This study of maternal early life factors potentially associated with ASD found a significant relationship with maternal late adolescent BMI, indicating an approximate doubling of the risk of having a child with an ASD among individuals with a BMI of 30 or greater at age 18. There was also an increased risk with early menarche (age ≤ 11)in the full study but not in the prospective sub-group. Our results suggest that overall, early life maternal menstrual cycle characteristics and OC use do not affect risk of ASD. However, the statistically significant associations we found with early menarche and high BMI at age 18 should be examined in other studies.
The strengths of this study include a large sample size, a geographically representative national sample, a wealth of high quality maternal information, and the ability to assess results in a prospective sub-group. A number of limitations should also be noted. All information utilized in this study is self-reported via mailed questionnaires; therefore miscategorization of menstrual cycle characteristics, OC use, body shape and BMI, and other model covariates is possible. However, this is a sample of nurses, who have previously been shown to provide highly reliable and valid self-reported health information (Colditz et al. 1986
; Troy et al. 1995
; Colditz et al. 1997
; Tomeo et al. 1999
). We also do not have diagnostic confirmation on ASD status; only maternal report was available. However, a number of studies have utilized maternally-reported diagnosis of ASD and other child conditions, and have demonstrated reliability of these reports as well as validity with measures of developmental problems (Faraone et al. 1995
; CDC 2006
); in addition, reporting may be even more accurate in this medically trained population. Misclassification of women who may later have gone on to have a child with an ASD as non-cases is also possible; however, by 2003 the majority of women in the cohort had completed their childbearing years (only 2.7% of the cohort gave birth after 2000); thus, any such misclassification is expected to be small. Further, the inclusion of borderline ASD cases would tend to weaken rather than strengthen observed associations. While additional information on the affected child, such as date of birth and gender, would be useful in detailed analyses of exposures during and immediately prior to pregnancy, our analyses focused on early life factors and utilized the mother as the case unit (i.e. women who later go on to have a child with an ASD).
Because we focused our analyses on factors occurring early in the mother’s life, unmeasured confounding by other maternal factors is unlikely. We did not have information on use of steroid hormones or hormone levels prior to the exposures under study; however, use of steroids would need to occur prior to menarche or age 18 to act as a confounder for the relationships studied here. Genetic factors may be involved in both the development of ASD and the exposures under study; however, this would need to be addressed in a separate study. The potential for recall bias is possible in our full study; however, utilization of the prospective sub-group reduces this concern.
While there has not been extensive research into the potential developmental effects of pre-gravid OC use on subsequent offspring, the research of Mucci and colleagues suggested that OC use does positively affect birth weight and fetal growth (Mucci et al. 2004
). Taken together with the decreased risk of certain cancers among past OC users (Mucci et al. 2004
), these results implicate a persistent effect of OCs on hormone levels and suggest that pregravid use is of interest when considering maternal hormonal factors. We did not find an association between maternal use of oral contraceptives prior to first birth and risk of having a child with ASD; nor was longer duration of use associated with ASD in our full study. These results are consistent with one study that found no association between maternal OC use and ASD (Torrey et al. 1975
), and are in line with the overall findings of the safety of OC use with regard to child health (Rothman 1977
; Vessey et al. 1979
). Our result in the prospective sub-group of a somewhat decreased risk associated with longer duration of OC use just reached significance and could be due to chance, but needs to be assessed further in other studies. One small retrospective case control study found significantly less use of oral contraceptives in cases compared to controls (Juul-Dam et al. 2001
), which is contradictory to our prospective group finding; it is possible that differences in timing of use or sample size across studies may account for the discrepancy.
Our results suggest that maternal menstrual cycle characteristics earlier in life such as length and regularity do not have a role in the development of ASD. However, we did see a significant relationship between early age at menarche in the full study but not prospective group. There is evidence to suggest that the average age at menarche in the USA has decreased, while the prevalence of ASD has increased over the past few decades (Fombonne 2005
; Golub et al. 2008
). Adjustment of the estimated effects for age should account for any confounding due to these secular trends. The lack of consistency of this association in the prospective sub-group could be due to the fact that statistical power to detect the moderate association was somewhat reduced with the smaller case size (70% in the prospective sub-group to detect OR of 2 comparing top category to lowest vs 99% in the full study) (Chapman and Nam 1968
), or possibly residual confounding in the full study.
This is the first study to suggest that there may be an association between maternal late adolescent BMI and risk of having a child with ASD. We found nearly a doubling of the risk associated with obesity at age 18, and this association was also observed in the full cohort with an additional measure of body fatness, a body shape diagram (“Appendix
”), at age 20 (data not shown). However, we did not find a significant association with BMI at baseline, either in the full or the prospective sub-group. Nor does it appear that the association with BMI or body fatness extends to very early in the mother’s life, as we assessed as early as the nurses’ birthweight (data not shown) and body shape at age 10, and no significant associations were seen with these factors. While the BMI at 18 result could be due to chance, the consistency of the association in our full study and prospective sub-group supports the finding, and future work should investigate the role of maternal late adolescent BMI in association with ASD.
Maternal BMI itself has not, as of this writing, been directly studied in association with ASD, although one previous study assessed maternal morphology during pregnancy and identified a positive association between both maternal weight and height and having a child with autism (Wilkerson et al. 2002
). Our results suggest that late adolescent/early adulthood obesity, rather than BMI at a closer time to pregnancy with the affected child, may increase risk of ASD. Such an association could be due to longer-term effects of obesity beginning earlier in life which may have an impact on hormone levels, reproductive processes, or the perinatal environment.
Overall, our results do not support the hypothesis that underlying maternal hormone levels prior to pregnancy have a strong effect on risk of ASD. However, certain factors, such as BMI or menarche, may influence risk through a hormonal pathway. The potential effect of early age at menarche found in our full study could lend support to an altered hormone profile; however, as discussed, the lack of consistency of this result in the prospective sub-group limits the strength of the observed association. Our findings regarding BMI at age 18 could also be due to a hormonally-mediated pathway. There is evidence in other fields to suggest that differing hormone profiles can have a lasting effect on disease risk. Specifically, individuals who develop breast cancer may be exposed to different hormone levels than those who do not, as suggested by both direct analyses of hormone levels and associations with BMI and body shape (Carmichael and Bates 2004
; Tworoger et al.2006
). Among women in this cohort, adult and adolescent BMI has been inversely associated with risk of breast cancer, which has been suggested to be due to a hormonally-mediated pathway (Baer et al. 2005
; Tworoger et al.2006
). Further, in a study of 592 premenopausal NHS II participants, aged 33–52 years old, a positive association has been found between BMI at age 18 and in adulthood and free testosterone levels (Tworoger et al. 2006
). This study measured plasma concentrations of hormone levels from timed follicular and luteal phase blood samples in order to measure levels of sex hormones—including testosterone, estradiol, and progesterone—as compared to BMI, waist-to-hip ratio, and body shape. However, the study only assessed hormone levels at one time point; further work is needed to determine whether body mass is consistently related to hormone levels over time. Further, given the association Tworoger and colleagues found between BMI in adulthood and hormone levels, future work should also further investigate maternal adult BMI despite our null finding for BMI at baseline.
Other factors not available here, such as maternal psychiatric history, could potentially confound the BMI relationship; in particular, obesity has been linked with depression (Mather et al. 2009
), which has also been found to be elevated in parents of children with ASD (Larsson et al. 2005
; Lauritsen et al. 2005
). While we do not have information on maternal depression history at age 18, information on antidepressant medications was collected in later questionnaire years (1993, 1997, 2001 and 2003), although the diagnosis of clinical depression was not assessed until 2003. Depression in any of these postbaseline years could be related to depression at age 18, since depression commonly reoccurs throughout the lifespan (Bos et al. 2005
). However, adjustment for reporting of depression or antidepressant use in any of these years did not significantly alter the association. Future work investigating maternal BMI should also include more detailed psychiatric information.
Autism is known to have strong genetic influences; however, there is research suggesting that fetal development may play a role in the expression of ASD (London and Etzel 2000
). We were interested in determining whether factors extending even earlier than pregnancy in the mother’s life may be involved. Our results suggest that many early life maternal reproductive factors, including menstrual cycle length and regularity, body shape in childhood, and pre-gravid OC use, do not influence the risk of having a child with an ASD, but that maternal age at menarche and late adolescent BMI may be associated with ASD in the offspring. Further work will be needed to determine the consistency and nature of the potential associations between maternal early menarche, late adolescent BMI, and risk of having a child with ASD. If these associations do hold in other studies, it would be suggestive of a role for maternal hormones in risk of having a child with autism.