Brain development begins close to conception and continues beyond early childhood. Micro- and macronutrients have been shown to affect neurocognitive development in early childhood, with long-term consequences for academic achievement, employment, and health status [1
]. Of many micronutrients studied, roles have been emerging for gestational 25-hydroxyvitamin D [25(OH)D] status with respect to cognitive and language development during early childhood [3
Vitamin D is a pleiotropic secosteroid hormone, which is ingested or produced by skin exposure to ultraviolet light [6
]. Vitamin D exerts classical actions on bone mass and non-classical effects on non-skeletal tissues throughout life [8
]. While thresholds for vitamin D sufficiency have been debated recently [10
], low circulating levels of 25(OH)D have been associated with increased mortality in black and white older adults dwelling in North America [12
]. Recent studies have suggested significant impact by gestational 25(OH)D status upon placental function [13
] and early brain development [14
Compared to European-Americans (EA), lower 25(OH)D levels during pregnancy and altered 25(OH)D skeletal metabolism in African-Americans (AA) are well documented [10
]. Biologically, higher melanin pigment in dermis of individuals with African ancestry contributes to lower production of 25(OH)D, yet higher bioavailable 1,25(OH)2
-D circulates due to lower levels of vitamin D binding protein (VDBP) [10
]. Unlike hepatocytes, neurons express megalin [20
]. Megalin is a transmembrane protein that internalizes free 25(OH)D from its circulating complexes with VDBP and albumin. This transport mechanism, along with other renal and skeletal adaptations by AA [21
], may buffer the developing brain from ambient vitamin D insufficiency. However, numerous factors have been associated with neurocognitive development and could possibly confound an association between gestational 25(OH)D and neurocognitive development, including preterm birth, socioeconomic status, and gestational exposures to maternal use of alcohol or tobacco [24
The Conditions Affecting Neurocognitive Development and Learning in Early Childhood (CANDLE) Study is a contemporary birth cohort representing a racially and socioeconomically diverse, North American, metropolitan population [27
]. The primary aim of the CANDLE study was to determine the pre- and postnatal factors that impact neurocognitive development by age 3. This report examines the hypothesis that gestational 25(OH)D status is associated with cognitive and language development at age 2 years. A secondary aim is to examine whether differences in gestational 25(OH)D and other socio-economic indicators could explain some well-recognized differences in neurocognitive scores between AA and EA [28
To our knowledge, this is the largest prospective study to date examining the potential impact of gestational 25(OH)D status upon early childhood cognitive and language development in a racially diverse population. We found that higher gestational 25(OH)D status significantly associated with higher scaled scores for receptive language in offspring at 2 years of age. Similarly, positive trends were suggested between gestational 25(OH)D status and the scaled scores for cognition and expressive language, although observations did not reach statistical significance. Controlling for maternal IQ, race, or use of tobacco products during pregnancy reduced the effect size of 25(OH)D on scaled scores at age 2 years, underscoring the importance of capturing and evaluating multiple exposures when studying neurocognitive development.
Neurocognitive assessments in early childhood assess brain development, which begins with formation of the neural plate 12 days post-conception and continues throughout early adulthood. Brain structure and function can be affected by nutrient insufficiencies or environmental insults at critical time points. Animal studies provide biological plausibility for the hypothesis that in utero
vitamin D exposure affects human brain development [39
]. Expression of vitamin D receptors (VDR) in mammalian brain occurs as early as day 12 of gestation, then increases throughout pregnancy [40
]. The presence of 1α-hydroxylase activity within numerous brain regions suggests that brain tissue can locally produce metabolically active vitamin D, further implicating a role for vitamin D in brain development. Murine dams maintained on vitamin D deficient diets before and during pregnancy display alterations in murine brain morphology that could contribute to deficits in memory, learning, and attention processing [39
In humans, 25(OH)D levels during pregnancy have been positively associated with neurocognitive development in some [14
] but not all studies [40
]. Our observations support that language development is positively associated with gestational 25(OH)D levels as early as 2 years, while Whitehouse et al.
, showed an effect as late as age 7 [14
]. Other groups have not reported strong associations with gestational 25(OH)D [16
], using a composite score to represent overall development. Combining cognitive and language development into one index could mask an association if it exists for a sub-scale. Our study and that of Whitehouse et al.
] independently examined the associations with cognitive and language, and the available findings support a role for gestational 25(OH)D status in early language development.
In the current study, the small effect sizes for the association between gestational 25(OH)D status and receptive language scaled scores agree with reports by other investigators [40
]. While stringent statistical significance was lacking for the cognitive score, the associations for all scaled scores trended consistently positive. In practical terms, an effect estimate of 0.07 for cognitive scaled score per 10 ng/dL increase in mid-gestational 25(OH)D translates to an increase of 0.35 IQ point for the 2-year-old child [44
]. Similarly, an effect estimate of 0.24 for receptive and 0.12 for expressive language scaled score per 10 ng/dL increase in 25(OH)D translates into an increase of 1.0 and 0.6 IQ point respectively by age 2. The reduction in lifetime earnings has been estimated at $22,000 lost for each 1 IQ point decrease [44
]. If the observed deficit in receptive language is maintained as the CANDLE children develop, the collective loss of total earnings would be estimated at $19 million. If the deficit in receptive language scaled scores increases as these children age, the total loss of earnings could be considerably more.
For optimal development, brain cells require a complement of macro- and micronutrients [2
]. When specific nutrients are limiting, the brain has priority use at the expense of other organs [2
]. Since the current inquiry focused on gestational 25(OH)D as a biomarker of nutrient exposure, it is possible that a peripheral marker is limited as an assessment of the availability of 25(OH)D for brain development. Animal data suggest the genetic expression of Fox2
, a regulator of speech and language, was lower in those on the vitamin D deficiency diet at 14.5 days but expression was higher at 17.5 days, compared to those on a vitamin D sufficient diet [39
]. These findings suggest that some compensatory mechanisms may exist to minimize effects of vitamin D deficiency.
Brain development from conception to age 5 varies across neurocognitive domains [46
]. During sensitive stages of development, nutrient deficiency has adversely affected early child development [2
] with persistent effects into adulthood [3
]. The interaction of nutrient influences and socio-economic factors can be difficult to parse out in developmental studies [47
]. We chose to use insurance as a measure of income because 9% of our sample did not report income, but did report insurance status. 99% of those not reporting income reported to be on Medicaid (TennCare) insurance. Income was collected in categories with the highest income category as ≥$75,000. Limitations due to how the income variable was obtained prevented the calculation of income as a continuous function of the federal poverty level. SES is a complex variable comprised of income, wealth, education, marital status, location of primary residence, and other factors. Based on the data available, we evaluated SES through a combination of variables (i.e.
, insurance as a proxy for income, marital status, education, and maternal IQ). Our final multivariate models support that SES is an important component of cognitive and language development. With regard to cognitive and expressive language, SES accounted for the most of the variability in the model, resulting in little association with 25(OH)D.
Two other limitations exist for this study design. First, tobacco use and alcohol use during pregnancy information were collected as dichotomous variables (“yes/no”) by questionnaire, which could have affected the low rates. Using dichotomous variables merges low and moderate-heavy users. Effects of low levels of smoking and alcohol use by many women may mask effects of moderate to heavy use by a smaller group of users. In multivariate modeling of this CANDLE sample, alcohol was not found to be a significant covariate, after accounting for race, education, mother’s IQ, and mother’s tobacco use. Thus, the variability associated with alcohol use was taken up by these other variables and not important in the final model. To determine whether smoking and alcohol use during pregnancy affect subsequent cognitive outcomes, future studies could collect these data using continuous measures of exposure (e.g., how many cigarettes did the mother use per day and how much alcohol did she consume?).
Second, the Bayley Scales detect general milestones and are less sensitive to subtle effects than narrow band assessments, such as continuous measures of recognition memory, symbolic play, acuity, or information processing speed. Despite these limitations, impressive effects on receptive language were observed. These findings suggest that this may be the tip of the iceberg and that other effects might be detected if more sensitive narrow band tests are used.
In our preliminary evaluation, we were unable to find any association with diet-based nutrient influences on neurocognitive development. Our bivariate data suggest that 25(OH)D sufficiency, maternal education level, ability to buy health insurance, use of tobacco products, premature birth, and marital status all affected the cognitive and language development at age 2. When multivariate analyses include the possible confounders, the effect size of gestational 25(OH)D on language was reduced in our study, in agreement with other reports [16
]. The positive associations between in utero
25(OH)D and language development remained, underscoring the consistency of findings across multiple cohorts.
Unlike other studies, our cohort was racially diverse, thereby allowing us to examine the effects of 25(OH)D in a population more representative of the USA. Differences in cognitive and language scaled scores were present between our AA and EA children. IQ scores favoring EA over AA have been documented [49
]. Our 1-point difference in scaled score between AA and EA is equivalent to that reported by other groups [51
]. Although controversial, the discussion often focuses on the “nurture versus
nature” that may lead to this difference [49
]. We recently reported [57
] that DNA methylation is jointly modulated by ancestry and gestational vitamin D levels, observations which support multidimensional factors. Despite these arguments, our data suggest the association between gestational 25(OH)D and receptive language are not different between AA and EA, when adjusted for other SES and maternal characteristics.
Strengths of our study include its contemporary design, large size, high retention rate (75% of the total CANDLE cohort attended the 2-year exam), permanent residence limited to one county (to reduce potentially confounding effects of latitude on sunlight exposure), high number and high proportion of African-American participants, and uniformity of data collection techniques throughout the study. Due to phlebotomy limits in infants and the burden of other CANDLE procedures, this CANDLE sub-study lacks repeated measures of 25(OH)D status for participants and lacks individual assessments of parathyroid hormone status or vitamin D receptor (VDR) genotype. The design also could not control for potential confounding by differences in sunscreen use, sun exposure (e.g., air pollution or outdoor lifestyles), culture, skin cover by clothing, or post-natal vitamin D intake via diet. Lower gestational 25(OH)D levels perhaps reflected dietary intake during pregnancy or less intense sun exposure in Memphis (latitude range 35° N–35°24′ N) compared to more equatorial latitudes.