OBJECTIVES. The objectives of this study were to compare the birth weight and gestational age distributions and prevalence rates of autism with those of other developmental disabilities and to estimate the birth weight–and gestational age–specific risks for autism.
METHODS. For the first objective, a retrospective cohort of children born in Atlanta, Georgia, in 1981–1993 who survived to 3 years of age was identified through vital records. Children in the cohort who had developmental disabilities (autism, mental retardation, cerebral palsy, hearing loss, or vision impairment) and were still residing in metropolitan Atlanta at 3 to 10 years of age were identified through the Metropolitan Atlanta Developmental Disabilities Surveillance Program. A nested case-control sample from the cohort was used for the second objective; all cohort children identified with autism were case participants, and control participants were cohort children who were not identified as having developmental disabilities or receiving special education services.
RESULTS. The prevalence of autism in low birth weight or preterm children was markedly lower than those of other developmental disabilities. In multivariate analyses, birth weight of <2500 g and preterm birth at <33 weeks' gestation were associated with an approximately twofold increased risk for autism, although the magnitude of risk from these factors varied according to gender (higher in girls) and autism subgroup (higher for autism accompanied by other developmental disabilities). For example, a significant fourfold increased risk was observed in low birth weight girls for autism accompanied by mental retardation, whereas there was no significantly increased risk observed in low birth weight boys for autism alone.
CONCLUSIONS. Gender and autism subgroup differences in birth weight and gestational age, resulting in lower gender ratios with declining birth weight or gestational age across all autism subgroups, might be markers for etiologic heterogeneity in autism.
Autism is one of the most common childhood neurodevelopmental disorders, with an estimated prevalence for all autism spectrum disorders of ∼6 cases per 1000 children.1–6 Although there is strong evidence for a genetic contribution to autism pathogenesis, the association with prenatal exposures such as thalidomide,7 valproic acid,8,9 and congenital viral infections10–13 suggests an etiologic role for adverse environmental exposures during pregnancy. If prenatal features, such as unusual birth weight or length of gestation, are associated with autism, then they might help to identify etiologic subsets, for which these features might be part of the phenotypic expression of autism in utero (eg, impaired fetal growth or length of gestation accompanying an underlying neurologic problem) or markers of prenatal insults contributing to increased risk.
Investigation of pregnancy risk factors in autism has been conducted in a variety of studies, but the associations between autism and birth weight, prematurity, and the related measure of small-for-gestational-age status are inconsistent, in part because of methodologic differences and limitations in some studies, such as small, clinic-based samples, and lack of control for confounding factors.13–36 In other studies, birth weight and prematurity were not considered separately but were included as components in summary measures of obstetric optimality.13,21,22,37,38 Collectively, these studies suggest that a history of suboptimal pregnancy conditions occurs for some children with autism, but the strength of the associations between autism and low birth weight (LBW) or preterm birth specifically, as well as the biological significance, is unclear.
Another consideration regarding pregnancy risk factors that was addressed inadequately in previous studies is the phenotypic diversity within the autism spectrum, because it is well known that autism is commonly accompanied by other developmental, medical, and genetic conditions, especially mental retardation (MR).5,19,39,40 The etiologic links between autism and these coexisting conditions, including MR, are not fully understood, and the coexisting conditions might be associated with LBW and prematurity.
To address these limitations, we used a large, population-based study to compare the birth weight–and gestational age–specific distributions and prevalences of autism with those of other select developmental disabilities (DDs) and to estimate the birth weight–and gestational age–specific risks for autism. For the latter, a main focus was multivariate analysis of the relationships between birth weight, gestational age, and autism, considering the presence or absence of MR and other select DDs accompanying autism. In this article, we use the term autism to refer to autism disorder, Asperger disorder, and pervasive developmental disorder not otherwise specified.
Study Design and Population
A retrospective cohort study design was used to determine the birth weight–and gestational age–specific distributions and prevalences of autism and select other DDs. The study cohort included all children identified through vital records who were born in metropolitan Atlanta, Georgia, in 1981–1993, survived to 3 years of age, and had recorded birth weight and gestational age data. Children in the cohort who had DDs and still resided in metropolitan Atlanta at 3 to 10 years of age were identified through the Metropolitan Atlanta Developmental Disabilities Surveillance Program (MADDSP). From the study cohort, a nested, case-control study sample was selected for estimation of the birth weight–and gestational age–specific risks for autism; the sample included case children with autism, on the basis of MADDSP data, and control children who were not identified in the MADDSP or were not receiving special education services.
Through 1996, the MADDSP was an active, population-based, surveillance system for children 3 to 10 years of age who had select DDs (cerebral palsy, MR, autism, hearing loss, and vision impairment) and resided in the 5-county metropolitan Atlanta area. Affected children were identified through active review of records from multiple sources, including special education and state service programs and hospitals and clinics providing pediatric medical care.41 The MADDSP multiple-source approach was designed explicitly to enhance completeness and to reduce ascertainment bias, which may arise if ascertainment is limited to single sources.42 Of the children with autism in this study, 54% were identified through both educational and noneducational sources, 40% were identified through educational sources only, and 6% were identified through noneducational sources only. Although this study includes only children who were both born in and currently resided in metropolitan Atlanta when identified by the MADDSP, this distribution according to source type is comparable to that reported by Yeargin-Allsopp et al5 for the full sample of both inborn and outborn Atlanta children with autism. The MADDSP autism case definition5 was based on the criteria for autism spectrum disorders in the Diagnosticand Statistical Manual of Mental Disorders, 4th Revision.43
We initially identified 434091 survivors to 3 years of age among metropolitan Atlanta births in 1981–1993. Children with cerebral palsy, MR, hearing loss, or vision impairment were identified at 3 to 10 years of age through the MADDSP for the 1991–1994 study years (ie, they were born in 1981–1991). Children with autism were identified at 3 to 10 years of age through the MADDSP for the 1996 study year (ie, they were born in 1986–1993). Children with no birth weight (0.1%) or gestational age (4.6%) information were excluded.
Nested Case-Control Study
We initially identified 617 children with autism in the MADDSP. The control children were randomly selected from the study cohort and frequency-matched to children with autism according to year of birth. Control children were not in the MADDSP or were not receiving special education services during the 1996 study year.
Information on DDs (including information on select coexisting medical conditions, such as epilepsy, obtained through record review) was obtained from the MADDSP. Birth weight, gestational age, parity, multiplicity, maternal age, maternal education, maternal race, and child gender data were obtained from birth certificates. Information on birth defects for children with autism and control children was obtained from the Metropolitan Atlanta Congenital Defects Program.44 We excluded 91 children (52 children with autism and 39 control children) from the nested case-control study analyses because of missing information on ≥1 of the perinatal or sociodemographic factors; lack of gestational age data accounted for 66% of the exclusions attributable to missing data (corresponding to 5.5% of children with autism and 4.2% of control children). Birth weight values of <500 g and gestational ages of <20 weeks were considered erroneous and were set as missing values for analytic purposes.
For the first objective, cumulative birth weight distributions in 250-g intervals were calculated for each DD, as well as for the entire study cohort (1981–1993). Birth weight–specific prevalence estimates were determined separately for each DD in 6 categories: <1000, 1000 to 1499, 1500 to 2499, 2500 to 2999, 3000 to 3999, and >3999 g. For gestational age, cumulative distributions in 1-week increments were calculated for each DD and for the entire study cohort. Gestational age-specific prevalence estimates for each DD were calculated for 5 gestational age categories: 20 to 23, 24 to 28, 29 to 32, 33 to 36, and ≥37 weeks. For each birth weight–or gestational age–specific estimate, the numerator was the number of children with the specific DD and the denominator was the total number of 3-year survivors in the corresponding study cohort in each birth weight or gestational age category.
For the nested case-control analyses, children with autism (n = 565) were examined as a whole, as well as divided into 3 mutually exclusive subgroups, that is, children with autism but no MR or any other DD surveyed in the MADDSP (autism/no DD; n = 216), children with autism and MR only (autism/MR only; n = 299), and children with autism and ≥1 other DD instead of or in addition to MR (autism/DD; n = 50). These subgroups were based on the availability and reliability of information on these conditions collected through the MADDSP; MADDSP methods did not permit diagnostic subtyping (eg, autism disorder versus Asperger disorder). The autism/DD group was heterogeneous in terms of the combinations of coexisting conditions, but in this sample the majority had ≥3 conditions (ie, autism plus MR plus ≥1 other DD).
To estimate the strength of the association between birth weight or gestational age and autism, adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were computed by using unconditional logistic regression techniques. Birth weight, gestational age, and all other perinatal and sociodemographic data were included in the logistic regression models. Primiparous status, singleton birth, no birth defect, children's age of 6 to 10 years in 1996, female gender, maternal white race, maternal age of 20 to 29 years, and maternal education equal to 12 years were used as reference categories.
Stratified analyses according to gender were also performed for all children with autism, as well as for each autism subgroup. Additional stratified analyses according to both gender and gestational age status (2 strata, ie, term birth, ≥37 weeks, and preterm birth, <37 weeks) were performed for all children with autism, to assess the impact of differences in birth weight on the risk for autism within these broad gestational age strata; these stratified analyses were not performed for each autism subgroup because of small sample sizes.
Birth Weight–and Gestational Age–Specific Distributions and Prevalence Rates
The cumulative birth weight and gestational age distribution curves for each of the DDs and the study cohort as a whole are presented in Figs 1 and 2, respectively; the numbers of children can be found in Tables 1 and 2. Both distribution curves for children with autism followed the distribution curves for the study cohort as a whole more closely than did the curves for each of the other DDs. Whereas the proportions of LBW (<2500 g) in the study cohort as a whole and for children with autism specifically were 8% and 12%, respectively, the LBW proportions were 24% for hearing loss, 26% for MR, 36% for vision impairment, and 50% for cerebral palsy. Similarly, the proportions of preterm birth (<37 weeks) in the study cohort and for children with autism were 12% and 14%, respectively, whereas the proportions were 23% for hearing loss, 27% for MR, 37% for vision impairment, and 49% for cerebral palsy.
The prevalence estimates (Tables 1 and 2) also revealed differences between autism and the other DDs. For cerebral palsy, MR, hearing loss, and vision impairment, the prevalence in each birth weight category of <2500 g or gestational age category of <37 weeks was 6 to 48 times higher than the corresponding overall prevalence rate. In contrast, the prevalence estimates for autism in the lowest birth weight or gestational age groups were at most threefold higher than the overall prevalence estimate.
Birth Weight–and Gestational Age–Specific Risks for Autism
Among all children, there was a significant 2.3-fold increased risk for autism among children with birth weights of <2500 g, which was significant after multivariate adjustment (unadjusted OR: 2.5 [95% CI: 1.6–3.8]) (Table 3); the nearly twofold increased risk among children born at <33 weeks of gestation was not significant (unadjusted OR: 2.6 [95% CI: 1.3–5.2]). With stratification according to gender (Table 3), the trends in birth weight or gestational age and autism among all children seemed to be largely attributable to the significant 3.5-fold increased risk for autism among LBW girls and the significant 5.4-fold increased risk for autism among early preterm girls (<33 weeks of gestation). In contrast, the adjusted ORs for autism were <2.0 and not significant among boys who were either of LBW or born at <33 weeks of gestation.
With stratification according to gestational age (preterm: <37 weeks; term: ≥37 weeks) (Table 4), the risks for autism in preterm children (overall or according to gender) weighing <2000 g were elevated, especially among girls, but results were not statistically significant. Among term children, however, the increased risk for autism was sevenfold and significant among girls of <2500 g.
Table 5 presents the adjusted ORs from birth weight or gestational age for the specific subgroups of autism, with stratification according to gender. LBW was associated with a higher risk of autism/MR only or autism/DD than autism/no DD for both genders but with higher and significant ORs for girls. For boys, there was no significant association between preterm birth and autism alone (autism/no DD) or accompanied by MR (autism/MR only); for girls, there was an approximately sixfold increased risk attributable to preterm birth at <33 weeks for both autism/no DD and autism/MR only, which approached significance for autism/MR only. The risk for autism/DD with preterm birth at <33 weeks was elevated twofold to threefold for both boys and girls, but the sample sizes were very small and the ORs were not significant.
Apparently, the proportions of LBW or preterm births were greater for girls, compared with boys, in all autism subgroups. In addition, there were greater proportions of individuals in each of the subgroups with autism accompanied by other DDs for girls, compared with boys. The effects of these differences in gender-specific proportions across the autism-, birth weight–and gestational age–specific subgroups on the autism gender ratio are illustrated in Table 6. Boys were overrepresented among all children with autism, yielding a gender ratio of 3.8, but the gender ratio was highest among children with autism/no DD at 5, declined to 3.5 among children with autism/MR, and was lowest in the autism/DD group at 2.5. This pattern was consistent within each birth weight/gestational age group. The gender ratio also declined with decreasing birth weight or decreasing gestational age for all children with autism, as well as within each autism subgroup. Therefore, the greatest male bias was for children with the highest birth weights (gender ratio: 6.5 for children of ≥4000 g) or gestational ages (gender ratio: 5.2 for children of ≥37 weeks) who were subsequently identified as having autism and no other DD (autism/no DD). The male bias declined markedly for children with the lowest birth weights (gender ratio: 1.6 for children of <2500 g) or gestational ages (gender ratio: 1.3 for children of <33 weeks) who were subsequently identified with autism accompanied by a DD other than or in addition to MR (autism/DD).
Overall, LBW was associated with an approximately twofold increased risk for autism, independent of other birth and demographic risk factors, but the magnitude and statistical significance of the risk varied markedly according to gender and whether autism was accompanied by other DDs. The risk for autism was consistently higher for LBW girls than for LBW boys. For all LBW children, the risk for autism accompanied by MR or other DDs was consistently higher than the risk for autism alone. Therefore, the risk was at least fourfold and significant for LBW girls for autism accompanied by MR (autism/MR only) or other DDs (autism/DD) but was only ∼1.5-fold and not significant for LBW boys for autism and no other DD (autism/no DD). We also observed a trend for an increased risk for autism from LBW in both term and preterm children (ie, a consistently higher risk for girls than boys, which was significant for LBW girls born at term); the potential association between small-for-gestational-age status and autism should be investigated further.
Considering gestational age, there was a nearly twofold increased risk for autism from early preterm birth (<33 weeks of gestation) overall, but this was primarily attributable to a significant, more than fivefold increased risk in early preterm girls. In additional stratified analyses according to the 3 autism subgroups, early preterm girls seemed to be at increased risk for all of the subgroups of autism, but the sample sizes in these strata were very small and none of the subgroup estimates was significant. The weaker overall associations between gestational age and autism than between birth weight and autism may reflect more instances of nondifferential misclassification according to gestational age than according to birth weight in our data (because birth certificate errors in gestational age data are more common than errors in birth weight45–48), which would bias risk estimates toward the null.
The foregoing patterns of risk for autism reflect a greater downshift in the birth weight and gestational age distributions for girls with autism than for boys, as well as for children of both genders with autism accompanied by other DDs (ie, our autism/MR only and autism/DD subgroups) than for children with autism alone (autism/no DD). These gender- and autism subgroup-specific distributions resulted in lower autism gender ratios with decreasing birth weight or gestational age, especially in autism accompanied by other DDs. Therefore, in comparison with boys, girls with autism were more likely to have reduced fetal size or length of gestation and to have additional DDs.
A few population-based studies examined the associations between autism and birth weight or gestational age in multivariate analyses.27,29,31,36 Although the autism diagnoses varied across those studies, in other respects the methods were fairly similar, and the results were consistent with the current study in describing increased risk (approximately twofold when adjusted ORs were reported) associated with different measures of preterm birth, small-for-gestational-age status, or lower birth weight. Except for the study by Eaton et al,29 however, those studies did not examine gender- or autism subgroup-specific differences in the association between autism and birth weight or gestational age.
Glasson et al34 reported that girls with autism were significantly shorter in length at birth and had greater gestational ages than did boys with autism; in their study, however, female case subjects were not compared with female control subjects. In the data reported by Lord et al,21 LBW/small-for-gestational-age status was more common among high-functioning girls with autism (20 of 23 girls) than among their female siblings (3 of 28 girls); there were no differences in frequency between male case subjects (9 of 23 boys) and their male siblings (11 of 26 boys). It has also been noted that the autism gender ratio is lower in individuals with MR than in individuals with normal cognitive functioning2 and lower in individuals with significant dysmorphologic features or microcephaly (“complex autism”) than in those without (“essential autism”).49 Furthermore, it was noted in some but not all studies that, in autism, the mothers of individuals with a lower IQ or more autism symptoms (eg, meeting criteria for autism disorder, compared with Asperger syndrome) had more pregnancy complications, compared with the mothers of control or case participants with a higher IQ or fewer autism symptoms.17–19,21,34 The current study extends these previous observations by reporting consistent trends for increasing proportions of girls with autism (both with and without comorbidities) and increasing proportions of individuals with comorbid conditions (both male and female) accompanying autism with declining birth weight and gestational age. Furthermore, across all birth weight and gestational age groups, greater proportions of girls with autism had comorbid conditions, compared with boys.
In the autism literature, both the high gender ratio and gender differences in the phenotypic profile for autism fuel the hypothesis that there are gender differences in the etiologic pathways to autism, at least for some individuals. Different proposed mechanisms to account for the gender differences include greater risk in boys through nongenetic or epigenetic mechanisms that act independently of genetic liability,50 which might include autism gene expression influenced by prenatal sex hormone exposure,51,52 a gender-specific threshold of risk, including a proposal for protective, imprinted, X-linked genes that increase the phenotypic threshold in girls,53 or gender-specific genes (including protective genes in girls) or gender-specific penetrance of some susceptibility genes.54,55
The results from this study might support such hypothetical gender differences. LBW and preterm birth are nonspecific markers of an adverse prenatal course (ie, they could be phenotypic manifestations of an impaired fetus or adverse exposures acting on a susceptible fetus). In autism, it is possible that either of the latter etiologic scenarios could be occurring, albeit in different subsets of individuals. Specifically, we hypothesize that some girls are more likely than boys to require a prenatal insult, such as reduced growth, preterm birth, or broader developmental insults, in the causal pathway leading to autism. Similarly, individuals with autism accompanied by other DDs (eg, the autism/MR only and autism/DD subgroups) are more likely then individuals with autism unaccompanied by other DDs (eg, the autism/no DD group) to have experienced a prenatal insult, such as reduced growth or preterm birth. In either instance, the birth weight or gestational age features might act as an added risk on an underlying genetic vulnerability to autism that otherwise might not be expressed (eg, in individuals with an insufficient “genetic load” or a protective genetic mechanism). In contrast, we hypothesize that boys, as well as individuals with autism and no other DD, are more likely to display a somewhat lower birth weight or length of gestation then those without autism, but as an epiphenomenon or phenotypic manifestation of an impaired fetus whose impairment leading to autism arises primarily through genetic mechanisms (eg, in individuals with a sufficient genetic load). The latter hypothetical group might, in the current study, be made up of the 28% of all autism cases involving term, normal birth weight boys with autism and no other DD (the proportion is 34% if we also include term, normal birth weight girls with autism and no other DD).
The relationships among prenatal factors including genotype, fetal growth, and postnatal/adult morbidity have been explored in a variety of contexts, for example, after the apparent association between LBW/intrauterine growth retardation and adult cardiovascular disease, stroke, and type 2 diabetes mellitus,56–59 and include evidence for shared genetic mechanisms of fetal growth and later disease.60 Distinguishing the relationships is complicated by the nonspecificity of LBW and preterm birth,61 however, and our results and foregoing hypotheses need to be explored in larger studies with more-precise measures of prenatal exposure and fetal status.
The strengths of this study include the population-based cohort design and availability of both sociodemographic and perinatal data collected prospectively in birth certificates to be included in a multivariate analyses. MADDSP prevalence rates and the distributions of autism according to gender and associated DDs are comparable to rates reported in other population-based studies, which suggests that MADDSP ascertainment is relatively complete and representative. There might be some ascertainment bias with respect to children with milder forms of DDs who did not need special education services or medical care. The birth weight and gestational age patterns among such children might resemble more closely the patterns observed for the autism/no DD subgroup, compared with the other subgroups in this study. If so, and if inclusion of such children in future studies is possible, then it might reduce the level of association between autism and birth weight or gestational age overall. Because of the small numbers of children, especially girls, in the lower birth weight and gestational age categories, these analyses should be replicated in larger studies. Also, MADDSP methods did not permit diagnostic subtyping (eg, autism disorder versus Asperger disorder).
We considered the implications of gender differences in survivorship (there are more male deaths than female deaths before 3 years of age) for our interpretation of gender differences in autism risk. In fact, because the number of deaths is relatively small, compared with the number of births, the actual number of “potential cases” that are removed, through death, from the study population of survivors is very small, and “loss” of these cases has little impact on our observed gender differences in autism distribution according to birth weight or gestational age among survivors. For example, we estimated the numbers of boys and girls of <2500 g with autism we might have lost through death (applying the same gender–and birth weight–specific rates for autism for deaths as we observed for survivors); if we included these lost cases in our sample, then the gender ratio for children with autism who were <2500 g changed to 2.3 from 2.1.
Children identified with autism more recently might have a different risk factor profile, compared with children in earlier cohorts. If changes in autism prevalence are associated with novel prenatal insults leading to autism, then this might be reflected in a shift in the birth weight distribution for autism spectrum disorders in more-recent births.
The magnitude of risk for autism resulting from LBW or preterm birth was dramatically lower than the levels associated with cerebral palsy, MR, hearing loss, or vision impairment. Failure to assess or to recognize autism features in LBW or preterm children with motor, cognitive, or sensory disabilities might account for some of this apparent difference in risk. Clearly, additional work is needed to confirm the magnitude of risk, to determine the proportion of children with autism associated with LBW or preterm birth, and to elucidate the role of these features in autism pathogenesis and phenotype, especially their role in gender-specific or other forms of autism subgroup variability.
- Accepted September 26, 2007.
- Address correspondence to Diana Schendel, PhD, National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, 1600 Clifton Rd, Mail Stop E-86, Atlanta, GA 30333. E-mail:
The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Suboptimal pregnancy conditions arise with autism, but the strength and biological significance of associations with LBW or preterm birth are unclear, including associations in autism subgroups with coexisting conditions that might be associated with LBW and prematurity.
What This Study Adds
This study compares the birth weight–and gestational age–specific distributions and prevalence rates of autism with those of other developmental disabilities and estimates birth weight–and gestational age–specific autism risks, with subgroup analyses considering other developmental disabilities that accompany autism.
- Bertrand J, Mars A, Boyle C, Bove F, Yeargin-Allsopp M, Decoufle P. Prevalence of autism in a United States population. Pediatrics.2001;108 (5):1155– 1161
- ↵Moore SJ, Turnpenny P, Quinn A, et al. A clinical study of 57 children with fetal anticonvulsant syndromes. J Med Genet.2000;37 (7):489– 497
- Deykin EY, MacMahon B. Viral exposure and autism. Am J Epidemiol.1979;109 (6):628– 638
- Knobloch H, Pasamanick B. Some etiologic and prognostic factors in early infantile autism and psychosis. Pediatrics.1975;55 (2):182– 191
- Mason-Brothers A, Ritvo ER, Pingree C, et al. The UCLA-University of Utah epidemiologic survey of autism: prenatal, perinatal, and postnatal factors. Pediatrics.1990;86 (4):514– 519
- Comi AM, Zimmerman AW, Frye VH, Law PA, Peeden JN. Familial clustering of autoimmune disorders and evaluation of medical risk factors in autism. J Child Neurol.1999;14 (6):388– 394
- Juul-Dam N, Townsend J, Courchesne E. Prenatal, perinatal, and neonatal factors in autism, pervasive developmental disorders-not otherwise specified, and the general population. Pediatrics.2001;107 (4). Available at: www.pediatrics.org/cgi/content/full/107/4/e63
- Klug MG, Burd L, Kerbeshian J, Benz B, Martsolf JT. A comparison of the effects of parental risk markers on pre- and perinatal variables in multiple patient cohorts with fetal alcohol syndrome, autism, Tourette syndrome, and sudden infant death syndrome: an enviromic analysis. Neurotoxicol Teratol.2003;25 (6):707– 717
- Indredavik MS, Vik T, Heyerdahl S, Kulseng S, Fayers P, Brubakk AM. Psychiatric symptoms and disorders in adolescents with low birth weight. Arch Dis Child Fetal Neonatal Ed.2004;89 (5):F445– F450
- ↵Larsson HJ, Eaton WW, Mardsen KM, et al. Risk factors for autism: perinatal factors, parental psychiatric history, and socioeconomic status. Am J Epidemiol.2005;161 (10):916– 925
- ↵Gillberg C, Steffenburg S, Schaumann H. Is autism more common now than ten years ago? Br J Psychiatry.1991;158 :403– 409
- ↵Boyle C, Yeargin-Allsopp M, Doernberg N, Holmgreen P, Murphy CC, Schendel DE. Prevalence of selected developmental disabilities in children 3–10 years of age: the Metropolitan Atlanta Developmental Disabilities Surveillance Program, 1991. MMWR Surveill Summ.1996;45 (2):1– 14
- ↵American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association; 1994
- ↵Piper JM, Mitchel EF, Snowden M, Hall C, Adams M, Taylor P. Validation of 1989 Tennessee birth certificates using maternal and newborn hospital records. Am J Epidemiol.1993;137 (7):758– 768
- ↵Barron-Cohen S, Hammer J. Is autism an extreme form of the male brain? Adv Infancy Res.1997;11 :193– 217
- ↵Barker DJ. The fetal and infant origins of adult disease. BMJ.1990;301 (6761):1111
- Barker DJ. Fetal origins of cardiovascular disease. Ann Med.1999;31 (suppl 1):3– 6
- Copyright © 2008 by the American Academy of Pediatrics