting before phlebotomy (37, 38); separate weights are provided by the NCHS for this subsample. The tabular and graphic results are adjusted means (sometimes called predicted margins) derived from multiple linear regression and multiple logistic regression models (52).

RESULTS

Non-Hispanic whites had higher family PIR and education than did non-Hispanic blacks and Mexican Americans (Table 1). Supplement use over the month preceding the MEC examination was more frequent among non-Hispanic white children and adolescents. Self-report of asthma, diabetes, or high blood pressure was more frequent among non-Hispanic black children and adolescents. Mean serum cotinine was higher in non-Hispanic blacks and mean serum C-reactive protein was higher in Mexican Americans relative to other race-ethnic groups (Table 1).

TABLE 1.

Sociodemographic, lifestyle, and other characteristics of 2–19-y-old respondents with measured serum vitamin and lipid concentrations: NHANES 2003–2006¹

	n	Non-Hispanic white (n = 1792)	Non-Hispanic black (n = 2195)	Mexican American (n = 2075)
Age group (%)
2–5 y	1154	17.3 ± 1.12	18.3 ± 1.0	23.2 ± 0.9
6–11 y	1529	31.8 ± 1.6	33.0 ± 1.4	36.1 ± 0.9
12–19 y	3379	50.9 ± 1.9	48.7 ± 1.8	40.7 ± 1.1
Sex (%)
Male	3078	52.6 ± 1.4	51.1 ± 1.4	51.7 ± 1.1
Female	2984	47.4 ± 1.4	48.9 ± 1.4	48.3 ± 1.1
Poverty-income ratio (%)
<130%	2663	20.8 ± 2.3	51.7 ± 2.5	50.6 ± 2.1
130–349%	2186	38.8 ± 2.1	34.9 ± 1.7	37.4 ± 1.6
>349%	1213	40.3 ± 2.5	13.5 ± 1.7	12.0 ± 1.4
Household head educational level (%)
<12 y	2029	9.5 ± 1.2	31.1 ± 1.8	53.1 ± 2.6
12 y	1498	27.3 ± 2.3	25.6 ± 2.1	21.6 ± 1,8
Some college	1711	38.0 ± 1.8	30.8 ± 1.5	18.7 ± 1.8
College degree or higher	824	25.2 ± 2.3	12.5 ± 1.6	6.6 ± 1.1
Used dietary supplement in past 30 d (%)	1468	37.5 ± 1.8	17.5 ± 1.3	24.1 ± 2.1
Self-reported diabetes, asthma, or high blood pressure (%)	1073	16.9 ± 0.9	21.5 ± 1.0	12.7 ± 0.7
Month of MEC examination (%)
November to April	3166	35.5 ± 4.5	48.5 ± 8.3	76.5 ± 6.1
May to October	2896	64.5 ± 4.5	51.5 ± 8.3	23.5 ± 6.1
BMI for sex-age percentile (%)
<5th	178	3.2 ± 0.5	3.1 ± 0.4	3.0 ± 0.4
5th to <85th	3676	65.4 ± 1.9	60.5 ± 1.4	58.4 ± 1.3
85th to <95th	982	16.5 ± 1.0	15.4 ± 1.0	17.8 ± 0.6
≥95th	1145	14.9 ± 1.4	21.0 ± 1.1	20.8 ± 1.4
Hours of fasting before phlebotomy3	6062	6.9 ± 0.2	7.8 ± 0.2	7.7 ± 0.2
Serum cotinine (ng/mL)	5471	0.18 (0.13, 0.25)4	0.30 (0.23, 0.37)	0.06 (0.04, 0.07)
Serum CRP (mg/dL)	5889	0.043 (0.040, 0.046)	0.050 (0.046, 0.054)	0.062 (0.057, 0.067)

¹CRP, C-reactive protein; MEC, mobile examination center.

²Weighted percentage ± SE (all such values) unless otherwise indicated.

³Values are means ± SEs.

⁴Geometric mean; 95% CI in parentheses (all such values).

Serum biomarker concentration

Interactions of race-ethnicity, family income, and education for predicting mean biomarker concentration

There was an interaction between age and race-ethnicity for vitamins D and A (P ≤ 0.0006), an interaction of race with PIR for vitamin A (P = 0.004), and an interaction of race with education for lutein/zeaxanthin and β-cryptoxanthin (P ≤ 0.005) (Figure 1). Mean serum vitamin D concentration was higher in non-Hispanic whites than in non-Hispanic blacks and Mexican Americans at every age (P < 0.0001). Serum vitamin D concentrations declined with age in all race-ethnic groups (P < 0.0001); however, the declines were steeper in the minority groups than in non-Hispanic whites.

FIGURE 1.

Interactions of race-ethnicity with age [serum vitamin D (A) and retinol (B)], poverty-income ratio (serum retinol; C), and education [serum lutein + zeaxanthin (D) and β-cryptoxanthin (E)] in US children and adolescents, P < 0.01 (Satterthwaite adjusted F test): NHANES 2003–2006. The estimates are adjusted means (and 95% CIs) from multiple linear regression models with each biomarker as a log-transformed dependent variable and included age (2–19 y), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, Mexican American), family poverty-income ratio (<130%, 130–349%, >349%), educational level of household head (<12 y, 12 y, some college, college degree or higher), month of mobile examination center examination (November–April, May–October), self-reported history of chronic disease (yes or no), hours of fasting (continuous), history of supplement use in past month (yes or no), BMI percentile for age and sex (continuous), serum cotinine, C-reactive protein, serum total cholesterol (for carotenoids), and hours of television and computer use (for vitamin D).

Mean serum retinol concentration was higher at age 12–19 y in all race-ethnic groups (P < 0.0001); however, the race-ethnic differences in serum retinol concentration were significant only in 12–19-y-olds (P = 0.0001). There were race-ethnic differences in serum retinol concentration in all PIR categories (P ≤ 0.0001).

Mean concentrations of lutein + zeaxanthin and β-cryptoxanthin differed by race-ethnicity at all categories of education (P < 0.0001); however, the education effect was positive and greater in non-Hispanic whites only (P ≤ 0.05). Among Mexican Americans, higher education predicted lower mean serum β-cryptoxanthin concentration (P = 0.02).

Race-ethnic differences in mean biomarker concentrations

With the exception of β-carotene and total and LDL cholesterol, race-ethnicity was a significant independent predictor of all examined biomarker outcomes (Table 2).

TABLE 2.

Multiple variable–adjusted geometric means and 95% CIs of biomarkers by race-ethnicity in US children and adolescents: NHANES 2003–2006¹

Serum biomarker	n	Age	Non-Hispanic white	Non-Hispanic black	Mexican American	P (race-ethnicity)2
		y
Water-soluble vitamins
Vitamin C (mg/dL)	4684	6–19	0.96a (0.91, 1.01)	1.16c (1.12, 1.21)	1.07b (1.01, 1.14)	<0.0001
Pyridoxal phosphate (nmol/L)3	2672	2–19	56.1b (53.8, 58.6)	49.3a (46.4, 52.4)	56.3b (51.0, 62.1)	0.01
Serum folate (ng/mL)	5388	2–19	13.8b (13.4, 14.3)	12.6a (12.2, 13.0)	13.0a (12.5, 13.5)	<0.0001
Red blood cell folate (ng/mL)	5389	2–19	257c (251, 263)	214a (210, 218)	242b (235, 250)	<0.0001
Vitamin B-12 (pg/mL)	5382	2–19	594a (579, 610)	704b (686, 723)	600a (580, 620)	<0.0001
Fat-soluble vitamins and carotenoids (μg/dL)
α-Carotene	4635	6–19	1.95b (1.80, 2.10)	1.63a (1.47, 1.81)	2.22c (2.02, 2.44)	0.0001
β-Carotene	4427	6–19	10.7 (10.2, 11.3)	10.6 (9.9, 11.3)	10.9 (10.2, 11.5)	0.8
trans Lycopene	4637	6–19	21.0b (20.4, 21.7)	23.7c (22.8, 24.8)	19.5a (18.8, 20.1)	<0.0001
Vitamin E	4637	6–19	803b (790, 815)	770a (759, 781)	797b (781, 814)	0.0001
Serum lipids (mg/dL)
Total cholesterol	4654	6–19	160 (158, 162)	160 (158, 162)	158 (156, 161)	0.6
Total cholesterol (morning subsample)4	1441	12–19	156 (153, 160)	156 (153, 160)	157 (152, 162)	1.0
HDL cholesterol	4654	6–19	51a (50, 52)	57b (56, 58)	51a (50, 53)	<0.0001
HDL cholesterol (morning subsample)4	1441	12–19	50a (48, 51)	56c (54, 57)	52b (50, 54)	<0.0001
LDL cholesterol (morning subsample)4	1437	12–19	85 (83, 88)	85 (82, 89)	84 (80, 88)	0.8
Triglycerides (morning subsample)4	1441	12–19	84a (79, 89)	61b (58, 64)	84a (80, 89)	<0.0001

¹The adjusted values are back-transformed predicted margins from multiple linear regression models with each biomarker as a log-transformed dependent variable and included age (2–5, 6–11, 12–19 y), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, Mexican American), family poverty-income ratio (<130%, 130–349%, >349%), educational level of household head (<12 y, 12 y, some college, college degree or higher), month of mobile examination center examination (November–April, May–October), self-reported history of chronic disease (yes or no), hours of fasting (continuous), history of supplement use in the past month (yes or no), BMI percentile for age and sex (continuous), serum cotinine, and C-reactive protein. Models for carotenoids and vitamin E also included serum total cholesterol, and models for vitamin D and serum lipids included hours of television and computer use. Adjusted values in each row not sharing a common superscript letter are significantly different, P < 0.05 (Satterthwaite adjusted F test).

²Satterthwaite adjusted F test for differences between race-ethnic groups.

³NHANES 2005–2006.

⁴Limited to the subsample with morning examination and >8.5 h of fasting.

Water-soluble vitamins

Serum vitamins C and B-12 concentrations were higher in non-Hispanic black children than in non-Hispanic whites and Mexican Americans (P < 0.0001). However, serum and RBC folate concentrations were lower in non-Hispanic black and Mexican American children compared with non-Hispanic whites (P < 0.01). The mean serum vitamin B-6 concentration in non-Hispanic black children was also lower than that in non-Hispanic whites and Mexican Americans.

Fat-soluble vitamins and lipids

Serum concentrations of α-carotene were lower and those of trans lycopene were higher in non-Hispanic blacks than in non-Hispanic whites and Mexican Americans (P < 0.0001). Serum vitamin E concentrations in non-Hispanic blacks were lower than in non-Hispanic whites. Non-Hispanic blacks had higher mean HDL cholesterol and significantly lower mean triglyceride concentrations compared with the other 2 race-ethnic groups (P < 0.0001).

Family PIR and education differences in mean biomarker concentrations

Serum vitamin D concentrations increased with increasing PIR (P = 0.01). For all other biomarkers, PIR was not a significant independent predictor (Table 3). RBC folate concentration differed among categories of family education; however, the 95% CIs overlapped in most cases (Table 4). Serum α-carotene concentration increased with increasing years of education (P = 0.009).

TABLE 3.

Multiple variable–adjusted geometric means and 95% CIs of biomarkers by categories of PIR in US children and adolescents: NHANES 2003–2006¹

			PIR
Serum biomarker	n	Age	<130%	130–349%	>349%	P (PIR)2
		y
Water-soluble vitamins
Vitamin C (mg/dL)	4684	6–19	1.02 (0.97, 1.08)	0.99 (0.94, 1.04)	1.01 (0.96, 1.07)	0.6
Pyridoxal phosphate (nmol/L)	2672	2–19	52.5 (49.7, 55.6)	56.5 (52.9, 60.3)	55.2 (52.2, 58.7)	0.2
Serum folate (ng/mL)	5388	2–19	13.9 (13.4, 14.5)	13.7 (13.2, 14.2)	13.0 (12.4, 13.6)	0.1
RBC folate (ng/mL)	5389	2–19	245 (239, 252)	250 (245, 256)	244 (238, 251)	0.1
Vitamin B-12 (pg/mL)	5382	2–19	622 (596, 649)	614 (597, 633)	602 (585, 619)	0.4
Fat-soluble vitamins and provitamins
α-Carotene (μg/dL)	4635	6–19	1.94 (1.78, 2.11)	1.92 (1.73, 2.13)	1.91 (1.75, 2.09)	0.9
β-Carotene (μg/dL)	4427	6–19	11.0 (10.5, 11.6)	10.6 (10.0, 11.2)	10.6 (9.9, 11.4)	0.4
β-Cryptoxanthin (μg/dL)	4606	6–19	8.15 (7.67, 8.66)	7.81 (7.30, 8.35)	7.81 (7.39, 8.24)	0.4
Lutein/zeaxanthin (μg/dL)	4639	6–19	12.0 (11.4, 12.5)	11.5 (11.2, 11.09)	11.7 (11.1, 12.3)	0.4
trans Lycopene (μg/dL)	4637	6–19	20.9 (20.3, 21.6)	21.7 (21.1, 22.4)	21.0 (19.9, 22.2)	0.2
Vitamin D (ng/mL)	5325	2–19	23.6a (22.7, 24.5)	24.0a (23.1, 25.0)	24.7b (23.7, 25.9)	0.01
Vitamin E (μg/dL)	4637	6–19	797a,b (779, 814)	787a (773, 801)	807b (795, 820)	0.05
Serum lipids (mg/dL)
Total cholesterol	4654	6–19	160 (158, 163)	158 (156, 160)	161 (158, 163)	0.1
Total cholesterol (morning subsample)3	1441	12–19	157 (152, 161)	156 (152, 160)	157 (152, 162)	0.9
HDL cholesterol	4654	6–19	52 (51, 53)	52 (51, 53)	52 (51, 53)	0.3
HDL cholesterol (morning subsample)3	1441	12–19	50 (48, 52)	50 (48, 52)	52 (51, 53)	0.2
LDL cholesterol (morning subsample)3	1437	12–19	85 (81, 89)	85 (82, 88)	85 (81, 89)	1.0
Triglycerides (morning subsample)3	1441	12–19	83 (77, 90)	81 (76, 86)	76 (70, 83)	0.2

¹The adjusted values are back-transformed predicted margins from multiple linear regression models with each biomarker as a log-transformed dependent variable and included age (2–5, 6–11, 12–19 y), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, Mexican American), family PIR (<130%, 130–349%, >349%), educational level of household head (<12 y, 12 y, some college, college degree or higher), month of mobile examination center examination (November–April, May–October), self-reported history of chronic disease (yes or no), hours of fasting (continuous), history of supplement use in the past month (yes or no), BMI percentile for age and sex (continuous), serum cotinine, and C-reactive protein. Models for carotenoids and vitamin E also included serum total cholesterol, and models for vitamin D and serum lipids included hours of television and computer use. Adjusted values in each row not sharing a common superscript letter are significantly different, P < 0.05 (Satterthwaite adjusted F test). PIR, poverty-income ratio; RBC, red blood cell.

²Satterthwaite adjusted F test for differences between PIR groups.

³Limited to the subsample with morning examination and >8.5 h of fasting.

TABLE 4.

Multiple variable–adjusted geometric means and 95% CIs of biomarkers by categories of educational level of household head in US children and adolescents: NHANES 2003–2006¹

			Educational level
Serum biomarker	n	Age	<12 y	12 y	Some college	College degree or higher	P (education)2
		y
Water-soluble vitamins
Vitamin C (mg/dL)	4684	6–19	0.99 (0.93, 1.05)	0.99 (0.93, 1.06)	1.02 (0.96, 1.08)	1.03 (0.97, 1.09)	0.7
Pyridoxal phosphate (nmol/L)	2672	2–19	57.1 (50.5, 64.4)	54.1 (49.9, 58.7)	54.7 (51.9, 57.7)	54.7 (49.8, 60.1)	0.8
Serum folate (ng/mL)	5388	2–19	13.4 (12.9, 13.9)	13.4 (12.9, 14.0)	13.4 (12.9, 14.0)	13.9 (13.2, 14.6)	0.6
Red blood cell folate (ng/mL)	5389	2–19	250a,b (243, 256)	246a,b (239, 254)	242a (235, 248)	255b (246, 263)	0.04
Vitamin B-12 (pg/mL)	5382	2–19	627 (606, 649)	603 (580, 627)	606 (589, 624)	621 (594, 650)	0.3
Fat-soluble vitamins and provitamins
Retinol (μg/dL)	4639	6–19	41.0 (39.1, 42.9)	41.6 (40.4, 42.8)	42.3 (41.3, 43.6)	42.9 (41.4, 44.5)	0.3
α-Carotene (μg/dL)	4635	6–19	1.86a (1.68, 2.06)	1.76a (1.62, 1.93)	1.88a (1.76, 2.01)	2.32b (1.96, 2.75)	0.009
β-Carotene (μg/dL)	4427	6–19	10.6 (9.8, 11.5)	10.4 (9.9, 11.0)	10.6 (10.2, 11.1)	11.4 (10.3, 12.5)	0.2
trans Lycopene (μg/dL)	4637	6–19	20.8 (19.9, 21.8)	22.1 (21.3, 22.9)	21.0 (20.1, 22.0)	20.9 (20.0, 21.8)	0.1
Vitamin D (ng/mL)	5325	2–19	24.9 (23.7, 26.1)	24.1 (22.9, 25.4)	23.9 (23.0, 24.8)	23.9 (22.9, 24.8)	0.2
Vitamin E (μg/dL)	4637	6–19	791 (775, 806)	800 (783, 817)	797 (783, 812)	796 (777, 815)	0.8
Serum lipids (mg/dL)
Total cholesterol	4654	6–19	158 (155, 161)	160 (157, 164)	161 (159, 164)	158 (155, 161)	0.2
Total cholesterol (morning subsample)3	1441	12–19	155 (149, 161)	159 (154, 165)	156 (151, 160)	155 (148, 163)	0.6
HDL cholesterol	4654	6–19	51 (50, 52)	52 (51, 53)	52 (50, 53)	52 (50, 54)	0.8
HDL cholesterol (morning subsample)3	1441	12–19	51 (49, 53)	51 (49, 52)	51 (49, 53)	50 (47, 53)	0.9
LDL cholesterol (morning subsample)3	1437	12–19	85 (81, 90)	87 (82, 92)	84 (80, 88)	84 (78, 89)	0.7
Triglycerides (morning subsample)3	1441	12–19	74 (69, 79)	85 (78, 93)	78 (73, 83)	82 (74, 91)	0.1

¹The adjusted values are back-transformed predicted margins from multiple linear regression models with each biomarker as a log-transformed dependent variable and included age (2–5, 6–11, 12–19 y), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, Mexican American), family poverty-income ratio (<130%, 130–349%, >349%), educational level of household head (<12 y, 12 y, some college, college degree or higher), month of mobile examination center examination (November–April, May–October), self-reported history of chronic disease (yes or no), hours of fasting (continuous), history of supplement use in the past month (yes or no), BMI percentile for age and sex (continuous), serum cotinine, and C-reactive protein. Models for carotenoids and vitamin E also included serum total cholesterol, and models for vitamin D and serum lipids included hours of television and computer use. Adjusted values in each row not sharing a common superscript letter are significantly different, P < 0.05 (Satterthwaite adjusted F test).

²Satterthwaite adjusted F test for differences between education groups.

³Limited to the subsample with morning examination and >8.5 h of fasting.

Race-ethnicity, family income, and education differentials in percentage of the population at risk by using biomarker cutoffs

The number of respondents with low serum concentrations of folate, vitamin B-12, and retinol was small [serum folate concentration <2 ng/mL (n = 1), RBC folate concentration <95 ng/mL (n = 27), vitamin B-12 concentration <200 pg/mL (n = 24), and serum retinol concentration of <20 μg/dL (n = 22)]. Therefore, further analyses were not performed for these biomarkers. For the remaining biomarkers, no examined interactions met the criteria for statistical significance; therefore, the main effects are presented in Table 5. Relative to non-Hispanic whites and Mexican Americans, a smaller percentage of non-Hispanic black children had vitamin C concentrations <0.2 mg/dL (P < 0.001). More than 50% of all non-Hispanic black children and nearly 30% of Mexican American children had serum vitamin D concentrations <20 ng/mL compared with ∼12% of non-Hispanic white children (P < 0.0001). Compared with non-Hispanic white children, Mexican Americans were less likely to have serum cholesterol concentrations ≥200 mg/dL (P = 0.01). The percentage of non-Hispanic black children with serum HDL cholesterol ≤40 mg/dL or triglyceride concentrations ≥130 mg/dL was smaller than the corresponding percentage of non-Hispanic white children (P < 0.0001). Family income and education were not independently associated with the risk of lower or higher concentrations of the examined biomarkers.

TABLE 5.

Multiple covariate–adjusted percentages and 95% CIs of the 2–19-y-old US children and adolescents below or above the at-risk cutoff: NHANES 2003–2006¹

Serum biomarker2	n	Age	Non-Hispanic white	Non-Hispanic black	Mexican American	P (race-ethnicity)3	P (PIR)34	P (education)3
		y	%	%	%
Water-soluble vitamins
Vitamin C	4684	6–19	2.83b (1.66, 3.99)	0.19a (0.01, 0.37)	1.19b (0.30, 2.07)	0.001	0.8	0.3
<0.2 mg/dL	64	—	—	—	—
Pyridoxal phosphate	2672	2–19	2.49 (1.12, 3.81)	3.72 (2.57, 4.86)	3.02 (1.04, 5.01)	0.4	0.5	0.1
<20 nmol/L	99	—	—	—	—
Fat-soluble vitamins
Vitamin D	5325	2–19	12.9a (10.8, 15.0)	53.6c (48.0, 59.3)	29.5b (25.3, 33.6)	<0.0001	0.4	0.4
<20 ng/mL	2112	—
Vitamin E	4637	6–19	1.69 (0.98, 2.40)	1.89 (1.06, 2.71)	1.63 (0.70, 2.55)	0.9	0.2	0.9
<500 μg/dL	80	—	—	—	—
Serum lipids
Total cholesterol	4654	6–19	11.8b (9.8, 13.9)	9.49b (7.9, 11.1)	6.75a (5.1, 8.3)	0.01	0.3	0.6
≥200 mg/dL	449	—	—	—	—
Total cholesterol (morning subsample)5	1441	12–19	11.2b (7.7, 14.6)	7.7a,b (5.6, 9.9)	6.2a (3.1, 9.3)	0.05	0.2	0.4
≥200 mg/dL	123	—	—	—	—
HDL cholesterol	4654	6–19	13.3b (10.7, 15.9)	5.81a (4.7, 6.9)	12.0b (9.8, 14.3)	<0.0001	0.3	0.8
<40 mg/dL	513	—	—	—	—
HDL cholesterol (morning subsample)5	1441	12–19	15.5b (11.2, 19.7)	7.7a (5.8, 9.7)	13.2b (9.6, 16.9)	0.009	0.4	0.4
<40 mg/dL	174	—	—	—	—
LDL cholesterol (morning subsample)5	1437	12–19	7.47 (5.14, 9.79)	6.03 (4.02, 8.04)	5.16 (2.01, 8.31)	0.3	0.7	0.6
≥130 mg/dL	95	—	—	—	—
Triglycerides (morning subsample)5	1441	12–19	18.7b (13.3, 24.0)	4.94a (3.02, 6.85)	17.6b (12.6, 22.6)	<0.0001	0.1	0.2
≥130 mg/dL	180	—	—	—	—

¹Multiple logistic regression model with each biomarker as a dichotomous outcome included age (2–5, 6–11, 12–19 y), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, Mexican American), family poverty-income ratio (<130%, 130–349%, >349%), educational level of household head (<12 y, 12 y, some college, college degree or higher), month of mobile examination center examination (November–April, May–October), self-reported history of chronic disease (yes or no), hours of fasting (continuous), history of supplement use in the past month (yes or no), BMI percentile for age and sex (continuous), serum cotinine, and C-reactive protein. Models for vitamin E also included serum total cholesterol, and models for vitamin D and serum lipids included hours of television and computer use. Value estimates in each row not sharing a common superscript letter are significantly different, P < 0.05 (Satterthwaite adjusted F test).

²Numbers were too small for formal testing for serum folate <2 ng/mL (n = 1), red blood cell folate <95 ng/mL (n = 27), vitamin B-12 <200 pg/mL (n = 24), and retinol <20 μg/dL (n = 22).

³Satterthwaite adjusted F test for differences between race-ethnic groups, PIR categories, and education categories.

⁴PIR, poverty-income ratio.

⁵Limited to the morning subsample and >8.5 h of fasting.

Dietary micronutrient intakes

Interactions of race-ethnicity, family income, and education for predicting mean dietary intake of selected micronutrients

A race-by-education interaction was noted for vitamin C intake (P = 0.004), and an age-by-PIR interaction was noted for dietary vitamin B-12 (P < 0.0005) (Figure 2). Race differences in dietary vitamin C intake were noted in all categories of education, with higher intakes in non-Hispanic blacks and Mexican Americans. However, education was unrelated to dietary vitamin C intake in non-Hispanic whites. In non-Hispanic blacks, higher education predicted higher vitamin C intake, but in Mexican Americans it predicted lower mean intake (P = 0.01). There were no age differences in dietary vitamin B-12 intake in the lowest PIR category; however, age differences were noted in the other 2 PIR categories (P ≤ 0.006). In 2–11-y-olds, dietary vitamin B-12 intake declined with increasing PIR (≤0.01); this relation was not observed in 12–19-y-olds.

FIGURE 2.

Interactions of race-ethnicity with education (dietary vitamin C intake; A) and age with poverty-income ratio (dietary vitamin B-12 intake; B) in US children and adolescents, P < 0.01 (Satterthwaite adjusted F test): NHANES 2003–2006. The estimates are adjusted means (and 95% CIs) from multiple linear regression models with each dietary variable as a dependent variable and included age (2–19 y), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, Mexican American), family poverty-income ratio (<130%, 130–349%, >349%), educational level of household head (<12 y, 12 y, some college, college degree or higher), month of mobile examination center examination (November–April, May–October), day of recalled intake (Monday–Thursday, Friday–Sunday), self-reported history of chronic disease (yes or no), and BMI percentile for age and sex (continuous).

Race-ethnic differentials in mean dietary nutrient intake

The reported 24-h mean intakes of all examined nutrients (except for folic acid, β-carotene, lycopene, and vitamin E) differed by race-ethnicity (Table 6). Compared with non-Hispanic whites, among non-Hispanic blacks the reported dietary intakes of total folate, vitamin B-12, retinol (μg), total vitamin A (retinol activity equivalent), α-carotene, and percentage of energy from saturated fat were lower and intakes of β-cryptoxanthin and lutein + zeaxanthin were higher. Among Mexican Americans, compared with non-Hispanic whites, the mean reported dietary intakes of most micronutrients were either not different [total folate, vitamin B-12, retinol (μg), vitamin A (retinol activity equivalent), α-carotene, and lutein + zeaxanthin], were lower (% of energy from total and saturated fat), or were higher (vitamin B-6 and β-cryptoxanthin).

TABLE 6.

Self-reported 24-h dietary energy and nutrient intake (multiple covariate–adjusted means and 95% CIs) of US children and adolescents by race-ethnicity: NHANES 2003–2006¹

Dietary nutrient	n	Age	Non-Hispanic white	Non-Hispanic black	Mexican American	P (race-ethnicity)2
		y
Energy (kcal)	6448	2–19	2127 (2078, 2176)	2065 (1988, 2142)	2093 (2011, 2174)	0.3
Water-soluble vitamins
Vitamin B-6 (mg)3	3286	2–19	1.74a,b (1.64, 1.84)	1.69a (1.54, 1.84)	1.87b (1.77, 1.97)	0.02
Total folate (μg)	6448	2–19	394b (377, 412)	365a (345, 385)	403b (384, 423)	0.05
Folic acid (μg)	6448	2–19	236 (221, 251)	221 (206, 236)	232 (215, 249)	0.3
Vitamin B-12 (μg)	6448	2–19	5.33b (5.06, 5.61)	4.72a (4.47, 4.96)	5.41b (5.11, 5.71)	0.0006
Fat-soluble vitamins and carotenoids
Retinol (μg)	5153	6–19	499b (469, 528)	399a (372, 426)	466b (435, 497)	<0.0001
Vitamin A (μg RAE4)	5153	6–19	604b (571, 637)	495a (458, 532)	580b (533, 626)	<0.0001
α-Carotene (μg)	5153	6–19	233b (191, 276)	147a (109, 185)	266b (191, 341)	0.008
β-Carotene (μg)	5153	6–19	1095 (972, 1218)	998 (808, 1188)	1139 (901, 1376)	0.5
β-Cryptoxanthin (μg)	5153	6–19	105a (88, 123)	140b (117, 163)	175b (141, 209)	0.0009
Lutein + zeaxanthin (μg)	5153	6–19	758a (684, 832)	955b (784, 1126)	706a (615, 797)	0.01
Lycopene (μg)	5153	6–19	5544 (5047, 6040)	5242 (4605, 5879)	6116 (4977, 7255)	0.4
Vitamin E (mg α-tocopherol)	5153	6–19	6.3 (6.0, 6.6)	6.1 (5.8, 6.4)	6.2 (5.7, 6.6)	0.6
Dietary fats (% of energy)
Total fat	5152	6–19	33.3b (32.8, 33.9)	33.9b (33.3, 34.5)	32.1a (31.5, 32.7)	0.007
Saturated fat	5152	6–19	11.8b (11.6, 12.0)	11.3a (11.1, 11.5)	11.1a (10.8, 11.3)	0.0001

¹The adjusted values are predicted margins from multiple linear regression models with each dietary variable as a continuous dependent and included age (2–5, 6–11, 12–19 y), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, Mexican American), family poverty-income ratio (<130%, 130–349%, >349%), educational level of household head (<12 y, 12 y, some college, college degree or higher), month of mobile examination center examination (November–April, May–October), day of recalled intake (Monday–Thursday, Friday–Sunday), self-reported history of chronic disease (yes or no), and BMI percentile for age and sex (continuous). Adjusted values in each row not sharing a common superscript letter are significantly different, P < 0.05 (Satterthwaite adjusted F test).

²Satterthwaite adjusted F test for differences between race-ethnic groups.

³NHANES 2005–2006.

⁴RAE, retinol activity equivalents.

PIR and education differentials in mean dietary nutrient intake

Dietary vitamin C intakes differed by categories of PIR (P = 0.03) (see Supplemental Table 1 under “Supplemental data” in the online issue) and intakes of β-carotene differed by categories of education (P = 0.01) (see Supplemental Table 2 under “Supplemental data” in the online issue). There was no independent association of family income or education for all other examined dietary outcomes.

DISCUSSION

The results of our study underscore the following: 1) race-ethnic differences in concentrations of micronutrient and lipid biomarkers remained after family income and education were accounted for, 2) race-ethnicity (but not family income and education) was a strong predictor of both biomarker and diet outcomes, and 3) the evidence of modification of the race-ethnic effect by family income and education was noted for very few outcomes. Overall, these findings indicate a possibly stronger influence of foodways associated with ethnicity relative to resource or knowledge constraints assessed as family income and education for the examined outcomes. The race-ethnic differentials in biomarkers may reflect the joint contribution of multiple interacting factors that include the following: cultural preferences for certain foods, their combinations, and their methods of preparation; biological influences on nutrient bioavailability and metabolism and supplement use; and social/neighborhood/environmental contexts that relate to food availability and demand, as effectors. The general lack of consistent family income associations with diet and biomarkers in children and adolescents may reflect the success of federal need-based food assistance and child nutrition programs in closing the income gap in micronutrient availability.

The observed race-ethnic gradients in serum concentrations of retinol and carotenoids in US children are consistent with previous reports (28–30, 53). Our results provide evidence for modification of the effect of race-ethnicity by family PIR or education for retinol, lutein + zeaxanthin, and β-cryptoxanthin. For these 3 biomarkers, the race-ethnic differences were present in every category of PIR (retinol) or education (lutein + zeaxanthin and β-cryptoxanthin). However, the magnitude of race-ethnic differences decreased with increasing PIR or education because the direction of the association of the biomarker with income or education differed among race-ethnic groups (Figure 1). For example, in non-Hispanic blacks and Mexican Americans, there appeared to be a positive (but nonsignificant) association of serum retinol and PIR; in non-Hispanic whites, this association was inverse. The expected positive association of education with serum concentrations of lutein + zeaxanthin and β-cryptoxanthin were significant only in non-Hispanic whites; in Mexican Americans, these associations were inverse. Although we did not observe race-by-PIR or -education interactions for self-reported dietary intake of vitamin A or lutein + zeaxanthin and β-cryptoxanthin (analogous to those observed for serum concentrations), we did find that higher education predicted lower dietary vitamin C intake in Mexican Americans (Figure 2). Collectively, these findings suggest possibly adverse effects of higher education on food selection of Mexican Americans. A possible explanation may be that higher education is related to greater dietary acculturation and less adherence to traditional foodways among Mexican Americans. In Mexican Americans, greater dietary acculturation has been reported to relate to adverse dietary outcomes (54, 55), and there is reported evidence of higher cardiovascular risk with higher education and SEP (56). Therefore, the nutritional advantages and healthful characteristics of traditional dietary patterns should be reinforced in all Mexican Americans irrespective of SEP.

Consistent with prior reports (32–36), we found race-ethnic differences in mean serum vitamin D concentration, which widened with increasing age. The widening of the race-ethnic differential with age may reflect greater age-associated decline in milk intake in non-Hispanic blacks relative to other ethnic groups (57) or a possible decrease in outdoor physical activity. The recent Institute of Medicine report (49) recognized the black-white differences in calcium economy and vitamin D nutritional status but did not provide race-specific values of vitamin D to indicate at-risk thresholds.

It has been suggested that race-ethnic differences in serum vitamin B-12 and folate concentrations may reflect race differentials in bioavailability and metabolism of these nutrients (58, 59). The results of our study provide some evidence in support of this argument. For example, non-Hispanic blacks had the lowest mean dietary intake but the highest mean serum vitamin B-12 concentrations (along with the lowest dietary, serum, and RBC folate). Similarly, although mean total dietary folate (and folic acid) intakes of Mexican Americans did not differ from those of non-Hispanic whites, the serum and RBC folate concentrations were significantly lower.

Our findings of significantly lower mean concentrations of vitamin C in non-Hispanic white children compared with non-Hispanic black and Mexican American children differ from those observed among adults in the NHANES III (19) and the NHANES 2003–2004 (20). In these reports, non-Hispanic white adults had higher serum ascorbate concentrations and family PIR was a positive predictor. However, non-Hispanic white children and adolescents in our study had the lowest mean ascorbate concentration, and a higher proportion was at risk for concentrations indicating marginal status. Moreover, neither PIR nor education was related with serum ascorbate after adjustment for race-ethnicity in multivariate-adjusted models. These results suggest that dietary vitamin C intake and sources may change during transition to adult years, and the assumption about relevance of ethnic foodways during adult years may not apply to all dietary components similarly.

Although summary estimates of micronutrient intake from self-reported diets are known to be measured with error (11, 12), our results show that non-Hispanic black and white differences in serum concentrations of several micronutrients (vitamin C, folate, vitamin A, α-carotene, β-cryptoxanthin, and lutein zeaxanthin) paralleled these differences in dietary intakes of corresponding micronutrients. This is all the more remarkable given the higher prevalence of supplement use among non-Hispanic white children. The correspondence between serum concentrations and dietary micronutrients in Mexican American and non-Hispanic white differentials was limited to only a few nutrients (vitamin C, vitamin E, and β-carotene). We can speculate that these ethnic differences in biomarker and dietary intake correspondence may possibly reflect ethnic differentials in reliability of the dietary assessment method used in the NHANES, completeness and accuracy of the database used to estimate micronutrient content of ethnic foods, differential bioavailability of micronutrients because of ethnic differences in food sources, and dietary supplement use. With the availability of total nutrient intakes from diet and supplements in recent surveys (60), future such analyses may provide a more nuanced interpretation of race-ethnic differentials in correspondence between biomarker and nutrient intakes.

Our study findings suggest that the serum lipid profile of non-Hispanic whites was more atherogenic than that of the comparison minority groups. Overall, non-Hispanic blacks had the most favorable lipid profile of all ethnic groups. The finding of higher mean serum HDL-cholesterol concentration and lower triglycerides in non-Hispanic black children and adolescents is in accord with prior such reports (6, 61–66). However, our result of null findings for association of serum lipids with SEP differs from an earlier report of effect modification by socioeconomic factors in black 9–11-y-olds (67) but concurs with the findings of Winkleby et al (8), who found that neither ethnicity nor family education was related to “non-HDL cholesterol concentration” in 4–24-y-olds in the NHANES III. We note that the race-ethnic differences in serum lipid profile observed in children and adolescents in our study have also been reported for US adults, which persisted after adjustment for SEP (61, 68).

Strengths of our study include a nationally representative sample of sufficient size with available information on family income and education to allow an assessment of their independent and joint associations with the outcomes examined in the study and the assessment of micronutrient status with the use of biomarkers, which avoids reliance on self-reported intakes.

In conclusion, there were distinct race-ethnic differentials, independent of family income and ethnicity, in serum concentrations and dietary intake of micronutrients and lipids in US children and adolescents. Although non-Hispanic blacks differed from non-Hispanic whites in several outcomes, significantly higher risk of marginal status was evident only for vitamin D. With the current safety net of food assistance and child-nutrition programs in place, overall results suggest little independent or interaction effect of SEP for the majority of the examined outcomes. Any interventions to affect a desired change will need to target ethnic groups.