Recently, a great deal of attention has been directed toward identifying determinants of circulating levels of vitamin D metabolites in humans.1, 2 There is sparse information regarding such measures in young children. Even less information is available for minority children, a group widely considered to be at higher risk for vitamin D deficiency. In a large urban cohort of infants and toddlers from the northeast United States (n > 750), we reported on demographic, nutritional, and biochemical variables associated with circulating vitamin D metabolite levels.3 Our findings indicated that significantly greater 25-hydroxyvitamin D (25-OHD) levels in this cohort were present in children of younger age, with lesser skin pigmentation, in those receiving formula feeds, during the summer or fall, and in association with lower circulating parathyroid hormone (PTH) levels.
Because there has been substantial concern in urban communities with respect to the risk of vitamin D deficiency, we have extended this analysis to identify whether ancestral background or specific genetic effects may also serve as determinants of the circulating 25-OHD level. Indeed, our cohort is genetically heterogeneous, but largely comprised of Hispanic and African American families. These groups are generally perceived to be at increased risk for developing overt skeletal manifestations of vitamin D deficiency, namely rickets.4 Further analysis of our cohort data therefore focuses on the potential role of common variants in genes identified to be determinants of circulating 25-OHD. Others have used similar approaches to the identification of genetic determinants of circulating 25-OHD levels in adult cohorts. Among these, the common single nucleotide polymorphisms (SNPs) in the GC gene encoding vitamin D binding protein (DBP) have been established as strong determinants of circulating 25-OHD in several genomewide association studies5–8 and in other cohorts9–11 as well, but all have largely been limited to adults. To address the substantial admixture present in our cohort, we derived quantitative scores of African, European, or Native American ancestry. These scores, based upon a validated set of more than 100 ancestry informative markers (AIMs) were used as a covariate in our extended analyses of 25-OHD levels and their genetic determinants. This study focuses on the common missense SNPs (the D432E and T436K variants) in the GC gene encoding biochemically polymorphic DBP species. These variants have previously been shown to be strongly associated with circulating 25-OHD. In addition, we looked at circulating GC levels in individuals in the cohort, in an effort to clarify the relationships between GC genotype and circulating DBP concentrations, and the potential contribution of each of these factors to the circulating concentrations of 25-OHD. Finally, we have constructed a multivariate model to examine these effects, taking into account the nutritional and demographic variables established in our initial study.3
Subjects and Methods
Subjects and Study Design
Over 750 healthy children were recruited during well-child visits to one of four neighborhood health clinics in a midsized northeastern U.S. urban community (New Haven, CT, USA). Recruitment was limited to subjects 6 to 36 months old, the age range in which we have most commonly observed nutritional rickets,4 and were products of pregnancies greater than 28 weeks gestational age. Children were excluded from the study if there was any prior history of vitamin D or mineral metabolic disorders or if medications known to affect vitamin D metabolism (eg, systemic glucocorticoids, pharmacologic vitamin D metabolites, or vitamin D supplements in excess of 400 IU/d) were being administered. The study was approved by the Yale University Human Investigation Committee and written informed consent was obtained from the appropriate parent or guardian.
Self-reported ancestry was determined by asking for the origin of the four grandparents of each subject. When three grandparents were similarly categorized as Hispanic, Caucasian, or African American, the subject was reported to be of that ancestry, and if there were not three grandparents of the same group, then the ancestry was designated as “other,” based on the likely admixture.
Skin type assessment employed the Fitzpatrick skin pigmentation scale.12 One of four grades of pigmentation was recorded, based on comparison with a color chart simultaneously examined at the interview. A blood sample was obtained for collection of DNA and serum for measurement of vitamin D metabolites, and other measures, as reported.3
The p.D432E (rs7041) and p.T436K (rs4588) GC SNPs were genotyped, with phase assignment based on allele-specific amplification of the p.T436K site, followed by restriction endonuclease digestion of the p.D432E site, as described.9 Allele-specific amplification was carried out in a 20-µL reaction mixture containing 1× PCR buffer (Qiagen, Toronto, Canada), 0.2 mM each of the four deoxynucleoside triphosphates (dATP, dGTP, dTTP, and dCTP), 50 ng genomic DNA, 0.5 U HotStarTaq (Qiagen), and 0.3–1 µM of the following primers: 5′-GGCATGTTTCACTTTCTGATCTC-3′ (forward), 5′-ACCAGCTTTGCCAGTACCG-3′ (wild-type reverse), and 5′-GCAAAGTCTGAGTGCTTGTTATGCAGCTTTGCCAGTTGCT-3′ (mutant reverse). The underlined bases in the primer sequences are mismatched nucleotides introduced to avoid cross-priming. After the initial DNA denaturation and HotStarTaq activation at 95°C for 15 minutes, the amplification went through 35 cycles of denaturation at 94°C for 20 seconds, annealing at 58°C for 20 seconds, and extension at 72°C for 20 seconds with an increment of 1 second after each subsequent cycle and a final extension at 72°C for 5 minutes at the finish. Eight microliters of the amplified products were run in a 2% NuSieve gel containing ethidium bromide and then visualized using UV illumination. The p.T436K wild-type allele produced a 246–base pair (bp) band and the mutant allele a 270-bp band. Another 8 µL of amplicon was digested with HaeIII (New England BioLabs) at 37°C overnight and the digested products were analyzed by repeat electrophoresis in 2% NuSieve gel. The DNA bands containing the wild-type p.D432E allele remained unchanged whereas those with p.D432E mutant allele were cut, producing 221-bp fragments.
Given the absence of any recombinants between the two polymorphic loci, assignment of a diplotype for each subject, based on the three haplotype alleles—wild-type (electrophoretic variant 1f), mutant 432E (electrophoretic variant 1s), and mutant 436K (electrophoretic variant 2)—was unambiguous.
Estimation of ancestry proportions using AIMs
Ancestry proportions were estimated with a panel of 107 AIMs. These markers are distributed throughout the genome, and were selected based on their high allele frequency differences between European, Native American, and African populations. This panel has been used in previous admixture studies in African American and Hispanic populations.13–16 Genotyping was carried out at the Biomedical Genomics Center of the University of Minnesota. Individual ancestry proportions were estimated with the program ADMIXMAP (http://homepages.ed.ac.uk/pmckeigu/admixmap), which is a general purpose program for modeling population admixture with genotype data, based on a combination of Bayesian and classical methods. For this study, the samples were modeled as formed by admixture between European, Native American, and African populations. Allele frequency data from unadmixed individuals were used as a prior distribution for the parental allele frequencies.
Serum 25-OHD was measured by radioimmunoassay kit methodology (DiaSorin, Stillwater, MN, USA). Results of samples analyzed in our assay for serum 25-OHD are consistently found to agree with the midrange of outcomes of those using this assay and participating in the international Vitamin D External Quality Assessment Scheme (DEQAS) standardization system (http://www.deqas.org).17 The inter- and intraassay coefficients of variation for the 25-OHD assay are 9.6% and 6.6%, respectively. Serum PTH was measured as described.3
Immunonephelometry was performed with a Behring Nephelometer (Behring Diagnostics, Inc., Westwood, MA, USA) as described by Wians and colleagues.18 Briefly, a working Gc (DBP) standard solution (400 µg/mL) was prepared by manually diluting a 1 mg/mL solution of purified DBP protein (Calbiochem, La Jolla, CA, USA) 1:2.5 with N-Diluent (Behring Diagnostics, Inc.). The working standard solution and subjects' serum samples were assayed in duplicate with rabbit antihuman DBP antibody (Ab) (Dako, Carpinteria, CA, USA), supplementary precipitation reagent (Suppl.R. “P”/SRP; Behring Diagnostics, Inc.). Light scattered by antigen (Ag)-Ab complexes was measured in light-scattering intensity units or bits. DBP concentration was calculated by automated data reduction by using the logit-log function and linear regression analysis. All samples were assayed in duplicate. Interassay variation was 6.8%. An estimate of free circulating 25-OHD was calculated from the measured total 25-OHD level, and the serum level of DBP, using a formula derived for this purpose and previously reported affinity constants of 25-OHD for DBP and albumin.19 We assumed a fixed level of the serum albumin concentration for all subjects, as previously validated for calculation of serum free testosterone.20
Descriptive statistics were used to summarize the data. Chi-square tests were performed to test the distribution of genotype among different ancestral groups, especially between African Americans and others. Pearson correlation was used to evaluate the associations among 25-OHD, DBP, and AIMs scores. Analysis of variance and/or regression were performed to assess the relative contributions of AIMs score, DBP concentration and GC genotype to the circulating 25-OHD level. Multivariate regression analysis was performed to determine specific significant genetic determinants of circulating 25-OHD. Analyses were adjusted for demographic and nutritional factors found to be significantly associated with 25-OHD as determined.3 We applied a similar model, adding the interaction of GC genotype and 25-OHD to explore the role of genotype with respect to the relationship of 25-OHD to PTH levels. We used a 0.05 significance level and two-sided tests in all cases without adjustments for repeated tests. SAS (version 9.2; SAS Institute, Cary, NC, USA) was employed for all analyses.
GC Variants and Self-Reported Ancestry
Consistent with our expectations from earlier reports, there was a significant association between self-reported ancestry and GC alleles. Table 1 shows the frequency of each diplotype, haplotype, and individual SNP variants at the D432E and T436K positions in the DBP protein. African Americans were significantly enriched with the DT (1f) allele (71%, p < 0.001), and the DDTT diplotype (homozygotes for the 1f allele; 53%, p < 0.001), compared to the others. Homozygosity for the D432 allele at position rs7041 was greatest in African Americans (71%, p < 0.001), and homozygosity for the T allele at position rs4588 was also greatest in this group (77%, p = 0.015). Among DDTT-identified individuals, African Americans were the most frequently identified group compared with Hispanics (p < 0.001).
Table 1. Frequency of rs7041/rs4588 Diplotypes, Haplotypes, and Alleles of the GC Gene Among Self-Reported Ancestral Groups
White n (%)
Hispanic n (%)
African American n (%)
Other n (%)
GC diplotypes (rs7041 + rs4588)
1f-1f (DDTT) ttcc
1f-1s (DETT) gtcc
1f-2 (DDTK) ttca
1s-1s (EETT) ggcc
1s-2 (DETK) gtca
2-2 (DDKK) ttaa
1f (DT) tc
1s (ET) gc
2 (DK) ta
Allele (amino acid)
Allele (amino acid)
Ancestry Assessment by AIMs
In order to quantitatively analyze the effect of ancestry in this analysis we determined individual admixture proportions with the program ADMIXMAP, as described above in Estimation of Ancestry Proportions Using AIMs. Figure 1 depicts the individual admixture proportions estimated for each subject, classified (by color) according to self-reported ancestry. As expected, self-reported African Americans showed, on average, a high proportion of African ancestry, but some variation is evident in the extent of European contributions. Similarly, self-reported Caucasians primarily had European ancestry, although there was also evidence of African and Native American contributions in some individuals. Finally, self-reported Hispanics had the broadest distribution of individual ancestry proportions. On average, self-reported Hispanics showed a relatively high Native American ancestry, but there is considerable dispersion in the relative Native American, European, and African contributions in this group.
Correlations with Circulating 25-OHD
We next performed a correlative analysis to assess the associations between 25-OHD, circulating DBP, and the African and European admixture proportions. (As the three parental scores sum to 1.00, only two of the three AIMs scores need be included in the analysis; Table 2). Pearson correlations indicate that 25-OHD levels are positively correlated with circulating DBP (R = 0.25, p < 0.001); the regression of this relationship is shown in Fig. 2, and is defined by the equation: 25-OHD (nmol/L) = 37.3 + 0.138 DBP (mg/L). Circulating 25-OHD was positively correlated with European ancestry (R = 0.19, p < 0.001), and modestly correlated with African ancestry in an inverse manner (R = −0.09, p = 0.016).
Table 2. Pearson Correlations Among Circulating Levels of Vitamin D Metabolites, DBP, and AIMs Scores
AIM West African
AIM Native American
DBP = vitamin D binding protein; AIM = ancestry informative marker; 25-OHD = 25-hydroxyvitamin D.
We then examined the impact of GC diplotype (eg, possible combinations of rs7041 and rs4588 variants) on the circulating DBP level. Mean ± SD DBP for the entire cohort was 197 ± 29 mg/L. There was no effect of vitamin D intake or season of sampling on circulating DBP levels. As shown in Fig. 3, circulating DBP levels (mean ± SD) were greatest in subjects with the EETT diplotype (homozygotes for the allele 1s) (204 ± 31 mg/L), and values were similar for the DDTT (homozygotes for the allele 1f) and DETT (1f/1s heterozygotes) diplotypes (203 ± 30 and 202 ± 27 mg/L, respectively). However, individuals with the 436 K variant (GC2 allele) had lower circulating DBP, with DETK (1s/2 heterozygotes) (191 ± 25 mg/L) > DDTK (1f/2 heterozygotes) (181 ± 25 mg/L) > DDKK (homozygous for the 2 allele) (173 ± 32 mg/L).
Thus, when examining the effect of the SNPs independently there is no effect of the rs7041 (D432E) polymorphism on circulating DBP, but in contrast we find a significant effect of the rs4588 (T436K) marker, with progressive substitution of the K allele for the T allele resulting in lower DBP levels (p < 0.001). The adjusted mean ± SD of serum DBP levels for the TT, TK, and KK haplotypes are 203 ± 29, 185 ± 25, and 173 ± 32 mg/L, respectively. Pairwise comparisons revealed highly significant differences between TT and TK (p < 0.001), and TT and KK groups (p < 0.001), with a near significant difference between TK and KK (p = 0.077). The possibility of a D/E effect (resulting in an increase in circulating DBP) in the setting of the TK heterozygote is of interest in that DBP levels for the DETK diplotype (1s/2 heterozygote) are greater than for the DDTK genotype (1f/2 heterozygote) (p = 0.019), whereas D/E changes in the setting of TT homozygosity at position 436 are not (p > 0.5).
GC Genotypes and 25-OHD
We then examined whether the GC genotype served as a determinant of circulating 25-OHD, as has been described in several adult populations and in genomewide association studies.5–11 The observed circulating 25-OHD levels by diplotype are shown in Fig. 4. Circulating 25-OHD (mean ± SD) is greatest in diplotypes with no T- > K substitution at position 436: DDTT (1f/1f) (66.5 ± 34.4 nmol/L), DETT (1f/1s) (68.2 ± 15.2 nmol/L), and EETT (1s/1s) (69.0 ± 17.5 nmol/L); lesser values for 25-OHD were evident in the setting of the presence of the 436 K (2) allele: DDTK (1f/2) (58.7 ± 15.7 nmol/L), DETK (1s/2) (63.4 ± 14.5 nmol/L), and DDKK (2/2) (62.7 ± 11.3 nmol/L). Effects of both rs7041 (D432E) and rs4588 (T436K) polymorphisms on 25-OHD were investigated after adjustment for circulating DBP and AIMs scores. We found no effect of the rs7041 (D432E) polymorphism on circulating 25-OHD (p = 0.53), but a significant effect of the rs4588 (T436K) SNP (p = 0.002).
Moreover, we calculated putative free 25-OHD levels based on the circulating serum DBP, assuming a fixed serum albumin level of 4.3 g/dL, and affinity constants as previously reported 19, 20. Free 25-OHD was strongly correlated with total circulating 25-OHD (R = 0.93, p < 0.0001). In contrast to the effects of genotypes on total 25-OHD, a comparable analysis revealed no effect of genotype on the free 25-OHD (p ≥ 0.2). We also explored the functional role of 25-OHD with respect to the relationship of total or free 25-OHD to circulating PTH as an entire group, and among different GC genotypes. Indeed, both total and free 25-OHD were (inversely) correlated with PTH (R = −0.11 and −0.12, respectively; p = 0.0018), but there was no effect of GC genotype on this relationship (p > 0.6).
These analyses indicated significant effects of GC genotype on both the circulating DBP and on the circulating 25-OHD. Moreover, as shown in Fig. 2, circulating DBP is related to 25-OHD. Thus it must be considered whether the genotype effect serves as a determinant of 25-OHD entirely because of the genotype effects on circulating DBP, or whether the influence of the genotype may be greater than that explained by a relationship with DBP concentration. We therefore examined the magnitude of these separate effects by multivariate analysis, and demonstrated that only part of the genotype effect on circulating 25-OHD can be attributed to the concomitant effect on DBP levels, suggesting that the effect may be mediated by other mechanisms as well. To explore a more complete model, our multivariate analysis adjusted for the demographic and nutritional variables we previously identified as significant correlates of 25-OHD, including age, season, daily vitamin D intake, and degree of skin pigmentation. Data from this analysis are shown in Table 3. In the complete model, DBP diplotype, circulating DBP, and European ancestry are all significant determinants of the circulating 25-OHD level. Age, estimated vitamin D intake, and season of sampling remain as significant determinants of 25-OHD, when diplotype, circulating DBP, and ancestry are included in the analysis. The effect of genotype on circulating 25-OHD levels persists after correction of 25-OHD for circulating DBP level, which is also affected by genotype (Fig. 3). The specific effect of diplotype on the 25-OHD level is explained by variance of the rs4588 (T436K) SNP; the mean ± SD values of serum 25-OHD are 67.8 ± 24, 62.7 ± 11, and 60.9 ± 15 nmol/L for TT, TK, and KK, respectively (Fig. 5).
Table 3. Multivariate Analysis of Determinants of 25-OHD
The R2 for the model described is 0.25.
25-OHD = 25-hydroxyvitamin D; DBP = vitamin D binding protein; TT, TK, KK = haplotypes; AIM = ancestry informative marker.
For every milligram per liter (mg/L) increase in DBP, there is a 0.1078 nmol/L increase in 25-OHD.
For every month of age, there is a 0.2676 nmol/L decrease in 25-OHD, or annual decrease of 3.211 nmol/L.
Increasing dietary intake of vitamin D by 100 IU daily increases 25-OHD by 3.164 nmol/L.
The AIM score for each ancestry group can range between 0 and 1. The estimated 25-OHD level in individuals with European AIM score of 1 is 7.8962 nmol/L greater than those with a European AIM score of 0 (no European ancestry).
The current study extends our investigation of an urban cohort of minority infants and children to include an analysis of some genetic determinants for circulating 25-OHD. In particular we have focused on diplotypes of common SNPs (rs7041 and rs4588) of the GC gene, encoding DBP, which have been shown to be strong predictors of circulating 25-OHD in adults. We have shown that individuals self-identified as African American have a significantly higher proportion (53%) of the WT diplotype (DDTT; 1f homozygosity) compared to those of Hispanic ethnicity (13%), as posited by Speekaert and colleagues.21 The DDTT diplotype was evident in 6% of Caucasians, and 14% of the “Other” or mixed ethnicity groups, but the numbers of subjects were substantially smaller than in the African American and Hispanic identified groups. Indeed, the frequency of the other possible diplotypes ranged between 2% and 21% among the African American subjects. Moreover, the DT (1f) allele occurred in 71% of African Americans, but only 32% to 40% of the others (Table 1). The ET (1s) and DK (2) alleles were much less common in African Americans (16% and 13%, respectively). These data confirm that our sample reflects the genetic variance by reported ancestry as shown by others.9, 10 Our analysis of AIMs scores among the categories of self-reported ancestry verifies that the self-reporting of ancestry is relatively consistent with validated genetic markers of the specific ancestral background.
We identified a strong correlation between ancestry and circulating 25-OHD. Greater African AIMs scores were correlated with lower 25-OHD levels, whereas greater European AIMs scores were correlated with higher 25-OHD levels. Although the AIMs scores are necessarily intercorrelated (because all three scores for an individual must sum to 1.0), admixture resulted in relatively large variance for Native American AIMs scores (see Fig. 1), such that use of this score alone did not reveal significant correlations with the 25-OHD levels. Interestingly, none of the AIMs scores correlated with the circulating DBP level. Yet, as expected, 25-OHD levels were strongly correlated with the circulating DBP level (Table 2, Fig. 2).
The GC genotype was also a strong determinant of circulating DBP level (p < 0.001, Fig. 3). The largest effect appears to be related to the T436K SNP, with TT individuals having the highest mean circulating DBP level, followed by TK heterozygotes, and KK homozygotes having the lowest values for this measure. The circulating DBP levels do not differ by D/K status at position 432 in the diplotypes homozygous for the T allele at position 436; however, there is a slight increase in the DETK (1s/2 heterozygote) as compared to the DDTK (1f/2 heterozygote) diplotype. The substitution of a lysine (K) for the threonine (T) at position 436 eliminates an O-glycosylation site from the molecule, and there is evidence that loss of glycosylation may affect DBP half-life. In addition, there may be trans effects, generated by a D->E substitution at the nearby 432 position that influence the extent of O-glycosylation at the 436 position, although this appears to be of significance only in the setting of double heterozygosity, consistent with the 436 allele being the major determinant. It is not known how such changes in the DBP molecule would affect its serum concentration, but the substitutions could result in altered rates of transcription, changes in mRNA stability, or as some have suggested, clearance of the protein itself.2
Genotype is also a strong determinant of the circulating 25-OHD level (p = 0.005, Fig. 4). As with effects on circulating DBP, the major effect appears to be associated with the T436K SNP. 25-OHD level decreases with progressive substitution of threonine (T) residues at position 436 with lysine (K) residues (Fig. 5), such that 25-OHD levels are ordered TT > TK > KK, as with DBP.
We used multivariate regression to determine whether the genotype effect on DBP could account for the effects on circulating 25-OHD. We included all significant determinants of the circulating 25-OHD level available in the current analysis (436 T/K haplotype, circulating DBP level, and the West African/European AIMs scores), together with the demographic and nutritional variables previously found to be significant determinants of circulating 25-OHD (age, degree of skin pigmentation, daily vitamin D intake, and season of sampling).3 Our regression model indicates that genotype is a significant determinant of circulating DBP and 25-OHD, an effect that is, at least in part, independent of possible direct DBP effects on 25-OHD. Overall, the estimated magnitude of the observed predictors for 25OHD are greatest for genotype, ancestry, and season, all of which have relatively larger estimated effects than age and circulating DBP level.
In the currently developed model with GC genotype incorporated as a determinant of the circulating 25-OHD level, the effect of daily vitamin D supplementation is moderately impressive in this age group. Quantitatively, a daily vitamin D intake of 100 IU may account for a 3.16 nmol/L increase in the circulating 25-OHD level. However, the relative importance of the T436K genetic effect appeared greater still. Mean circulating 25-OHD in the TT genotype group was 3.3 ± 1. 5 nmol/L greater than that for the TK genotype, and 7.3 ± 4.1 nmol/L greater than in the KK group (p = 0.025). Thus progressive substitution of the T allele correlates with the circulating 25-OHD level.
Our findings are consistent with the notion that DBP concentration and diplotype influence the total circulating reservoir of 25-OHD, but the regulated compartment may be free 25-OHD. Indeed recent examination of bioavailability of vitamin D metabolites by Chun and colleagues,22 ascertained by mathematical modeling, suggests that the DBP variants studied here have differences in binding affinity for both 25-OHD and 1,25(OH)2D. In the context of our finding that total, but not free, 25-OHD is determined by GC diplotype, it appears that homeostatic mechanisms favor regulating circulating levels of free 25-OHD, independent of genotype. Moreover Chun and colleagues22 predict that the TK variants have an ordered effect on affinity for 25-OHD, in keeping with the effects on circulating levels of this metabolite we observed. It is of interest that these GC SNPs are not situated within the vitamin D binding site of the protein,23 suggesting that secondary or higher order structure may be important in determining the binding affinity. It should also be considered that GC genotype may influence 25-OHD levels beyond its effects on DBP concentration: because DBP serves as the ligand for the megalin/cubulin complex that mediates entry of bound 25-OHD into renal tubular cells,24 the protein may also be involved in regulating bioavailability of the active metabolite, 1,25(OH)2D. Other mechanisms worth further investigation include the possibility that genotype may affect DBP clearance from the circulation, binding affinities with other cell surface receptors related to tissue delivery,25 metabolizing hydroxylase enzymes, or the vitamin D receptor itself.
All authors state that they have no conflicts of interest.
We acknowledge the support and cooperation of the following individuals who made this project possible: Drs. Elizabeth Bailey and Mary Ellen Flaherty-Hewitt at the Hospital of St. Raphael, Drs. Robert Windom and Steven Updegrove at the Hill Health Center, Dr. Brian Forsyth of the Yale-New Haven Hospital Primary Care Clinic, and Dr. Semeon Tsalbins of the Fair Haven Health Center, all in New Haven, CT, USA. We thank Dr. Francisca Herreros and Ms. Esther Torrealba-Fox for their dedicated assistance, and Dr. Hong-Yu Zhao for his advice and fruitful discussion. Finally, we thank the children of New Haven and their families for their generosity in participating in the study. These studies were carried out with the support of the Gerber Foundation, and made possible by the Yale Center for Clinical Investigation (supported by CTSA UL1 RR024139 from the National Center for Research Resources of the NIH), and the Dairy Farmers of Canada (to DECC, LF, and BY-LW). Support for data analysis was provided in part by the Thrasher Research Fund Award 02829-4 (to TOC) and the National Institute of Child Health and Development of the NIH (ARRA Award 5RC1HD063562 to TOC).
Authors' roles: Study design: TOC and DECC. Study conduct and data collection: EP, BKE, CS, WML, JB, LF, and BY-LW. Data analysis and interpretation: JHZ, EP, TOC, and DECC. Drafting manuscript: TOC. Revising manuscript and approving final content: JHZ, EP, WML, JB, and DECC. Primary responsibility for content of the manuscript: TOC.