Maternal Prenatal Smoking and Autism Spectrum Disorder in Offspring: A California Statewide Cohort and Sibling Study

Abstract

We examined associations between maternal smoking and autism spectrum disorder (ASD) in children in a statewide population-based cohort and sibling-comparison design using California birth records (n = 2,015,104) with information on maternal smoking, demographic factors, and pregnancy (2007–2010). ASD cases (n = 11,722) were identified through California Department of Developmental Services records with diagnoses based on the Diagnostic and Statistical Manual of Mental Disorders–IV-TR. We estimated odds ratios for ASD with and without intellectual disability in the full cohort using logistic regression and in a sibling comparison using conditional logistic regression. In the full cohort, the adjusted odds ratio for ASD and maternal smoking 3 months before/during pregnancy compared with nonsmoking was 1.15 (95% confidence interval (CI): 1.04, 1.26), and it was similar in cases with (odds ratio = 1.12, 95% CI: 0.84, 1.49) and without intellectual disability (odds ratio = 1.15, 95% CI: 1.04, 1.27). Heavy prenatal smoking (≥20 cigarettes/day in any trimester) was related to an odds ratio of 1.55 (95% CI: 1.21, 1.98). In the sibling comparison, the odds ratio for heavy smoking was similarly elevated but the confidence interval was wide. Our findings are consistent with an increased risk for ASD in offspring of mothers who smoked ≥20 cigarettes/day during pregnancy; associations with lighter smoking were weaker.

Abbreviations

ASD

autism spectrum disorder

CI

confidence interval

DDS

Department of Developmental Services

SES

socioeconomic status

Editor’s note: An invited commentary on this article appears on page 738, and the authors’ response appears on page 742.

Autism spectrum disorder (ASD) is a severe and heterogeneous developmental disorder characterized by atypical socialization, restricted interests, and repetitive behavior. Prenatal maternal smoking has been suspected to be a risk factor for ASD but previous studies have resulted in equivocal findings. Several studies found no associations (1), while others suggested small to moderate increases in risk for ASD (2–4) related to maternal smoking. A recent meta-analysis reported a pooled odds ratio of 1.02 (95% confidence interval (CI): 0.93, 1.12); however, there was large heterogeneity in study designs, covariate adjustments, classification of exposure, case definitions, and effect estimate sizes (5). A number of studies reported associations between ASD and prenatal air pollution exposures that contain many combustion-derived toxics also found in cigarette smoke (6–8). Biological mechanisms relating maternal smoking to ASD include mitochondrial dysfunction, epigenetic changes, (9, 10) gene expression (11), and fetal testosterone concentrations (12). Additionally, maternal smoking has been related to impaired fetal brain development (13, 14), adverse behavioral outcomes (15), attention deficit hyperactivity disorder (16), and neuropsychological function related to inhibitory control (17).

Research has established adverse impacts of maternal smoking on fetal brain development, including impairments in cognitive function (18). Several biological mechanisms that are thought to affect brain function and morphology (13, 19), including inflammatory and epigenetic mechanisms (20, 21), might also be involved in the pathology of ASD (22). However, previous studies have not considered whether ASD phenotypes with intellectual disability (which is considered a marker of severity of impairment in ASD) are differentially affected or particularly susceptible to maternal prenatal smoking. There is ongoing debate as to whether associations between maternal smoking and child neurobehavioral outcomes might be confounded by shared familial and genetic factors. Women who smoke in pregnancy might be affected by neuropsychological conditions affecting smoking behavior and might carry genetic traits potentially associated with neuropsychiatric outcomes in offspring, including autism (23–27). Therefore, to examine associations between maternal smoking in pregnancy and child’s diagnosis of ASD, we conducted a California-statewide population-based cohort study of all births, and we embedded a sibling comparison aiming to assess the role of shared familial confounding factors. We consider comorbid intellectual disability diagnosis, which is a marker of ASD severity and developmental impairment.

METHODS

We used a cohort registry linkage design including all births during 2007–2010 in California; data were retrieved from Office of Vital Statistics birth rolls. ASD cases were identified using records maintained at the California Department of Developmental Services (DDS) and collected by 21 regional centers through December 31, 2013. We included only cases with a primary diagnosis of “autistic disorder” based on the Diagnostic and Statistical Manual of Mental Disorders (code 299.00 in DSM IV-TR—the standard diagnostic instrument until December 31, 2013) (28), as reported on the DDS Client Development Evaluation Report. Validation studies have established the reliability and validity of the Client Development Evaluation Report in California (29). Eligibility for DDS services does not depend on citizenship or financial status (i.e., services are available to all children). We linked DDS case records to California birth records using a probabilistic linkage based on child and parental identifiers (e.g., first/last name, birth date, and sex) to estimate the probability/likelihood that 2 records are for the same person, assigning total scores for a linkage as the sum of scores generated from matching individual fields using the National Program of Cancer Registries Link Plus Software (30) (linkage rate: 86.3%). We manually checked pairs with borderline scores; nonlinkage was mainly due to missing information on records. We excluded 455 case and 82,100 noncase records with missing or implausible/nonviable gestational ages (included: 147–322 days) or birth weights (included: 500–7,000 g); after restricting to singleton births, this yielded a final sample of 11,722 cases and 2,003,382 noncases. Among these, siblings (born 2007–2010) were identified using a similar approach based on child and maternal identifiers (mandatory) and paternal when available, to estimate the probability/likelihood that 2 birth records relate to the same mother/parents.

We classified ASD cases based on DDS evaluation records by “intellectual disability” status (recorded as “mental retardation” at the time of study, diagnosed according to Diagnostic and Statistical Manual of Mental Disorders–IV criteria corresponding to International Classification of Diseases, Ninth Revision) applying the last evaluation record for each child. We determined age at diagnosis in the CDER; the later recorded diagnosis date was used for 44 cases with 2 recorded dates.

This research was approved by the University of California, Los Angeles, Office of the Human Research Protection Program and the California Committee for the Protection of Human Subjects, and was exempted from informed consent requirements because there was no contact with human subjects.

Smoking exposure and pregnancy variables

Information on pregnancy characteristics such as gestational age, birth weight, and pregnancy complications, as well as sociodemographic data, were retrieved from birth records. Smoking information was recorded (since 2007) on California birth certificates as number of cigarettes smoked per day 3 months prior to pregnancy and in each trimester.

Statistical analysis

Odds ratios and 95% confidence intervals were estimated for maternal smoking and ASD using logistic regression. Smoking was assessed according to developmental period. All models adjusted for year of birth and sex, and additionally for potential confounders selected based on prior knowledge (31–33), including maternal age, education, race/ethnicity, parity, and pregnancy complications (definitions displayed in Table 1). We conducted sensitivity analyses adjusting for additional covariates, including preterm birth, birthweight, type of insurance (as proxy for socioeconomic status (SES) (34)), maternal birth place (United States/non–United States), and DDS Regional Center catchment area (in a subsample); because none of these variables changed the estimates of interest by >5%, they were not retained in final models. Further we stratified by ASD with and without intellectual disability comorbidity to examine phenotypes by severity. We also conducted analyses restricted to term births and to diagnosis received at or before age 3 years, and not adjusting for pregnancy complications.

Table 1.

Basic Characteristics in Autism Spectrum Disorder Cases and Noncases in the Full Cohort and Case-Sibling Cohort^a Births, California, 2007–2010

	Full Cohort				Sibling Cohort a
Characteristic	Case (n = 11,722)		Noncase (n = 2,003,382)		Case (n = 2,705)		Noncase (n = 2,639)
	No.	%	No.	%	No.	%	No.	%
Sex
Male	8,720	74.4	1,024,163	51.1	1,987	73.5	1,282	48.6
Female	3,002	25.6	979,206	48.9	718	26.5	1,357	51.4
Missing	0	0.0	13	0.0	0	0.0	0	0.0
Birth year
2007	3,066	26.2	523,897	26.2	717	26.5	734	27.8
2008	3,394	29.0	511,530	25.5	753	27.8	533	20.2
2009	3,061	26.1	490,423	24.5	683	25.2	569	21.6
2010	2,201	18.8	477,532	23.8	552	20.4	803	30.4
Maternal age, years
≤18	416	3.5	113,262	5.7	106	3.9	107	4.1
19–25	3,360	28.7	615,843	30.7	943	34.9	873	33.1
26–30	3,161	27.0	543,822	27.1	759	28.1	772	29.3
31–35	2,889	24.6	456,801	22.8	589	21.8	585	22.2
>35	1,896	16.2	273,620	13.7	308	11.4	302	11.4
Missing	0	0.0	34	0.0	0	0.0	0	0.0
Maternal race/ethnicity
Whiteb	3,072	26.2	532,259	26.6	750	27.7	739	28.0
Latina	5,745	49.0	1,043,033	52.1	1,327	49.1	1,277	48.4
African American	784	6.7	104,285	5.2	157	5.8	172	6.5
Asian	1,392	11.9	205,861	10.3	327	12.1	316	12.0
Others	547	4.7	87,221	4.4	110	4.1	106	4.0
Unknown	182	1.6	30,723	1.5	34	1.3	29	1.1
Maternal educationc, years
≤8	714	6.1	169,649	8.5	134	5.0	133	5.0
9–11	1,702	14.5	336,292	16.8	452	16.7	429	16.3
12	3,054	26.1	518,328	25.9	716	26.5	731	27.7
13–15	3,088	26.3	446,226	22.3	690	25.5	656	24.9
≥16	2,788	23.8	467,872	23.4	635	23.5	617	23.4
Missing	376	3.2	65,015	3.2	78	2.9	73	2.8
Pregnancy complicationsd
Yes	6,878	58.7	1,178,691	58.8	1,586	58.6	1,535	58.2
No	4,843	41.3	824,440	41.2	1,119	41.4	1,104	41.8
Missing	1	0.0	251	0.0	0	0.0	0	0.0
Smokinge beforef or during pregnancy
Yes	464	4.0	71,366	3.6	89	3.3	84	3.2
No	11,035	94.1	1,898,355	94.8	2,563	94.8	2,513	95.2
Missing	223	1.9	33,661	1.7	53	2.0	42	1.6
Smokinge during pregnancy
Yes	301	2.6	48,110	2.4	64	2.4	58	2.2
No	11,217	95.7	1,923,831	96.0	2,591	95.8	2,543	96.4
Missing	204	1.7	31,441	1.6	50	1.8	38	1.4
No. of cigarettes/dayg
Prepregnancy	10 (5–20)		10 (4–20)		10 (4–20)		10 (4–20)
First trimester	10 (4–12)		6 (3–10)		8 (3–15)		6 (3–10)
Second trimester	6 (3–10)		5 (3–10)		5 (3–10)		5 (3–10)
Third trimester	5 (3–10)		5 (3–10)		5 (3–10)		5 (3–10)

^a Includes all siblings with at least 1 case among the siblings.

^b Non-Hispanic White.

^c Years of education numbering 12, 13–15, and ≥16 corresponded to high-school graduate, some degrees less than college, and college or more than college.

^d Pregnancy complications included any conditions recorded on birth certificates.

^e Smoking defined as 1 or more cigarettes/day.

^f Prepregnancy defined as 3 months prior to pregnancy.

^g Among smokers, reported as median, and interquartile range.

To examine the role of confounding by unmeasured shared familial factors, we implemented a sibling comparison restricting to the group of cases with siblings (“case families”). That is, from the original full cohort, we selected all ASD cases with a sibling (n = 2,705) and the nonaffected siblings of ASD cases (born 2007–2010; n = 2,639); these subjects were considered to belong to the same families (case families, n = 2,611) (Figure 1). Conditional logistic regression clustering on maternal identifier as the family indicator for sets of siblings was applied. We conducted sensitivity analyses to compare the findings from the case-family only analyses with those using all siblings (Figure 1).

Figure 1.

Flow chart of sample derivation for the maternal smoking and autism spectrum disorder study, among births in California, 2007–2010.

If associations seen in the full cohort persist in the sibling comparison subcohort, this supports the hypothesis that maternal smoking in pregnancy increases offspring’s ASD risk even after controlling for family-specific factors. To the extent that associations are attenuated, at least 2 explanations should be considered; first, the full cohort associations did at least in part result from familial confounding; second, nondifferential measurement error of the exposure (i.e., random misclassification of maternal smoking on the birth record). While the sibling comparison should adjust for shared familial factors by design (35), the approach has several limitations; it relies on within-family discordance, is prone to exposure misclassification and to unmeasured nonfamilial confounding, and usually has less power and generalizability (35, 36). Analyses were conducted using SAS, version 9.4 (SAS Institute Inc., Cary, North Carolina).

RESULTS

Characteristics of the full cohort, and of the case-sibling subcohort, are displayed by case status in Table 1 and by smoking status in Web Table 1 (available at https://doi.org/10.1093/aje/kwaa182). Case mothers were on average older than noncase mothers, more educated, and less frequently identified as Latina/Hispanic compared with mothers of noncases in the full cohort (Table 1). These demographic factors were similar for cases with and without a sibling born during the study period (Web Table 2). Among cases with a sibling, demographic characteristics were similar comparing autism cases with and without intellectual disability comorbidity (Web Table 3). Demographic characteristics of the “all siblings” noncases and the full cohort or the case-sibling controls were also similar (Web Table 4).

In the full cohort, we estimated an adjusted odds ratio for ASD in relation to smoking 3 months before or during pregnancy (“ever smokers”) of 1.15 (95% CI: 1.04, 1.26) compared with nonsmoking mothers. Odds ratios for smoking only prior to pregnancy, and prior and during pregnancy (“continuous smokers”), were similarly elevated (Table 2). The point estimates for maternal smoking and ASD in the sibling comparison were generally similar but less stable, and the 95% confidence intervals were wide due to the relatively small number of discordant sibling pairs (Table 2).

Table 2.

Maternal Smoking and Autism Spectrum Disorder in the Full Cohort and Case-Sibling-Cohort^a Births, California, 2007–2010

Smoking Status	Full Cohort				Sibling Cohort a
	No. of Cases	No. of Noncases	OR b	95% CI	No. of Cases	No. of Noncases	OR b	95% CI
Never smokers	11,035	1,898,355	1.00	Referent	2,563	2,513	1.00	Referent
Ever smokersc	464	71,366	1.15	1.04, 1.26	89	84	1.03	0.64, 1.68
Prepregnancy smokersd	172	23,952	1.21	1.04, 1.41	26	27	0.75	0.36, 1.58
Continued smokerse	282	45,718	1.11	0.98, 1.26	61	56	1.40	0.71, 2.73
During pregnancyf, no. of cigarettes/day
1–19	233	40,007	1.05	0.92, 1.20	52	50	1.30	0.66, 2.54
≥20	68	8,103	1.55	1.21, 1.98	12	8	1.97	0.53, 7.38
Pregnancy Periods
Prepregnancyg, no. of cigarettes/day
1–19	321	51,863	1.08	0.96, 1.21	60	63	0.97	0.58, 1.64
≥20	133	17,907	1.36	1.14, 1.62	27	20	1.52	0.64, 3.59
First trimester, no. of cigarettes/day
1–19	229	38,849	1.06	0.93, 1.21	52	49	1.29	0.66, 2.54
≥20	65	7,614	1.58	1.23, 2.03	10	8	1.60	0.46, 5.56
Second trimester, no. of cigarettes/day
1–19	173	30,538	1.04	0.89, 1.22	38	41	0.92	0.43, 2.00
≥20	30	3,731	1.46	1.00, 2.13	6	6	1.06	0.24, 4.76
Third trimester, no. of cigarettes/day
1–19	161	28,697	1.04	0.89, 1.22	36	34	1.02	0.44, 2.39
≥20	25	3,066	1.52	1.02, 2.28	7	5	1.21	0.21, 7.00

Abbreviations: CI, confidence interval; OR, odds ratio.

^a Includes all siblings with at least 1 case among the siblings.

^b Adjusted for birth year, sex, maternal age, maternal race/ethnicity, maternal education, parity, and pregnancy complications.

^c Smoked 3 months before pregnancy or any time during pregnancy.

^d Smoked 3 months before pregnancy but quit smoking during pregnancy.

^e Smoked 3 months before pregnancy and kept smoking during pregnancy.

^f Smoking was defined as maximum number of cigarettes/day during any trimester.

^g Prepregnancy was defined as 3 months prior to pregnancy.

For heavier smokers (≥20 cigarettes/day), we estimated elevated odds ratios in the full cohort and in the sibling comparison. Smoking 20 or more cigarettes/day during any trimester period during pregnancy was associated with greater odds (odds ratio = 1.55, 95% CI: 1.21, 1.98) for developing the disorder in the full cohort. The estimated association related to 1–19 cigarettes/day was weaker (Table 2).

Stratifying by intellectual disability comorbidity for ever-smoking, estimated odds ratios were similar for cases with intellectual disability (odds ratio = 1.12, 95% CI: 0.84, 1.49) and without (odds ratio = 1.15, 95% CI: 1.04, 1.27) in the full cohort (Table 3). In the sibling comparisons, numbers in some cells were too small for models to converge (among “with intellectual disability”) and generally 95% confidence intervals were wide (Table 3).

Table 3.

Maternal Smoking and Autism Spectrum Disorder According to Intellectual Disability Status in the Full Cohort and the Case-Sibling-Cohort^a Births 2007–2010, California.

Smoking Status	With ID				Without ID
	No. of Cases	No. of Noncases	OR b	95% CI	No. of Cases	No. of Noncases	OR b	95% CI
Full cohort
Never smokers	1,144	1,898,355	1.00	Referent	9,891	1,898,355	1.00	Referent
Ever smokersc	51	71,366	1.12	0.84, 1.49	413	71,366	1.15	1.04, 1.27
Prepregnancy smokersd	13	23,952	0.81	0.46, 1.43	159	23,952	1.26	1.07, 1.48
Continued smokerse	37	45,718	1.29	0.92, 1.81	245	45,718	1.09	0.95, 1.24
During pregnancyf, no. of cigarettes/day
1–19	32	40,007	1.27	0.89, 1.82	201	40,007	1.02	0.88, 1.18
≥20	7	8,103	1.42	0.67, 3.01	61	8,103	1.57	1.21, 2.03
Sibling cohorta
Never smokers	259	259	1.00	Referent	2,304	2,256	1.00	Referent
Ever smokersc	13	10	4.62	1.00, 21.44	76	74	0.82	0.50, 1.34
Prepregnancy smokersd	2	2	3.04	0.38, 24.52	24	25	0.87	0.43, 1.77
Continued smokerse	11	7	–g	–g	50	49	0.79	0.43, 1.46
During pregnancyf, no. of cigarettes/day
1–19	7	7	3.87	0.33, 46.10	45	43	0.83	0.45, 1.55
≥20	4	1	–g	– g	8	7	0.89	0.65, 1.53

Abbreviations: CI, confidence interval; ID, intellectual disability; OR, odds ratio.

^a Includes all siblings with at least 1 case among the siblings.

^b Adjusted for birth year, sex, maternal age, maternal race/ethnicity, maternal education, parity, and pregnancy complication.

^c Smoked 3 months before pregnancy or any time during pregnancy.

^d Smoked 3 months before pregnancy but stopped smoking during pregnancy.

^e Smoked 3 months before pregnancy and kept smoking during pregnancy.

^f Smoking was defined as maximum number of cigarettes/day during any trimester.

^g Models did not converge due to distribution of n resulting in small n in some cells.

In sensitivity analyses, restricting to term births or to cases diagnosed by age 3 years, or not adjusting for pregnancy complications, did not change our estimated associations appreciably (data not shown). We also evaluated the entire sample of siblings within the full cohort. Compared with case-family siblings, estimated odds ratios based on all siblings (by design not a “sibling comparison”) were generally similar; all confidence intervals were wide and overlapped (Web Table 5).

DISCUSSION

To our knowledge, this study is the first in the United States to examine maternal smoking during pregnancy in relation to ASD in offspring, considering intellectual disability comorbidity, in a statewide cohort involving all births. Additionally, we used a sibling-comparison design aimed at assessing the potential role of familial confounding. In the full cohort, with over 11,000 cases, we estimated a small to moderate increase in risk for ASD related to maternal prenatal smoking (defined as at least 1 cigarette/day), with more strongly elevated odds ratios related to smoking ≥20 cigarettes/day than for smoking <20 cigarettes/day. Associations were similar for cases with and without intellectual disability. In the sibling comparison, point estimates were quite similar, but confidence intervals were wide due to the small number of discordant siblings. The findings based on our full cohort involving over 2,000,000 births are consistent with the conclusion that heavy maternal smoking increases risk for ASD in offspring. There was little indication of differences in associations with smoking by phenotype (with or without intellectual disability). While the estimates using the sibling-comparison design are generally consistent with findings from the full cohort, the small number of discordant siblings resulted in imprecise estimations, thus limiting our ability to evaluate family-based confounding; if there is a maternal characteristic that determines both pregnancy smoking behavior and offspring autism risk, we cannot rule out this potential confounding.

Earlier studies examining ASD and maternal smoking were inconclusive and have limitations, including heterogeneous disease definitions (including milder cases) and diagnostic changes during long study periods that might have affected results (1, 3, 37, 38). Phenotype severity was rarely assessed, and sibling comparisons have also not been considered. Smoking was usually recorded as “yes/no” and not as number of cigarettes/day and by trimester. Of the earlier studies, most were Scandinavian registry-based studies using cohort designs that examined ASD related to maternal smoking recorded on birth records (1, 37, 38), similar to our study design. Two studies—from Finland and Denmark—reported no associations (37, 38). In Sweden, birth registry data (1984–2003), linked with multisource identification of ASD, suggested associations with smoking similar to our cohort findings, but after adjusting for SES-related variables associations became null in the Swedish study (1). We observed little change in our estimated odds ratios and confidence intervals when we included different indicators of SES measures, after adjustment for birth year and sex, maternal age, and parity. In another Swedish study with a nested case-control design (1987–1994) and infantile autism as the outcome (International Classification of Diseases, Ninth Revision, code 299A), the authors adjusted for pregnancy, maternal, and SES-related variables, and reported an odds ratio of 1.4 (95% CI: 1.1, 1.8) for “daily smoking” (3), similar to our findings. Two recent meta-analyses, including 15 studies, reported a pooled odds ratio for autism related to prenatal maternal smoking of 1.02 (95% CI: 0.93, 1.12) (5, 39). However, heterogeneity between studies was large, and calculating a pooled estimate might not have been appropriate. A systematic review including a subset of 7 studies that controlled for SES also suggested no overall increase in risk (40). In 2 of the studies included, a slightly increased odds ratio for ASD without intellectual disability but no increase for ASD with intellectual disability was reported. However, the diagnosis included milder cases on the autism spectrum (2), thus the case definitions are not directly comparable to ours. Previous research is equivocal with regard to maternal smoking and intellectual disability in offspring, with some studies suggesting impairment following heavy maternal smoking in pregnancy (41, 42) but other studies failing to corroborate this (43). We lack data to examine associations with intellectual disability separately. As a comorbidity with ASD, our findings suggest that both ASD phenotypes, with and without intellectual disability comorbidity, might be affected by maternal smoking.

There is biological evidence to support in utero impacts of maternal smoking on autism phenotypes (44). Interestingly, maternal smoking–related epigenetic changes in offspring were found at the same sites for smoking and more general impairment of child developmental (45), which points to possible epigenetic mechanisms that connect smoking and ASD. Effects of maternal smoking were shown on trophoblast-related processes suggesting dysregulation of nicotine receptors and metalloproteinase activity (46) and on umbilical-cord-tissue gene expression related to immune development (11), mechanisms suggested to be relevant to autism development (47–49). In addition, young adults exposed in utero to maternal smoking showed altered activity in certain brain regions (17), and prenatal smoking exposure was associated with changes of the precentral and superior frontal cortices, which are related to affective problems (50). Further support comes from animal data indicating behavioral changes in offspring (13), decreased myelin gene expression (51), attenuated γ-aminobutyric acid release (52), and decreased brain-derived neurotrophic factor/tyrosine kinase receptor B (53) after in utero exposure to maternal smoking. Overall, there is biological evidence supporting associations between maternal smoking and ASD, and future translational research should help to link epidemiologic and experimental data.

Several limitations need to be considered when interpreting our study findings. Even though we included all births occurring over 4 years statewide in California (over 2,000,000), our data is limited for implementing sibling comparisons to assess the role of family-level confounding. Smoking in pregnancy is now a rather rare behavior among Californian women, and autism is a rare outcome. Also, the sibling design has less statistical power and requires a relatively large number of discordant pairs; thus, sample size limitations did not allow us to assess the role of familial confounding as intended. While it remains important to consider shared familial factors, which might influence both smoking behavior of the mother and autism in offspring (as suggested, e.g., for hyperactivity (23, 54)), the sibling-comparison design has several weaknesses, including the inability to assess nonfamilial confounding. As demonstrated by Frisell et al. (35), within–sibling pair estimates will not be confounded by factors perfectly shared by siblings, but the estimates might be more severely biased due to nonshared confounders than estimates from an unpaired design. In addition, within-sibling exposure-effect estimates might be more attenuated than nonsibling pair–based estimates due to nondifferential exposure misclassification (35). Among discordant pairs there is also possible bias (“carryover”) if the outcome (autism) of the first birth influences exposure (smoking) or outcome of the subsequent birth (55). If smoking in an earlier pregnancy had long-term impacts that are related to the risk of autism in subsequent pregnancies, this would also result in bias (55). However, no such long-term biological mechanisms are known. In our full cohort, we addressed potential nonfamilial confounders by controlling for known and suspected confounders and by conducting a number of sensitivity analyses, which did not appreciably influence the estimated associations between maternal smoking and offspring autism. We cannot, however, rule out family-level confounding.

Further limitations include the self-reported nature of maternal smoking on birth certificates. However, the specificity of self-reported prenatal smoking was shown to be high in validation studies comparing pregnancy self-report with cotinine measurement (56), and reports on birth certificates have been confirmed by data from other sources (57). Moreover, the percentages of smokers on California birth certificates, in the California Maternal and Infant Health Survey (58), and in our own survey data, are similar (34, 59). However, we are still likely misclassifying some smokers who are underreporting smoking behavior in pregnancy, and differential misclassification cannot be ruled out. Women with high SES who are ≥30 years of age when giving birth are more likely to underreport prenatal smoking (60), which would lead to overall underestimation of risk, given that mothers of cases tend to be older than mothers of noncases. We do not have information on intellectual disability unrelated to ASD, but future research addressing this outcome would be informative. It is hard to examine a “most critical exposure period” for any smoking-related associations because those who smoke later in pregnancy also smoked already in the first trimester. Strengths of our study include the registry-based design including all births (over 2,000,000) in California in the study period, which prevents participation bias due to self-selection. Importantly, mothers’ report of smoking was recorded at the time of birth—thus, before a diagnosis of autism—and all cases were diagnosed according to the same criteria (using Diagnostic and Statistical Manual of Mental Disorders–IV).

In conclusion, there is a long, ongoing debate on how to assess causal inference between maternal prenatal smoking and offspring neuropsychiatric health, and triangulation via sibling design approaches as we attempted herein has been suggested (27, 35, 36, 61). A critical limitation of this approach in population-based studies is that very large data sets are required with detailed information about maternal smoking behavior to ensure sufficient numbers of discordant siblings, which is a challenge for rare neurodevelopmental conditions such as autism. Overall our full population cohort findings support the hypothesis that rather heavy maternal smoking is associated with greater odds for ASD and suggests weaker impacts for lighter smoking.