Our findings are consistent with those recently reported from a large study of members of a California health maintenance organization (4
) that found the risk of ASD to be positively and independently associated with both maternal and paternal age, with adjusted odds ratios nearly identical to those reported here. These findings contrast somewhat with 5 other recent epidemiologic studies that found only 1 or neither parent's age to be associated with ASD risk after controlling for the other parent's age (2
The lack of consistency across studies could be due to limitations of sample size and of population representation of previous studies as well as other methodological differences, including autism case definitions and inclusion criteria and the ability to control for important variables. The present study included a large sample of children with sufficient information to enable evaluation of separate and combined effects of each parent's age as well as birth order and other variables. With more than 1,200 cases, it included over 50% more cases and thus more statistical power than any of the previous studies examining independent effects of maternal and paternal age on ASD risk.
Another advantage of this study is the population-based nature and diversity of the cohort, allowing control for factors that may confound the association between parental age and ASD. Maternal education is 1 variable we considered to be a potentially confounding factor because it is associated with maternal age and has been observed to be related to ASD risk (15
). Our results, however, suggest that the association between advanced maternal education and ASD risk observed in unadjusted analysis may be spurious and due to confounding by parental age.
The results of this study also demonstrate the importance of controlling for birth order in evaluating independent effects of parental age on ASD risk. Because birth order increases with parental age and, in this and other studies, has been found to be negatively associated with ASD risk, failure to control for birth order may mask a positive association between parental age and ASD risk. Two of the previous studies reporting an association between advancing maternal age and ASD (2
) also had adjusted for birth order and, similar to the present study, found birth order to be negatively associated with ASD.
An additional advantage of this study is its restriction to a single birth year, thereby controlling for temporal trends in recent decades in both ASD prevalence and parental ages at the birth of their children. This feature of the study allows estimation of the association between parental age and ASD risk independently of temporal trends in diagnostic practices or other factors.
Public health implications
The strength of the independent associations between maternal and paternal age and ASD risk, as indicated by the odds ratios in the range of 1.2–1.4 reported here, is modest. However, the observation that these effects are independent of each other and of low birth order raises the likelihood that the combined effects of parental age and birth order may have important public health implications. Mean maternal age in the United States has increased steadily since the 1970s, particularly for firstborn children, for whom mean maternal age at delivery increased by 3.8 years between 1970 and 2004 (16
). In addition, the proportion of births to women aged ≥35 years began increasing in the United States after 1980, when it was 5%; by 2004, it had increased to 14.2% (17
). During this same period, fertility rates for men aged ≥40 years also increased each year, while fertility among men aged <30 years declined (16
). With the decline in average family size in recent decades, we would also expect the proportion of children who are firstborn to have increased. Similar trends are occurring in other developed countries (7
). The results of this study raise the question of whether some portion of the recent rise in ASD prevalence (19
) may be linked to recent trends in parental age and family size. A further question is whether a modest increase in prevalence associated with advancing parental age and low birth order may have contributed to a greater awareness of ASD and, in turn, increases in measured prevalence. The tendency for older parents of firstborn children to have higher levels of educational achievement and resources than other parents could further contribute to increased awareness and an expansion of the diagnosis of ASD.
Potential etiologic implications of parental age effects
Because we observed independent effects of the age of each parent on ASD risk, the possible mechanisms for these effects could include a broad range of processes associated with either or both maternal and paternal age. The observed paternal age effect independent of maternal age could point to a causal role of gene mutations in male germ cells, because the probability or selection of these mutations increases as men age (20
). The independent effect of maternal age, on the other hand, may point to age-related chromosome changes, pregnancy complications, or environmental exposures during pregnancy. Independent effects of 1 or both parents' ages also could point to a role of accumulated environmental exposures that may have mutagenic effects on gametes or could result from a combination of mechanisms (21
The association between advanced maternal and paternal age and ASD is also consistent with a potential role of infertility treatments or assisted reproductive technologies, the uses of which have increased in the past decade, especially by women and men of advanced reproductive age (23
). Numerous studies have found associations between these technologies and adverse pregnancy outcomes, including those due to epigenetic effects (24
), although a recent review found no evidence of elevated rates of autism among children born after in vitro fertilization techniques (28
). Even though we have no information about exposure to these treatments in our cohort, the observation that firstborn children of older parents had the highest ASD risk is consistent with a possible infertility treatment effect because women who give birth after infertility treatment are more likely to be primiparous than those represented in the general birth cohort. However, the association between multiple birth and ASD in this study was weak and not statistically significant (, unadjusted odds ratio), whereas assisted reproduction technologies are strongly associated with multiple birth (23
Another unmeasured factor in the present study potentially associated with both advanced parental age and ASD risk in offspring is psychopathology or behavioral traits of parents that may result in both delayed parenthood and genetic susceptibility to autism in offspring (14
The observation in this and at least 2 previous studies (2
) that the risk of developing ASD was highest for firstborn children and declined with increasing birth order is a pattern also observed for other childhood disorders, including type I diabetes and atopy, and is cited as support for the “hygiene hypothesis.” According to this hypothesis, firstborn children are exposed to fewer infections from other children early in childhood and, because of delayed immunologic challenge, may be more likely to develop autoimmune responses including those that may adversely affect neurodevelopment (29
). Another possible factor that could lead to the observed birth-order effect is exposure to potentially neurotoxic, fat-soluble chemicals accumulated in maternal tissue that have been passed to offspring transplacentally or through breast milk (30
). Because of accumulation over a lifetime, the load of such neurotoxins transmitted might be expected to be highest for firstborn children, particularly when combined with advanced maternal age. Another possible explanation for the observed birth order effect is “stoppage” or a tendency for parents of 1 child with ASD not to have subsequent children because of the demands of parenting a child with a disability or concerns about genetic susceptibility (31
), thus increasing the likelihood in the cohort as a whole that a child with ASD will have a low birth order. Information available for the present study did not allow examination of these hypotheses.
Another important limitation of this study is that the cohort available for analysis excludes births with missing paternal age information. Because this exclusion applied to both the ASD cases and the comparison group (), we would not expect it to have resulted in biased estimates of the association between ASD and parental age. In a separate analysis, we examined the association between maternal age and ASD without adjusting for paternal age and including the full birth cohort, and we found the association between maternal age and ASD to be the same as that observed in the subcohort with paternal age.
Another limitation is that the birth cohort comparison group includes about 1% of births of children who died postnatally in addition to an undetermined number who moved out of the study area between birth and the age of 8 years, whereas children who died postnatally and those moving out of the study area after birth are excluded from the case group. Because of this limitation, we could not estimate cumulative incidence of ASD. Nonetheless, this limitation is unlikely to have biased the estimated odds ratios reported in this study, particularly those adjusted for factors such as gestational age and birth weight for gestational age, which are strongly associated with postnatal mortality. Another possible explanation for the increase in ASD among offspring of older parents, but one we cannot evaluate with the data available, is that, compared with younger parents, older parents may be more aware of developmental abnormalities and better able to access diagnostic and special educational services. Other limitations are that parity pertains to only mothers and does not take into account the number of previous births fathered by the fathers represented in the cohort, potential for residual confounding by factors not measured in the present study, possible misclassification of ASD case status, and missing information on paternal education.
The results of this study provide the most compelling evidence to date that ASD risk increases with both maternal and paternal age and decreases with birth order. Further research involving large, well-characterized birth cohorts followed longitudinally will be required to confirm these findings and adequately evaluate the range of alternative genetic and environmental hypotheses that this and other studies raise regarding parental age and birth-order effects on ASD risk. Smaller, focused studies may also be useful, such as Crow's idea to look for mutations responsible for complex disorders of unknown etiology and with parental age effects by studying affected families with older parents (20