The observed prevalence of autistic spectrum disorders (ASD) has risen dramatically in the western world over the past 30 years. It is still unclear whether this is a true rise in incidence, or an artefact due to increased awareness and more inclusive diagnostic criteria. If the former, then an increased exposure to environmental influences (probably interacting with genetic variants) is likely to be responsible. The role of environmental factors in ASD is increasingly being acknowledged1 and the identification of modifiable environmental determinants offers potential for primary prevention of ASD. The diagnosis of ASD relies on behaviorally defined impairments in social interaction, communication, restricted patterns of interests and repetitive behaviors. Proposed revisions to the classification of ASD highlight the importance of studying it dimensionally; thus it is important to use study trait measures, as the various autistic traits (ATs) may have different aetiologies.2

Since features of ASD are evident in early life, processes leading to ASD are likely to begin during the perinatal period. Procedures such as induction of labour, epidural anaesthesia and caesarean section have also increased over time throughout the Western world. Thus features of the management of pregnancy, labour and delivery are plausible candidates for ASD risk; however high quality studies investigating these are lacking and may have resulted in misleading conclusions in recent meta-analyses, which did not take into account the quality of the data, or methodological issues such as identification of ASD cases, appropriate choice of controls, specificity of definitions of obstetric details, or sources of data.3,4 For example, the meta-analyses grouped together hypertension, pre-eclampsia (PE) and 'swelling'- probably to increase statistical power, but this strategy was far from biologically appropriate since PE has a major deleterious effect on the fetus, but neither hypertension without PE nor oedema (swelling) do. Other studies used obstetric optimality indices combining biological features such as parity, social circumstances and medical conditions. The resulting indices are unable to provide information on any specific factors or mechanisms and often reflect whether this was a first pregnancy rather than any clinical abnormality. No studies appear to have considered obstetric/neonatal factors in regard to measures of autistic traits.

Aims: The proposed study has the overall aim of carrying out a detailed assessment of the association between features of pregnancy and delivery with autism and autistic traits, to assess whether there is any evidence of a dose dependent relationship, and to determine whether there is any evidence of interactions between obstetric features and autistic traits contingent upon (a) other aspects of labour/delivery or (b) of genetic markers.

Methods: The study will take advantage of 2 data-rich cohorts in England.

The Avon Longitudinal Study of Parents and Children (ALSPAC):has collected detailed information on approximately 14000 mothers and children born 1991-2 in England.5 The mothers, who were enrolled in early pregnancy, completed detailed questionnaires at various times during pregnancy and subsequently. Information has been extracted from the medical records of both mother in pregnancy, delivery and postpartum, and of the infant in the neonatal period; the data abstracted includes the conditions arising, the treatments given, and biological measurements for 8300 mothers, comprising approximately 3000 variables. Thus, for example, the detailed blood pressure measures throughout pregnancy, together with gestation at which proteinuria was detected, allow a rigorous definition of pre-eclampsia, together with the gestation at onset.

Although 86 children with a confirmed diagnosis of ASD by age 11 years have been identified,6 the true strength of ALSPAC lies in the extensive autistic trait measurements throughout childhood.7 Of 93 individual trait measures available between age 6 months and 9 years, 4 have been identified that together explain 48% of the variance in diagnosis: the coherence subscale of the Children's Communication Checklist(CCC), the Social and Communication Disorders Checklist, repetitive behavior, and the sociability subscale of the EAS measure (38 m).7 We have shown that epidemiological and genetic associations with ASD are found in specific traits as predicted.7,8

The Twins Early Development Study(TEDS) is a prospective longitudinal study of a UK-based community sample of twins9 comprising all live twin births between January 1994 and December 1996 in England and Wales (n=16,810 families). Information about prenatal and neonatal problems were first collected at around 1 1/2 years of age using a questionnaire that asked about prenatal, birth and neonatal experience using the Obstetric Enquiry Schedule,10 a validated maternal interview on pregnancy and birth complications. Families have subsequently been followed and assessed using a variety of measures of development and behavior. Between the ages of 7-12 years of age, autistic traits were assessed using (CAST)11 and the CCC.12 Intellectual ability was assessed using the WISC.13 Mothers were interviewed about the pregnancy and birth.

Statistical analyses: Methods of analysis will be decided in view of the frequency distributions, and will use, where appropriate, apart from traditional analyses, statistical techniques that may further illuminate causal inferences including sibling-control design and propensity score matching . Careful consideration will be given to the multiple factors that may confound observed relationships, and a focus will be to disentangle any causal role of treatment from condition (confounding by indication). We will take account of the numbers of tests undertaken using the false discovery rate. The ALSPAC analyses will take advantage of the gene-environment interaction results identified in the project currently underway by Beate St Pourcain. Twin analyses as described by Purcell et al14 and van der Sluis et al15 will be used as a general test of gene-environment interaction; and candidate genes e.g. those concerned with the oxytocin / vasopressin system to test for specifc GxE interactions.

Unique features of proposed study:

1.Extensive prospectively recorded information concerning obstetric features in ALSPAC, blind to whether or not the child would develop ASD or autistic traits including detailed time related measures where appropriate.

2. ALSPAC and TEDS data-bases contain many genetic variants which will be used to determine interactions between obstetric factors with candidate genes. In addition ALSPAC will include epigenetic and CNV results, as they become available.

3. Both studies include information on a large number of potential socioeconomic, lifestyle and health-related confounders. Both include IQ measures on the child.

4. The ability to study various trait dimensions of autism in singletons in ALSPAC, and in twins in TEDS.

Expected results: The information generated should be of immediate clinical and policy importance, and impact on the current guidance related to obstetric treatment. The data should clarify whether features of obstetric management contribute to the development of autistic traits and whether the effects are contingent on genetic background.

Funding will be sought to enhance the obstetric data base by: [i] extract further details from medical records, focusing in particular on pregnancies resulting in the remaining cases of autism, and those scoring high on the 4 autistic traits, together with a control sample collected at random from the remaining records [total n=500+]; the records will be extracted by trained personnel, blind to whether a case or control. [ii] keying of the text data on the cases that have already been documented (n=~2000 records); [iii] coding of the various text categories, including the causes of admission(s), induction, caesarian section; the medication given in pregnancy and delivery; the characteristics of the neonate. All such factors will be categorized, documented and returned to the built files for use by others.

References

1. Hallmayer J et al. Arch Gen Psychiatry 2011; 68: 1095-1102. 2. Happe F et al. Nature Neuroscience 2006; 9: 1218-1220. 3. Gardener H et al. Brit J Psychiatry 2009; 195: 7-14. 4. Gardener H et al. Pediatrics 2011;128: 344-355. 5. Golding J et al.. Paediatr Perinat Epidemiol 2001; 15: 74-87. 6. Williams E, et al. Dev Med Child Neurol 2008; 50: 672-7. 7. Steer CD, et al., PLoS One 2010; 5: e12633. 8. Golding J et al. PLoS One 2010; 5: e9939. 9. Oliver BR, Plomin R. Twin research and Human Genetics 2007; 10: 96-105. 10. Bolton PF et al. J Amer AcadChild and Adolescent Psychiatry 1997; 36: 272-281. 11. Scott FJ et al. Autism 2002; 6: 9-31. 12. Bishop DVM. J Child Psychology and Psychiatry 1998;39: 879-891. 13. Weschler et al. The Psychological Corporation 1992.14. Purcell S. Twin Research. 2002; 5: 554-571.15. van der Sluis S et al. Behav Genet 2012; 42: 170-186.