Background
Autism, high-functioning autism, Asperger syndrome (AS) and pervasive developmental disorder not otherwise specified (PDD-NOS) are collectively referred to as autism spectrum conditions (ASC). Recent research has suggested that ASC represent the upper extreme of a collection of traits that are continuously distributed in the population [
1,
2]. This continuum view provides a shift away from the categorical diagnostic approach and towards a quantitative approach for measuring autistic traits.
The strong bias of ASC towards males is well established [
3], with a clear male to female ratio, estimated at 4:1 for classic autism [
4] and as high as 10.8:1 in individuals with AS [
5]. The extreme male brain (EMB) theory of autism proposes that ASC are an exaggeration of certain male-typical traits [
6,
7]. This theory has been extended to explain both cognition and neuroanatomy in individuals with autism [
8]. It has been suggested that prenatal exposure to testosterone may be a key biological mechanism for shaping sex differences in the brain, and may be involved in the biased sex ratio found in these conditions [
8,
9].
Although genetic sex is determined at conception, it is the gonadal hormones (that is, androgens, estrogens and progestins) that are responsible for differentiation of the male and female phenotypes in the developing human foetus [
10]. Direct sampling of foetal serum or manipulation of foetal hormone levels would be very dangerous; hence, researchers have used indirect methods of measuring prenatal hormone exposure to study effects on later development.
One such indirect measure is the ratio between the length of the second and fourth digits (2D:4D) of the hand. This ratio has been found to be sexually dimorphic, being lower in males than in females. The 2D:4D ratio is thought to be fixed by week 14 of foetal life, and has been found to reflect foetal exposure to prenatal sex hormones in early gestation [
11,
12]. Results from studies of 2D:4D ratios as a proxy for FT levels show that children with ASC have more masculinised digit ratios compared with typically developing boys. These patterns have also been observed in the siblings and parents of children with ASC, indicating the possibility of a link between genetically based elevated FT levels and the development of ASC [
13,
14]. Other evidence of a genetic link to ASC was provided by a recent study, which showed that genes regulating sex steroids are associated with autistic traits, as measured by scores on the Autism Spectrum Quotient (AQ), in a typical adult sample [
15]. A parallel study also showed that genes regulating sex steroids are associated with a diagnosis of AS in a case-control sample [
15].
The medical condition of congenital adrenal hyperplasia (CAH) leads to abnormally high androgen levels, and has provided researchers with a 'natural experiment' in which to examine the effects of elevated androgen exposure. Girls with CAH have more autistic traits (measured using the adult AQ) than their unaffected sisters [
16]. Given that this condition is usually treated after birth, this suggests that the higher AQ scores in such children reflect elevated prenatal androgen levels. However, these findings should be interpreted with caution, because CAH carries a number of related problems (and requires extensive treatment) which may affect the atypical cognitive profiles found in this population [
17,
18].
Some studies have also compared measurements of testosterone in umbilical cord (UC) blood with postnatal development. A recent study using UC blood testosterone measurements examined pragmatic language ability in girls followed up at 10 years of age. Results showed that the higher a girl's free testosterone level at birth, the higher the scores on a pragmatic language difficulties questionnaire [
19]. However, levels of FT are typically at very low levels from about week 24 of gestation, and UC samples can contain blood from the mother as well as the foetus (and hormone levels may vary due to labour itself) [
20], so UC blood testosterone does not allow testing of whether outcomes reflect FT
per se.
Currently, the best method to examine the effect of FT is to sample via amniocentesis the amniotic fluid surrounding the foetus. An advantage of amniotic fluid samples is that amniocentesis is often performed for routine clinical purposes within a relatively narrow time period that coincides with the hypothesised critical period for human sexual differentiation between weeks 8 and 24 of gestation [
21]. This is also more direct than the 2D: 4D method, as the hormones themselves can be assayed, rather than relying on a proxy for these.
Several studies have linked elevated levels of FT in the amniotic fluid with the masculinisation of certain behaviours, beginning shortly after birth. Elevated FT has been linked to reduced eye contact in infants, smaller vocabulary in toddlers, narrower interests at 4 years of age, less empathy at 4 and 8 years, and increased systemizing (the drive to analyse and construct systems) at 8 years [
9,
22‐
26]. In addition, FT levels have been found to be positively correlated with the number of autistic traits in older children (6-10 years old), using two independent dimensional measures of autistic traits (the AQ-Child version and the Childhood Autism Spectrum Test) [
27].
The current study aimed to extend our understanding of the effect of prenatal hormones on the development of autistic traits, using a new measure for evaluating autistic traits in toddlers. The Quantitative Checklist for Autism in Toddlers (Q-CHAT) is a parent-report questionnaire that evaluates autistic traits in toddlers between 18 and 24 months old [
28]. Scores on this measure have shown a near-normal distribution in a large general population sample (n = 779), suggesting that it is a useful measure of individual differences in autistic traits in toddlers [
28]. In the present study, the relationship between measurements of FT and scores on the Q-CHAT were examined.
Discussion
This study examined the relationships between amniotic measurements of FT and autistic traits measured using the Q-CHAT. Sex differences were observed, with boys scoring higher on the Q-CHAT than girls. These early sex differences are consistent with other studies in young children that have found a female advantage in eye contact [
26], vocabulary development [
25] and preference for looking at faces [
37,
38]. They are also consistent with sex differences on measures of autistic traits, both on the Childhood Autism Spectrum Test [
39], the Autism Spectrum Quotient (AQ)-Child version [
40], the AQ-Adolescent version [
41] and on the English version of the adult AQ [
1]. This pattern has also been observed cross-culturally using translated versions of the AQ [
42‐
45].
FT levels were found to be approximately 2.5 times higher in boys than girls, consistent with previous studies using measurements of testosterone levels in amniotic fluid [
22,
23,
27,
46‐
48]. FT level was the only variable that was significantly related to Q-CHAT scores, both when the sexes were combined and when girls and boys were examined separately. Sex and the sex/FT interaction were not retained in the final regression model, suggesting that this is an effect of FT, rather than sex. No significant association between FT levels and FE levels was found, nor was FE correlated with autistic traits, and nor was there any association between child sex and FE levels (
r = -0.02,
P > 0.05), providing evidence for a link between autistic traits and FT levels rather than FE levels. This replicates previous studies showing no link between FE levels and behaviour [
24‐
26,
49,
50].
The current study supports the idea that FT levels are associated with autistic traits in toddlers. Like previous research in older children showing a positive relationship with FT levels and autistic traits in 6 to 10-year-olds [
27], these results at 18 to 24 months of age suggest that FT levels are associated with aspects of early development measured by the Q-CHAT, including sociability, early play behaviour and joint attention. These findings across different ages in development also suggest that the association between FT and autistic traits is robust. However, owing to the correlational design of this study, these results cannot be used to infer that FT is one of the causes of autism. Causality can only be demonstrated by manipulation of FT levels, which in humans would be wholly unethical. In addition, these studies only provide evidence for the role of FT in the development of autistic traits in typically developing children. This association remains to be tested in clinical samples, and a large-scale collaboration is currently underway to obtain a sufficient sample size to compare FT levels in cases of ASC versus controls.
The majority of total circulating testosterone is bound to sex hormone-binding globulin, and to albumin- and cortisol-binding globulin, and only the remaining 1-2% (the free fraction) is active [
51]. In this study, the assay used can only measure total extractable testosterone and cannot be adapted to obtain measurements of free fraction testosterone levels. Whitehouse
et al. (in press) obtained measurements of both the total testosterone and the free fraction testosterone concentrations using UC blood samples, and investigated relationships to pragmatic language ability in girls [
19]. Although the only significant relationship was observed between the free fraction testosterone and pragmatic language ability, a significant correlation was also observed between total testosterone and the free fraction testosterone concentrations (
r = 0.64,
P < 0.01). This suggests that the measurements of total testosterone obtained in this study are still useful.
It is also known that hormones fluctuate during the day and between days, even in foetuses [
52,
53]. It is not possible to obtain repeated samples of FT because amniocentesis itself carries a risk of causing miscarriage (of about 1%) [
54,
55] so obtaining amniotic FT measures are opportunistic, when the procedure is being carried out for clinical reasons, with never more than a single measurement of FT at one time-point. The representativeness of a single sample of FT thus remains unclear, but would be difficult to explore in an ethical manner. Given the reported time course of testosterone secretion [
56], the most promising time to measure FT is probably at prenatal weeks 8 to 24 [
9,
21,
57], but this is still a relatively wide range. The inferences we can therefore draw about the single measurement of FT are necessarily limited.
This sample of children whose mothers had undergone amniocentesis is also not representative of the wider population (for example, maternal age is usually higher in such samples). In this study, no relationship between maternal age and Q-CHAT scores were observed, suggesting that the association found here may apply to all pregnant women. For the reasons mentioned above, it would be unethical to test this association in a randomly selected sample of the child-bearing population [
9].
Acknowledgements
This work was supported by grants to SBC from the Nancy Lurie-Marks Family Foundation and the MRC. BA was supported by a scholarship from Trinity College, Cambridge. We are grateful to the families who have taken part in this longitudinal study over many years and to Rebecca Knickmeyer, Bhismadev Chakrabarti, Carrie Allison, Liliana Ruta, Mike Lombardo, Melissa Hines, Ieuan Hughes, and Richard Nash for valuable discussions. This study was conducted in association with the NIHR CLAHRC for Cambridgeshire and Peterborough NHS Foundation Trust.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
BA conceived of and carried out the study, and drafted the manuscript. KT carried out the radioimmunoassays. GH participated in the recruitment of women who had an amniocentesis. SBC contributed to the study design, obtained funding for and coordinated the study, and helped to draft the manuscript. All authors read and approved the final manuscript.