Preterm or Early Term Birth and Risk of Autism

The Swedish Medical Birth Register contains prenatal and birth information for nearly all births in Sweden since 1973.¹³  Using this register, we identified 4 070 204 singleton infants born during 1973–2013 who survived to age 1 year. Singleton births were chosen to improve internal comparability, given the higher prevalence of preterm birth and its different underlying causes among multiple births. Because ASD is usually diagnosed after 1 year of age, survival to this age was required to enable an assessment of ASD prevalence among individuals who survived long enough to be diagnosed. We excluded 8409 (0.2%) births that had missing information for gestational age, leaving 4 061 795 births (99.8% of the original cohort) for inclusion in the study. This study was approved by the Regional Ethical Review Board in Lund, Sweden (2020/627). Participant consent was not required because this study used only pseudonymized, registry-based secondary data.

Gestational age at birth was identified from the Swedish Birth Register on the basis of maternal report of last menstrual period in the 1970s and ultrasound estimation starting in the 1980s and later (>70% of the cohort). This was examined alternatively as a continuous variable or categorical variable with 6 groups: extremely preterm (22–27 weeks), very to moderate preterm (28–33 weeks), late preterm (34–36 weeks), early term (37–38 weeks), term (39–41 weeks, used as the reference group), and postterm (≥42 weeks). In addition, the first 3 groups were combined to provide summary estimates for preterm birth (<37 weeks).

The study cohort was followed-up for the earliest registered diagnosis of ASD from birth through December 31, 2015 (maximum age 43.0 years; median 21.5). International Classification of Diseases, Ninth Revision, (ICD-9) introduced in 1987, was the earliest ICD version to include a code for ASD. In the current study, ASD was identified on the basis of at least 1 diagnosis in the Swedish Hospital or Outpatient Registers (which include all primary and secondary diagnoses) or Primary Care Register (which includes all primary diagnoses) by using ICD-9 codes 299A, 299B, 299W, and 299X during 1987–1996 and International Classification of Diseases, 10th Revision, (ICD-10) codes F84.0, F84.1, F84.3, F84.5, F84.8, and F84.9 during 1997–2015. The Hospital Register started in 1964 and had >99% nationwide coverage by 1987 and onward.¹⁴  The Outpatient Register started in 2001 and contains all outpatient diagnoses from all specialty clinics nationwide. The Primary Care Register initially included 2 populous counties covering 20% of the national population starting in 1998 and then was expanded to cover ∼90% of the national population by 2008 and onward.¹⁵  In a previous validation of ASD diagnoses in Swedes, researchers reported a positive predictive value of 96%.¹⁶  We performed a sensitivity analysis examining ASD on the basis of ≥2 (rather than ≥1) registered diagnoses by using the same diagnostic codes as above.

Other perinatal and sociodemographic characteristics that may be associated with gestational age at birth and ASD were identified by using the Swedish Birth, Outpatient, and Hospital Registers and national census data, which were linked by using a pseudonymized version of the personal identification number. Covariates included birth year (modeled simultaneously as a continuous variable and categorical variable by decade), sex, birth order (1, 2, and ≥3), maternal and paternal age at birth of the offspring (5-year intervals), highest maternal and paternal education level achieved (≤9, 10–12, and >12 years), maternal BMI at the beginning of prenatal care (<18.5, 18.5–24.9, 25.0–29.9, and ≥30), preeclampsia (International Classification of Diseases, Eighth Revision [ICD-8]: 637; ICD-9: 624.4–624.7; and ICD-10: O14–O15), other hypertensive disorders during pregnancy (ICD-8: 400–404; ICD-9: 401–405, 642.0–642.3, and 642.9; and ICD-10: I10–I15, O10–O11, O13, and O15–O16), maternal diabetes mellitus (ie, gestational or pregestational types 1 or 2; ICD-8: 250; ICD-9: 250, 648.0, and 648.8; and ICD-10: E10–E14 and O24), and family history of ASD in a first-degree relative (ie, parents or siblings, identified by using the Total Population Register and ASD diagnoses on the basis of inpatient and outpatient ICD codes as above). Maternal BMI, preeclampsia, other hypertensive disorders, and diabetes were included because they have been associated with preterm birth¹⁷  and are reported risk factors for ASD in the offspring.¹⁸  All variables were >98% complete, except for maternal BMI, which was recorded starting in 1982 (60.1% of the cohort). Missing data were modeled as a separate category. In addition, a sensitivity analysis was performed after restricting to all individuals with complete data (N = 2 409 928).

Poisson regression with robust SEs was used to compute prevalence ratios (PRs) and 95% confidence intervals (CIs) for associations between gestational age at birth and ASD, relative to term birth. Analyses were conducted both unadjusted and adjusted for covariates (as above). Prevalence differences (PDs) and 95% CIs, attributable fraction in the exposed (AF_e), and population attributable fraction (PAF) also were computed for each gestational age group compared to term. Sex-specific differences were assessed by performing sex-stratified analyses and formally testing for interaction between gestational age at birth and sex on the additive and multiplicative scale. Poisson model goodness of fit was assessed by using deviance and Pearson χ² tests, which revealed a good fit in all models.

Cosibling analyses were performed to assess the potential influence of unmeasured shared familial (genetic and/or environmental) factors on the observed associations.⁷  A total of 3 151 723 individuals had ≥1 full sibling and were included in these analyses. Poisson regression fixed-effects models were performed at the maternal and paternal level to account for all factors shared by full siblings, including genetic and early-life environmental factors, thus controlling for their shared exposures even if not specifically measured. In addition, these analyses were adjusted for the same covariates as in the primary analyses.

In secondary analyses, we also explored (1) associations between gestational age at birth and ASD either with or without intellectual disability (ICD-8: 310–315; ICD-9: 317–319; and ICD-10: F70–F79); (2) spontaneous versus medically indicated preterm birth in relation to ASD, relative to term birth; (3) interactions between gestational age at birth and fetal growth (small for gestational age [SGA], <10th percentile for gestational age] versus appropriate for gestational age [AGA], 10th to 90th percentile]) on the additive and multiplicative scale; and (4) associations between gestational age at birth and ASD risk after stratifying by birth decade. All statistical tests were 2 sided and used an α-level of 0.05. All analyses were conducted by using Stata version 15.1 (Stata Corp, College Station, TX).

In Table 1, we report offspring and parental characteristics by gestational age at birth. Preterm infants were more likely than term infants to be male or first born or have a family history of ASD; their mothers and fathers were more likely to be at the extremes of age or have low education level; and their mothers were more likely to have high BMI, preeclampsia, other hypertensive disorders, or diabetes.

ASD was identified in 58 404 (1.4%) persons. The median age for the entire cohort at the end of follow-up was 21.5 years (mean 20.8 ± 13.1). The median age at the earliest registration of ASD was 14.9 years (mean 15.8 ± 8.6). ASD diagnostic codes were unavailable before 1987, and the earliest available outpatient data started in 1998. Consequently, age at diagnosis was not precisely known and does not necessarily correspond to age at earliest registration of ASD.

Gestational age at birth was inversely associated with ASD risk in the entire cohort. ASD prevalences were 6.1% for extremely preterm, 2.6% for very to moderate preterm, 1.9% for late preterm, 2.1% for all preterm, and 1.6% for early term, compared with 1.4% for term birth (Table 2). Unadjusted PRs for ASD among those born extremely preterm, any preterm, or early term were 4.54 (95% CI, 4.08–5.06), 1.57 (1.52–1.62), and 1.19 (1.16–1.21), respectively, relative to those born term. After adjusting for covariates, most PRs were moderately reduced, but all remained significantly elevated. Adjusted PRs associated with extremely preterm, all preterm, or early term birth were 3.87 (95% CI, 3.48–4.31), 1.40 (1.36–1.45), and 1.12 (1.10–1.14), respectively, compared with term (P < .001 for each). Each additional week of gestation was associated with a 5% lower prevalence of ASD on average (adjusted PR per additional week, 0.95; 95% CI, 0.95–0.96; P < .001).

An estimated 83.7%, 37.7%, and 18.9% of ASD cases among individuals born extremely preterm, any preterm, or early term, respectively, were attributable to extremely preterm, any preterm, or early term birth (Table 2) (AF_e). In the entire population, an estimated 0.7%, 3.7%, and 4.4% of ASD cases were attributable to extremely preterm, any preterm, or early term birth, respectively (Table 2) (PAF).

Attention-deficit hyperactivity disorder (ICD-9 340.9 and ICD-10 F90) was identified as a comorbidity in 23 447 (40.7%) persons diagnosed with ASD. Its prevalence by gestational age at birth in the entire cohort was 8.4% for extremely preterm, 4.7% for moderate to very preterm, 3.6% for late preterm, 4.0% for all preterm, and 3.2% for early term, compared with 2.8% for term.

Preterm and early term birth were associated with increased risk of ASD among boys and girls (Table 2). In boys, the adjusted PRs for ASD associated with extremely preterm, all preterm, and early term birth were 3.72 (95% CI, 3.27–4.23), 1.35 (1.30–1.40), and 1.11 (1.08–1.13) and in girls were 4.19 (3.45–5.09), 1.53 (1.45–1.62), and 1.16 (1.12–1.20), respectively, compared with term (P < .001 for each comparison). PRs for ASD by gestational age at birth for boys and girls are shown in Figure 1A and 1B, respectively.

Interactions between gestational age at birth and sex are reported in Table 3. Regardless of gestational age, ASD diagnosis was substantially more common in boys than in girls. Preterm birth and male sex had a positive additive interaction (P = .004) but negative multiplicative interaction (P < .001) (ie, the combined effect of preterm birth and male sex on ASD risk exceeded the sum but was less than the product of their separate effects). The positive additive interaction indicates that preterm birth accounted for more ASD cases among boys than among girls.

Compared with the primary analyses, cosibling analyses to control for unmeasured shared familial factors resulted in only modest attenuation of risk estimates (Table 4). For example, comparing preterm with term births, the adjusted PRs for ASD in the entire cohort were 1.40 (95% CI, 1.36–1.45; P < .001) in the primary analysis versus 1.32 (1.29–1.36; P < .001) in the cosibling analysis; among boys were 1.35 (1.30–1.40; P < .001) vs 1.31 (1.26–1.35; P < .001); and among girls were 1.53 (1.45–1.62; P < .001) vs 1.51 (1.43–1.59; P < .001).

Preterm and early term birth were associated with ASD either with or without intellectual disability (Supplemental Table 5). However, they were more strongly associated with ASD with intellectual disability (adjusted PR, preterm 2.02 [95% CI, 1.89–2.17; P < .001] and early term 1.25 [1.19–1.31; P < .001]) than without intellectual disability (preterm 1.28 [1.23–1.32; P < .001] and early term 1.10 [1.07–1.12; P < .001]).

We further examined ASD risk after stratifying by spontaneous (71.4%) or medically indicated (28.6%) preterm birth, which was systematically recorded starting in 1990 (N = 2 415 158). Both spontaneous and medically indicated preterm birth were associated with an increased risk of ASD compared with term birth (adjusted PR, 1.34 [95% CI, 1.28–1.41; P < .001] and 1.52 [95% CI, 1.44–1.61; P < .001], respectively), but medically indicated preterm birth was the stronger risk factor (P < .001 for difference in PRs).

Both preterm birth and SGA were associated with increased risk of ASD after adjusting for each other as well as other covariates (PR, preterm versus term 1.40 [95% CI, 1.36–1.45; P < .001] and SGA versus AGA 1.29 [1.26–1.33; P < .001]), but preterm birth was the stronger risk factor (P < .001 for difference in PRs). These factors had a significant positive interaction on the additive but not multiplicative scale (ie, the combined effects of preterm birth and SGA on ASD risk exceeded the sum of their separate effects; P = .001 [Supplemental Table 6]). The positive additive interaction indicates that preterm birth accounted for more ASD cases among individuals who were also SGA compared with those who were AGA.

The main analyses were repeated after stratifying by birth decade (1970s, 1980s, 1990s, 2000s, and 2010s). Most results were similar across different birth cohorts (Supplemental Table 7). For example, comparing preterm versus term birth, the adjusted PR by birth decade was 1.44 (95% CI, 1.27–1.63) for the 1970s, 1.37 (1.28–1.47) for the 1980s, 1.37 (1.30–1.44) for the 1990s, 1.41 (1.33–1.49) for the 2000s, and 1.55 (1.30–1.85) for the early 2010s (P < .001 for each). Because ICD-9 was the first International Classification of Diseases version to include codes for ASD and was introduced in 1987, the similar findings for persons born in the 1970s (for whom ASD diagnoses would be recorded at later ages) suggest that the main results were not strongly influenced by the unavailability of ASD codes during the earliest years of follow-up. In addition, primary care and specialty outpatient diagnoses were unavailable before 1998 and 2001, respectively. The similar results among persons born in the 2000s and 2010s compared with earlier decades suggest that the findings were not strongly influenced by the unavailability of outpatient data in earlier years.

In sensitivity analyses, the main analyses were repeated after (1) identifying ASD on the basis of ≥2 (rather than ≥1) diagnoses or (2) restricting to individuals with complete data (N = 2 409 928). All risk estimates were nearly identical to the main findings, and the conclusions were unchanged. For example, the adjusted PR for ASD comparing preterm with term birth was 1.40 (95% CI, 1.35–1.45; P < .001) when identifying ASD on the basis of ≥2 diagnoses and 1.40 (1.34–1.45; P < .001) in a complete case analysis.

In this large national cohort, preterm and early term birth were associated with significantly increased risks of ASD in boys and girls. Persons born extremely preterm had an approximately fourfold risk of ASD. These associations were largely independent of covariates as well as shared genetic or environmental determinants of preterm or early term birth and ASD within families, consistent with a potential causal relationship.

To our knowledge, this is the largest study to date of gestational age at birth in relation to ASD, and 1 of the first in which researchers investigate sex-specific differences, early term birth, or the influence of shared familial factors. In a previous Swedish study of 3.3 million persons born in 1973–2008 who overlapped with the present cohort, researchers reported a threefold risk of ASD among those born extremely preterm (23–27 weeks) compared with those born at 37 to 42 weeks (unadjusted hazard ratio, 3.2; 95% CI, 2.6–4.0) but did not examine sex-specific risks or early term birth.⁹  In a US study of 195 021 births in 2000–2007, researchers reported a 2.7-fold risk (95% CI, 1.5–5.0) of ASD among those born at 24 to 26 vs 37 to 41 weeks.¹⁰  In a US study of 7876 births in 1994–2000, researchers reported nonsignificant increased risks of ASD among those born extremely preterm (adjusted hazard ratio, 1.75; 95% CI, 0.41–7.40) or early term (1.16; 0.83–1.62) compared with those born at 39 to 40 weeks, but the study was likely underpowered.¹¹  In a recent meta-analysis of 10 studies with a total of 6 485 081 participants, researchers reported a pooled relative risk of 1.31 (95% CI, 1.16–1.48; P < .001) among those born preterm versus those born term (variably defined).⁵ 

In the current study, we extend this evidence by providing population-based risk estimates in a cohort of 4 million persons with >58 000 ASD cases, which enabled well-powered analyses of sex-specific differences, narrowly defined gestational age groups, and more complete adjustment for potential confounders. We found that preterm birth was associated with a higher relative risk of ASD in girls (adjusted PR, 1.53) than in boys (1.35), likely because of a lower background prevalence of ASD diagnosis in girls. However, preterm birth was associated with a larger number of ASD cases among boys (additive interaction, P = .004). Furthermore, early term birth (37–38 weeks) was associated with a modestly increased risk of ASD in girls (adjusted PR, 1.16; P < .001) and boys (1.11; P < .001). Early term birth has previously been associated with other chronic medical disorders^19–24  and premature mortality from infancy into early adulthood,^7,12  supporting a redefinition of term birth.²⁵  However, to our knowledge, in the current study, we provide the first well-powered assessment of ASD risk associated with early term birth, which comprises 17% to 30% of all births.^7,8,26  Consistent with a previous study,⁹  our findings appeared largely independent of shared familial (genetic and/or environmental) factors. Preterm and early term birth were associated with increased risk of ASD either without or (more strongly) with intellectual disability, which was overall consistent with a smaller overlapping cohort study from Stockholm.²⁷  In addition, preterm birth was associated with increased risk of ASD across different birth decades from the 1970s through 2000s, consistent with a previous study in Denmark.²⁸ 

Previous evidence has indicated that ASD is highly (∼80%) but not entirely heritable.²⁹  Preterm birth and other early-life environmental exposures may also be important contributors to ASD risk. The mechanisms are not established but may potentially involve inflammatory pathways.^30–32  Increased proinflammatory cytokine levels, specifically interleukin 1, interleukin 6, and tumor necrosis factor α, have been associated with uterine activation and the timing and initiation of preterm birth.^33,34  Upregulation of proinflammatory cytokines in preterm infants has been reported to persist across the developmental trajectory in childhood.^32,35  Elevated inflammatory marker levels also have been detected in the brain and cerebrospinal fluid of individuals with ASD and may play a key role in its pathogenesis.^36,37  Inflammatory-driven activation of microglia and synaptogenesis may lead to alteration or loss of neuronal connections during critical periods of brain development,^36–38  which may be central to the development of ASD as well as other neurodevelopmental disorders.³¹  Intensive research into these mechanisms is ongoing and may potentially reveal new targets for intervention at critical windows of neurodevelopment to improve the disease trajectory.

The current study’s findings suggest that preterm and also early term birth should be recognized as independent risk factors for ASD. ASD may be significantly underdiagnosed in adult populations.³⁹  In the present cohort, the median age at ASD diagnosis was 15 years, partly due to the unavailability of ASD codes before 1987, so that persons born in earlier years were often diagnosed in adolescence or adulthood. Our findings provide further evidence that gestational age at birth should be routinely included in history taking and medical records for patients of all ages to help identify in clinical practice those born preterm or early term.⁴⁰  Such information can provide additional valuable context for understanding patients’ health and may facilitate earlier evaluation for ASD and other neurodevelopmental conditions in those born prematurely.

A key strength of the current study was its large national cohort design, which afforded the high statistical power needed to examine ASD risk in narrowly defined gestational age groups and other relevant subgroups. Highly complete nationwide birth and medical registry data helped minimize potential selection or ascertainment biases. The availability of outpatient as well as inpatient diagnoses enabled more complete ascertainment of ASD, rather than only the most severe cases, which may improve generalizability of the findings. Sweden’s national health system also facilitates a more complete capture of diagnoses for the entire population by removing barriers to health care access. The overall prevalence of ASD in this cohort (1.4%) was commensurate with those based on clinical assessments in other populations.²  We were able to control for multiple potential confounders and assess the influence of unmeasured shared familial factors using cosibling analyses.

This study also had several limitations. Detailed clinical records needed to validate ASD diagnoses were unavailable. However, ASD and other neuropsychiatric diagnoses in the Swedish registries have been validated previously and found to be highly reliable.^14,16  Outpatient diagnoses were available starting in 2001, resulting in underreporting during earlier years. However, a sensitivity analysis suggested that this was unlikely to have a major influence on the findings. It is possible that ASD is more likely to be diagnosed in preterm children because of greater contact with the health care system (ie, detection bias). However, this may be less likely among those born late preterm or early term, in whom increased risks also were found.

In this large population-based cohort, preterm and early term birth were associated with significantly increased risks of ASD in boys and girls. These findings were largely independent of measured covariates and unmeasured shared familial factors, consistent with a potential causal relationship. Persons born prematurely need early evaluation and long-term follow-up to facilitate timely detection and treatment of ASD, especially those born at the earliest gestational ages.

AF_e

attributable fraction in the exposed

AGA

appropriate for gestational age

ASD

autism spectrum disorder

CI

confidence interval

ICD-8

International Classification of Diseases, Eighth Revision

ICD-9

International Classification of Diseases, Ninth Revision

ICD-10

International Classification of Diseases, 10th Revision

PAF

population attributable fraction

PD

prevalence difference

PR

prevalence ratio

SGA

small for gestational age