Journal of Autism and Developmental Disorders

, Volume 43, Issue 7, pp 1495–1504

Maternal Vitamin D Levels and the Autism Phenotype Among Offspring

Authors

    • Telethon Institute for Child Health Research, Centre for Child Health ResearchUniversity of Western Australia
    • School of PsychologyUniversity of Western Australia
  • Barbara J. Holt
    • Telethon Institute for Child Health Research, Centre for Child Health ResearchUniversity of Western Australia
  • Michael Serralha
    • Telethon Institute for Child Health Research, Centre for Child Health ResearchUniversity of Western Australia
  • Patrick G. Holt
    • Telethon Institute for Child Health Research, Centre for Child Health ResearchUniversity of Western Australia
  • Prue H. Hart
    • Telethon Institute for Child Health Research, Centre for Child Health ResearchUniversity of Western Australia
  • Merci M. H. Kusel
    • Telethon Institute for Child Health Research, Centre for Child Health ResearchUniversity of Western Australia
Original Paper

DOI: 10.1007/s10803-012-1676-8

Cite this article as:
Whitehouse, A.J.O., Holt, B.J., Serralha, M. et al. J Autism Dev Disord (2013) 43: 1495. doi:10.1007/s10803-012-1676-8

Abstract

We tested whether maternal vitamin D insufficiency during pregnancy is related to the autism phenotype. Serum 25(OH)-vitamin D concentrations of 929 women were measured at 18 weeks’ pregnancy. The mothers of the three children with a clinical diagnosis of autism spectrum disorder had 25(OH)-vitamin D concentrations above the population mean. The offspring of 406 women completed the Autism-Spectrum Quotient in early adulthood. Maternal 25(OH)-vitamin D concentrations were unrelated to offspring scores on the majority of scales. However, offspring of mothers with low 25(OH)-vitamin D concentrations (<49 nmol/L) were at increased risk for ‘high’ scores (≥2SD above mean) on the Attention Switching subscale (odds ratio: 5.46, 95 % confidence interval: 1.29, 23.05). The involvement of maternal vitamin D during pregnancy in autism requires continued investigation.

Keywords

Autism spectrum disorderAutistic-like traitsVitamin DPrenatalPregnancyEnvironment

Introduction

Low levels of maternal vitamin D during pregnancy are known to be associated with a range of adverse health outcomes in offspring, including intrauterine growth-restriction (Morley et al. 2006), reduced bone-mineral accrual (Javaid et al. 2006) and recurrent wheeze (Camargo et al. 2007). A recent proposal is that maternal vitamin D insufficiency during pregnancy may also increase the risk for autism spectrum disorder (ASD) among offspring (Cannell 2008). Epidemiological observations have provided the primary support for this hypothesis, with a meta-analysis finding that ASD is more common among the offspring of pregnancies in which the first and second trimesters coincided with the winter and spring months (Gardener et al. 2009). Approximately 90 % of human vitamin D stores come from direct sunlight (Holick 1987), and consequently vitamin D insufficiency is most common during the winter months (Levis et al. 2005). Because the developing fetus is completely reliant on maternal vitamin D stores (Holick 2007), low levels of maternal vitamin D during the first and second trimesters, which are known to be a critical period of vulnerability for the developing nervous system (Rice and Barone 2000), may underpin the link between the timing of pregnancy and offspring ASD. However, literature in this area has been inconsistent, and there have been a number of notable failures to replicate this association (Kolevzon et al. 2006; Landau et al. 1999). Furthermore, it is impossible from these observational studies to investigate sunlight exposure and consequent vitamin D levels in isolation from other seasonal factors that may influence prenatal development, such as changes in temperature, and maternal nutrition and infection.

Vitamin D from the skin and diet is metabolized in the liver to 25(OH)-vitamin D, which can then be measured in blood samples as a more direct investigation of Vitamin D status (Holick 2007). While rodent studies have linked low maternal 25(OH)-vitamin D during pregnancy with atypical behavior among pups (Eyles et al. 2003; O’Loan et al. 2007), two studies only have examined this association in humans, with mixed results. Gale et al. (2008) measured the circulating 25(OH)-vitamin D levels of 466 pregnant women, and reported no statistically significant association with offspring behavior and verbal IQ at 9 years of age. However, the study was limited by sample attrition of approximately 60 %, which considerably reduced the statistical power to identify significant effects. Furthermore, maternal blood samples were obtained during the third trimester of pregnancy (median 32.6 weeks; range 28–42 weeks), and therefore the 25(OH)-vitamin D concentrations may not have reflected circulating levels during critical phases of neurodevelopment earlier in gestation. A more recent study investigated maternal 25(OH)-vitamin D concentrations during the second trimester of pregnancy (18 weeks’ pregnancy) and the behavioral, emotional and language development of offspring enrolled in the Western Australian Pregnancy (Raine) Cohort Study (Whitehouse et al. 2012). No association was observed between maternal 25(OH)-vitamin D levels and offspring behavioral and emotional development. However, women with vitamin D insufficiency during the second trimester had close to twice the risk of having a child with a language impairment in middle childhood (measured by the Peabody Picture Vocabulary Test-Revised) compared to women with high 25(OH)-vitamin D levels. Language impairment is often a prominent feature of the ASD behavioral profile, and therefore this study may indicate the involvement of maternal vitamin D insufficiency in the etiology of ASD. However, there is a clear need for studies to investigate the influence of maternal vitamin D levels on the broader ASD behavioral profile in offspring, such as social difficulties and repetitive behaviors.

One methodological approach that is obtaining increasing traction in ASD research—but is yet to be utilized in the context of maternal vitamin D concentrations—is to examine autistic-like traits in the broader community. There is accumulating evidence that autistic-like traits are on a continuum in the general population, with ASD representing the extreme end of the distribution (Lundström et al. 2012; Robinson et al. 2011). Understanding developmental factors that are associated with variation in autistic-like traits in the general population may provide important insights into the biological mechanisms that underpin the clinical condition. Here, we report a follow-up investigation of the Raine cohort, providing the first investigation of the prospective relationship between maternal vitamin D status during pregnancy and the ASD phenotype. A small number of offspring were diagnosed with clinical ASD, and a measure of autistic-like traits was obtained from the remainder of available participants. We hypothesized that decreased concentrations of maternal Vitamin D at 18 weeks pregnancy would be associated with increased ASD-like traits in the broader sample, and that those individuals with a clinical diagnosis of ASD would have mothers who had particularly low concentrations of serum vitamin D during pregnancy.

Methods

Participants

The Raine Study recruited pregnant women from the public antenatal clinic at King Edward Memorial Hospital or surrounding private clinics in Perth (Australia) between May 1989 and November 1991 (n = 2,900). The inclusion criteria were a gestational age between 16 and 20 weeks, English language skills sufficient to understand the study demands, an expectation to deliver at King Edward Memorial Hospital, and an intention to remain in Western Australia to enable future follow-up of their child (Newnham et al. 1993). Participant recruitment and all follow-ups of the study families were approved by the Human Ethics Committees at King Edward Memorial Hospital and/or Princess Margaret Hospital for Children in Perth, Western Australia. The current study included those mother–child dyads where maternal blood was collected at 18 weeks’ gestation and data on autistic-like traits were obtained from the offspring in early adulthood. Informed written consent was obtained from all mothers and offspring who participated in this study.

Maternal 25(OH)-Vitamin D

The procedure for maternal blood collection and analysis for concentrations of 25(OH)-vitamin D has been discussed in detail in Whitehouse et al. (2012). In brief, from 1989 to 1991, venous blood was obtained at 18 weeks’ pregnancy in 929 randomly selected pregnant women. Randomisation was based on a sealed-envelope technique using computer-generated numbers. The blood was immediately centrifuged, and serum collected and stored at −80 °C. In June 2011, Serum 25(OH)-vitamin D levels were measured using an enzyme immunoassay kit from Immunodiagnostic Systems Ltd. (Scottsdale, Arizona, USA). Vitamin D concentrations in stored sera have been shown to remain stable for over three decades (Corder et al. 1993; Nomura et al. 1998). A subset of samples (n = 28) were also measured using isotope-dilution liquid chromatography-tandem mass spectrometry. The correlation of 25(OH)-vitamin D concentrations for samples assayed by both techniques was strong (r2 = 0.87) and confirmed that there were no molecules (vitamin D metabolites or otherwise) in sera of 18-week pregnant women that interfered with the immunoassay of 25(OH)-vitamin D.

Offspring ASD Diagnosis

At the 5-, 8-, 10-, 14- and 17-year follow-ups of the Raine cohort, parents were asked whether their child had ever received a diagnosis of ASD by a health professional. Diagnosis of these conditions in Western Australia mandates consensus by a team comprising a Pediatrician, Psychologist and Speech-Language Pathologist under DSM-IV guidelines (American Psychiatric Association 1994). See Whitehouse et al. (2011) for more detail on the ascertainment methods.

Offspring Autistic-Like Traits

In early adulthood, the Raine cohort was contacted specifically for the purpose of completing the Autism Spectrum Quotient (AQ). The AQ is a self-report questionnaire that provides a quantitative measure of autistic-like traits in the general population (Baron-Cohen et al. 2001). Individuals are provided with 50 statements and asked to indicate on a 4-point scale how well that statement applies to them (strongly agree, agree, disagree, strongly disagree). The statements are divided into five subscales: social skills (‘I prefer to do things with others rather than on my own’), communication (‘People often tell me that I keep going on and on about the same thing’), imagination (‘I don’t particularly enjoy fictional stories’), attention to detail (‘I am fascinated by numbers’), and attention switching (‘I enjoy doing things spontaneously’). Items within each subscale are then summed to provide a quantitative measure of that particular autistic-like trait, with higher scores denoting increased symptomatology. A total AQ is calculated by tallying the scores from the five subscales. The scale is known to have good test–retest (r = 0.7) and interrater reliability. Adults with a known diagnosis of any intellectual disability or ASD were not asked to complete the AQ due to ethical concerns.

Confounders and Covariates

Sociodemographic variables that may influence AQ scores were also investigated. Maternal education was indexed by whether the mother had completed secondary school (i.e., year 12) at the time of pregnancy, and family income during pregnancy was dichotomized according to whether the minimum household income exceeded the ‘poverty line’ defined by the Australian Government at the time of recruitment (i.e., $24,000 AUD). Maternal race (Caucasian vs. non-Caucasian) and age at conception (years) were also considered, as were maternal smoking and alcohol intake during pregnancy, parity and the sex of the offspring. Infant health at birth was indexed by birthweight (grams), gestational age (weeks), and Apgar scores 5 min after delivery. Serum 25(OH)-vitamin D levels are known to vary based on the season of collection (Whitehouse et al. 2012), and thus this variable was also considered.

Statistical Analyses

Chi-square analyses were used to examine the influence of participant attrition by comparing sociodemographic, antenatal and obstetric variables between the participants with both maternal 25(OH)-vitamin D data and AQ data available (and were therefore included in the study), and the remainder of the Raine cohort. Independent-samples t tests were then used to compared AQ scores between the current study participants, and the Raine Study participants who completed the AQ, but who had no maternal 25(OH)-vitamin D data available.

We then examined the 25(OH)-vitamin D concentrations of the mothers who reported that their offspring received a clinical diagnosis of ASD during childhood. Raw scores were also converted to Z-scores (calculated based on all Raine Study participants with maternal 25(OH)-vitamin D data) to enable a comparison of individual data points with the distribution of the broader sample. The analyses the turned to the broader cohort, for which the association between maternal 25(OH)-vitamin D concentrations at 18 weeks’ pregnancy and the scores on the AQ were examined as continuous and categorical data. The continuous data were investigated using Pearson’s correlations. Significant correlations were further examined using hierarchical multivariate linear regression. Regression modelling followed a three-step procedure. Step 1 investigated the effect of maternal 25(OH)-vitamin D concentration on the outcome variable; Step 2 entered any sociodemographic, antenatal or obstetric variable which correlated with the predictor or outcome variable of interest at the conservative alpha level of p < .2; and Step 3 entered the variable denoting the season in which maternal blood was collected, in order to determine whether any effect was specific to maternal 25(OH)-vitamin D levels, rather than potentially other seasonal factors. For the linear regression model, we report unstandardized Beta coefficient (B), the standard error of B, and the standardised beta coefficient (β).

We then investigated the data when expressed categorically to determine whether maternal vitamin D status was associated with high scores on the AQ and its subscales. ‘High’ scores were defined as scores that were two or more standard deviations from the mean on a particular scale (to the nearest whole number). Maternal 25(OH)-vitamin D concentrations were grouped into tertiles, which was a division of data that power analyses revealed provided optimal statistical power for identifying differences in the proportion of offspring with high AQ scores. Chi-square analyses were conducted between the predictor and outcome variables. Any significant difference on any scale was followed up with multivariate logistic regression, following the same three-step procedure outlined for the multivariate linear regression. For the logistic regression model, odds ratios (OR) and 95 % Confidence Intervals (95 % CI) are reported. The alpha level for all analyses was p < .05.

Results

There were 929 offspring for whom blood was collected from their mothers at18 weeks’ pregnancy and later analyzed for 25 (OH)-vitamin D levels. Among these participants, 406 offspring (male = 149; female = 257) completed the AQ in early adulthood (M age 19.81 years; SD = 0.77 years). Table 1 shows the characteristics of the Raine Study participants who were and were not included in the current study. Participants included in the current study were more likely to have mothers who were older at conception, who had completed secondary education and who drank alcohol during pregnancy. Similarly, the current participants were more likely to be born into households with an income above the poverty level of $24,000 AUD at the time of pregnancy. Furthermore, the current study oversampled for female offspring and for women who had their blood collected during the summer months. However, there was no difference between the current participants and the remainder of the Raine cohort in birthweight, gestational age, sex, parity, and Apgar scores 5 min after birth.
Table 1

Characteristics of participants in the Raine cohort who were (N = 406) and were not (N = 2462) included in the current study

 

Included

Not included

p value

N

n (%)

N

n (%)

Maternal age at conception

397

 

2,370

 

<.01

 <20

 

22 (5.5)

 

251 (10.6)

 

 20–24

 

71 (17.9)

 

521 (22.0)

 

 25–29

 

107 (27.0)

 

730 (30.8)

 

 30–34

 

136 (34.3)

 

571 (24.1)

 

 35+

 

61 (15.4)

 

297 (12.5)

 

Maternal race

406

 

2,460

 

.38

 Caucasian

 

364 (89.7)

 

2,168 (88.1)

 

 Non-Caucasian

 

42 (10.3)

 

292 (11.9)

 

Season in which maternal blood was collected at 18 weeks’ pregnancy

406

 

2,460

 

.03

 Summer

 

85 (20.9)

 

423 (17.2)

 

 Autumn

 

87 (21.4)

 

573 (23.3)

 

 Winter

 

101 (24.9)

 

756 (30.7)

 

 Spring

 

133 (32.8)

 

208 (28.8)

 

Maternal education at pregnancy

397

 

2,406

 

.02

 Completed secondary school

 

177 (44.6)

 

918 (38.2)

 

 Did not complete secondary school

 

220 (55.4)

 

1,488 (61.8)

 

Family income below poverty line

392

 

2,311

 

.001

 Yes

 

140 (35.7)

 

1,042 (45.1)

 

 No

 

252 (64.3)

 

1,269 (54.9)

 

Maternal smoking in pregnancy

397

 

2,201

 

.09

 None

 

311 (78.3)

 

1,629 (74.0)

 

 1–10 cigarettes daily

 

49 (12.3)

 

284 (12.9)

 

 11+ cigarettes daily

 

37 (9.3)

 

288 (13.1)

 

Maternal alcohol intake during pregnancy

397

 

2,201

 

<.01

 None

 

219 (55.2)

 

1,406 (63.9)

 

 Once a week or less

 

155 (39.0)

 

681 (30.9)

 

 Several times a week or more

 

23 (5.8)

 

114 (5.2)

 

Gestational age

397

 

2,372

 

.79

 <32 weeks

 

6 (1.5)

 

35 (1.5)

 

 32–37 weeks

 

72 (18.1)

 

397 (16.7)

 

 >37 weeks

 

319 (80.4)

 

1,940 (81.8)

 

Birthweight

406

 

2,450

 

.67

 <2,500 g

 

39 (9.6)

 

213 (8.7)

 

 2,500–4,000 g

 

335 (82.5)

 

2,017 (82.3)

 

 >4,000 g

 

32 (7.9)

 

220 (9.0)

 

Sex

406

 

2,462

 

<.01

 Males

 

149 (36.7)

 

1,305 (53.0)

 

 Females

 

257 (63.3)

 

1,157 (47.0)

 

Parity

406

 

2,460

 

.61

 1

 

195 (48.0)

 

1,179 (47.9)

 

 2

 

123 (30.3)

 

700 (28.5)

 

 >2

 

88 (21.7)

 

581 (23.6)

 

Apgar scores 5 min after birth

383

 

2,365

 

.63

 Generally normal (8–10)

 

383 (96.7)

 

2,266 (95.8)

 

 Fairly low (4–7)

 

13 (3.3)

 

97 (4.1)

 

 Critically low (0–3)

 

0 (0)

 

2 (0.1)

 

p values are for Chi-square comparisons

The AQ was collected on 1278 participants from the broader Raine cohort. Table 2 shows that there no differences in AQ scores between participants who also had available data on maternal 25 (OH)-vitamin D levels (and thus included in the current study, n = 406), and those with no available data on maternal 25 (OH)-vitamin D levels (and therefore not included in the current study, n = 872).
Table 2

Scores on the Autism-Spectrum Quotient (AQ) and its subscales for those Raine cohort participants according to who did and did not have maternal 25 (OH)-vitamin D levels available

 

Available

(n = 406)

Not available

(n = 872)

p value

Total AQ

15.32 (5.99)

15.33 (5.32)

.97

Social skills

1.73 (1.86)

1.71 (1.79)

.85

Communication

2.05 (1.88)

2.10 (1.69)

.68

Attention to detail

5.33 (2.14)

5.23 (2.09)

.42

Attention switching

3.86 (2.03)

3.92 (1.94)

.63

Imagination

2.34 (1.75)

2.38 (1.72)

.75

p values are for independent-samples t tests

The mean maternal 25 (OH)-vitamin D concentration of the current sample (n = 406) was 58.02 nmol/L, with a standard deviation of 19.04 nmol/L and a range from 15 to 114 nmol/L. Maternal 25 (OH)-vitamin D concentrations and all AQ scales were found to be normally distributed (skewness statistic <1). A one-way ANOVA and Bonferroni post hoc tests found that mothers who had serum collected during the summer (M = 67.32 nmol/L, SD = 20.17 nmol/L) and autumn (M = 64.36 nmol/L, SD = 18.74 nmol/L) months had significantly higher concentrations of 25 (OH)-vitamin D concentrations than women who had blood collected during the winter (M = 50.60 nmol/L, SD = 17.74 nmol/L) and spring (M = 54.29 nmol/L, SD = 17.74 nmol/L) months, F (3, 402) = 17.77, p < .01.

Maternal Vitamin D and Clinical ASD

Among the 929 offspring for whom maternal vitamin D data were available, three had received a diagnosis by a health professional of ASD. The maternal 25 (OH)-vitamin D concentration at 18 weeks pregnancy for these three males were 78 nmol/L (Z score = 1.04; diagnosis of Autistic Disorder), 63 nmol/L (Z score = 0.26; diagnosis of Pervasive Developmental Disorder-Not Otherwise Specified) and 65 nmol/L (Z score = 0.37; diagnosis of Asperger’s Syndrome).

Maternal Vitamin D and AQ Variation

Among the 406 offspring without ASD who completed the AQ, there was no significant correlation between maternal 25 (OH)-vitamin D levels at 18 weeks’ pregnancy and scores on Total AQ scores (r = −0.08, p = .10), the Social Skills subscale (r = −0.04, p = .48), the Communication subscale (r = −0.05, p = .34), the Attention to Detail subscale (r = 0.01, p = .92), and the Imagination subscale (r = −0.05, p = .34). However, there was a significant correlation with the Attention Switching scale (r = −0.13, p = .01), and this scatter-plot is displayed in Fig. 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs10803-012-1676-8/MediaObjects/10803_2012_1676_Fig1_HTML.gif
Fig. 1

Correlation between maternal 25(OH)-vitamin D concentration at 18 weeks’ pregnancy and scores on the Attention Switching subscale of the Autism-Spectrum Quotient in early adulthood

Step 1 of the hierarchical multivariate linear regression found an association between maternal 25 (OH)-vitamin D levels and scores on the Attention Switching subscale of the AQ (B = −0.01, SE B = 0.01, β = −0.13, p = .01). Step 2 of the regression model included the variables that correlated with maternal 25 (OH)-vitamin D levels (gestational age at birth, parity), scores on the Attention Switching subscale (maternal education, household income, Apgar score 5 min after birth), or both (maternal race), at the level of p < .2. The effect of maternal 25 (OH)-vitamin D levels on Attention Switching scores remained statistically significant (B = −0.01, SE B = 0.01, β = −0.12, p = .02). This model accounted for 5 % of variance in Attention Switching scores, 2 % of which was due to maternal 25 (OH)-vitamin D concentrations. When the variable denoting season of maternal blood collection was entered into the regression model in the Step 3, the association between maternal 25 (OH)-vitamin D levels and offspring scores on the Attention Switching subscale of the AQ was no longer statistically significant (B = −0.01, SE B = 0.01, β = −0.07, p = .17). In this final model, significant positive predictors of offspring Attention Switching scores were increasing Apgar Scores 5 min after birth (B = −1.21, SE B = 0.57, β = −0.11, p = .03), having a non-Caucasian mother (B = 0.71, SE B = 0.34, β = 0.11, p = .04), and season of maternal blood collection (B = 0.22, SE B = 0.09, β = 0.12, p = .02).

Maternal Vitamin D and High AQ Scores

Tertile markers for maternal 25 (OH)-vitamin D concentrations were ≤49, 50–66, and ≥67 nmol/L for Tertiles 1–3, respectively. Importantly, the lower tertile threshold (≤49 nmol/L) corresponded well with the most widely-used definition of vitamin D insufficiency as a 25(OH)-vitamin D concentration less than 50 nmol/L (Holick 2007). High AQ scores (2 standard deviations above sample mean) were ≥27 for the Total AQ, ≥5 for the Social Skills subscale, ≥6 for the Communication subscale, ≥9 for the Attention Switching subscale, ≥8 for the Attention to Detail subscale, and ≥6 for the Imagination subscale. Chi-square analyses, presented in Table 3, revealed that offspring of mothers with 25 (OH)-vitamin D concentrations in the lower tertile during pregnancy were more likely to have high scores on the Attention Switching subscale (8.4 %), than offspring of mothers in the middle (1.5 %) and upper (2.3 %) tertiles, χ2 = 9.95, df = 2, p = < .01. Comparisons on all other AQ scale did not reach statistical significance.
Table 3

Number (%) of offspring with a high score on the AQ according to tertiles of maternal serum 25(OH)-vitamin D concentration at 18 weeks’ pregnancy

 

Maternal 25(OH)-vitamin D concentration

p value

Tertile 1

(lowest)

Tertile 2

Tertile 3

(highest)

Total AQ

7 (4.9)

2 (1.5)

5 (3.8)

.29

Social skills

12 (8.4)

10 (7.5)

11 (8.4)

.95

Communication

10 (7.0)

5 (3.7)

6 (4.6)

.44

Attention to detail

14 (9.8)

4 (3.0)

10 (7.6)

.08

Attention switching

12 (8.4)

2 (1.5)

2 (2.3)

.01

Imagination

9 (6.3)

5 (3.7)

6 (4.6)

.60

Table 4 presents the findings from the hierarchical multivariate logistic regression analysis. Step 2 revealed that offspring of Tertile 1 women were more than three as likely to have high scores on the Attention Switching Subscale relative to offspring of Tertile 3 women, odds ratio: 3.84, 95 % CI: 1.06, 13.94, p = .04. Step 2, which adjusted for covariates and confounders, identified a stronger effect of ‘low’ maternal 25(OH)-vitamin D levels (i.e., Tertile 1) on high Attention Switching scores, OR = 6.06, 95 % CI = 1.54, 23.90, p < .05. This effect remained after further adjusting for season of maternal blood collection (Step 3), OR = 5.46, 95 % CI = 1.29, 23.05, p < .05. Increasing parity was the only other variable in the final model that significantly predicted a high score on the Attention Switching subscale, OR = 2.44, 95 % CI = 1.24, 4.82, p = .01.
Table 4

Logistic regression models showing the association between maternal 25(OH)-vitamin D concentration at 18-weeks pregnancy and offspring high scores (≥2 standard deviations above mean) on the Attention Switching subscale of the Autism-Spectrum Quotient

Maternal 25(OH)-vitamin D concentration

Step 1a

Step 2b

Step 3c

Odds ratio (95 % CI)

p

Odds ratio (95 % CI)

p

Odds ratio (95 % CI)

p

Tertile 3 (highest)

1.00 (reference)

 

1.00 (reference)

 

1.00 (reference)

 

Tertile 2

0.68 (0.11, 4.13)

.67

0.91 (0.15, 5.72)

.92

0.87 (0.14, 5.53)

.88

Tertile 1 (lowest)

3.84 (1.06, 13.94)

.04

6.06 (1.54, 23.90)

.01

5.46 (1.29, 23.05)

.02

Odds ratios and 95 % confidence intervals (95 % CI) are presented

aUnadjusted model

bThe following variables entered into model: maternal race, maternal alcohol intake during pregnancy, maternal education, family income, offspring gestational age at birth, offspring parity, and offspring Apgar scores 5 min after birth

cSeason of maternal blood collection entered into model

Post hoc analyses

Sample attrition was biased towards variables indexing lower socioeconomic strata. As a final set of post hoc analyses, we examined whether Total AQ scores varied according to these factors in the Raine Cohort participants who had completed the AQ. Independent-samples t-tests found that Total AQ scores were significantly higher for those adults whose mother was living below the poverty line during pregnancy (below poverty line: n = 414, M = 16.08, SD = 5.66; above poverty line: M = 14.94, SD = 5.50; p < .01), and whose mother had not completed secondary school at the time of pregnancy (not completed secondary school: n = 645; M = 15.74, SD = 5.56; completed secondary school: n = 607; M = 14.95, SD = 5.49; p = .01).

Discussion

This prospective study provides the first direct investigation of the association between maternal vitamin D levels during pregnancy and the ASD phenotype among offspring. Three children in the current sample had received a clinician-based diagnosis of ASD. There was no evidence that the mothers of these children had low levels of maternal 25(OH)-vitamin D concentrations at 18 weeks pregnancy, with all three values above the broader population mean. A measure of autistic-like traits, the AQ, was collected from 406 of the offspring without ASD in early adulthood. Maternal 25(OH)-vitamin D concentrations did not significantly correlate with the Total AQ of the offspring, nor the majority of AQ subscales. However, there was a weak, inverse correlation with the Attention Switching subscale. The significant association remained after adjustment for a range of sociodemographic, antenatal and obstetric covariates, though was rendered non-significant after adjusting for the season at which maternal blood was collected. When the data were expressed categorically, we found that high scores on the Attention Switching subscale were significantly more likely among offspring of mothers with low levels of Vitamin D during pregnancy (≤49 nmol/L) compared to women with 25(OH)-vitamin D levels greater than 67 nmol/L. The statistically significant effect remained after adjusting for a range of confounders and covariates, including the season at which maternal blood was collected. Maternal 25(OH)-vitamin D levels was not associated with high scores on any other AQ scale.

Before exploring the biological plausibility of this single association, it is important to discuss the broader study design. Strengths of the study include a moderate to large participant sample, a follow-up period that spanned approximately 20 years, and the direct measurement of maternal 25 (OH)-vitamin D serum levels using enzyme immunoassay and validated by liquid chromatography-tandem mass spectrometry. A limitation of the study was the considerable sample attrition over the two decades, with only 14 % of offspring from the original cohort included in the current study. While the AQ scores of the current sample did not significantly differ from scores of the broader Raine cohort (Table 2), the attrition did appear to bias the loss of individuals from lower socioeconomic strata (Wolke et al. 2009). Post hoc analyses found that the Total AQ was significantly higher among offspring from households with low income and from mothers with less maternal education, which raises the concern that the attrition may have underestimated any effect of maternal 25 (OH)-vitamin D concentrations on AQ scores. However, computer simulations using data from the Avon Longitudinal Study of Parents and Children (United Kingdom) have found that selective dropout in cohort studies only marginally affect regression coefficients, if participant selection occurs according to predictor variable(s). Replication studies using data from other pregnancy cohorts less affected by sample attrition will build on the findings presented here.

There were three offspring in the current sample who were diagnosed with an ASD, one each with Autistic Disorder, Pervasive Developmental Disorder-Not Otherwise Specific and Asperger’s Syndrome. Ascertainment of these ASD cases within the Raine cohort was based upon parent-report of a clinician-based diagnosis (by a Pediatrician, Psychologist and Speech Pathologist) according to DSM-IV guidelines at any of the 5-, 8-, 10-, 14- or 17-year follow-ups. While it is commonplace in ASD research to confirm clinical diagnoses using ASD-specific behavioral observation and/or parent interview assessments, this was not possible within the Raine cohort. However, it is important to note that a previous investigation of developmental data obtained prior to 5 years of age (Whitehouse et al. 2011), found that each ASD case in the current study demonstrated behaviors consistent with ASD (e.g., poor eye contact, delayed language, absence of pretend play). In the current study, the mothers of all three offspring with ASD had 25(OH)-vitamin D concentrations at 18 weeks pregnancy (78, 63 and 65 nmol/L) that were above the sample mean (58.02 nmol/L). While these findings appear to discount low levels of maternal vitamin D during pregnancy as a contributing cause to all cases of ASD, the very small number of clinical cases leads us to caution against extrapolating the results to the broader ASD population. Research into potential antenatal pathways contributing to ASD is hampered by the relatively low population prevalence of the condition. One possible research design that may facilitate a better understanding of the link between maternal 25 (OH)-vitamin D serum concentrations and offspring ASD is to study the pregnancies of women who have an existing child with ASD. Sibling fetuses are at increased risk for ASD (Ozonoff et al. 2011), and may provide opportunities to elucidate how maternal vitamin D relates to offspring phenotype, perhaps in interaction with genetic factors.

The measurement of autistic like traits within the general population provided a further opportunity to investigate the possible effects of maternal vitamin D levels during pregnancy on the ASD phenotype. Twin studies have reported no difference in heritability estimates of autistic symptomatology between the extremes of the distribution and normal variation (Lundström et al. 2012; Robinson et al. 2011), suggesting that clinical ASD and autistic-like traits in the general population are etiologically linked. In the current study, maternal 25 (OH)-vitamin D concentrations were unrelated to scores on the Total AQ and four of the five subscales. The one statistically significant correlation, between maternal vitamin D levels during pregnancy and offspring scores on the Attention Switching subscale, was weak in magnitude (r = −0.13), and when interpreted in the context of the null findings for all other AQ scales, raises the possibility that this may be a spurious association. After adjusting for confounding variables, regression analyses found that maternal 25 (OH)-vitamin D concentrations accounted for a very small, though statistically significant, amount of variance (2 %) in offspring Attention Switching scores. While the association was reduced to non-significance when the season of maternal serum collection was entered into the model, we argue that this does not discount a relationship, given the strong association of this variable with maternal 25 (OH)-vitamin D concentrations.

Analyses of the categorical data revealed that the association between maternal 25 (OH)-vitamin D concentrations and offspring Attention Switching scores was driven by a significantly greater proportion of ‘high’ scores from offspring of mothers in the lower tertile of vitamin D concentrations (8.4 %), compared to offspring of mothers in the middle (1.5 %) and upper (2.3 %) tertiles. After accounting for covariates, logistic regression revealed that this represented a five- to six-fold increase in ‘high’ Attention Switching scores for offspring of mothers in the lower tertile compared to offspring of mothers in the upper tertile. Importantly, the lower tertile threshold (≤49 nmol/L) corresponded well with the most widely-used definition of vitamin D insufficiency as a 25(OH)-vitamin D concentration less than 50 nmol/L (Holick 2007). This pattern of findings is consistent with the findings of a number of large population-based twin studies, which have found largely independent causes (genetic and/or environmental) operating on different aspects of ASD symptomatology (for a review, see Happé and Ronald 2008). In accordance with this ‘fractionable’ view of ASD, low maternal 25 (OH)-vitamin D levels may contribute to the etiological substrates underpinning certain ASD behaviors, such as attention switching difficulties, while exerting minimal influence over other aspects of the behavioral phenotype. However, given that this was a single significant association among six outcome variables investigated, we urge caution in this interpretation until replication studies have been conducted.

Vitamin D performs a number of biological functions that are fundamental to neurodevelopment, including a signalling role in neuronal differentiation (Brown et al. 2003; Marini et al. 2010), a regulation role in the metabolism of neurotrophic factors (Neveu et al. 1994; Wion et al. 1991) and neurotoxins (Garcion et al. 2002), and a protective role during brain inflammation (Atif et al. 2009; Cekic et al. 2009). Vitamin D may also be indirectly involved in fetal brain growth through its role in a number of endocrine functions (Eyles et al. 2003). Reduced levels of Vitamin D may disrupt one or more of these functions during critical phases of neurodevelopment and lead to behaviors consistent with aspects of the ASD phenotype.

In conclusion, the current study found little evidence that maternal vitamin D levels during pregnancy are related to the ASD phenotype among offspring. One significant association was observed between low maternal 25(OH)-vitamin D concentrations and an increase in the proportion of offspring with ‘high’ scores on the Attention Switching subscale of the AQ. While the effect was striking—a five to sixfold increase in ‘high’ scores—and remained after adjusting for confounding variables, we emphasise that this was the only one of six outcome variables examined that yielded significant effects. This pattern of findings suggests that low maternal vitamin D is not deterministic of ASD, and any possible association is complex. Studies that seek to replicate these findings in independent birth cohorts, in addition to investigations of animal models of maternal vitamin D insufficiency (e.g., Eyles et al. 2003; Zosky et al. 2011) as well as genetic interactions in humans (Wilkinson et al. 2000), will provide important extensions of this research.

Acknowledgments

The authors would like to acknowledge the National Health and Medical Research Council (NHMRC) for their long term contribution to funding the study over the last 20 years. Core Management of the Raine study has been funded by the University of Western Australia (UWA), Curtin University, the UWA Faculty of Medicine, Dentistry and Health Sciences, the Raine Medical Research Foundation, the Telethon Institute for Child Health Research, and the Women’s and Infants Research Foundation. AJOW is funded by a Career Development Fellowship from the NHMRC (#1004065). This study was partly funded by NHMRC Project Grant #1003424. These funders had no further role in study design, analysis, data interpretation or manuscript writing and submission. The authors are extremely grateful to all of the families who took part in this study and the whole Raine Study team, which includes the Cohort Manager, Data Manager and data collection team.

Copyright information

© Springer Science+Business Media New York 2012