Autism risk in small- and large-for-gestational-age infants

Abstract

OBJECTIVE:

We sought to determine whether small-for-gestational age (SGA) and large-for-gestational age (LGA) birthweights increase autism risk.

STUDY DESIGN:

This was a retrospective cohort analysis comparing children with autism (n = 20,206) within a birth cohort (n = 5,979,605). Stratification by sex and birthweight percentile (SGA, <5th or 5–10th percentile; appropriate size for gestational age [GA], >10th to <90th percentile; LGA, either 90–95th or >95th percentile) preceded Cochran-Mantel-Haenszel analysis for GA effect, and multivariate analysis.

RESULTS:

Autism risk was increased in preterm SGA (<5th percentile) infants 23–31 weeks (adjusted odds ratio [aOR], 1.60; 95% confidence interval [CI], 1.09–2.35) and 32–33 weeks (aOR, 1.83; 95% CI, 1.16–2.87), and term LGA (>95th percentile) infants 39–41 weeks (aOR, 1.16; 95% CI, 1.08–1.26), but was decreased in preterm LGA infants 23–31 weeks (aOR, 0.45; 95% CI, 0.21–0.95).

CONCLUSION:

SGA was associated with autism in preterm infants, while LGA demonstrated dichotomous risk by GA, with increased risk at term, and decreased risk in the premature infants. These findings likely reflect disparate pathophysiologies, and should influence prenatal counseling, pediatric autism screening, and further autism research.

Autism is a neurodevelopmental disorder involving impaired communication skills, limited social interactions, restricted interests, and stereotypical behaviors. Autism is often associated with abnormalities in cognitive functioning, learning, attention, and sensory processing,^1,2 and typically presents in early childhood. The autism spectrum disorders are considered to be a significant public health issue, affecting 0.9% of US children.³

Despite increased public awareness and research efforts, the etiology of autism remains largely uncertain. Based on the myriad of genetic, environmental, perinatal, and immunologic associations to date, it is likely that either there is a multifactorial causal pathway (eg, an underlying genetic susceptibility triggered by an exogenous stressor)^4,5 or that autism is the common end point of multiple causal pathways.² Adverse perinatal conditions are common targets of inquiry, and several conditions, including prematurity^5–10 and low birthweight (BW)^7,11–14 have been associated with the later development of autism.

Prior analyses evaluated the role of abnormal BW percentiles, with conflicting results. A recent comprehensive meta-analysis of perinatal and neonatal risk factors for autism reported that small-for-gestational age (SGA) but not large-for-gestational age (LGA) BW were significantly associated with autism.¹² Within a Swedish case-control study, LGA BW were associated with increased rates of autism (odds ratio [OR], 1.7; 95% confidence interval [CI], 1.0–2.7).¹⁴ An analysis of autism risk by BW appropriateness stratified by gestational age (GA) at birth has not been published. It is conceivable that the combination of intrauterine and postnatal stressors associated with growth restriction and prematurity, or alternately fetal overgrowth met with placental senescence at term, may be more detrimental to neurodevelopment than anticipated. The goal of this analysis was to determine whether SGA or LGA infants are at increased risk for autism relative to appropriate-for-GA infants, and whether the association differs by GA at birth.

Materials and Methods

This was a population-based cohort study, approved by the California Protection of Human Subjects Committee; the Office of Statewide Health Planning and Development (OSHPD); and the University of California, Davis Human Subjects Committee.

We utilized a database constructed by our group in 2007 to evaluate perinatal risk factors for developmental disabilities. An 11-year birth cohort (Jan. 1, 1991, through Dec. 31, 2001) was identified within a database provided by the California OSHPD, which had previously merged birth records from the Linked Vital Statistics Birth and Infant Death File published by the California Department of Health Services with maternal and infant hospital discharge records from the entire state. The analysis was limited to infants who survived to 1 year of age, without exclusion of children with comorbid congenital or neurodevelopmental abnormalities. This database provided perinatal data that was collected prospectively (at the time of hospital discharge, and filing of birth certificate data), such as maternal demographics and comorbidities, prenatal care, delivery type, and infant BW and GA at delivery. We then identified children with autism within the California Department of Disabilities database who were born in California from Jan. 1, 1991, through Dec. 31, 2001, and linked their information to the maternal and infant birth records within the OSHPD/Department of Health Services files using linkage analysis. The overall rate of successful linkage was 85%, with a range of 81% (in 1991) to 90% (in 2001).

In the current analysis, the predictive variable was defined as BW appropriateness for GA. All births in the cohort were separated by year of birth and sex, and were then assigned to a GA group based on completed weeks of gestation at the time of birth. Regarding determination of GA, the basis for dating criteria was not a variable included within the California OSHPD administrative birth files. For each year, we calculated the threshold values for male and female BW appropriateness by GA within the annual birth cohort (at the 5th, 10th, 90th, and 95th BW percentiles). Each birth according to sex and year was identified as SGA (either <5th or 5–10th percentile), appropriate for GA (>10th to <90th percentile), or LGA (either 90–95th or >95th percentile). Annualized BW percentile thresholds for each GA were tabulated and averaged for descriptive purposes (Table 1). Trend tests demonstrated no significant time trend in percentile thresholds over the 1991 through 2001 time period.

TABLE 1.

Birthweight percentiles by gestational age at birth

Sex	Birthweight percentile
	5th	10th	90th	95th
	Birthweight, g
Male
Gestational age
23	539	569	1147	1302
24	543	599	1138	1289
25	587	648	1236	1447
26	633	712	1431	1655
27	695	779	1622	1892
28	774	875	1886	2168
29	865	967	2262	2505
30	963	1105	2642	2828
31	1118	1278	2960	3107
32	1319	1502	3208	3346
33	1532	1735	3412	3557
34	1765	1964	3585	3747
35	1994	2188	3722	3931
36	2193	2380	3755	3977
37	2411	2602	3844	4055
38	2630	2797	3962	4157
39	2772	2934	4080	4267
40	2853	3021	4174	4360
41	2888	3060	4249	4438
42	2838	3016	4240	4433
43	2807	2984	4180	4376
Female
Gestational age
23	491	520	1130	1253
24	523	547	1110	1263
25	558	600	1181	1405
26	596	656	1325	1593
27	652	723	1600	1915
28	687	795	1908	2189
29	800	913	2322	2534
30	933	1063	2679	2849
31	1065	1221	2987	3124
32	1219	1411	3186	3339
33	1451	1643	3373	3526
34	1677	1873	3521	3697
35	1891	2089	3627	3837
36	2091	2287	3655	3871
37	2320	2500	3725	3928
38	2530	2700	3827	4017
39	2681	2829	3925	4111
40	2762	2913	4015	4194
41	2790	2951	4084	4267
42	2758	2918	4081	4269
43	2724	2883	4041	4233

For each year within the 1991–2001 birth cohort, annual birthweight percentiles were calculated and used to determine birthweight appropriateness for children born within that year. The analysis included children who survived to 1 year of age. Gestational age refers to completed gestational age at birth (23w1d = 23 weeks).

Moore. Autism risk by birthweight percentile. Am J Obstet Gynecol 2012.

The outcome of the current analysis was a diagnosis of autism. Within California, 75–80% of children with autism are followed up by the California Department of Developmental Services (DDS),¹⁵ which provides services for people with autism, epilepsy, cerebral palsy, and mental retardation, without regard for income. The Client Development Evaluation Report (CDER) database held by DDS was utilized to identify cases of autism by: (1) an autistic level of “one” (full syndrome autism) on any CDER report; or (2) an International Classification of Diseases, Ninth Revision (ICD-9) code of 299.0 (autistic disorder), 299.8, or 299.9.¹⁶ As specific autism ICD-9 codes were not routinely included in the CDER during the time period queried, we do not have precise information regarding the number of cases of each autism spectrum disorder included without our cohort. We suspect that the majority of cases were autistic disorder, which is the only autism spectrum disorder eligible for services through DDS, in the absence of significant disability. While autism is typically diagnosed by age 3 years, the analysis included cases identified by DDS through Nov. 30, 2006, at which time the youngest member of our cohort was 4 years and 11 months old,¹⁷ leaving time for most of the children with a delayed diagnosis to be included in the analysis.

To estimate the strength of association between BW percentile and autism, OR was calculated with 95% CI. A multivariate logistic regression analysis was designed to account for perinatal risk factors previously associated with both BW percentiles extremes (SGA or LGA) and autism. We screened covariates for association with BW percentile (either SGA or LGA) using previously published studies, and confirmed an association with autism through the univariate analysis, using a 95%CI entirely >1.0 or <1.0 as a threshold for inclusion. We included chronic hypertension (, Clinical Modification [ICD-9-CM] 642.0–642.2, 642.7),^14,18,19 preeclampsia (ICD-9-CM 642.5),^20,21 maternal diabetes (ICD-9-CM 250, 648.0, 648.8),^22–24 maternal age,^15,25,26 twin gestation (ICD-9-CM 651.0),^15,27–29 birth order (derived from maternal parity),^{15,26,30–32} race,^33–35 and interpregnancy interval (derived from a search for previous deliveries in mothers with a parity >0).^36,37

To determine whether GA at birth influenced the association between autism risk and BW appropriateness, the analysis was stratified by GA: very preterm 23–27 weeks 6 days and 28–31 weeks 6 days; midpreterm 32–33 weeks 6 days; late preterm 34–36 weeks 6 days; early term 37–38 weeks 6 days; term 39–41 weeks 6 days; and postdates >42 weeks. Effect modification was ascertained for GA strata using a Cochran-Mantel-Haenszel test for difference of OR without a continuity correction.

Results

Within the cohort of 5,979,605 children born in California from 1991 through 2001, 21,717 children with autism were identified. The remainder of the birth cohort served as the control group (n = 5,957,888). Male sex, advanced maternal and paternal ages, Asian race, chronic hypertension, preeclampsia, any form of diabetes, high birth order (≥3), short interpregnancy interval, and twin gestations were significantly associated with autism (Table 2).

TABLE 2.

Demographic data of the 1991–2001 California birth cohort according to autism status

Variable	No autism		Autism		Analysis
	Deliveries, n	%	Deliveries, n	%	OR	95% CI
Total	5,957,888		21,717
Sex of child
Male	3,040,131	51	18011	83	4.66	4.50–4.84
Female	2,917,757	49	3706	17	1.00	Reference
Age of mother
≤20	960,822	16	1753	8	0.77	0.68–0.86
21–25	1,499,201	25	4263	20	1.00	Reference
26–30	1,633,158	27	6081	28	1.28	1.21–1.36
30–35	1,242,483	21	5927	27	1.67	1.57–1.77
35–40	530,653	9	3107	14	2.02	1.89–2.16
≥41	90,664	2	583	3	2.15	1.91–2.43
Age of father
≤20	408,807	7	668	3	0.73	0.61–0.87
21–25	1,149,714	19	2838	13	1.00	Reference
26–30	1,480,484	25	4958	23	1.30	1.21–1.39
30–35	1,339,188	22	5741	26	1.63	1.52–1.74
35–40	757,722	13	3953	18	1.98	1.84–2.12
≥41	408,036	7	2449	11	2.29	2.12–2.48
Maternal race/ethnicity
Non-hispanic white	2,082,149	35	8789	40	1.00	Reference
African American	421,764	7	1764	8	1.02	0.95–1.10
Hispanic	2,759,541	46	7809	36	0.69	0.66–0.73
Asian	593,146	10	3049	14	1.33	1.25–1.41
Other race	75,398	1	206	1	0.69	0.57–0.84
Chronic hypertension
No	5,919,730	99	21516	99	1.00	Reference
Yes	38,158	1	201	1	1.45	1.26–1.67
Preeclampsia
No	5,700,487	96	20468	94	1.00	Reference
Yes	257,401	4	1249	6	1.42	1.29–1.57
Diabetes (any)
No	5,747,210	96	20624	95	1.00	Reference
Yes	210,678	4	1093	5	1.42	1.34–1.52
Parity
Unknown	8,336	<1	23	28	0.14	0.01–1.50
Nulliparous	2,314,526	39	9194	42	1.00	Reference
1	1,851,911	31	7810	36	0.29	0.04–2.04
2	1,003,761	17	2937	14	0.20	0.03–1.41
3	440,443	7	1087	5	0.15	0.02–1.08
4	181,708	3	370	2	0.13	0.02–0.93
≥5	157,203	3	296	1	0.13	0.02–0.91
Birth order
Unknown	772,322	13	3185	41	1.16	1.11–1.20
1	5,162,747	87	18,414	36	1.00	Reference
2	22,466	<1	114	5	1.4	1.18–1.71
≥3	353	<1	4	11	3.18	1.19–8.52
Multiple gestation
No	5,882,776	99	21234	98	1.00	Reference
Yes	75,112	1	483	2	1.81	1.65–1.98
Months since last live birth
Unknown	2,721,666	46	10992	51	1.5	1.39–1.61
0–1.5 y	367,914	6	1673	8	1.70	1.60–1.81
1.5–2.5 y	796,301	13	3310	15	1.53	1.45–1.62
2.5–4 y	850,760	14	2328	11	1.00	Reference
4–6 y	604,500	10	1383	6	0.82	0.76–0.88
≥6 y	616,747	10	2031	9	1.19	1.12–1.26
Birthweight percentile
Unknown	436,800	7	1516	7	0.97	0.92–1.02
<5%	260,587	4	1090	5	1.17	1.10–1.24
5–10%	280,316	5	1070	5	1.07	1.00–1.13
>10 to <90%	4,414,624	74	15828	73	1.00	Reference
90–95%	290,242	5	1083	5	1.04	0.98–1.11
>95%	275,319	5	1130	5	1.15	1.08–1.22

Data includes only infants that survived to 1 year of age.

CI, confidence interval; OR, unadjusted odds ratio.

Moore. Autism risk by birthweightpercentile. Am J Obstet Gynecol 2012.

Within the entire population, autism risk was significantly higher at BW percentile extremes, with a ~10% increase in autism risk in the SGA (<5^th percentile) and LGA (>95th percentile) groups. The trend toward increased autism risk was also seen at both the 5–10th percentile and 90–95th percentile BW groups, although the results did not reach significance (Table 3).

TABLE 3.

Autism risk by birthweight percentile after multivariate analysis

Variable	aOR	95% CI	95% Wald
Birthweight percentile
<5%	1.10	1.04–1.18	1.04	1.18
5–10%	1.04	0.98–1.11	0.98	1.11
>10 to <90%	1.00	Reference
90–95%	1.04	0.97–1.10	0.97	1.10
>95%	1.12	1.05–1.19	1.05	1.19
Multivariate analysis
Maternal preeclampsia	1.11	0.95–1.29	0.95	1.29
Maternal hypertension	1.09	0.95–1.25	0.95	1.25
Maternal diabetes	1.22	1.15–1.30	1.15	1.30
Maternal age
≤20	0.58	0.55–0.61	0.55	0.61
21–25	1.00	Reference
26–30	1.3 5	1.30–1.40	1.30	1.40
30–35	1.76	1.69–1.83	1.69	1.83
35–40	2.19	2.09–2.30	2.09	2.30
≥41	2.42	2.22–2.65	2.22	2.65
Twin gestation	1.56	1.41–1.73	1.41	1.73
Birth order
1	1.0	Referenee;
2	0.62	0.50–0.76	0.50	0.76
≥3	1.13	0.42–3.05	0.44	3.19
Unknown	1.18	1.14–1.23	1.14	1.23
Race
Non-hispanic white	1.00	Reference	Reference
Hispanic	0.83	0.80–0.86	0.80	0.86
African American	1.19	1.13–1.25	1.13	1.25
Asian	1.15	1. 10–1. 19	1. 10	1. 19
Other	0.72	0.63–0.83	0.63	0.83
Time since last live birth
Missing	1.74	1.67–1.82	1.67	1.82
0–1 5 y	1.89	1.78–2.02	1.78	2.02
1.5–2.5 y	1.60	1.52–1.69	1.52	1.69
2.5–4 y	1.00	Reference	Reference
4–6 y	0.81	0.75–0.86	0.75	0.86
≥6 y	0.99	0.94–1.06	0 94	1.06

aOR, adjusted odds ratio; CI, confidence interval.

Moore. Autism risk by birthweight percentile. Am J Obstet Gynecol 2012.

After stratification by GA at birth, the test for difference of OR among different GA groups was significant (P <.001). The association between BW appropriateness and autism reached significance in the 23- to 31-week and 32- to 33-week GA groups. While an association between SGA (<5th percentile) and autism was noted in all GA strata, the effect reached significance only in infants born <34 weeks (23–31 weeks: adjusted OR, 1.60; 95% CI, 1.09–2.35; and 32–33 weeks: adjusted OR, 1.83; 95% CI, 1.16–2.87) (Table 4 and Figure). The risk for autism with an LGA BW depended on GA at birth. Above 37 weeks, there was a positive correlation between LGA BW and autism, significant in the 39-to 41-weekgroup, where LGA BW (>95th percentile) was associated with a 16% increase in autism risk. The effect was protective however, in the setting of prematurity (23–31 weeks), where LGA BW (90–95th or >95th percentile) was associated with a 51% and 55% reduction, respectively, in autism risk (Table 4). The number of children in each BW percentile and GA strata is outlined by autism status in Table 5.

TABLE 4.

Autism risk by gestational age and birthweight percentile

Gestational age groups	All (23–43 wks)		23–31 wks		32–33 wks		34–36 wks		37–38 wks		39–41 wks		≥42 wks
	aOR	95% CI	aOR	95% CI	aOR	95% CI	aOR	95% CI	aOR	95% CI	aOR	95% CI	aOR	95% CI
Birthweight percentile
SGA: <5%	1.10	1.04–1.18	1.60	1.09–2.35	1.83	1.16–2.87	1.07	0.86–1.34	1.10	0.97–1.25	1.09	1.00–1.18	1.24	0.98–1.58
SGA: 5–10%	1.04	0.98–1.11	1.36	0.91–2.02	1.00	0.57–1.78	1.12	0.91–1.38	1.01	0.89–1.15	1.03	0.95–1.12	1.17	0.92–1.48
AGA: >10 to <90%	1.00	Reference	1.00	Reference	1.00	Reference	1.00	Reference	1.00	Reference	1.00	Reference	1.00	Reference
LGA: 90–95%	1.04	0.97–1.10	0.49	0.25–0.95	0.56	0.25–1.26	1.04	0.84–1.30	0.98	0.86–1.11	1.07	0.99–1.16	1.16	0.92–1.46
LGA: >95%	1.12	1.05–1.18	0.45	0.21–0.95	1.00	0.52–1.89	0.97	0.77–1.22	1.10	0.97–1.24	1.16	1.08–1.26	1.11	0.88–1.41

The odds ratios were adjusted for maternal age, race, hypertension, preeclampsia, diabetes, birth order, twin gestation, and months since last live birth. The analysis included infants that survived to 1 year of age. 95% confidence internals (CIs) and adjusted odds ratios (aOR) are reported.

AGA, appropriate-for-gestational age; LGA, large-for-gestational age; SGA, small-for-gestational age.

Moore. Autism risk by birthweight percentile. Am J Obstet Gynecol 2012.

FIGURE. Autism risk by birthweight percentile and gestational age.

Within 1991 through 2001 California birth cohort, small-for-gestational age (SGA) birthweights (BW) (blue lines) were associated with increased autism risk <34 weeks, relative to appropriate-for-gestational age (GA) BW. Large-for-gestational age (LGA) BW (red lines) were protective against autism in preterm period but were associated with increased autism risk at 39–41 weeks. Analysis included infants that survived to 1 year of age. Multivariate analysis was performed for maternal age, race, hypertension, preeclampsia, birth order, twin gestation, and months since last live birth.

TABLE 5.

Numerical comparison of birthweight percentile and gestational age groups

Gestational age groups	23–31 wks				32–33 wks				34–36 wks				37–38 wks				39–41 wks				≥42 wks
	No autism		Autism		No autism		Autism		No autism		Autism		No autism		Autism		No autism		Autism		No autism		Autism
	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%
Birthweight percentile
SGA <5%	2423	5	30	8	2971	5	23	10	17,513	4	87	5	58,053	5	256	5	160,921	5	620	5	18,706	5	74	6
SGA 5–10%	2722	5	28	8	3079	5	13	5	18,533	5	95	6	63,558	5	252	5	172,256	5	609	5	20,168	5	73	6
AGA >10 to <90%	40,470	79	293	80	51,279	80	185	78	319,225	80	1325	79	1,018,370	80	3855	79	2,673,571	80	9184	78	311,709	80	986	77
LGA 90–95%	3028	6	9	2	3730	6	6	3	21,771	5	85	5	66,445	5	246	5	174,857	5	658	6	20,411	5	79	6
LGA >95%	2641	5	7	2	3427	5	10	4	20,704	5	77	5	63,750	5	273	6	165,432	5	689	6	19,365	5	74	6

AGA, appropriate-for-gestational age; LGA, large-for-gestational age; SGA, small-for-gestational age.

Moore. Autism risk by birthweight percentile. Am J Obstet Gynecol 2012.

Comment

Our large birth-cohort analysis describes autism risk by BW percentile, stratified by GA at birth. Within the entire cohort, both SGA and LGA BW were associated with autism. Stratification by GA revealed that only preterm SGA infants (<34 weeks) and term LGA infants (39–41 weeks) had a significantly increased risk for autism, while preterm LGA infants (<32weeks) were protected against autism.

The strengths of this analysis lie primarily in the large number of children with autism diagnosed using standardized criteria within a single analysis, and a population-based birth cohort with systematic and prospectively collected perinatal data allowing for determination of risk and limitation of recall bias. One of the potential limitations inherent to utilization of a large administrative database is the dependence on previously conducted diagnostic evaluations, as the sheer number of cases included precludes confirmation of diagnostic criteria for each individual case. It is fortunate that the California DDS uses a strict and uniform diagnostic criteria for autism throughout the state, distinguishing cases of autistic disorder from other disorders along the autism spectrum (pervasive developmental disorder–not otherwise specified and Asperger disorder), which are not eligible for services in the absence of significant disability. Regarding the possibility of incomplete inclusion of children with autism born within our birth cohort, DDS provides services to people with autism without regard to financial means or citizenship, and there isa75–80% rate of service provision by the DDS to children with autism.¹⁵ Intrinsic BW percentile tables were created for this analysis. We suspect that the tables are slightly skewed toward higher BW due to the increased mortality rates at the lowest BW,³⁸ and the inclusion within the database of only children who survived to 1 year of age. As the BW percentiles that reached significance in our study (<5th and >95th) are likely to be within the SGA and LGA criteria used in practice (<10th and >90th), we believe that our results will be clinically useful.

Our finding that SGA BW is associated with increased autism risk within the entire cohort confirms the results of 2 recent studies describing perinatal and neonatal risk factors for autism.^10,12 The current study provides certainty and strength to this association within a single study population using a large birth cohort. In addition, the current analysis demonstrates that this association is significant only in preterm infants, which has not been previously reported in a GA-stratified fashion. We also demonstrate a GA-dependent relationship between LGA BW and autism or of a protective effect of LGA BW in premature children. The increased risk for autism seen in term LGA infants is consistent with the results of a nested Swedish case-control study in which BW >2 SD was associated with autism (OR, 1.7; 95% CI, 1.0–2.7).¹⁴ The association was again noted but without attainment of significance in a comprehensive metaanalysis of perinatal risk factors for autism (OR, 1.11; 95% CI, 0.85–1.46).¹² LGA may not be a more commonly reported finding due to 2 factors. Principally, the increase in risk we found was small (12% prior to stratification by GA), and may not have reached significance in smaller analyses. Secondly, the opposing term and preterm relationships may have cancelled out any significant correlations between LGA BW and autism in analyses that did not stratify results by GA at birth.

The association between SGA BW and autism may reflect an insult to neurodevelopment that occurred in the prenatal or postnatal time period. Within the prenatal state, it is conceivable that the pathophysiology that limited fetal growth also compromised neurologic development. Of primary interest is intrauterine growth restriction (IUGR) due to placental insufficiency, in which a fetus does not reach its growth potential due to limited transport of nutrients and oxygen. IUGR of placental insufficiency is a condition associated with chronic hypoxia, metabolic derangements, and eventually acidosis.^39,40 Neonatal consequences of intrauterine acidosis include hypoxic ischemic encephalopathy, intraventricular hemorrhage, and periventricular leukomalacia.⁴¹ Not surprisingly, IUGR is associated with poor neurodevelopmental outcomes^42,43 including behavioral and school difficulties,⁴⁴ cognitive deficits,⁴⁵ and cerebral palsy.⁴⁶ It is thus conceivable that IUGR may contribute to the manifestations of autism seen in SGA children. Aside from IUGR from placental dysfunction, SGA can also be a consequence of intrauterine infection⁴⁷ or genetic abnormality,⁴⁸ both conditions associated abnormal neurodevelopment.^43,49 The association between SGA BW and autism may also reflect complications after birth such as intraventricular hemorrhage, which is more commonly seen in SGA infants of all GA.⁵⁰ Within our study population, the greatest risk of autism was noted within the 23- to 31-week and 32- to 33-week age groups of SGA infants (<5th percentile), which confirmed our hypothesis that the highest risk would be seen at the earliest GA, based on prior associations between autism and prematurity.^5–10 Finally, speculation into the nature of the association between SGA BW and autism must take into consideration the accuracy of the variables used to calculate BW percentiles. BW are a more certain variable than GA at birth, which can be derived from either menstrual record or ultrasound history. GA definitions using last menstrual period alone can underestimate the degree of prematurity by up to 35%.⁵¹ In fact, a common explanation for SGA BW is incorrect dating. Thus, low BW may reflect a greater degree of prematurity, which is a known risk factor for autism.^5–10

We found a dichotomous relationship between LGA BW and autism in preterm and term infants. The increased risk for autism seen in term LGA infants may reflect direct exposure to maternal metabolic abnormalities such as diabetes and obesity, inflammatory changes associated with maternal metabolic abnormalities, complications of labor, birth trauma, or postnatal metabolic derangements. It is notable that maternal diabetes and obesity are both risk factors for delivery an LGA child,^20–21 and have both been associated with increased autism risk.^9,24 While diabetes was a significant risk factor for autism in our overall analysis (OR, 1.42; 95% CI, 1.34–1.52), obesity was not included due to the poor coding of obesity within the OSHPD database, with a previously demonstrated sensitivity of 14%.⁵² The nature of the relationship between maternal diabetes and obesity with autism may be related to the altered inflammatory milieu seen in both maternal diabetes and obesity. Higher levels of circulating interleukin-6 levels were found in mothers with gestational diabetes (independent of obesity),⁵³ and systemic increases in circulating inflammatory cytokines and acute phase proteins were demonstrated in obese women.⁵⁴ Inflammatory mediators (including interleukin-6), have been shown to cross the placenta and to alter neurodevelopment in animal models.⁵⁵ Aside from the potential for a direct effect on fetal neurodevelopment, it is hypothesized that exposure to in utero inflammation primes the fetal immune system for lifelong immune dysfunction, a condition previously demonstrated in both children and adults with autism (including elevated proinflammatory cytokine profiles in the cerebrospinal fluid and blood), and that may contribute to the underlying aberrant neurodevelopment and behavioral abnormalities seen in autistic disorder.⁵⁶ Alternately, the relationship between maternal diabetes or obesity and autism may reflect permanent metabolic alterations via fetal programming. While the association may be incidental, a recent case-control study supported the role of fetal programming in autism, as people affected by autism were more likely to develop hyperlipidemia inadulthood.⁵⁷

Perinatal events may also contribute to the connection between LGA BW and autism seen in term infants. Birth trauma including shoulder dystocia is more common in LGA infants, as the risk appears to be proportional to weight at birth.⁵⁸ Shoulder dystocia is associated with neonatal encephalopathy⁵⁹ (which is a risk factor for poor neurologic outcome),⁶⁰ and birth trauma was significantly associated with autism in a recent metaanalysis of perinatal risk factors for autism (OR, 4.9; 95% CI, 1.41–16.94).¹² While one half of shoulder dystocias occur in infants of normal BW (2500–4000 g), the incidence of shoulder dystocia is 10-fold higher in infants weighing >4000 g.⁶¹ Within our population, a BW of 4000 g is not reached at the 90th percentile until >38 weeks in boys and 39 weeks in girls (Table 1), which coincides with the increased autism risk seen in term LGA children >39 weeks. LGA BW are associated with prolonged second and third stages of labor that impart greater risk for intraamniotic infection, and an increased likelihood of undergoing augmentation of labor; however, neither prolonged labor nor augmentation of labor were associated with autism in the recent metaanalysis of Gardener et al¹² on perinatal risk factors for autism. Why LGA BW would be protective against autism in children born <32 weeks is unclear. As preterm LGA infants have both higher rates of survival and higher rates of survival without major morbidity (grade III or IV intraventricular hemorrhage or periventricular leukomalacia, retinopathy of prematurity, necrotizing enterocolitis, or chronic lung disease),³⁸ more LGA preterm infants may survive infancy without significant neurologic pathology.

Conclusion

Autism risk is significantly increased at both extremes of BW appropriateness. Stratification by GA yielded a more complex relationship, with increased risk seen only in preterm SGA infants and term LGA infants. Unexpectedly, LGA BW correlated with a significantly lower autism risk in children born <32 weeks. The nature of the associations between SGA and LGA BW and autism risk remains uncertain. We hope to further investigate perinatal conditions surrounding term delivery of LGA infants, and better characterize the relationship between maternal metabolic disorders and autism. The association between LGA BW and autism risk at term adds further momentum to the current public health emphasis on appropriate prepregnancy weight and weight gain in pregnancy.