Association between gestational weight gain and autism spectrum disorder in offspring: a meta-analysis

Abstract

Objective:

To quantitatively examine the association between gestational weight gain (GWG) and risk of autism spectrum disorder (ASD) in offspring.

Methods:

Electronic databases were searched for studies of excessive or inadequate GWG, as compared to recommended GWG, in relation to the risk of ASD in offspring. Measures of the association from primary studies were pooled using a meta-analytic approach and expressed as weighted odds ratios (ORs) with 95% confidence intervals (95% CIs).

Results:

Nine studies were identified, including 323,253 participants with 4,135 cases of ASD from five cohort studies, and 1,462 cases and 3,265 controls from four case-control studies. Evidence from cohort studies indicates that both excessive and inadequate GWG were significantly associated with a higher risk of ASD in offspring. The pooled OR of ASD was 1.10 (95% CI: 1.02–1.18) for excessive GWG and 1.13 (95% CI: 1.04–1.24) for inadequate GWG using recommended GWG as the reference. Evidence from case-control studies suggests that excessive GWG (1.38 [95% CI: 1.19–1.62]), but not inadequate GWG (0.87 [95% CI: 0.72–1.04]) was significantly associated with a higher risk of ASD.

Conclusions:

The accumulated evidence supports that gaining weight outside the recommended GWG is associated with a higher risk of ASD in offspring.

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental disorder commenced from early childhood that is manifested by persistent social interaction deficits in parallel with repetitive and stereotypic activities (1). ASD is a life-long disorder. Most of the children with ASD may not work full-time or even live independently after they reach adulthood, which may bring substantial economic burden to their families, healthcare system, and society (2). Since ASD is estimated to affect approximately 1% of people globally (3), identifying modifiable risk factors for ASD is of great public health significance.

Gestational weight gain (GWG) is one of the few modifiable factors related to maternal and neonatal health outcomes (4). However, many women neither received consistent advice on GWG from healthcare providers nor realized the risks of inappropriate GWG to maternal and fetal health (5, 6). According to a recent review of more than 1 million women worldwide, approximately 47% of pregnant women gained excessive weight, and 23% of them gained inadequate weight during pregnancy (4). Both excessive and inadequate GWG are indicators of suboptimal nutritional and metabolic status, and can lead to an unfavorable intrauterine environment for the development of the fetus (7). It has been suggested that gaining gestational weight outside the current recommended guidelines may play a critical role in triggering the manifestations of ASD phenotypes in predisposing individuals during the prenatal period. However, existing studies have provided inconsistent evidence. Some reported a moderate association between excessive GWG and ASD risk in offspring, but some others found no association (8–16). Findings for inadequate GWG are even more controversial (8–12, 14–16).

A recent meta-analysis summarized data from two cohort studies and three case-control studies on this topic (17), however, this meta-analysis missed three large cohort studies and one case-control study. Of note, the authors pooled results from cohort and case-control studies together due to the limited number of primary studies, which is not an ideal analytic strategy and may lead to biased results. In addition, the data from one cohort study (9) used a different reference group for excessive and inadequate weight gain, which might bias the pooled analysis. Therefore, we conducted this meta-analysis that includes de novo results from the primary studies to provide the most comprehensive data to the literature.

Methods

Study selection

This meta-analysis was conducted based on the criteria of Preferred Reporting Items for Systematic Reviews and Meta Analyses statement (Supporting Information Table S1). Studies that examined the association between GWG and ASD risk in offspring and published before March 19, 2020, were identified in PubMed and Embase databases. The literature search was based on the following terms: (maternal OR gestational OR prenatal OR pregnant OR pregnancy) AND (weight gain OR weight growth OR weight change) AND (autistic OR autism OR autism spectrum disorder OR ASD OR Asperger). Bibliographies of relevant studies and Google Scholar were searched for additional eligible studies. Published studies with English or Chinese languages were reviewed.

Two independent investigators (L.S. and C.C.) reviewed the titles, abstracts, and full-texts of potentially eligible studies. Disagreements were resolved by group discussion. Studies were considered eligible if they met the following inclusion criteria: 1) cohort, case-control, or cross-sectional studies that investigated excessive GWG or inadequate GWG in relation to risk of ASD in offspring, compared with recommended GWG; and 2) the relative risk (RR), hazard ratio (HR), or odds ratio (OR) with corresponding 95% confidence intervals were reported, or such information can be derived from published results.

The detailed study selection process is shown in Figure 1. From 206 non-duplicate articles identified from the databases, 93 irrelevant studies, 53 animal studies, and 23 non-original studies (reviews or letter-to-editors) were excluded. After reviewing the full-texts, 29 articles were further excluded because the association of interest was not reported (n=26), or the available data could not be combined with other studies (n=3). In order to appropriately combine data from the primary studies, we requested for de novo results from three studies (12, 18, 19), and obtained the new results from Xiang et al.(12). Therefore, 9 studies (5 cohort and 4 case-control studies) met the inclusion criteria and were included in the present meta-analysis.

Figure 1.

Flow diagram of the study selection process

Quality assessment

A modified Newcastle-Ottawa Scale (NOS) was used to assess the quality of the included studies (20). All primary studies scored equal to or above 5 out of 7. The mean score of the five cohort studies was 5.7 out of 7, while the mean score of the four case-control studies was 5.9 out of 7. Thus, the identified studies were considered to have medium to high quality and thus were included in the current meta-analysis (Supporting Information Table S2–5).

Data extraction

Information on the primary studies was extracted by L.S. and verified by C.C. independently. By using a standardized data extraction form, the following information was retrieved: the first author’s last name and published year, the region of study, number of cases and participants (or number of cases and controls in case-control studies), offspring age, study period (for cohort studies), GWG assessment method, ASD confirmation, guidelines of GWG, risk estimates compared with the recommended GWG, and adjusted covariates.

Eight included studies (8, 10–16) categorized GWG based on the 2009 Institute of Medicine (IOM) guidelines, which were endorsed by the American College of Obstetricians and Gynecologists (21). These studies reported the multivariable-adjusted ORs or HRs that compared excessive or inadequate GWG to the recommended GWG. In Dodds et al. (9), the authors used the 1990 IOM guidelines and calculated the multivariable-adjusted RR of excessive GWG (GWG ≥18 kilograms) by comparing them to the rest participants including both inadequate GWG (GWG < 7 kilograms) and recommended GWG, and vice versa for the multivariable-adjusted RR of inadequate GWG. Since the results presented in Dodds et al. (9) cannot be directly combined with that of other primary studies, we calculated the crude ORs of excessive and inadequate GWG as compared to the recommended GWG based on the available data presented in the study.

Statistical analysis

HR was considered equivalent to OR in this meta-analysis. OR and HR were transformed into their natural logarithms (ln) and the corresponding 95% CIs were used to calculate their standard errors. The association between inadequate or excessive GWG, compared to recommended GWG, and risk of ASD in offspring were expressed as weighted ORs in cohort and case-control studies. Fixed-effect models were used in all analyses since heterogeneities across studies were found to be low-to-moderate when evaluated using I² statistics along with Cochran’s Q test. Random-effects models were used in a sensitivity analysis. In addition, the influence of each study on the overall association was assessed by excluding one study each time. Publication bias was assessed by using Begger’s regression asymmetry test. A two-sided P value ≤0.05 was considered statistically significant. All analyses were conducted with STATA software (Version 16.0, STATA Corporation LP, College Station, TX).

Results

Study characteristics

Nine studies were included in this meta-analysis: 323,253 participants with 4,135 ASD cases in five cohort studies (8–12), and 1,462 cases and 3,265 controls in four case-control studies (13–16). One study was conducted in Europe (10), four in North America (8, 9, 12, 16), and the rest four in Asia (11, 13–15). Other characteristics of the study populations are shown in Table 1 and 2.

Table 1.

Characteristics of the cohort studies

Author (year)	Region	Cases/participants	Offspring age	Study period	GWG assessment	ASD confirmation	GWG guidelines	Risk estimates compared with recommended GWG (referent)		Adjusted covariates
								Excessive GWG	Inadequate GWG
Brumbaugh (2020)	USA	266/7,876	Median, 6.6 yearsa (Interquartile range, 4.3–9.2 years)	1994–2000	Birth certificate data	Research-identified ASD cases based on DSM-IV	Modified 2009 IOM guidelines: reference GWG is 11–41 lbs for singleton gestation; 25–55 lbs for multiple gestation	HR (95% CI): 0.94 (0.63–1.42)	HR (95% CI): 1.50 (0.86–2.63)	History of maternal psychiatric disorder, sex of infant, gestational age, size for gestational age, Apgar at 5 min, maternal age, maternal unmarried status, maternal education level, tobacco use during pregnancy, parity, mode of delivery
Sun (2016)	China	290/3,663	51.5±5.58 Months	2008–2010	Self-report	SAT ascertained by CABS	2009 IOM guidelines	OR (95% CI): 0.95 (0.70–1.28)	OR (95% CI): 1.54 (1.10–2.18)	Maternal education level and vomiting in the first trimester
Gardner (2015)	Sweden	1,947/113,469	4–27 years	1984–2007	Medical records	ASD ascertained by ICD-9, ICD-10, DSM-IV	2009 IOM guidelines	OR (95% CI): 1.12 (1.10–1.25)	OR (95% CI): 1.17 (1.04–1.31)	Maternal BMI, gestational age at birth, gender of child, birth year of the child, maternal age, parity, paternal age, maternal country of birth, socioeconomic status factors, and parental history of psychiatric treatment
Xiang (2015)b	USA	708/68,512	Median, 4.0 years (Interquartile range, 3.2–5.0 years)	1995–2009	Medical records	ASD ascertained by ICD-9	2009 IOM guidelines	HR (95% CI): 0.95 (0.80–1.13)	HR (95% CI): 0.87 (0.71–1.07)	Pre-pregnancy BMI, birth year, maternal age, parity, education, household income, race/ethnicity, history of comorbidity other than diabetes, smoking during pregnancy, preeclampsia/eclampsia, diabetes status, sex of child
Dodds (2011)c	Canada	924/129,733	1–17 years	1990–2002	Medical records	ASD ascertained by ICD-9, ICD-10	Modified 1990 IOM guidelines: referent GWG is 7–18 kg	OR (95% CI): 1.26 (1.08–1.48)	OR (95% CI): 1.16 (0.90–1.49)	None

Abbreviations: 95% CI, 95% confidence interval; ASD, Autism spectrum disorder; BMI, body mass index; CABS, Clancy Autism Behavior Scale; DSM, Diagnostic and Statistical Manual of Mental Disorders; GWG, gestational weight gain; ICD, International Classification of Diseases; IOM, Institute of Medicine; OR, odds ratio; RR, relative risk; SAT, sub-threshold autism trait.

^aMedian age of ASD-R cases are reported.

^bDe nova results offered by Xiang et al.

^cCalculated crude results from data in published paper.

Table 2.

Characteristics of the case-control studies

Author (year)	Region	Cases/controls	Offspring agea, y	GWG assessment	ASD confirmation	GWG guidelines	Risk estimates compared with recommended GWG (referent)		Adjusted covariates
							Excessive GWG	Inadequate GWG
Qiu (2018)	China	36/72	1.5–5	Self-report	ASD ascertained by DSM-IV	2009 IOM guidelines	OR (95% CI): 1.11 (0.36–3.27)	OR (95% CI): 0.73 (0.29–1.80)	Parents with overweight/obesity, maternal age, birth weight, gestational age, and folic acid supplementation
Shen (2018)	China	705/2,236	2–9	Self-report	ASD ascertained by DSM-IV	Modified 2009 IOM guidelines for Chinese population	OR (95% CI): 1.33 (1.02–1.73)	OR (95% CI): 0.86 (0.69–1.06)	Gender of child, age of child, parental age, and family annual income
Windham (2018)	USA	540/776	2–5	Self-report	ASD identified by previous diagnosis; with a SCQ score of 11 or higher; ascertained by clinician during inperson assessment	2009 IOM guidelines	OR (95% CI): 1.29 (1.00–1.66)	OR (95% CI): 0.92 (0.63–1.34)	Maternal age, parity, race/ethnicity, and smoking
Ling (2015)	China	181/181	1–5	Self-report confirmed by hospital data monitor system or prenatal health handbook	ASD ascertained by DSM-IV	2009 IOM guidelines	OR (95% CI): 1.64 (1.21–2.21)	Not reported	Maternal education level, maternal age, number of births, threatened miscarriage, prenatal nutritional status, history of diseases, and drug use.

Abbreviations: 95% CI, 95% confidence interval; ASD, Autism spectrum disorder; BMI, body mass index; CABS, Clancy Autism Behavior Scale; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders; GWG, gestational weight gain; ICD, International Classification of Diseases; IOM, Institute of Medicine OR, odds ratio; SCQ, Social communication questionnaire.

^aRange of age was reported.

Excessive GWG with ASD risk

Excessive GWG was significantly associated with a higher risk of ASD in offspring (Figure 2). Compared with the children whose mothers’ weight gain during pregnancy within the recommended range, the pooled OR of ASD for children whose mothers had excessive GWG was 1.10 (95% CI: 1.02–1.18) in cohort studies and 1.38 (95% CI: 1.19–1.62) in case-control studies. The heterogeneity was moderate (I²=44.5%, p=0.13) in cohort studies and low (I²=0.0%, p=0.62) in case-control studies. Sensitivity analysis indicated that excluding the study of Gardner et al. (10) (pooled OR: 1.08, 95% CI: 0.97–1.19) or the study of Dodds et al. (9) (pooled OR: 1.05, 95% CI: 0.97–1.15) slightly attenuated the overall association in cohort studies. Using the random-effects model, the combined association (pooled OR: 1.08, 95% CI: 0.96–1.21) in cohort studies was also attenuated. The association in case-control studies did not appreciably change by using the random-effects model (pooled OR: 1.38, 95% CI: 1.19–1.62). In addition, evidence on publication bias was not found in either cohort studies (p=0.62) or case-control studies (p=0.50).

Figure 2. Meta-analysis of excessive gestational weight gain in relation to the risk of autism spectrum disorder in offspring.

Abbreviations: GWG, gestational weight gain; IOM, Institute of Medicine; OR, odds ratio; 95% CI, 95% confidence interval.

Inadequate GWG with ASD risk

In comparison with the children whose mothers had recommended GWG, the pooled OR of ASD for children whose mothers had inadequate GWG was 1.13 (95% CI: 1.04–1.24) in cohort studies (Figure 3). Of note, this association was not seen when combining case-control studies (pooled OR: 0.87, 95% CI: 0.72–1.04). The heterogeneity was medium in cohort (I²=62.8%, p=0.03) and low in case-control studies (I²=0.0%, p=0.89). Sensitivity analysis indicated that excluding the study of Gardner et al. (10) attenuated the association (pooled OR: 1.08, 95% CI: 0.94–1.24) in cohort studies. Using the random-effects model, the pooled point estimated in cohort studies (pooled OR: 1.16, 95% CI: 0.97–1.38) was higher, but 95% confidence interval was wider. The pooled OR was not appreciably changed in case-control studies (pooled OR: 0.87, 95% CI: 0.72–1.04) using the random-effects model. Evidence of publication bias was not observed in either cohort (p=0.33) or case-control studies (p=0.60).

Figure 3. Meta-analysis of inadequate gestational weight gain in relation to the risk of autism spectrum disorder in offspring.

Abbreviations: GWG, gestational weight gain; IOM, Institute of Medicine; OR, odds ratio; 95% CI, 95% confidence interval.

Discussion

This meta-analysis indicates a higher risk of ASD in offspring if pregnant women’s GWG are outside the recommended ranges. The association for inadequate GWG is not consistently seen in cohort and case-control studies.

Maternal weight status may influence the neurodevelopment of offspring. A growing number of studies have investigated the relationship between maternal obesity and ASD in offspring (22). However, previous studies mainly focused on maternal BMI at the first antenatal visit as a proxy of pre-pregnancy maternal weight status (23, 24). Recently, GWG was suggested to be a more important risk factor for ASD compared to pre-pregnancy obesity (10, 15, 18). In a large cohort study with 17,860 mothers and 333,057 children, Gardner et al. (10) reported that both pre-pregnancy obesity and GWG were significantly associated with increased risk of ASD in offspring. However, in their matched-siblings analysis, which is less likely to be affected by confounding from genetic susceptibility and living environment, the association with pre-pregnancy obesity was attenuated to non-significance, while the association with GWG remained. Similarly, two case-control studies found a modest association between excessive GWG and ASD risk in offspring, but no association with pre-pregnancy BMI (15, 18). These studies suggest that, compared to pre-pregnancy obesity, GWG may play more critical roles in the development of ASD in offspring or serve as a reliable indicator for this disorder.

Although the underlying mechanisms by which GWG affects neurodevelopment of offspring are still unclear, it has been suggested that excessive GWG may induce disturbed blood leptin signaling in offspring, which may consequently lead to adverse neurobiological conditions (25). Leptin is a proinflammatory cytokine produced by adipose tissue and placenta, and its receptors are widely expressed in brain regions that are related to behavior regulation (26). A proper level of leptin has neurotrophic effects and stimulates neuronal differentiation during fetal neurodevelopment (27). Previously, a positive correlation between GWG and blood leptin levels in pregnant women has been observed (28, 29). Neonates whose mothers gained excessive GWG had significantly elevated leptin levels in their cord blood (30). In addition, a few cohort and case-control studies found that leptin levels were associated with a higher risk of ASD (31–33). Therefore, the observed association between excessive GWG and ASD risk may be partially explained by the dysregulation of leptin signaling during the fetal neurodevelopment period.

How inadequate GWG may affect the neurodevelopment of children is rarely studied. Inadequate GWG may be caused by pregnancy complications such as anorexia nervosa, hyperemesis gravidarum, and malabsorption. It may be considered as a marker of suboptimal nutritional status of the fetus, and nutrient deficiencies have been linked to poor neurodevelopment and ASD risk in children (34, 35). In animal models, maternal protein malnutrition in rats can induce autism-like symptoms in their offspring (36). Although a smaller proportion of pregnant women has inadequate GWG compared to excessive GWG, it has been suggested that women with insufficient GWG benefit more from interventions during pregnancy (37).

Of note, the 2009 IOM guidelines for GWG was designed for the American population (21), and thus may not be suitable for Asian women, similar to the definition of metabolic syndrome and obesity (38, 39). Misclassification of GWG is possible in Asian populations using IOM guidelines, and may lead to biased results towards the null in the primary studies. We performed a stratified analysis by study regions (Supporting Information Figure S1, S2). The association of excessive GWG with ASD was generally consistent when comparing Western to Asian studies. However, the association of inadequate GWG with ASD persisted in Western, but was not seen in Asian studies. A modified recommendation of GWG for Asian populations merits further investigation.

Some limitations need to be acknowledged. First, a strong association between inadequate GWG and the risk of ASD in offspring was not consistently seen when we examined cohort studies and case-control studies separately. The inherent limitations of case-control study design, e.g., potential selection and/or recall bias may be one explanation for this inconsistency. Second, the association found in this meta-analysis was slighted attenuated in our sensitivity analyses using the “leave-one-out” approach or the random-effects models. The attenuated association in “leave-one-out” sensitivity analysis may be due to the fact that some studies have substantially larger sample sizes than others, and excluding them may change the overall results. Because of the low-to-moderate heterogeneities tested across primary studies, we chose to use the random-effects models only in the sensitivity analysis. There is also little evidence that the underlying mechanism between GWG and ASD in offspring vary across different populations. The point estimates of ORs in random-effects models were similar or slightly higher than the fixed-effect models, but the confidence intervals were wider, suggesting a lower statistical power possibly due to the limited number of primary studies. Third, GWG was self-reported in the majority of the primary studies. GWG is better directly measured and collected by trained study personnel. Self-reported GWG is prone to measurement error, though it is commonly used in the previous studies (40). However, one study that examined the correlation between self-reported and clinically measured GWG found that the proportions of women that gained excessive, recommended, and inadequate GWG were essentially identical using two approaches (41). Fourth, the definitions of ASD cases were not completely identical across the primary studies, which might cause misclassification in ASD. Clinically diagnosed ASD was examined in seven studies with different diagnostic assessments including ICD-9, ICD-10, and DSM-IV (9, 10, 12–16). One study investigated research-identified ASD (ASD-R) by the epidemiological approach (8). Another study determined sub-threshold autism traits (SAT), which is the broader phenotype of autism that not yet meets the clinical diagnosis of ASD (11). However, when we excluded these two studies in a sensitivity analysis, the pooled association remained. Fifth, covariates’ adjustments were different across primary studies. Important confounders, such as genetic susceptibility indicated by maternal psychiatric disorders, were considered in only two studies (8, 10). Thus, the possibility of confounding from unmeasured or unknown factors in primary studies cannot be completely ruled out. Finally, the association between GWG and ASD may differ depending on weight status. For example, Shen et al. (15) found that women with overweight or obesity who gain excessive GWG had a higher risk of ASD in offspring than normal-weighted women with excessive GWG. However, most of the primary studies did not report BMI-stratified results. We were not able to explore the potential effect modification by pre-pregnancy weight status due to limited data. Thus, future prospective studies with large sample sizes, using clear ASD diagnostic criteria and reporting the association between GWG and ASD risk in offspring stratified by BMI categories are needed.

Conclusion

This meta-analysis combines the most comprehensive literature and suggests that gaining gestational weight outside the current recommended guidelines is associated with a higher risk of ASD in offspring. Women are generally highly motivated to improve diet and lifestyle during pregnancy for the sake of babies. Thus, promoting optimal GWG could be a cost-effective preventive strategy for neurodevelopment disorders such as ASD.

Study importance questions.

What is already known about this subject?

• More than half of pregnant women gained weight outside the recommended guidelines worldwide.

• Existing studies provided inconsistent evidence for the association between suboptimal gestational weight gain and autism spectrum disorder.

What are the new findings in your manuscript?

• This is an updated meta-analysis regarding the association between gestational weight gain and risk of autism spectrum disorder in offspring.

• Gaining gestational weight outside of the current recommended guidelines, including both excessive and inadequate gestational weight gain, is associated with a higher risk of autism spectrum disorder in offspring.