Question Is maternal prenatal cannabis use during early pregnancy associated with child early developmental delays (ie, speech and language disorders, motor delays, global delays)?
Findings In this cohort study of 119 976 mother-child dyads, maternal cannabis use during early pregnancy was not associated with child early developmental delays in children aged 5.5 years or younger.
Meaning These findings suggest that maternal cannabis use in early pregnancy was not associated with an increased risk of child early developmental delays, but additional research on cannabis use throughout pregnancy, mode of administration, and product strength should be conducted.
Importance Maternal prenatal cannabis use is associated with adverse neonatal health effects, yet little is known about its association with child developmental outcomes.
Objective To evaluate associations between maternal prenatal cannabis use in early pregnancy and child early developmental delays.
Design, Setting, and Participants This cohort study included 119 976 children born to 106 240 unique individuals between January 2015 and December 2019 and followed up to aged 5.5 years or younger (through December 31, 2021) at Kaiser Permanente Northern California. Individuals were screened for prenatal cannabis use via self-report and urine toxicology at entrance into prenatal care (approximately 8- to 10-weeks' gestation). Data were analyzed from February 2023 to March 2024.
Exposure Maternal prenatal cannabis use defined as any use (self-reported or by urine toxicology testing) and use frequency.
Main Outcomes Early developmental delays (speech and language disorders, motor delays, global delays) in children up to age 5.5 years defined by International Statistical Classification of Diseases and Related Health Problems, Ninth Revision and Tenth Revision diagnoses codes ascertained from electronic health records.
Results In this cohort of 119 976 pregnancies among 106 240 unique pregnant individuals, there were 29 543 Hispanic pregnancies (24.6%), 6567 non-Hispanic Black pregnancies (5.5%), 46 823 non-Hispanic White pregnancies (39.0%), 12 837 pregnancies (10.7%) to individuals aged 24 years or younger, and 10 365 pregnancies (8.6%) to individuals insured by Medicaid. Maternal prenatal cannabis use was documented for 6778 pregnancies (5.6%). Daily maternal prenatal cannabis use was reported for 618 pregnancies (0.5%), weekly for 722 pregnancies (0.6%), and monthly or less for 1617 pregnancies (1.3%). No association was observed between maternal prenatal cannabis use and child speech and language disorders (HR, 0.93; 95% CI, 0.84-1.03), global developmental delays (HR, 1.04; 95% CI, 0.68-1.59), or motor delays (HR, 0.86; 95% CI, 0.69-1.06). No association was detected between the frequency of maternal prenatal cannabis use and child early developmental delays.
Conclusions and Relevance In this cohort study, maternal prenatal cannabis use was not associated with an increased risk of child early developmental delays. Future research is needed to assess different patterns of cannabis use throughout pregnancy. Given the association between maternal prenatal cannabis use and other adverse outcomes, pregnant individuals should be educated on those risks.
Cannabis use during pregnancy has been increasing in the US. A growing number of states across the US have legalized cannabis and studies have documented an increase in accessibility and acceptance of use. Thus, understanding how exposure during pregnancy affects child development is important.
A growing body of literature has documented associations between maternal prenatal cannabis use and adverse birth outcomes, including preterm birth and low birthweight. Additionally, cannabinoids can cross the placenta and enter the fetal bloodstream potentially disrupting fetal neurodevelopment. However, there has been little research assessing the association of maternal prenatal cannabis use and childhood developmental outcomes, specifically early developmental delays such as speech and language disorders, motor delays, and global delays. Studies evaluating speech and language outcomes report conflicting results with a study reporting worse verbal ability and others documenting no association with language and verbal ability. Similarly, inconsistent findings have been reported for motor skills. One study reported superior motor performance, another found worse processing speed and interhemispheric coordination on fine motor tasks, and others documented no association with motor ability or performance or psychomotor speed or eye-hand coordination. To our knowledge, global delays, such as delayed milestones, have not been evaluated in the context of maternal prenatal cannabis use. Furthermore, nearly all prior studies rely solely on self-reported cannabis use, which may underestimate the true prevalence of use during pregnancy and parent-reported developmental outcomes or the use of screeners rather than clinical diagnoses which may result in misclassification. This study evaluated the association between maternal cannabis use during early pregnancy and delays diagnosed in early childhood (speech and language, motor, and global) in a population universally screened for maternal prenatal cannabis use and early childhood developmental delays.
The study was set within Kaiser Permanente Northern California (KPNC), an integrated health care delivery system, which provides health care to approximately 4.6 million members with insurance coverage by Kaiser Foundation Health Plans, Medicaid, and/or Medicare Advantage. Clinical data are maintained within administrative and electronic health records (EHR). Pregnant patients are universally screened for prenatal substance use by both self-report (via a self-administered questionnaire) and a urine toxicology test to which they provide consent. As part of standard pediatrics care, children are screened for age-appropriate development at each well-baby or well-child visit. The institutional review board at KPNC approved the study and waived consent because data were deidentified. This cohort study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
This retrospective birth cohort study included pregnant individuals with a singleton pregnancy and their children born between January 1, 2015, and December 31, 2019. Eligibility criteria included continuous KPNC health plan membership (ie, gaps of less than 3 months) from 1 year prior to pregnancy onset through delivery date, a minimum of 1 prenatal care visit, and a response to the cannabis use during pregnancy screening question and a prenatal toxicology test for tetrahydrocannabinol (THC) (eFigure 1 in Supplement 1). Pregnant individuals with a prenatal prescription fill for a teratogenic, antineoplastic, or antiepileptic drug (eAppendix 1 in Supplement 1), missing data on parity or address during pregnancy, and infants not enrolled in the health plan within 1 year of birth date or who died within 1 month of birth were excluded. Data for this study were obtained from KPNC's administrative databases and EHR and California State Birth Certificates.
Maternal cannabis use during early pregnancy (primary exposure) was ascertained at entry into prenatal care (mean [SD] 8.2 (3.0) weeks' gestation) and defined as maternal self-reported cannabis use since pregnancy and/or a positive urine toxicology test for cannabis. Frequency of prenatal cannabis use (secondary exposure) was ascertained from the self-report questionnaire and categorized as never, monthly or less, weekly, or daily. For individuals with a positive urine toxicology test for cannabis who self-reported never using cannabis since becoming pregnant, frequency of use was categorized as unknown frequency (eAppendix 1 in Supplement 1).
International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) diagnosis codes and Common Procedure Terminology (CPT) codes were ascertained from the EHR to define 3 early developmental outcomes: speech and language disorders (ICD-10: F80.x, R47.X; CPT: 92507), motor delay (ICD-10: F82), and global developmental delay (ICD-10: F88, F89). Speech and language disorders were defined as at least 2 diagnoses on different dates by age 5.5 years, or 2 or more speech therapy sessions on different dates between age 1 year and 5.5 years. Motor delay and global developmental delays were defined as at least 2 diagnoses on 2 different dates by age 5.5 years.
Maternal sociodemographic characteristics were ascertained from the EHR and birth certificates. Age at pregnancy onset (18 years or younger, 18 to 24 years, 25 to 30 years, 31 to 35 years, 36 years or older), parity (0, 1, more than 1, unknown), and insurance type (Medicaid vs other) were extracted from the EHR. The Neighborhood Deprivation Index (NDI), an indicator of neighborhood-level socioeconomic position, was calculated based on the earliest address in pregnancy. Self-reported race or ethnicity (Asian or Pacific Islander, Hispanic, non-Hispanic Black, non-Hispanic White, and other or unknown [including American Indian, Alaska Native, and multiracial individuals]) and self-reported educational attainment (high school or less, some college or technical school, college graduation, graduate school, unknown) were ascertained from the EHR and supplemented with birth certificate data when available. Race and ethnicity were included as a social construct due to known differences in the prevalence of cannabis use and incidence of child developmental delays by race and ethnicity.
Other noncannabis prenatal substance exposure included individual variables for alcohol, nicotine, opioids, stimulants, and anxiety or sleep medications, defined by maternal self-report of any use since pregnancy at entry to prenatal care, a positive toxicology test at entry to prenatal care, and/or pharmacy dispensation during pregnancy before the first prenatal visit date or prior to pregnancy with supply lasting through pregnancy onset (eAppendix 1 in Supplement 1). Adequacy of prenatal care was assessed using the Kotelchuck Month of Initiation Index and was categorized as inadequate (month 7 or more), intermediate (month 5 to 6), adequate (month 3 to 4), or adequate plus (month 1 to 2).
Maternal comorbidities were defined by ICD-9 and ICD-10 codes (eAppendix 2 Supplement 1) and included diabetes (diagnosed in the 2 years prior to pregnancy onset); asthma, mood or anxiety disorders, other psychiatric disorders, chronic pain, and substance use disorders excluding cannabis-related disorders (diagnosed in the year prior to pregnancy onset through the first prenatal visit date); and nausea or vomiting during pregnancy (diagnosed from pregnancy onset through the first prenatal visit date). Antidepressant medication use was defined as a pharmacy dispensing during pregnancy before the first prenatal visit date or prior to pregnancy with supply lasting through pregnancy onset (eAppendix 1 in Supplement 1).
We fit Cox proportional hazards regression models to examine the association of maternal prenatal cannabis use with each developmental delay. Follow-up time began at birth, with age of the child (in months) as the time scale. Children were followed up until the outcome (first diagnosis or therapy session) or until censored at the end of health plan membership (ie, more than a 3-month gap in enrollment), the time of a missed well-child visit, death, end of the study period (December 31, 2021), or the maximum follow-up age of 66 months (5.5 years).
Attendance at well-child visits (part of standard pediatric care) was required for cohort follow-up to ensure equal opportunity for ascertainment of the outcome. Prior analyses identified differential attendance of well-child visits by maternal prenatal cannabis use. To account for the potential impact of informative censoring, we fit Cox models with inverse probability of censoring weights (IPCW). Time-varying stabilized weights were included for each time interval that the children were followed (eAppendix 3 in Supplement 1 for time intervals). To account for correlation of outcomes among children born to the same mother in separate pregnancies during the study, we used a marginal Cox model with a cluster term at the maternal level and robust standard errors. To determine the set of adjustment covariates, we developed a directed acyclic graph (DAG) with the DAGitty web application, using the best available evidence from the literature and subject matter expertise. The minimally sufficient set of variables identified by the DAG were included in the fully adjusted models (model 5) (eFigure 2 in Supplement 1).
We fit models sequentially to examine the degree of confounding by each set of covariates. All models adjusted for child age by using child age as the time scale of the Cox model. Model 1 did not include any confounders. Model 2 was adjusted for maternal sociodemographic characteristics. Model 3 was additionally adjusted for maternal noncannabis prenatal substance use. Model 4 was additionally adjusted for month of prenatal care initiation. Model 5 (fully adjusted) was additionally adjusted for maternal comorbidities. Analyses were conducted from February 2023 to March 2024 using R version 4.0.2 (R Project for Statistical Computing). Tests were 2-sided, and P < .05 was considered statistically significant.
We conducted 3 additional analyses to assess the sensitivity of our exposure. The sensitivity of our exposure was assessed by defining prenatal cannabis as determined only by self-reported use, and then only by urine toxicology results. The third analysis excluded pregnancies with non-cannabis prenatal substance use. To assess the potential effects of selection bias, we conducted a fourth sensitivity analysis including all pregnancies with any enrollment at KPNC during the pregnancy period (eFigure 5 in Supplement 1).
The cohort included 119 976 pregnancies (among 106 240 unique pregnant individuals). The sample included 32 793 Asian or Pacific Islander pregnancies (27.3%), 29 543 Hispanic pregnancies (24.6%), 6567 non-Hispanic Black pregnancies (5.5%), and 46 823 non-Hispanic White pregnancies (39.0%); 12 837 pregnancies (10.7%) were to individuals aged 24 years or less; 25 568 (21.3%) had at least 2 previous live births, and 10 365 (8.6%) were insured by Medicaid (Table 1). Maternal prenatal cannabis use was identified (via self-report or urine toxicology) for 6778 (5.6%) pregnancies. Of those, 1785 (26.3%) were positive by both self-report and toxicology results, 3821 (56.4%) by toxicology test only, and 1172 (17.3%) by self-report only. A total of 618 pregnancies (0.5%) identified with maternal prenatal cannabis use were to individuals who reported daily use, 722 (0.6%) who reported weekly use, and 1617 (1.3%) who reported monthly or less use. Urine toxicology tests were conducted at mean (SD) of 9.4 (3.6) weeks' gestation; 5606 pregnant individuals (4.7%) tested positive for THC. Overall, 115 363 pregnancies (96.2%) were screened in the first trimester. Maternal sociodemographic and clinical characteristics by prenatal cannabis use are shown in Table 1 and eTable 1 in Supplement 1.
The median (IQR) age of follow-up was 39 (26-46) months (39 [26-47] months for children not exposed to maternal prenatal cannabis use and 29 [15-42] months for those exposed). The cumulative incidence of speech and language disorders among children up to age 5.5 years was 11.1% (95% CI, 10.9%-11.4%), with a first diagnosis or therapy session at mean (SD) age of 25 (9) months. The cumulative incidence of global developmental delay was 0.5% (95% CI, 0.5%-0.6%) among children up to age 5.5 years, with a first diagnosis at mean (SD) age of 23 (11) months. The cumulative incidence of motor delay up to age 5.5 years was 2.2% (95% CI, 2.1%-2.3%), with a first diagnosis at mean (SD) age of 13 (11) months. Cumulative incidences by prenatal cannabis use are presented in Table 2.
Maternal prenatal cannabis use was not associated with speech and language disorders in model 1 (hazard ratio [HR]: 1.01; 95% CI, 0.91-1.10) (Table 2). Results were similar as covariates were added in models 2 to 5 (model 5: HR, 0.93, 95% CI, 0.84-1.03). Self-reported frequency of maternal prenatal cannabis use was not associated with speech and language disorders (Figure). The sensitivity analysis defining prenatal cannabis use with only self-report data produced results similar to the main analysis (eTable 2 in Supplement 1). The sensitivity analysis defining prenatal cannabis use with only urine toxicology test data produced a modest inverse association with speech and language disorders (HR, 0.88; 95% CI, 0.78-0.99) (eTable 3 in Supplement 1). The sensitivity analysis restricted to pregnancies with no noncannabis substance use produced an inverse association with self-reported monthly or less cannabis use and speech and language disorders (HR, 0.68; 95% CI, 0.46-1.00) (eFigure 3 in Supplement 1).
Although a suggestive increased risk of global developmental delay was found with any prenatal cannabis use in model 1 (HR, 1.38; 95% CI, 0.94-2.03), after adjusting for maternal demographic characteristics the association was attenuated and remained not statistically significant (model 2: HR, 1.08; 95% CI, 0.71-1.64). Adjustment for all potential confounders also demonstrated no statistically significant association (model 5: HR, 1.04; 95% CI, 0.68-1.59) (Table 2). There was no association between self-reported frequency of cannabis use and global developmental delay (Figure). Similarly, no significant associations were observed for global developmental delay in sensitivity analyses defining prenatal cannabis use by self-report data only and by urine toxicology test data only (eTables 2 and 3 in Supplement 1) and in the sensitivity analysis restricting to pregnancies with no noncannabis substance use (eFigure 3 in Supplement 1).
Although a suggestive decreased risk of motor delay was found with any maternal prenatal cannabis use in model 1 (HR, 0.84; 95% CI, 0.69-1.00), after adjustment for all potential confounders the association was not statistically significant (model 5: HR, 0.86; 95% CI, 0.69-1.06) (Table 2). There was no association between self-reported frequency of cannabis use and motor delay (Figure). No statistically significant associations were observed for motor delay in the sensitivity analyses (eTables 2 and 3 and eFigure 3 in Supplement 1). The sensitivity analysis with no continuous enrollment requirement in the year prior to or during pregnancy produced results generally similar to the main analyses (eFigures 4 and 5 in Supplement 1).
In this large, population-based longitudinal birth cohort study, maternal cannabis use during early pregnancy was not associated with speech and language disorders, global delay, or motor delay. These associations were relatively stable across multiple sets of adjustment for potential confounders. Sensitivity analyses also largely support findings from the main analyses, except for a suggested modest inverse association with speech and language disorders when defining cannabis use based only on the urine toxicology results. Understanding the associations between maternal prenatal cannabis use and these early developmental outcomes is important because they may be early indicators of other salient developmental outcomes including autism spectrum disorder (ASD), attention-deficit/hyperactivity disorder (ADHD), and intellectual disability.
Our findings are consistent with the majority of previous studies evaluating maternal prenatal cannabis use and various characteristics of speech and language development. The previous research evaluating speech and language outcomes was based on data from the Ottawa Prenatal Prospective Study (OPPS) which ascertained prenatal cannabis use from maternal self-report. The study population was recruited in 1978 representing a time of lower potency cannabis products and the largest analytic sample evaluating speech and language disorders included 272 parent-child dyads. Maternal prenatal cannabis use was not associated with several speech and language outcomes (eg, verbal ability, language ability) at age 12 and 24 months, and 3, 4, or 5 to 6 years. Contrary to our findings, an association between heavy cannabis use and poor verbal ability at 4 years of age was documented in an analysis.
An association between maternal prenatal cannabis use and superior motor performance in children aged 3 years was documented in an OPPS analysis, in contrast with the majority of the research evaluating motor developmental outcomes, including ours, which did not find an association. Four analyses conducted on data from the OPPS and the Maternal Health Practices and Child Development (MHPCD) studies found no association between maternal prenatal cannabis use and various motor skill traits including fine motor skills, psychomotor speed and eye-hand coordination, and motor ability. In contrast, an analysis from the MHPCD study reported worse interhemispheric coordination on fine motor tasks. This study evaluated outcomes at age 16 years, suggesting poorer motor outcomes may not arise until adolescence.
Future research is needed to understand the inverse association with speech and language disorders that emerged with a positive urine toxicology test. Our dose-response analyses of self-report data and speech and language disorders showed null associations, and the estimate for self-reported daily use was in the positive direction, contrasting the urine toxicology findings. Because our routine toxicology screening only documents yes or no use rather than concentration of THC, we were unable to investigate the dose-response association of toxicology results. Our prior study comparing self-report and toxicology results found that toxicology screening detected 65.8% of women who self-reported cannabis use, with greater detection by toxicology for self-reported daily (83.9%) and weekly (77.4%) than monthly or less use (54.1%). Future studies with more detailed data on frequency and duration of use and concentration of THC may be able to determine whether there is any association or whether the association is null.
Maternal prenatal cannabis use has been associated with adverse neonatal outcomes and increased child psychopathology. Furthermore, the underlying neurobiology and animal models suggest potential health risks. Animal models indicate that THC exposure may disrupt normal brain development and function by interfering with the endocannabinoid system resulting in long-lasting neurodevelopmental and behavioral abnormalities. Epidemiologic research has documented changes to brain structure and function with prenatal cannabis exposure. Despite our findings of no increased risks for early developmental delays, past evidence for adverse fetal and neonatal outcomes indicate that American College of Obstetrics and Gynecology and American Academy of Pediatrics recommendations for pregnant individuals and those considering pregnancy to discontinue cannabis use should be followed.
We note several strengths. This study included the largest number of pregnancies exposed to maternal cannabis use, and 1 of the largest cohorts to evaluate the association between maternal prenatal cannabis use and early developmental delays. The population of pregnant individuals was universally screened for cannabis use at entrance to prenatal care via urine toxicology and self-report, and children were routinely screened at multiple time points for developmental delays reducing recall bias and misclassification. Our study used diagnosis codes prospectively recorded in electronic medical records to identify early developmental delays. Diagnoses codes represent a more rigorous assessment of delay than used in previous research, which has relied on developmental screeners and parental report which are less accurate and informative for clinical care. To our knowledge, this analysis is among the first to assess infant global delay in relation to maternal prenatal cannabis use. We used rigorous methods through the development of a DAG to reduce potential bias and residual confounding and included many potential confounders captured through EHR data. We conducted sensitivity analyses to address variance in exposure classification by self-report and urine toxicology and confounding by cooccurrence of other maternal prenatal substance use which all produced similar results. Finally, we highlight the study sample was racially, ethnically, and geographically diverse.
This study has limitations. Maternal prenatal cannabis use was measured at a single time point (ie, entry into prenatal care). While research indicates the prevalence of cannabis use decreases by nearly half over the course of pregnancy (5.3% in the first trimester to 2.5% in the second and third trimester), we highlight this is an avenue for future research. Similarly, evaluation of exposure in the first trimester is relevant given it is widely recognized as a critical period for brain development and may reflect an important period for early developmental delays. We also did not have information on cannabis potency, mode of cannabis consumption (eg, smoking, vaping), or the concentration of THC detected in the positive toxicology screens. Children of individuals who reported cannabis use during pregnancy are less likely to attend well-child visits in early childhood and more likely to end their insurance coverage with KPNC, reducing their opportunity to be diagnosed with an early developmental delay. While we used rigorous statistical techniques, including censoring and applying IPCW, it may not have adequately addressed differential follow-up. If individuals who used cannabis during pregnancy are less likely to identify and/or report delays leading to an underdiagnosis of early developmental delays in their children, it could result in residual confounding and biased estimates. Finally, California legalized cannabis for adult use in 2016 and findings may not be generalizable to patients or outside of the state. Yet, we highlight the US population prevalence of maternal prenatal cannabis use (3% to 7%) is similar to the prevalence documented in the current study.
This study did not find evidence of an association between maternal cannabis use in early pregnancy and child developmental delays up to age 5.5 years. Additional studies are needed to evaluate cannabis use throughout pregnancy, mode of administration and product strength, as well as potential factors that may mitigate adverse associations and neurodevelopmental outcomes that may emerge later in childhood. Given the association with other adverse neonatal outcomes and documented changes to brain structure and function, pregnant individuals and those considering pregnancy should be educated on the risks of prenatal cannabis use.
Corresponding Author: Lyndsay A. Avalos, PhD, MPH, Division of Research, Kaiser Permanente Northern California, 2000 Broadway, Pleasanton, CA 94612 ([email protected]).
Author Contributions: Dr Avalos and Ms Oberman had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Critical review of the manuscript for important intellectual content: All authors.
Conflict of Interest Disclosures: Dr Chambers reported receiving grants from the Gerber Foundation outside the submitted work. No other disclosures were reported.
Funding/Support: This study was supported by grant R01DA048033 and co-funded by the National Institute on Drug Abuse (NIDA) and the Office of the Director of the National Institutes of Health and grants R01DA047405 and K01DA043604 from the NIDA. A portion of the data were obtained through the Kaiser Permanente Northern California Division of Research's Perinatal Research Unit's Perinatal Obstetric Database.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.