Unmetabolized Folic Acid Is Detected in Nearly All Serum Samples from US Children, Adolescents, and Adults^,,,

Abstract

Background: Serum total folate consists mainly of 5-methyltetrahydrofolate (5-methylTHF). Unmetabolized folic acid (UMFA) may occur in persons consuming folic acid–fortified foods or supplements.

Objectives: We describe serum 5-methylTHF and UMFA concentrations in the US population ≥1 y of age by demographic variables and fasting time, stratified by folic acid–containing dietary supplement use. We also evaluate factors associated with UMFA concentrations >1 nmol/L.

Methods: Serum samples from the cross-sectional NHANES 2007–2008 were measured for 5-methylTHF (n = 2734) and UMFA (n = 2707) by HPLC–tandem mass spectrometry.

Results: In supplement users compared with nonusers, we found significantly higher geometric mean concentrations of 5-methylTHF (48.4 and 30.7 nmol/L, respectively) and UMFA (1.54 and 0.794 nmol/L, respectively). UMFA concentrations were detectable (>0.3 nmol/L) in >95% of supplement users and nonusers, regardless of demographic or fasting characteristics; concentrations differed significantly by age and fasting time, but not by sex and race-ethnicity, both in supplement users and nonusers. The prevalence of UMFA concentrations >1 nmol/L was 33.2% overall and 21.0% in fasting (≥8 h) adults (≥20 y of age). Using multiple logistic regression analysis, UMFA concentrations >1 nmol/L were associated with being older, non-Hispanic black, nonfasting (<8 h), having smaller body surface area, higher total folic acid intake (diet and supplements), and higher red blood cell folate concentrations. In fasting adults, a decrease in the mean daily alcohol consumption was also associated with increased odds of UMFA concentrations >1 nmol/L.

Conclusions: UMFA detection was nearly ubiquitous, and concentrations >1 nmol/L were largely but not entirely explained by fasting status and by total folic acid intake from diet and supplements. These new UMFA data in US persons ≥1 y of age provide much-needed information on this vitamer in a fortified population with relatively high use of dietary supplements.

See corresponding commentary on page 387

Introduction

Serum and RBC “total” folates are traditional biomarkers of folate status and have been measured since the mid 1970s, covering the time from pre– to post–folic acid fortification in a representative sample of the US population participating in the cross-sectional NHANES (1, 2). The specific vitamer composition of serum “total” folate, however, requires sensitive chromatography-based procedures that became available more recently, but are limited to more specialized research laboratories (3). Serum 5-methyltetrahydrofolate (5-methylTHF)⁸ is the main circulating form (4), generally making up >80% of total folate (5).

Dietary folate is an umbrella term used to represent food folate and folic acid from fortified foods and dietary supplements. The human gut has limited ability to reduce and methylate folic acid (6). Administration of a one-time bolus dose that exceeds ∼260 μg or consecutive small doses of folic acid (100 μg) within a day have been shown to result in the appearance of folic acid in serum, i.e., unmetabolized folic acid (UMFA) (7, 8). UMFA has been measured in observational studies (9–13) or intervention trials (14, 15), most of small sample size, but there is a lack of nationally representative data for various demographic subgroups.

Serum 5-methylTHF and UMFA concentrations were measured by use of HPLC with electrochemical detection in American seniors (≥60 y of age) in the NHANES 1999–2002 as part of a Tufts University surplus specimen project to assess the relation to anemia, macrocytosis, and cognitive test performance (16). The NHANES 2001–2002 data were also used to explore the relation of UMFA to total folate intakes from diet and supplements (17). In that analysis, the presence of UMFA was not easily explained by folic acid intakes alone. The NHANES 2007–2008 was the first survey cycle where these 2 folate forms were measured in serum for the entire US population ≥1 y of age; this cycle was also the first instance where a sensitive and specific HPLC–tandem MS (HPLC-MS/MS) method was used, and where serum and RBC total folate data measured by microbiologic assay were also available (2).

In this paper, we describe distributions of serum 5-methylTHF and UMFA concentrations in the US population participating in the NHANES 2007–2008. We assess how these 2 folate forms correlate with serum and RBC total folate and with dietary folate intake. Lastly, we evaluate what variables explain the presence of UMFA concentrations >1 nmol/L. These data allow the assessment of exposure to folic acid in the current era of folic acid fortification, as recommended by a 2010 roundtable of experts that discussed folate methodologies in the NHANES (18).

Methods

Study participants and study design.

The NHANES collects cross-sectional data on the health and nutritional status of the civilian noninstitutionalized US population. The survey is conducted by the National Center for Health Statistics (NCHS) at the CDC and has a stratified, multistage, probability sample design. Survey participants are first interviewed in their homes to collect information on demographic characteristics, dietary supplement use, and some health-related issues. Participants then undergo a physical examination and blood draw in a mobile examination center (MEC) ∼1–3 wk after the household interview, where they also complete a 24-h dietary recall. The sample design in 2007–2010 changed with respect to the oversampling of certain demographic groups; adolescents were no longer oversampled and Hispanic persons were oversampled instead of just Mexican Americans (19, 20). This design change makes the following 2 Hispanic subgroups appropriate for analysis: Mexican Americans or all Hispanics, which includes Mexican Americans and all other Hispanics. The unweighted response rates for participants ≥1 y of age in 2007–2008 were 77.9% for the interview component and 75% for the examination component (21). All respondents gave their informed consent, and the NHANES protocol was reviewed and approved by the NCHS Research Ethics Review Board.

Biochemical methods.

The NHANES 2007–2008 serum samples (n = 2734) from a random one-third subset of all participants ≥1 y of age were analyzed for the first time for folate forms by use of HPLC-MS/MS by the CDC laboratory (22–24). Results were reported for 5-methylTHF and UMFA [fewer results were available for UMFA (n = 2707)]. Results for 3 minor non-methyl folate forms [5-formyltetrahydrofolate (5-formylTHF), tetrahydrofolate, and 5,10-methenyltetrahydrofolate] were not part of the public release data file because of a measurement inaccuracy; the 5-formylTHF peak represented the sum of 2 compounds, 5-formylTHF and the biologically inactive oxidation product of 5-methylTHF known as MeFox (pyrazino-s-triazine derivative of 4α-hydroxy-5-methylTHF) (25). However, the non-methyl folate constitutes a small fraction of the total folate (generally <5%). We used Westgard-type quality control multirules to judge assay performance in the NHANES 2007–2008 (26). Long-term quality control CVs for serum 5-methylTHF and UMFA were <5% (6.34–46.8 nmol/L) and 9.5–11.8% (0.71–8.14 nmol/L), respectively. The limits of detection (LODs) were 0.5 nmol/L for 5-methylTHF and 0.3 nmol/L for UMFA. Other method performance characteristics were described elsewhere (18). More recently, the National Institute of Standards and Technology (NIST) released standard reference material 1950 (27). Our HPLC-MS/MS procedure matched the certified value (uncertainty) for 5-methylTHF [26.7 vs. 26.9 (0.70) nmol/L] and the reference value for UMFA [4.03 vs. 3.42 (1.02) nmol/L].

Other biochemical variables determined earlier as part of the NHANES 2007–2008 and used in this analysis were serum and RBC total folate by microbiologic assay (28) and serum cotinine by HPLC-MS/MS (29).

Folate intake.

All NHANES examinees were eligible for two 24-h dietary recall interviews: one collected in person in the MEC and a second interview collected by telephone 3–10 d later. The following dietary variables, calculated by the NCHS before the data release, were available for each recall: dietary folic acid (μg), food folate (μg), and total dietary folate [μg and expressed in dietary folate equivalents (DFEs)] (30). We only used dietary folic acid and total dietary folate (expressed in DFEs) data from the first recall because it preceded the MEC blood collection. Dietary supplement information was collected as part of the NHANES Dietary Supplement Questionnaire, which recorded detailed information about the type, consumption frequency, duration of use, and amount taken for each reported dietary supplement. To calculate a person’s total mean daily intake of folic acid (μg) from all dietary supplement products consumed over the past 30 d, we first calculated the mean daily intake of folic acid for each reported product separately. We then divided the participant’s reported amount taken by the serving size and multiplied by the quantity of folic acid recorded on the product label. This amount was multiplied by the proportion of days out of 30 the participant reported taking the product. We then summed each mean daily intake of folic acid for each reported product for each participant. All participants who reported never taking a dietary supplement over the past 30 d were assigned a zero, whereas all participants who did not respond to this question were assigned a missing value. Supplement users with incomplete information (quantity or duration missing) may also have a missing value assigned. We combined the mean folic acid from supplemental sources with dietary folic acid to reflect total folic acid intake (μg). In the United States, the dietary folic acid can come from mandatory fortified cereal-grain foods and voluntary fortified products, mainly ready-to-eat cereals.

Covariates.

For descriptive analyses, we categorized the demographic variables as follows: age (1–5 y, 6–11 y, 12–19 y, 20–39 y, 40–59 y, and ≥60 y; in some instances, 1–11 y was used); sex (males, females); and race-ethnicity [all Hispanic (Mexican American + other Hispanic), non-Hispanic black (NHB), non-Hispanic white (NHW); other racial-ethnic groups were included in overall estimates; in descriptive tables we also reported separate estimates for Mexican Americans]. Other variables were fasting time (<3 h, 3 to <8 h, ≥8 h); folic acid–containing supplement use (yes, no); total folic acid intake (<100 μg, 100 to <200 μg, 200 to <400 μg, ≥400 μg); RBC total folate (<800 nmol/L, 800 to <1200 nmol/L, 1200 to <1500 nmol/L, ≥1500 nmol/L); body surface area [(BSA) calculated as square root of (height in cm × weight in kg/3600); <1.5 m², 1.5 to <1.8 m², 1.8 to <2.0 m², ≥2.0 m²] (31); smoking [serum cotinine ≤10 μg/L (nonsmoker), >10 μg/L (smoker)] (32); and alcohol intake [mean daily number of “standard” drinks (1 drink ∼15 g alcohol): no drinks, <1 (not 0) drink/d, 1 to <2 drinks/d, ≥2 drinks/d; only available for participants ≥18 y] (33). We also used these variables (same categories as above unless specified) to develop multiple logistic regression models: age (continuous), sex, race-ethnicity, fasting status [≥8 h (fasting), <8 h (not fasting)], total folic acid intake (continuous), RBC total folate (to adjust for folate status; continuous), BSA (to adjust for body size; continuous), smoking, and alcohol consumption (continuous).

Statistical analyses.

Statistical analyses were performed with use of SAS (version 9.2; SAS Institute) and SUDAAN (version 11.0.1; RTI) software. We excluded pregnant and lactating women from all analyses (n = 31 for 5-methylTHF and n = 30 for UMFA) and 1 individual because of a sample weight of zero. Otherwise we used pairwise deletion for missing data to minimize unnecessary removal of valid data and thus the introduction of bias. This resulted in an analytic sample of 2702 participants for 5-methylTHF and 2676 participants for UMFA (Figure 1) and in different sample sizes depending on the scope of the analysis. We used the one-third subsample weights to account for differential nonresponse or noncoverage and to adjust for oversampling of some groups. We calculated the proportion of detectable UMFA results by demographic, fasting, or supplement use characteristics. CIs for the prevalence estimates were calculated using a logit transformation with the SEM of the logit prevalence based on the delta method. We calculated the mean absolute and percent contribution of UMFA and 5-methylTHF in serum to the sum of 5-methylTHF and UMFA (for all, fasting persons, fasting nonusers, and users of folic acid–containing dietary supplements) by weighted quartiles of 5-methylTHF concentration derived from the entire population. We calculated cumulative frequency distributions for UMFA overall and by fasting status. We explored bivariate associations between geometric mean folate biomarker concentrations (serum 5-methylTHF, UMFA, and total folate; RBC total folate) or median folic acid intake amounts (diet, supplements, and total folic acid) and demographic variables (age, sex, race-ethnicity) or fasting status, stratified by folic acid supplement use. Geometric means were compared across the categories on the basis of Wald F tests with and without controlling for additional covariates (age, sex, race-ethnicity, and fasting time). Sample sizes for the folate biomarkers and folic acid intakes stratified by folic acid supplement use are presented in Supplemental Tables 1 and 2, respectively. We assessed pairwise correlations between folate biomarkers and folate intake by calculating Spearman correlation coefficients. We assessed the distributions [geometric means and selected percentiles (95% CIs)] of serum 5-methylTHF and UMFA among participants ≥1 y of age fasting from food and supplements for ≥8 h by demographic variables or folic acid supplement use.

FIGURE 1.

Study participants ≥1 y of age with serum folate vitamer data in the NHANES 2007–2008 one-third subsample. Serum 5-methylTHF and UMFA were measured by LC–tandem MS. Serum and RBC total folate were measured by microbiologic assay. UMFA, unmetabolized folic acid; 5-methylTHF, 5-methyltetrahydrofolate.

To evaluate what variables explained UMFA concentrations >1 nmol/L, we first explored bivariate associations between the prevalence of UMFA concentrations >1 nmol/L and variables of interest specified in the covariates section and then performed multiple logistic regression analysis to provide adjusted ORs (ORs_adj). We selected a UMFA cutoff of 1 nmol/L for several reasons: it represented approximately the 75th percentile for fasting (from food and supplements) individuals (supplement users and nonusers together) in our dataset (0.957 nmol/L) and was therefore not too unusual; it was reasonably close to the 1.35 nmol/L cutoff used by Kalmbach et al. (10), which represented the 85th percentile for fasting UMFA concentrations in the prefortification Framingham Offspring Cohort; and it represented a round number that will not fluctuate from one dataset to another as would be the case with a specific percentile derived from our dataset.

We used a backward elimination approach with listwise deletion for missing values to arrive at the final multivariate logistic regression model and made some preliminary analytic decisions to conserve the limited df and to allow simultaneous modeling of several variables and investigation of pairwise interactions (Supplemental Methods). Once we developed a final model for the entire study population (≥1 y of age, model 1), we fit a subsequent model limited to fasting (≥8 h) persons ≥20 y of age (model 2) to evaluate whether similar factors explained UMFA concentrations >1 nmol/L in a more homogeneous sample. In model 2, we used all relevant variables included in model 1 plus mean daily alcohol consumption, which was only publically available among adults ≥18 y of age. We reported OR_adj (95% CI) and Wald F test P values for each pairwise comparison to the selected reference category. We also evaluated model sensitivity at 2 other UMFA cutoff values using data for everybody: 2 nmol/L and 5 nmol/L, corresponding to the 95th and 98th percentile for fasting persons in our dataset, respectively. All statistical comparisons were evaluated at a 2-sided significance level of α = 0.05.

Results

Characteristics of the study population and of 5-methylTHF and UMFA concentrations.

Among US persons ≥1 y of age in the unweighted one-third subsample of NHANES 2007–2008, 26% were children (1–11 y), 12% adolescents (12–19 y), and 62% adults (≥20 y), nearly half were female (49%), more than one-third were NHW (41%), 38% were fasting for ≥8 h, and 27% reported using a folic acid–specific dietary supplement in the last 30 d (Supplemental Table 3). The ranges of serum 5-methylTHF and UMFA concentrations were 5.37–190 and <LOD (0.3)–397 nmol/L, respectively. UMFA concentrations were detectable in >95% of samples regardless of demographic, fasting, or supplement use characteristics (Supplemental Table 4). When we grouped samples according to weighted 5-methylTHF quartiles, the absolute mean UMFA concentration in quartile 4 (5.60 nmol/L) was significantly higher than in quartiles 1–3 (1.81, 1.41, and 3.41 nmol/L, respectively) (Supplemental Table 5), but the relative UMFA contribution to the sum of 5-methylTHF + UMFA was fairly constant (3.1–5.0%) across the quartiles. When we limited the samples to fasting (≥8 h) persons, the relative UMFA contribution appeared somewhat lower (2.2–3.5%). We observed the most distinction in the relative UMFA contribution across quartiles when we limited the samples to fasting nonusers (3.6, 2.3, 1.9, and 1.5%, respectively).

Bivariate associations between folate biomarkers or folic acid intake and variables of interest.

For users of folic acid–containing dietary supplements (SUP+) compared with nonusers (SUP−), the overall geometric mean serum concentrations of 5-methylTHF (48.4 nmol/L vs. 30.7 nmol/L) and UMFA (1.54 nmol/L vs. 0.794 nmol/L) were significantly higher (Table 1). Similarly, the overall median total folic acid intake was significantly higher in the SUP+ vs. SUP− group (508 μg vs. 145 μg), whereas the dietary folic acid intake was comparable (135 μg vs. 145 μg) in the 2 groups (Table 2). Age was a significant factor for 5-methylTHF, UMFA, serum total folate, and RBC total folate in both the SUP+ and SUP− groups and we noted a generally U-shaped age pattern (Table 1). Age was also a significant factor for folic acid intake, but here we noted an inverse U-shaped age pattern for dietary folic acid intake (Table 2). Although females (SUP+ and SUP−) had significantly higher 5-methylTHF and serum total folate concentrations, there were no significant sex differences for UMFA and RBC folate concentrations (Table 1). Females had significantly lower dietary folic acid intake, but significantly higher supplemental folic acid intake (Table 2). We found significant associations between race-ethnicity and the folate biomarkers in the SUP+ and SUP− groups with the exception of 5-methylTHF and UMFA in the SUP− group (Table 1). We found significant associations between race-ethnicity and dietary folic acid intake, and supplemental and total folic acid intake, with NHW persons having higher supplemental folic acid intakes (Table 2). As expected, fasting status was significantly associated with lower UMFA concentrations in the SUP+ and SUP− groups, with larger concentration differences between persons who were not fasting (<3 h) and those who were fasting (≥8 h) in the SUP+ group (1.17 nmol/L) compared with the SUP− group (0.269 nmol/L) (Table 1). We found no significant association between fasting status and folic acid intake, except for dietary folic acid intake in nonusers (Table 2). After we controlled for selected covariates (age, sex, race-ethnicity, and fasting status), the interpretation of associations based on statistical significance remained unchanged except for serum UMFA in users being no longer different by race-ethnicity (Table 1); serum total folate in nonusers being no longer different by fasting time (Table 1); and folic acid intake from diet in users being no longer different by race-ethnicity (Table 2).

TABLE 1.

Folate biomarker concentrations by variable categories and stratified by folic acid–containing dietary supplement use in the US population ≥1 y of age, NHANES 2007–2008¹

	Nonusers				Users
	Serum			RBC	Serum			RBC
Variables	5-methylTHF, nmol /L	UMFA, nmol/L	Total folate, nmol/L	Total folate, nmol/L	5-methylTHF, nmol/L	UMFA, nmol/L	Total folate, nmol/L	Total folate, nmol/L
All	30.7 (29.3, 32.1)	0.794 (0.735, 0.857)	33.4 (31.8, 35.0)	1020 (974, 060)	48.4* (46.6, 50.1)	1.54* (1.42, 1.66)	54.0* (51.9, 56.1)	1440* (1390, 1480)
Age, y
1–5	45.5 (42.9, 48.3)	1.09 (0.960, 1.23)	51.3 (47.9, 54.8)	1060 (993, 1130)	50.1* (45.1, 55.6)	2.12* (1.13, 3.99)	60.0 (49.9, 72.0)	1160 (1030, 1310)
6–11	46.5 (45.0, 48.2)	1.19 (0.996, 1.43)	50.9 (48.2, 53.8)	1120 (1070, 1180)	51.6* (48.0, 55.5)	1.25 (0.900, 1.74)	56.2 (51.4, 61.5)	1120 (1050, 1200)
12–19	33.5 (30.7, 36.6)	0.877 (0.762, 1.01)	36.5 (33.1, 40.2)	991 (902, 1090)	46.2* (40.3, 53.0)	0.882 (0.648, 1.20)	49.6* (43.6, 56.5)	1160* (1010, 1340)
20–39	26.1 (24.0, 28.5)	0.698 (0.635, 0.766)	28.5 (26.3, 30.8)	935 (888, 984)	39.1* (35.1, 43.6)	1.06* (0.888, 1.27)	42.9* (38.2, 48.1)	1250* (1140, 1360)
40–59	27.5 (25.9, 29.3)	0.716 (0.644, 0.796)	29.8 (27.8, 32.1)	1030 (966, 1110)	44.6* (41.4, 48.0)	1.34* (1.05, 1.72)	49.0* (44.5, 54.0)	1420* (1350, 1500)
≥60	32.7 (29.8, 35.7)	0.805 (0.731, 0.885)	35.0 (31.6, 38.9)	1130 (1030, 1240)	62.2* (55.3, 70.0)	2.70* (2.29, 3.18)	72.1* (66.7, 77.9)	1870* (1770, 1980)
P value	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
P value adjusted	<0.0001	<0.0001	<0.0001	<0.0001	0.0002	<0.0001	<0.0001	<0.0001
Sex
Male	29.8 (28.3, 31.4)	0.773 (0.688, 0.868)	32.6 (31.1, 34.2)	1030 (985, 1070)	45.8* (43.6, 48.2)	1.40* (1.20, 1.65)	51.5* (49.0, 54.1)	1390* (1330, 1460)
Female	31.6 (30.1, 33.1)	0.817 (0.770, 0.866)	34.2 (32.3, 36.3)	1010 (956, 1060)	50.6* (47.9, 53.6)	1.66* (1.49, 1.85)	56.3* (52.8, 60.0)	1480* (1420, 1540)
P value	0.0106	0.28	0.0325	0.34	0.0157	0.12	0.0403	0.06
P value adjusted	0.0150	0.41	0.0208	0.13	0.0076	0.14	0.0386	0.22
Race-ethnicity2
All Hispanic	30.7 (28.2, 33.4)	0.756 (0.666, 0.859)	34.8 (33.3, 36.4)	1000 (943, 1060)	42.4* (39.3, 45.7)	1.07* (0.868, 1.31)	46.1* (42.3, 50.2)	1180* (1100, 1270)
MA	30.5 (28.2, 33.0)	0.693 (0.620, 0.775)	34.9 (33.0, 36.9)	1020 (943, 1090)	44.0* (39.9, 48.5)	1.11* (0.821, 1.49)	47.0* (41.0, 53.8)	1180* (1070, 1310)
NHB	27.4 (25.0, 30.0)	0.894 (0.779, 1.03)	30.3 (28.2, 32.6)	907 (877, 938)	36.7* (32.5, 41.3)	1.16* (0.924, 1.45)	40.5* (36.6, 44.8)	1100* (1020, 1190)
NHW	31.3 (29.4, 33.3)	0.792 (0.735, 0.854)	33.4 (30.9, 36.2)	1050 (991, 1120)	50.3* (48.2, 52.6)	1.66* (1.53, 1.80)	56.4* (54.0, 59.0)	1510* (1460, 1560)
P value	0.06	0.23	0.0042	0.0003	0.0006	0.0002	0.0001	<0.0001
P value adjusted	0.0116	0.12	0.0003	0.0003	0.0050	0.05	0.0017	0.0004
Fasting time, h
<3	32.5 (30.4, 34.6)	0.936 (0.837, 1.05)	35.5 (33.1, 38.2)	1030 (966, 1100)	48.8* (45.8, 52.0)	2.14* (1.73, 2.64)	56.0* (51.5, 60.9)	1380* (1300, 1470)
3 to <8	29.4 (27.6, 31.3)	0.835 (0.740, 0.941)	32.1 (29.9, 34.4)	999 (959, 1040)	48.3* (43.9, 53.2)	2.24* (1.65, 3.05)	55.4* (49.4, 62.1)	1500* (1410, 1600)
≥8	29.7 (28.2, 31.2)	0.667 (0.631, 0.706)	32.1 (30.7, 33.6)	1020 (967, 1070)	48.0* (45.2, 50.9)	0.966* (0.841, 1.11)	51.6* (48.8, 54.6)	1460* (1400, 1520)
P value	0.0196	<0.0001	0.0156	0.60	0.93	<0.0001	0.29	0.24
P value adjusted	0.0195	<0.0001	0.23	0.69	0.47	<0.0001	0.25	0.95

¹Values are weighted geometric means (95% CIs). Serum 5-methylTHF and UMFA were measured by LC-tandem MS. Serum and RBC total folate were measured by microbiologic assay. For sample sizes, see Supplemental Table 1. Pregnant and lactating women were excluded; otherwise pairwise deletion was used for missing data. P values [with and without controlling for additional covariates (age, sex, race-ethnicity, and fasting time)] are based on the Wald F test, which tests whether at least 1 of the means across the categories is significantly different. *Significant difference between users and nonusers of folic acid–containing dietary supplements for that subgroup, P < 0.05. MA, Mexican American; NHB, non-Hispanic black; NHW, non-Hispanic white; UMFA, unmetabolized folic acid; 5-methylTHF, 5-methyltetrahydrofolate.

²All Hispanic subgroup represents the sum of MA and other Hispanic (not shown) ethnicity. P value represents the comparison of all Hispanic, NHB, NHW, and other (not shown).

TABLE 2.

Folic acid intake from different sources by variable categories and stratified by folic acid–containing dietary supplement use in the US population ≥1 y of age, NHANES 2007–2008¹

	Nonusers	Users
Variables	Diet, μg/d	Diet, μg/d	Supplements, μg/d	Total, μg/d
All	145 (137, 156)	135 (126, 148)	400 (379, 400)	508 (482, 539)
Age, y
1–5	126 (103, 149)	86.2 (62.5, 121)*	198 (181, 262)	345 (306, 389)
6–11	183 (157, 218)	141 (115, 206)	183 (138, 259)	373 (320, 458)
12–19	161 (145, 200)	187 (97.3, 279)	159 (106, 389)	479 (282, 602)
20–39	137 (121, 167)	149 (126, 195)	325 (266, 381)	490 (436, 592)
40–59	146 (130, 168)	135 (96.7, 178)	385 (380, 390)	542 (479, 597)
≥60	127 (104, 144)	126 (101, 165)	400 (400, 400)	589 (534, 626)
P value	<0.0001	0.0069	<0.0001	<0.0001
P value adjusted	<0.0001	0.0414	<0.0001	0.0018
Sex
Male	170 (154, 181)	164 (142, 181)	390 (325, 392)	497 (474, 547)
Female	128 (117, 140)	116 (91.4, 137)	400 (400, 400)	515 (481, 550)
P value	<0.0001	0.0108	0.0029	0.20
P value adjusted	<0.0001	0.0058	0.0429	0.50
Race-ethnicity2
All Hispanic	146 (134, 163)	152 (124, 203)	283 (222, 391)	491 (407, 576)
MA	133 (115, 161)	143 (109, 207)	254 (174, 391)	451 (356, 602)
NHB	125 (115, 135)	112 (89.3, 135)	267 (201, 354)	423 (374, 461)
NHW	145 (133, 160)	135 (122, 155)	400 (400, 400)	520 (484, 562)
P value	0.0165	0.0230	0.0013	0.0003
P value adjusted	0.0361	0.10	0.0108	0.0065
Fasting time, h
<3	166 (140, 179)	127 (107, 162)*	390 (270, 393)	479 (436, 528)
3 to <8	145 (130, 160)	113 (82.3, 180)	383 (367, 393)	540 (499, 605)
≥8	141 (129, 152)	148 (135, 170)	400 (348, 400)	536 (484, 588)
P value	0.0036	0.19	0.10	0.15
P value adjusted	0.0019	0.15	0.38	0.60

¹Values are weighted medians (95% CIs). “Diet” represents information from first 24-h dietary recall. “Supplements” represents mean daily use during the past 30 d. For sample sizes, see Supplemental Table 2. Pregnant and lactating women were excluded; otherwise pairwise deletion was used for missing data. P values [with and without controlling for additional covariates (age, sex, race-ethnicity, and fasting time)] are for comparison of geometric means (not medians) and are based on the Wald F test, which tests whether at least 1 of the means across the categories is significantly different. *Significant difference in dietary folic acid intake between users and nonusers of folic acid–containing dietary supplements, P < 0.05. MA, Mexican American; NHB, non-Hispanic black; NHW, non-Hispanic white.

²All Hispanic subgroup represents the sum of MA and other Hispanic (not shown) ethnicity. P value represents the comparison of all Hispanic, NHB, NHW, and other (not shown).

Correlations between folate biomarkers and folate intake.

Spearman correlations were significant for all pairwise comparisons of folate biomarkers (P < 0.0001 in all cases) (Supplemental Figure 1 and Supplemental Table 6). We observed moderate-to-strong correlations for 5-methylTHF with serum (r > 0.9) and RBC (r > 0.5) total folate overall in the SUP+ and SUP− groups. We noted moderate correlations for UMFA (r > 0.4) with 5-methylTHF and serum total folate overall in the SUP+ group, but weak correlations (r = 0.35) in the SUP− group. The correlation between UMFA and RBC total folate was weak (r = 0.30 overall, r = 0.36 in the SUP+ group, and r = 0.17 in the SUP− group).

Spearman correlations between folate biomarkers and total folic acid intake (from diet and supplements) were significant but generally weak (r ≤ 0.4; overall: P < 0.0001 in all cases; SUP+ group: < 0.0001 < P < 0.03; SUP− group: P ≤ 0.0001 in all cases). Spearman correlations between folate biomarkers and total dietary folate intake (expressed in DFEs, not including intake from supplements) were weaker yet (overall: < 0.0001 < P ≤ 0.0004; SUP− group: P ≤ 0.0001 in all cases) and not significant in the SUP+ group except for RBC folate (r = 0.08, P = 0.0105).

Distributions of serum 5-methylTHF and UMFA.

To allow a meaningful interpretation of distributions for 5-methylTHF (Table 3) and UMFA (Table 4), we limited our sample to fasting (≥8 h) persons to minimize the transient effects of recent consumption from meals or supplements. We observed distinct distributions for the SUP+ and SUP− groups for both folate forms. The overall central 95% reference intervals (2.5–97.5 percentile) for 5-methylTHF and UMFA were 11.5–92.9 nmol/L and <LOD (0.3)–3.08 nmol/L, respectively. The cumulative frequency distributions for UMFA (Supplemental Table 7) provide a means of estimating the proportion of concentrations above and below the reported concentration groupings. Although both the nonfasting (44%) and fasting (59%) persons had the majority of UMFA concentrations between 0.5 and <1 nmol/L, almost half of the nonfasting persons but only one quarter of the fasting persons had UMFA concentrations >1 nmol/L.

TABLE 3.

Serum 5-methylTHF concentration distributions by variable categories in the fasting US population ≥1 y of age, NHANES 2007–2008¹

			Percentiles, nmol/L
Variables	Samplesize, n	Geometric mean, nmol/L	5th	25th	50th	75th	95th
All	1066	34.9 (33.3, 36.6)	13.0 (11.9, 14.6)	24.0 (22.0, 26.4)	35.2 (32.7, 37.9)	52.7 (50.0, 54.8)	79.9 (75.5, 89.3)
Age, y
1–11	95	49.0 (46.0, 52.2)	26.8* (15.4, 29.3)	38.4 (33.8, 45.6)	54.3 (46.0, 59.5)	62.6 (58.5, 65.9)	74.4* (71.7, 87.6)
12–19	139	36.0 (32.0, 40.6)	15.1* (9.71, 18.3)	27.7 (23.7, 30.4)	36.0 (32.3, 42.6)	50.3 (42.1, 57.8)	77.8* (58.5, 95.3)
20–39	257	28.3 (25.3, 31.6)	11.7 (9.88, 13.5)	20.2 (16.6, 23.5)	28.6 (25.4, 32.6)	40.5 (36.5, 47.2)	59.3 (53.8, 106)
40–59	272	33.7 (30.4, 37.4)	13.0 (11.1, 14.1)	23.4 (20.7, 26.3)	34.3 (28.9, 39.7)	52.4 (44.6, 56.4)	75.5 (69.4, 85.4)
≥60	303	45.1 (41.0, 49.7)	17.6 (15.0, 19.8)	30.0 (25.9, 33.7)	47.0 (41.6, 53.1)	67.1 (60.2, 75.8)	103 (95.0, 129)
Sex
Male	546	33.5 (31.0, 36.2)	13.0 (11.2, 15.2)	24.4 (21.6, 27.6)	34.2 (30.3, 37.9)	48.3 (45.1, 52.5)	70.8 (67.5, 79.1)
Female	520	36.3 (34.7, 38.0)	13.1 (11.9, 15.2)	24.0 (21.9, 25.4)	36.8 (33.6, 40.9)	55.6 (53.3, 57.9)	89.3 (80.3, 96.1)
Race-ethnicity2
All Hispanic	324	30.4 (26.5, 34.8)	12.0 (7.92, 13.8)	21.4 (17.3, 25.3)	30.5 (26.9, 35.9)	44.7 (41.0, 50.1)	70.0 (62.7, 74.9)
MA	188	30.5 (26.4, 35.2)	11.3* (5.43, 14.9)	21.4 (16.8, 26.6)	30.9 (27.2, 36.4)	43.5 (40.8, 50.0)	73.5* (62.4, 79.7)
NHB	229	27.0 (24.8, 29.5)	11.5 (7.42, 13.0)	17.8 (16.6, 19.4)	27.8 (23.6, 31.9)	40.9 (36.6, 45.1)	58.2 (55.7, 66.2)
NHW	478	37.8 (35.3, 40.5)	14.6 (12.2, 16.7)	26.4 (23.2, 29.5)	37.8 (34.5, 42.0)	55.4 (52.6, 59.4)	87.5 (78.6, 96.2)
Folic acid–containing  supplement
No	717	29.7 (28.2, 31.2)	12.0 (11.1, 13.1)	21.4 (19.7, 22.9)	30.3 (28.4, 32.8)	42.3 (40.4, 45.3)	64.6 (60.4, 72.4)
Yes	283	48.0 (45.2, 50.9)	19.7 (11.1, 23.3)	35.7 (32.7, 37.6)	52.1 (48.1, 54.3)	66.2 (59.1, 78.0)	95.7 (87.6, 133)

¹Values are weighted geometric means and selected percentile concentrations (95% CIs) for fasting persons (no consumption of food and supplements for ≥8 h before blood draw). Pregnant and lactating women were excluded; otherwise pairwise deletion was used for missing data. *Estimate is subject to greater uncertainty because of small cell size (n < 224). MA, Mexican American; NHB, non-Hispanic black; NHW, non-Hispanic white; 5-methylTHF, 5-methyltetrahydrofolate.

²All Hispanic subgroup represents the sum of MA and other Hispanic (not shown) ethnicity.

TABLE 4.

Serum UMFA concentration distributions by variable categories in the fasting US population ≥1 y of age, NHANES 2007–2008¹

			Percentiles, nmol/L
Variables	Sample size, n	Geometric mean, nmol/L	5th	25th	50th	75th	95th
All	1056	0.758 (0.714, 0.804)	0.308 (<LOD, 0.366)	0.547 (0.510, 0.580)	0.696 (0.671, 0.723)	0.951 (0.896, 1.02)	1.98 (1.67, 2.45)
Age, y
1–11	95	0.920 (0.709, 1.19)	0.373* (<LOD, 0.527)	0.617 (0.600, 0.663)	0.743 (0.672, 0.897)	1.04 (0.865, 1.28)	2.89* (1.65, 42.3)
12–19	138	0.734 (0.645, 0.836)	0.348* (<LOD, 0.476)	0.579 (0.483, 0.615)	0.698 (0.617, 0.802)	0.939 (0.815, 1.18)	1.48* (1.31, 3.70)
20–39	255	0.673 (0.608, 0.745)	<LOD (<LOD, 0.369)	0.506 (0.453, 0.561)	0.639 (0.590, 0.701)	0.872 (0.788, 0.936)	1.62 (1.15, 2.75)
40–59	268	0.693 (0.660, 0.728)	0.301 (<LOD, 0.340)	0.519 (0.483, 0.568)	0.693 (0.671, 0.710)	0.876 (0.809, 0.955)	1.61 (1.41, 1.98)
≥60	300	1.01 (0.860, 1.18)	0.330 (<LOD, 0.422)	0.597 (0.563, 0.629)	0.863 (0.757, 0.953)	1.18 (1.09, 1.40)	7.58 (3.06, 35.5)
Sex
Male	541	0.696 (0.636, 0.762)	<LOD	0.496 (0.453, 0.546)	0.670 (0.606, 0.718)	0.899 (0.832, 0.992)	1.95 (1.48, 2.82)
Female	515	0.824 (0.762, 0.891)	0.379 (0.306, 0.427)	0.592 (0.561, 0.614)	0.713 (0.697, 0.760)	0.997 (0.927, 1.11)	2.10 (1.61, 4.60)
Race-ethnicity2
All Hispanic	324	0.642 (0.571, 0.723)	<LOD	0.483 (0.394, 0.552)	0.645 (0.595, 0.694)	0.847 (0.784, 0.933)	1.33 (1.19, 1.68)
MA	188	0.607 (0.523, 0.704)	<LOD*	0.441 (0.336, 0.534)	0.609 (0.538, 0.690)	0.837 (0.729, 0.936)	1.20* (1.17, 1.46)
NHB	225	0.797 (0.709, 0.896)	0.326 (<LOD, 0.453)	0.554 (0.501, 0.619)	0.780 (0.714, 0.874)	1.03 (0.927, 1.25)	2.06 (1.56, 3.10)
NHW	472	0.789 (0.728, 0.856)	0.352 (<LOD, 0.396)	0.563 (0.515, 0.600)	0.698 (0.670, 0.729)	0.959 (0.898, 1.05)	2.07 (1.64, 4.58)
Folic acid–containing  supplement
No	712	0.667 (0.631, 0.706)	0.302 (<LOD, 0.351)	0.518 (0.484, 0.556)	0.655 (0.619, 0.682)	0.856 (0.806, 0.919)	1.42 (1.31, 1.73)
Yes	281	0.966 (0.841, 1.11)	0.339 (<LOD, 0.433)	0.630 (0.535, 0.689)	0.825 (0.707, 0.918)	1.13 (1.05, 1.38)	5.44 (2.24, 40.4)

¹Values are weighted geometric means and selected percentile concentrations (95% CIs) for fasting persons (no consumption of food and supplements for ≥8 h before blood draw). Pregnant and lactating women were excluded; otherwise pairwise deletion was used for missing data. LOD for UMFA is 0.3 nmol/L. *Estimate is subject to greater uncertainty because of small cell size (n < 224). LOD, limit of detection; MA, Mexican American; NHB, non-Hispanic black; NHW, non-Hispanic white; UMFA, unmetabolized folic acid.

²All Hispanic subgroup represents the sum of MA and other Hispanic (not shown) ethnicity.

Bivariate associations between the prevalence of UMFA concentrations >1 nmol/L and variables of interest to explain UMFA concentrations >1 nmol/L.

The prevalence of UMFA concentrations >1 nmol/L was 33.2% overall and 21.0% in fasting (≥8 h) adults (Table 5). The prevalence varied significantly by age group, sex, race-ethnicity, consumption of a folic acid–containing supplements, dietary folic acid intake, alcohol intake (only available for adults), smoking status, and folate status (as represented by RBC folate concentrations) in the overall group and in fasting adults. In the overall group, the prevalence also varied significantly by BSA and fasting status. The prevalence of UMFA concentrations >2 and >5 nmol/L was 13.9% and 7.0%, respectively, in the overall group and 5.2% and 1.9%, respectively, in fasting adults (data not shown).

TABLE 5.

Prevalence of UMFA concentrations >1 nmol/L by variable categories in US participants ≥1 y of age, NHANES 2007–2008¹

	Fasting and nonfasting participants ≥1 y		Fasting (≥8 h) adults (≥20 y)
Variables	Sample size, n	Prevalence, %	Sample size, n	Prevalence, %
All	2676	33.2 (29.7, 36.9)	823	21.0 (17.7, 24.7)
Age, y
1–5	263	41.2 (33.7, 49.1)	—	—
6–11	337	44.8 (36.3, 53.6)	—	—
12–19	337	33.7 (26.4, 41.9)	—	—
20–39	536	22.7 (19.2, 26.6)	255	13.9 (9.1, 20.7)
40–59	560	29.0 (23.3, 35.4)	268	17.6 (13.5, 22.7)
≥60	643	48.3 (42.7, 54.0)	300	36.2 (29.8, 43.1)
P value		<0.0001		0.0001
Sex
Male	1378	30.1 (25.4, 35.3)	411	16.3 (11.4, 22.8)
Female	1298	36.3 (32.5, 40.2)	412	25.3 (21.2, 30.0)
P value		0.0277		0.0177
Race-ethnicity2
All Hispanic	875	25.7 (22.0, 29.8)	245	10.9 (7.0, 16.6)
NHB	563	35.8 (28.4, 44.0)	162	26.0 (17.6, 36.6)
NHW	1121	34.7 (30.5, 39.1)	391	22.6 (18.1, 27.9)
P value		0.0040		0.0159
Fasting time, h
<3	1047	41.9 (36.8, 47.1)	—	—
3 to <8	573	41.7 (35.8, 47.8)	—	—
≥8	1056	21.5 (18.2, 25.3)	—	—
P value		<0.0001		—
Folic acid–containing supplement3
Yes	688	51.1 (47.8, 54.3)	241	35.2 (28.0, 43.2)
No	1857	24.9 (20.8, 29.6)	523	12.9 (8.9, 18.3)
P value		<0.0001		0.0001
Total folic acid intake,4 μg/d
<100	596	18.3 (14.0, 23.6)	209	8.2 (4.1, 15.6)*
100 to <200	609	22.4 (17.5, 28.1)	168	12.5 (7.4, 20.4)
200 to <400	614	31.2 (27.3, 35.4)	157	15.0 (9.9, 22.0)
≥400	630	54.9 (50.1, 59.6)	214	38.7 (31.6, 46.3)
P value		<0.0001		0.0001
Alcohol intake,5 drinks/d
No drinks	—	—	253	24.0 (19.2, 29.5)
<1	—	—	404	21.1 (15.7, 27.8)
1 to <2	—	—	71	19.8 (11.6, 31.6)
≥2	—	—	39	8.2 (2.0, 28.5)**
P value		—		0.0403
Smoking status6
Nonsmoker	2029	36.4 (32.6, 40.3)	615	24.2 (20.7, 28.1)
Smoker	487	21.2 (16.0, 27.6)	207	11.7 (6.3, 20.9)
P value		0.0009		0.0131
Body surface area,7 m2
<1.5	612	42.8 (37.6, 48.2)	22	17.2 (4.5, 47.7)**
1.5 to <1.8	686	35.0 (29.0, 41.6)	238	26.4 (19.4, 34.9)
1.8 to <2.0	578	35.2 (28.7, 42.3)	229	21.2 (14.7, 29.7)
≥2	685	25.4 (21.1, 30.2)	316	16.0 (11.7, 21.5)
P value		0.0005		0.06
RBC folate,8 nmol/L
<800	504	16.3 (10.4, 24.6)	167	9.1 (5.2, 15.4)
800 to <1200	1051	26.5 (22.4, 31.1)	278	11.9 (7.0, 19.5)
1200 to <1500	555	34.5 (28.6, 41.0)	141	20.3 (10.8, 34.7)
≥1500	557	53.4 (48.2, 58.4)	234	37.3 (30.2, 44.9)
P value		<0.0001		<0.0001

¹Values are percentages (95% CIs). Pregnant and lactating women were excluded; otherwise pairwise deletion was used for missing data. Fasting adults did not consume any food or supplements for ≥8 h before blood draw. *30% > RSE ≤ 40%; **RSE >40%; all other values had an RSE ≤30%. NHB, non-Hispanic black; NHW, non-Hispanic white; RSE, relative standard error; UMFA, unmetabolized folic acid.

²All Hispanic subgroup represents the sum of Mexican American (not shown) and other Hispanic (not shown) ethnicity.

³Use of folic acid–containing supplement during the last 30 d.

⁴From folic acid–containing supplement use during the last 30 d and from dietary folic acid intake from first 24-h dietary recall.

⁵Calculated as the mean daily number of “standard” drinks [(quantity × frequency)/365.25]; 1 drink ∼15-g ethanol; only available for persons ≥20 y of age.

⁶“Smoker” defined by a serum cotinine concentration >10 μg/L.

⁷Calculated as the square root of [(height in cm × weight in kg)/3600] or the square root of [(height in inches × weight in pounds)/3131].

⁸RBC folate was measured by microbiologic assay.

Multiple logistic regression modeling to explain UMFA concentrations >1 nmol/L.

The following variables were significant in our model 1 for persons ≥1 y of age when we used a UMFA cutoff of >1 nmol/L: age (P = 0.0164), race-ethnicity (P = 0.0407), fasting status (P < 0.0001), total folic acid intake (from diet and supplements, P < 0.0001), RBC folate (P < 0.0001), and BSA (P = 0.028) (Table 6). After controlling for these variables, young (20–39 y) and middle-aged (40–59 y) adults were 2 times less likely of having UMFA concentrations >1 nmol/L compared with persons ≥60 y of age. NHB persons were 1.8 times more likely of having UMFA concentrations >1 nmol/L compared with NHW persons. Nonfasting persons were 3.5 times more likely of having UMFA concentrations >1 nmol/L compared with fasting persons. The odds of having elevated UMFA concentrations were 1.5 times higher for a 200-μg/d increase in total folic acid intake from diet and supplements and 1.2 times higher for a 200-nmol/L increase in RBC folate. Similarly, the odds were 2 times more likely for a 1-m² decrease in BSA [i.e., child with a BSA of 1 m² (e.g., height 48 in and weight 66 lbs) compared with an adult with a BSA of 2 m² (e.g., height 73 in and weight 175 lbs)]. We found a significant pairwise interaction between RBC folate × fasting status (P = 0.0151) that we did not fit because it did not change the overall interpretation of the model [OR_adj = 1.3 (nonfasting); OR_adj = 1.1 (fasting) for a 200-nmol/L increase in RBC folate]. A modification of model 1 with dietary folic acid intake and folic acid intake from supplements entered as 2 separate variables produced β coefficients that were not significantly different, indicating that it was reasonable to combine the 2 variables into 1 total folic acid intake variable. To verify whether our model would arrive at a different conclusion if we considered the overall folate intake from all sources (natural food folate, folic acid added to foods, and folic acid consumed through supplements), we evaluated a modification of model 1 with food folate intake as a new variable in addition to total folic acid intake. Food folate intake was not significant (β = −0.0002, P = 0.76). In addition, adding food folate had no impact on the statistical significance nor the OR_adj, which remained the same as in the original model.

TABLE 6.

Determinants of UMFA concentrations >1 nmol/L¹

	Model 1: Fasting and nonfasting participants ≥1 y of age		Model 2: Fasting (≥8 h) adults (≥20 y of age)
Variables	ORadj	P value	ORadj	P value
Age group,2 y
1–5	0.4 (0.2, 1.1)	0.08	—	—
6–11	0.7 (0.3, 1.5)	0.33	—	—
12–19	0.9 (0.5, 1.4)	0.58	—	—
20–39	0.5 (0.3, 0.7)	0.0008	0.5 (0.2, 1.1)	0.09
40–59	0.5 (0.3, 0.9)	0.0147	0.5 (0.3, 0.9)	0.0206
≥60	REF		REF
Race-ethnicity3
All Hispanic	0.9 (0.7, 1.1)	0.23	0.7 (0.4, 1.4)	0.32
NHB	1.8 (1.2, 2.8)	0.0127	2.1 (1.0, 4.3)	0.0393
NHW	REF		REF
Fasting ≥8 h
Yes	REF		—	—
No	3.5 (2.7, 4.5)	<0.0001	—	—
Total folic acid intake (any 200-μg/d increase)4	1.5 (1.4, 1.6)	<0.0001	1.5 (1.2, 1.7)	0.0001
RBC folate5 (any 200-nmol/L increase)	1.2 (1.2, 1.3)	<0.0001	1.2 (1.1, 1.3)	<0.0001
BSA6 (any 1-m2 decrease)	2.0 (1.1, 3.3)	0.028	2.3 (0.8, 6.7)	0.13
Alcohol intake7 (any 1-drink/d decrease)	—	—	1.3 (1.1, 1.6)	0.0138

¹Values are ORs_adj (95% CIs) from a multiple logistic regression model. Pregnant and lactating women were excluded. Model 1: n = 2343; model 2: n = 693; participants with missing data for 1 or several covariates were excluded via listwise deletion. P value is from the Wald F test. BSA, body surface area; NHB, non-Hispanic black; NHW, non-Hispanic white; OR_adj, adjusted OR; REF, reference; UMFA, unmetabolized folic acid.

²The final models, which included age with a quadratic term, were reported with use of a categorized version of the continuous variable to facilitate the interpretation of the effect of age.

³All Hispanic subgroup represents the sum of Mexican American and other Hispanic (not shown) ethnicity.

⁴From folic acid–containing supplement use during the last 30 d and from dietary folic acid intake from first 24-h dietary recall.

⁵RBC folate was measured by microbiologic assay.

⁶Calculated as the square root of [(height in cm × weight in kg)/3600] or the square root of [(height in inches × weight in pounds)/3131].

⁷Calculated as the mean daily number of “standard” drinks [(quantity × frequency)/365.25]; 1 drink ∼15 g ethanol; only available for persons ≥20 y of age.

When we limited our model 2 to fasting adults (UMFA >1 nmol/L) and added alcohol consumption as an additional variable, our findings were generally similar (Table 6). Total folic acid intake (P = 0.0001) and RBC folate (P < 0.0001) were still significant variables in model 2 and the β coefficients were generally comparable to model 1 (data not shown). Age (P = 0.07), race-ethnicity (P = 0.13), and BSA (P = 0.13) were no longer significant variables, possibly because of the loss of power in model 2. Alcohol consumption was significant (P = 0.0138). After controlling for these variables, middle-aged (40–59 y) adults compared with persons ≥60 y of age were 2 times less likely, and NHB persons compared with NHW persons were 2.1 times more likely, of having UMFA concentrations >1 nmol/L. The odds of having elevated UMFA concentrations were 1.5 times higher for a 200-μg/d increase in total folic acid intake from diet and supplements, 1.2 times higher for a 200-nmol/L increase in RBC folate, and 1.3 times higher for a 1-drink/d decrease in alcohol intake.

Our sensitivity analysis at 2 higher UMFA cutoff values (2 nmol/L and 5 nmol/L) identified similar predictors, with age, fasting status, total folic acid intake, and RBC folate still being significant variables; additionally, use of any dietary supplement in the last 30 d emerged as a significant variable (P = 0.0016 for UMFA >2 nmol/L; P = 0.0087 for UMFA >5 nmol/L; data not shown).

Discussion

To our knowledge, this paper presents the first data for serum 5-methylTHF and UMFA concentrations in a nationally representative sample of the US population ≥1 y of age where we have readily available information on dietary intake and supplement use. Our results raise questions as to what UMFA concentrations mean and whether they contribute useful information to the assessment of folate status. We confirmed previous findings that UMFA concentrations are associated with folic acid intake (9–14, 17) and that UMFA is closely associated (r = ∼0.3–0.6) with 5-methylTHF (9, 14) and total folate concentrations (11). UMFA was detected in nearly everyone, and concentrations >1 nmol/L were mainly associated with being nonfasting and having higher folic acid intake from diet and supplements. However, at the same total folic acid intake and fasting status, certain groups were more likely to have UMFA concentrations >1 nmol/L: NHB persons, older persons, persons with a smaller BSA, and persons with decreased alcohol consumption.

These findings raise the possibility that UMFA concentrations may be an indicator of a mismatch between supply (i.e., intakes) and cellular demand (i.e., requirements). For example, the lower serum and RBC folate concentrations for NHB persons compared with all Hispanics and NHW persons, also observed here, have always created a paradox with respect to our understanding of a causal relation between folate status and neural tube defects rates. NHB persons have both the lowest neural tube defects rates and the lowest blood folate concentrations compared with other race-ethnic groups (34). Although NHB persons in the SUP− group in our study had lower 5-methylTHF concentrations but comparable UMFA concentrations compared with NHW persons, being NHB almost doubled their odds of having UMFA concentrations >1 nmol/L. Whether this reflects a lower requirement for dietary folic acid in this population subgroup or genetic differences in how efficiently folic acid is metabolized is unknown. Human liver has been shown to have low and variable dihydrofolate reductase activity (35), and it has been suggested that a polymorphism in dihydrofolate reductase may limit folic acid assimilation into cellular folate stores at high and low intakes leading to increased plasma UMFA (36).

The higher serum UMFA concentrations in children or older adults compared with other age groups may also be a reflection of lower requirements relative to their intakes. For example, the estimated average requirements of folate for children 1–3 y and 4–8 y of age compared with adults ≥19 y of age are 120, 160, and 320 μg DFE/d, respectively (37), thus approximately one-third to one-half that of adults. Yet the folic acid intakes for children 1–11 y of age in the SUP+ group in this study were approximately two-thirds that of adults. Our finding that a smaller BSA was associated with higher UMFA concentrations seems plausible because children have a higher intake-to-BSA ratio compared with adults. In another example, older adults in the SUP+ group may have higher UMFA concentrations than younger adults because their intakes were higher although their estimated average requirements are the same. And finally, we found an inverse association between UMFA concentrations and alcohol consumption even though serum and RBC folate concentrations were similar across categories of increasing alcohol consumption (data not shown).This may reflect altered folate metabolism because of the known adverse interactions (inhibited absorption and metabolism and/or increased urinary losses) between alcohol consumption and folate utilization (38). Although the speculations described here are suggestive of a potential use for UMFA in folate status assessments, more research is needed to confirm that these hypotheses are in fact valid.

Challenging our understanding of folic acid biotransformation, a recent human intervention study showed that 15 min after a 220-μg dose of folic acid, 80% of labeled folate in the hepatic portal vein was unmodified folic acid (6). The human gut thus appears to have limited ability to reduce folic acid, whereas a 5-formylTHF dose was converted very efficiently to 5-methylTHF (4% unmodified 5-formylTHF in the portal vein). The authors suggested that “chronic exposure to folic acid in physiologic doses (as would be the case with mandatory fortification) may induce saturation and explain the observed systemic circulation of UMFA” (6). We detected UMFA in 96% and 98% of the SUP− and SUP+ groups, respectively. The higher UMFA geometric mean in the SUP+ group (1.54 nmol/L) compared with the SUP− group (0.794 nmol/L) suggested that some of the UMFA in the SUP+ group likely comes from consumption of fortified cereal-grain foods and/or ready-to-eat cereals.

Earlier studies that measured UMFA or 5-methylTHF did not represent demographically diverse populations. Kalmbach et al. (10) evaluated determinants of UMFA concentrations ≥1.35 nmol/L in adult participants of the Framingham Offspring Cohort before (n = 1103) and after (n = 600) fortification. Dietary folic acid intake, total folate intake, B vitamin supplement use, and plasma folate concentrations were positively associated with the prevalence of high UMFA concentrations, but age, sex, BMI, smoking, alcohol intake, caffeine intake, natural folate intake, and plasma concentrations or dietary intake of vitamins B-6 and B-12 were not.

Serum UMFA and 5-methylTHF concentrations appeared higher in the United States where there is fortification and approximately one-third of the population uses folic acid supplements [as shown in this study, Kalmbach et al. (10), and Bailey et al. (17)] compared to small convenience sample reports from countries with lower dietary supplement use and with voluntary fortification [Ireland, n = 135 (11); Germany, n = 25 (8) and n = 37 (13)], although this was merely an ecologic observation (Supplemental Figure 2). However, aside from the fact that these data come from different populations, caution should be used when comparing data from different studies because of method differences, including LOD differences. Previous studies reported different UMFA detection rates even with the same method (Tufts LOD = 0.18 nmol/L): 38% (16), 67% (9), and 78% (39). The Irish studies (11, 12) reported detection rates (85–94%, LOD = ∼0.07–0.15 nmol/L) similar to this study (>95%, LOD = 0.30 nmol/L). To assess the comparability of the Tufts and CDC method, we performed a method comparison on 311 surplus sera from American seniors (≥60 y of age) in the NHANES 2001–2002 (40). Serum 5-methylTHF concentrations were comparable (no bias, Pearson r = 0.89), whereas UMFA concentrations were not (CDC method detected UMFA in all samples; CDC results were on average 33% higher than Tufts results in a subset of 152 samples with detectable UMFA results by the Tufts method after excluding 11 extreme outliers, Pearson r = 0.80), thereby restricting the comparability of the current to the previous UMFA data in the NHANES.

Our study has some limitations. We were limited in the number of stratifications we could perform with a one-third subset 2-y NHANES data set. We did not estimate usual intake, which is preferred to describe the distribution of intakes in the population. Instead we focused on mean intakes collected in the MEC at the time of the blood draw so as to have these 2 indicators of folate status reflect the same time period. We did not investigate the association of UMFA with sources of folic acid intake (fortified cereal-grain foods, ready-to-eat cereals, supplements, and combinations thereof). The strengths of our analysis are several fold. The inclusion of all age groups provides the first nationally representative data for children and adolescents, in addition to adults, in a demographically diverse population. The availability of dietary intake and supplement use information as well as demographic and lifestyle characteristics allowed us to evaluate a multitude of factors that may be related to high UMFA concentrations. UMFA was detectable in nearly all samples, maximizing the power of our study. Our HPLC-MS/MS procedure matched the uncertainty limits of the NIST-certified value for 5-methylTHF and the reference value for UMFA, proving that this method yielded comparable results to NIST’s reference measurement procedure. In summary, our findings suggest that UMFA concentrations cannot be solely explained by total folic acid intake and fasting status. They may indicate a mismatch between supply and cellular demand, but some common genetic polymorphisms may also have an impact (36). Confirmation of our findings in other populations will be important to enhance our understanding of how to interpret folate forms in the context of folate status and health outcomes. Given that in the United States we have both folic acid fortification and a high rate of supplement use, it remains important to monitor the folate vitamer concentrations; this work serves to establish the first nationally representative data for such purposes.