Prenatal Tobacco and Cannabis Co-exposure and Offspring Obesity Development from Birth to Mid-childhood

Abstract

Background:

Although the association between prenatal tobacco exposure and child obesity risk is well established, less is known about co-exposure to tobacco and cannabis.

Objective:

Determine the relation between prenatal substance co-exposure and obesity risk.

Methods:

In a diverse sample of pregnant women, we examined the association between prenatal substance exposure (tobacco-only and co-exposure) and child BMI (kg/m²) trajectories from birth to mid-childhood (n=262), overweight/obese status based on BMI percentiles from toddlerhood (24 months) to mid-childhood (9–12 years), and adiposity outcomes at mid-childhood (fat mass [kg], fat mass [%], and fat free mass [kg]; n=128). Given that the major goal of this study was to examine associations between prenatal substance exposure and child outcomes, we oversampled pregnant women for substance use (with tobacco as the primary focus).

Results:

Multilevel models demonstrated that children in both exposure groups had a steeper increase in BMI trajectory from birth to mid-childhood and among co-exposed children, girls had a steeper increase than boys. Odds ratio of having obesity by mid-childhood was 12 times higher among those co-exposed than non-exposed. Co-exposure led to significantly greater fat mass and fat mass % compared with no exposure, but exposure to only tobacco was no different than no exposure.

Conclusions:

Results highlight potentiating effects of cannabis exposure in the context of maternal tobacco use in pregnancy on obesity risk and the importance of multi-method assessments of obesity.

INTRODUCTION

Over several decades, the US witnessed a dramatic rise in the prevalence of pediatric overweight and obesity as one third of its citizens between ages 2–19 now have unhealthy body weights.¹ Consuming processed foods and having little physical activity, a maternal pre-pregnancy BMI ≥ 25 kg/m², and/or excessive gestational weight gain consistently predicted childhood obesity,^2,3 though a compromised fetal environment became an increasing concern. According to the Developmental Origins of Health and Disease hypothesis and behavioral teratology models, environmental risks during sensitive periods may influence developmental outcomes and responses to environmental stimuli at older ages.⁴ In addition, harmful components in substances used during pregnancy are known to cross the placenta and enter a baby’s bloodstream,^5,6 which in turn may restrict intrauterine growth and program metabolic shifts for storing energy.^7,8 These individuals typically have low birthweights,⁸ but then experience catch-up growth and reduced insulin secretion as a result of their bodies’ attempt to compensate for previous deficiencies and adapt to postnatal environments.^9,10

Two of the most commonly used substances during pregnancy are cigarettes and cannabis. One meta-analysis and systematic review examining 236,687 women from several continents noted that an average of 19.1% smoked in pregnancy,¹¹ and it is estimated that one in five women of reproductive age will use tobacco by 2025.¹² Tobacco is often co-used with cannabis, with rates of cannabis use among pregnant tobacco users being as high as 20–30%.^13–15 Co-use of these two substances is associated with greater health risks among co-using adults¹⁶ and greater substance dependence.¹⁷ In the future, rates of prenatal tobacco and cannabis co-use are forecasted to climb rapidly because of mainstream beliefs that cannabis is not as dangerous as other substances and possesses therapeutic benefits (e.g., relieves stress, nausea, vomiting, and back pain occurring during pregnancy).¹⁸ This may be especially problematic given dramatic increases in the amount of Tetrahydrocannabinol (THC; the primary psychoactive compound in cannabis) in cannabis over the past two decades from ~2% prior to 1990s to 17% in 2017.¹⁹

When examining the impact of in utero exposure to substances on child obesity risk, there is an association with tobacco, yet little is known about cannabis or co-exposure to both tobacco and cannabis. Several meta-analyses have estimated that maternal smoking during pregnancy is associated with a 23–37% increase in odds of overweight and a 38–55% increase in odd of obesity, among offspring of 3 to 18 years of age. As to cannabis, one relevant investigation is the longitudinal Adolescent Brain and Cognitive Development Study.²⁰ Children ranging from 9–11 years displayed a higher BMI if mothers used cannabis while pregnant relative to matched counterparts,²⁰ though this was determined by retrospective reports. More recently, two prospective studies demonstrated that fetal exposure to cannabis could lead to a higher BMI, fat mass, fat-free mass index, adiposity and fasting glucose in the offspring compared to non-exposed individuals.^21,22 As far as we are aware, little is known about the prospective association between co-exposure to tobacco and cannabis during pregnancy and offspring body weight in early life, which is arguably of greatest concern given that both appear to accelerate weight gain and co-usage is becoming more widespread.

We examined difference in obesity development from birth to mid-childhood (9–12 years of age) among children prenatally exposed to tobacco only, to both tobacco and cannabis and a demographically similar non-exposed group using prospective measurement of substance use among pregnant women recruited in the 1^st trimester of pregnancy. Since the traditional and most frequently employed measurement of BMI does not distinguish between adipose and muscle tissues,²³ outcomes of interest included amounts of fat mass and fat free mass. We hypothesized that children in the tobacco-exposed groups would have steeper BMI trajectories from infancy to mid-childhood and exhibit greater adiposity in mid-childhood. Given the sparse literature on cannabis and none on co-exposure to tobacco and cannabis, we tentatively hypothesized steeper BMI trajectories in the co-exposure compared with tobacco-only and no-exposure groups, and greater adiposity among children in the co-exposure group compared with the other groups. We also hypothesized dose-response associations with heavier use of tobacco associating with higher risk for child obesity variables. Finally, we conducted exploratory analyses with dose-response measures of cannabis use and interactions of the amount of tobacco and cannabis use in pregnancy, and exploratory analyses examining if outcomes varied as a function of child sex. Findings will extend understanding of how prenatal tobacco and cannabis co-exposure affects later offspring health, with a focus on weight trajectories.

METHODS

All pregnant women presenting for prenatal care at an urban hospital located in Western New York completed a screening that included questions about substance use during pregnancy from Year 2007–2008 (see Eiden et al.²⁴ for details). We invited pregnant women to participate if they met initial eligibility and determined final eligibility after childbirth. Exclusion criteria included >20 weeks’ gestation, <18 years of age, expecting multiple births, using illicit substances other than cannabis, or were heavy drinkers after pregnancy recognition (>1 drink/day on average or a binge episode of 4+ drinks on one occasion). Women who only used cannabis in pregnancy were also excluded by study design. Pregnant women who agreed to participate were assessed at the end of each trimester and completed laboratory visits with their child at approximately 2, 9, 16, and 24 months, as well when the child entered kindergarten (~69 months) and during mid-childhood (~127 months). Written informed consent was obtained at the first prenatal appointment, with additional consents at kindergarten and mid-childhood. The University at Buffalo’s Institutional Review Board approved the study. Given that our aim was to examine prenatal tobacco exposure effects on developmental outcomes, women who were current smokers were recruited first by completing a health screener. At the end of each month of recruitment, the closest matching non-smoker (based on age and education) was included, with smokers oversampled to allow for a full range of light to heavy smokers. The final sample included in the current study consisted of 259 mother-child dyads (83 tobacco-exposed, 107 co-exposed to tobacco and cannabis, and 69non-exposed). The sample was diverse with mostly families of color (51% Black, 19% Hispanic, 8% other or mixed-race infants) with relatively low education (59.1% with high school education/equivalent or less) (Table 1). Fifty-seven percent of pregnant women were not working at a paid job at the time of their 1^st trimester appointment, and 47.5% were married or living with a partner. We obtained Federal Certificate of Confidentiality from the National Institutes of Health to protect maternal substance use data even from threat of subpoena.

Table 1.

Group differences in demographics and prenatal substance use (N = 259)

	No-use (n = 69)	Tobacco only use (n = 83)	Tobacco & cannabis co-use (n = 107)	F	p	Partial η2
Demographics
	n (%)	n (%)	n (%)
Maternal Race (white)	10 (14.49%)a	31 (37.35%)b	34 (31.78%)b	5.29**	0.006	0.04
	M (SD)	M (SD)	M (SD)
Education(years)	12.52 (1.89)	12.25 (1.86)	12.22 (1.88)	0.59	0.556	0.00
Age at birth	23.67 (4.99)	24.95 (5.13)	23.80 (4.72)	1.70	0.185	0.01
Parity	1.93 (2.22)	2.55 (2.34)	2.19 (2.46)	1.37	0.257	0.01
Psychological functioning
	M (SD)	M (SD)	M (SD)
Depression	9.81 (6.01)a	13.09 (7.05)b	11.79 (5.61)ab	5.29**	0.006	0.04
Anger/hostility	2.18 (0.54)a	2.51 (0.56)b	2.57 (0.59)b	10.30***	0.000	0.07
Other characteristics
	n (%)	n (%)	n (%)
Breastfeeding status (yes)	45 (65.22%)a	42 (50.60%)a	39 (36.45%)b	7.36***	0.001	0.05
	M (SD)	M (SD)	M (SD)
Prenatal stress	23.16 (7.49)a	26.19 (7.70)b	25.63 (6.77)b	3.59*	0.029	0.03
Mother’s BMI	30.29 (7.64)	32.40 (8.98)	31.12 (8.09)	0.85	0.427	0.01
Prenatal substance use
	M (SD)	M (SD)	M (SD)
Average cigarettes/day	0.00 (0.00)a	4.51 (4.88)b	5.60 (4.62)b	41.99***	0.000	0.25
Average joints/day	0.00 (0.00)a	0.00 (0.00)a	0.57 (0.83)b	35.28***	0.000	0.22
Average standard drinks/day	0.01 (0.04)a	0.05 (0.10)a	0.11 (0.24)b	7.10**	0.001	0.05
Child characteristics
	n (%)	n (%)	n (%)
Premature infants  (<37 weeks)	3 (4.35%)	6 (7.23%)	8 (7.48%)	0.38	0.687	0.00
	M (SD)	M (SD)	M (SD)
Gestational age (week)	39.16 (1.41)	38.89 (1.56)	38.76 (2.22)	1.02	0.363	0.01
Birth weight (kg)	3.36 (0.57)a	3.28 (0.55)ab	3.16 (0.48)b	3.23*	0.041	0.03
BMI at birth	13.03 (1.71)	12.97 (1.27)	12.81 (2.04)	0.38	0.682	0.00
BMI at 2-month-old	16.73 (1.62)	16.74 (1.48)	16.81 (1.70)	0.05	0.948	0.00
BMI at 9-month-old	18.21 (1.68)	18.28 (2.03)	18.81 (2.25)	1.87	0.157	0.02
BMI at 16-month-old	17.71 (1.56)	17.84 (2.10)	18.10 (1.98)	0.72	0.487	0.01
BMI at 24-month-old	17.71 (1.54)	17.77 (1.87)	17.93 (2.25)	0.22	0.800	0.00
BMI at kindergarten	16.76 (2.49)	17.13 (2.38)	17.09 (2.62)	0.39	0.679	0.00
BMI at mid-childhood	20.49 (4.17)	23.14 (4.99)	22.10 (5.46)	2.44	0.091	0.04
Mid-childhood fat mass (kg)	12.04 (6.88)	16.18 (9.18)	14.49 (8.24)	2.18	0.118	0.03
Mid-childhood fat mass (%)	24.78 (6.95)a	30.26 (8.74)b	27.84 (8.19)ab	3.87*	0.023	0.06
Mid-childhood fat free mass (kg)	34.04 (7.70)	34.14 (6.06)	34.13 (8.66)	0.00	0.998	0.00

^***p <.001

^**>p <.01

^*p <.05.

Note.

^{a, b}= the inclusion of the same alphabet indicates non-significant difference between the groups while totally different ones without sharing any common alphabet indicates significant mean difference between the groups by Tukey post-hoc test at p = .05 level.

p-values were round up to three decimal places.

Prenatal Substance Exposure

Multiple methods measured substance use in pregnancy. At the end of each trimester, we administered the Timeline Follow-Back Interview (TLFB),²⁵ a calendar-based method demonstrated as valid and reliable for measurement of substance use.²⁵ Average number of cigarettes, joints, and standard drinks per day across pregnancy and within each trimester were computed based on this interview. In addition to self-reports, oral fluid samples were analyzed by the US Drug Testing Laboratory (Des Plaines, IL) for cotinine, the primary nicotine biomarker, and for Δ9-tetrahydrocannabinol (THC), the primary psychoactive component of cannabis. Cotinine assays were conducted with enzyme-linked irnmunosorbent assay (ELISA) or liquid chroma-tography-tandem mass spectrometry (LC-MS/MS) at 10 ng/mL cutoff and ranged from 0 to 569 ng/mL. Assays for THC were conducted with immunoassay screening (4.0 μg/L cutoff) and GC-MS confirmation (4.0 μg/L cutoff). Infant meconium samples were collected across several days after delivery in one collection bottle, and were assayed with a validated LC-MSMS method^26,27 at 2.5 ng/g nicotine, 1 ng/g cotinine, and 5 ng/g OHCOT, and with a validated 2-dimensional GC-MS analytical method for THC, 11-hydroxy-THC; 8,11-di-hydroxy-THC; 11-nor-9-carboxy-THC (THCCOOH), and cannabinol (Gray et al., 2010b). Limits of quantification for cannabinoid meconium assays were 10 ng/g for all analytes, except 11-hydroxy-THC at 15 ng/g. Women were assigned to the tobacco exposure group if they self-reported smoking during pregnancy on the screener or the TLFB, if oral fluid samples were cotinine positive, or if infant meconium was positive for cotinine, nicotine, or trans-3’ hydroxycotinine (OHCOT). Women were assigned to the co-exposure group if they also self-reported cannabis use during pregnancy, if infant meconium was positive for cannabinoids, or if oral fluid was positive for THC in any of the 3 trimesters. We included pregnant women in the no-exposure group if self-report and biomarkers of substance exposure were negative for both tobacco and cannabis.

Child Anthropometric Measures

Anthropometrics were taken by staff blinded to substance exposure status. We used a scale to measure weight to the nearest 0.001 kg (SECA, Hamburg, Germany). Length was measured in the supine position at 2, 9, and 16 months using an infantometer (SECA, Hamburg, Germany). Height was measured in the standing position at 24 months and kindergarten using a stadiometer measured to the nearest 0.01 cm (SECA, Hamburg, Germany). At the mid-childhood assessment, a Tanita MC-780U multi-frequency, segmental body composition analyzer measured weight and fat and fat free mass. This method was validated in children with various levels of adiposity.²⁸ We calculated BMI (kg/m²) at all ages. BMI percentiles were calculated beginning at 24 months using the Centers for Disease Control and Prevention BMI calculator for children.

Covariates

Covariates included maternal age (during 1^st trimester), education, race/ethnicity, parity, prenatal stress, psychological functioning during pregnancy (e.g., depression), breastfeeding status, and maternal BMI. Maternal age, education, race/ethnicity, and parity were assessed during the 1^st trimester demographic interview. Breastfeeding status was noted at each appointment through 24 months child age (at approximately 2, 9, 16, and 24 months), and dummy-coded such that 0 indicated no breastfeeding at all and 1 indicated the mother breastfed exclusively or mixed fed (fed breast milk and infant formula). At the kindergarten assessment, maternal weight was measured to the nearest 0.001 kg (SECA, Hamburg, Germany) and maternal height was measured to the nearest 0.01 cm after removing shoes and socks using a calibrated stadiometer (SECA, Hamburg, Germany). Maternal BMI was calculated.

We assessed depression during the 2^nd and 3^rd trimesters of pregnancy using the Beck Depression Inventory II,²⁹ a self-report consisting of 21 statements reflecting depressive mood (e.g., pessimism, sadness, loss of interest). It has good internal consistency (Cronbach’s α=0.88 at the 2^nd trimester and α=0.86 at the 3^rd trimester). Maternal stress during the 2^nd and 3^rd trimesters was assessed using the Perceived Stress Scale,³⁰ global self-report of perceived stress. It includes 14 items and was internally consistent at each time point (Cronbach’s α=0.81 at 2^nd trimester and α=0.83 at 3^rd trimester).

Analytic strategy

To investigate the association between prenatal substance exposure and obesity development, longitudinal models were tested using child BMI measured at seven time points (i.e., birth, 2-months-old, 9-months-old, 16-months-old, 24-months-old, kindergarten age, and mid-childhood). We used a two-level multilevel model analysis, which included the repeated measure of child BMI at level 1 and time invariant predictors and covariates at level 2. Categorical variables indicating group differences based on prenatal substance exposure were key predictors. Specifically, we tested models reflecting different contrasts: (1) substance use (i.e., a combined group of tobacco and cannabis co-exposure and tobacco-only exposure) vs. non-exposed, (2) non-exposed as the referent group comparing tobacco and cannabis co-exposed vs. non-exposed as well as tobacco-only exposed vs. non-exposed, and (3) co-exposed as the referent group comparing tobacco-only vs. co-exposed in addition to non-exposed vs. co-exposed. We also examined continuous measures of average number of cigarettes per day and joints per day during pregnancy and in each trimester to examine dose-response relationships with obesity development.

With regard to outcomes, we also examined risk of having overweight (BMI percentile ≥85th) or obesity (BMI percentile ≥95th) at 24-months, kindergarten, and mid-childhood using logistic regression models. Finally, we employed multivariable linear regression models to examine associations of prenatal substance exposure with continuous adiposity outcomes measured at mid-childhood [fat mass (kg), fat mass (%) and fat free mass (kg)]. We tested two models: Model 1, adjusted for child age at time of each body composition measure and sex; and Model 2 utilized Model 1 plus an additional adjustment for covariates noted above (mothers’ education, age at birth, race/ethnicity, parity, prenatal stress, depression, anger/hostility, breastfeeding status, and maternal BMI). Covariates were included in the model if they correlated with the predictors or outcome variables at p < 0.10 and retained the covariates having a significant association at p < 0.05 with outcome variables in the final models based on the principle of model parsimony. All analyses were performed using STATA version 17.

Missing Data

As with all longitudinal cohorts, there were missed data across time. In addition, data were missing by design at mid-childhood since only a portion of participants were invited to participate at this assessment (see below). All participants (N=259) completed at least one or more assessments from birth to mid-childhood. Among 259 cases in total, 12 were missing at birth (but had prenatal data), 18 at 2 months, 57 at 9 months, 64 at 16 months, 72 at 24 months, 56 at kindergarten follow-up. At mid-childhood wave, we proportionally re-recruited 132 cases based on substance use in pregnancy (see below). There were no differences in demographic characteristics, mothers’ prenatal tobacco and cannabis use, and child BMI, and body fat measures between participants who had any missing data at any time point and the participants with complete data. In terms of missing data, there were no missing prenatal substance exposure variables, child sex, demographics, parity, and breastfeeding status. Missing data on prenatal stress (3.1%) and mothers’ BMI (31.7%) were handled by multiple imputation (20 imputations). At mid-childhood, we planned to invite 160 mothers from the larger cohort to participate, with 40 from the non-exposed group, 40 tobacco-only group, and 80 co-exposed group ranging from light to heavy co-exposure based on a rank ordering of participants. Thus, data missing at mid-childhood were initially planned missing. Given interruptions in data collection due to the pandemic and difficulties locating some participants, the final sample for the body composition data included 128 participants (n=34 for the no-use group, n=32 for the tobacco-only exposure group, and n=62 for the co-exposure group) who completed the data collection at mid-childhood. Of the 128 participants, missing values for mothers’ BMI (16.4%) and prenatal stress (3.1%) were handled by multiple imputation with 20 imputations in STATA 17.

RESULTS

Group differences in demographics, prenatal substance exposure, and child BMI trajectory

Descriptive statistics are presented in Table 1. There were no significant group differences in maternal education, age, parity, and BMI by prenatal substance exposure group. However, maternal race/ethnicity, prenatal stress, depression, anger/hostility, and breastfeeding status differed by substance exposure. There was a greater number of White pregnant women in substance exposed vs. non-exposed group (F=5.29, p <.01). Women in the tobacco-only and co-exposure groups reported higher levels of stress during pregnancy compared with women in the no use group (F=3.59, p <.05). In addition, women in the tobacco-only group during pregnancy reported higher symptoms of depression (F=5.29, p <.01) and women in both substance exposure groups reported greater symptoms of anger/hostility (F=10.30, p <.001) from pregnancy to 24-months child age compared with women in the no-exposure group. Women in the co-exposure group (n=39, 36.45%) or those in the tobacco-only group (n=42, 50.60%) were less likely to breastfeed than women in the no-exposure group (n=45, 65.22%) (F=7.36, p <.001). Associations among variables in the models are shown in Supplemental Table 1. Child BMI since 9-month child age was correlated with child’s fat mass % in mid-childhood and these associations tended to increase over time (r=.20-.33, p <.05 with child BMI in infancy/toddlerhood, r=.59, p <.001 with child BMI in kindergarten, r=.88, p <.001 with child BMI in mid-childhood). Fisher’s r to z transformation indicated that these associations strengthened over time (Fisher’s z=.20, .28, .34, .68, and 1.38 at 9-month, 16-month, 24-month, kindergarten, and mid-childhood, respectively). As would be expected, the average number of cigarettes, joints, and drinks per day during pregnancy were positively correlated with one another (r=.14-.20, p <.05).

Associations between prenatal substance exposure and child BMI trajectory

Results of the final longitudinal multilevel models with different contrasts of the predictor variables (non- and co-exposed as the referent groups) are presented in Tables 2 and 3. The intraclass correlation coefficient in the unconditional model revealed that 6.2% of variance is attributable to between-person variation whereas 93.8% is attributable to within-person variation. After adding child age to the model as a sole predictor, results demonstrated that child BMI increased over time (B=.49, SE=.02, p < .001). Child BMI increased across the early infancy waves of data collection (BMI of 12.9 at birth, 16.8 at 2-month postpartum, and 18.5 at 9-month postpartum), remained fairly stable from toddlerhood to kindergarten age (BMI of 17.0), and increased from kindergarten age to mid-childhood (BMI of 22.0). When a binary variable comparing no-use and tobacco use (combining tobacco-only and co-exposure) group and covariates were included in the model (Table 2), mothers’ prenatal tobacco use was associated with the increase of child BMI over time (B=.17, SE=.05, p <.01). In the second model, the two substance exposure groups were examined separately in comparison with the no-use group (Table 3), children in the tobacco-only (B=.22, SE=.07, p <.01) or co-exposure groups (B=.14, SE=.06, p <.05) were more likely to experience a steeper increase in BMI trajectory from birth to mid-childhood (Figure 1), compared with the no-use group. In the third model, we examined if the two exposure groups were different from each other, using the co-exposure group as the reference group. However, there were no differences between the two exposure groups on child BMI trajectories from birth to mid-childhood. We examined interactions with child sex and substance exposure in the final trajectory model with the substance exposure grouping variables. A significant sex difference in the BMI trajectory was found in the co-use group compared with no-use group (B=.27, SE=.12, p <.05) or tobacco-only use group (B=.31, SE=.12, p <.01) (Table 4). Among children in co-exposed group, females were more likely to experience a steeper increase in BMI trajectory than males (Figure 2). We also tested group differences using the no-use group as a reference group, but there were no sex differences found between the no-use and tobacco-only use groups. In terms of dose-response associations, longitudinal multilevel modeling demonstrated that there were no significant dose-response associations with prenatal tobacco or cannabis use and BMI trajectory. Rather, the number of standard drinks per day during pregnancy was significantly associated with a lower BMI gained over time (B=−.34, SE=.12, p <.01).

Table 2.

Longitudinal multilevel modeling about child BMI from birth to mid-childhood comparing prenatally tobacco-use group and no-use group (N = 259)

Predictor	B (SE)	p	95% CI
Tobacco Use vs. No-use	−.15 (.26)	0.554	−.65, .35
Mothers’ BMI	.00 (.02)	0.948	−.03, .03
Child Sex (1 = Male, 0 = Female)	.18 (.23)	0.418	−.26, .63
Child BMI slope
Child Age	.40 (.05)***	0.000	.30, .50
Tobacco Use x Child Age	.17 (.05)**	0.002	.07, .28
Mothers’ BMI x Child Age	.01 (.003)**	0.001	.01, .02
Child Sex x Child Age	−.07 (.05)	0.131	−.17, .02

^***p <.001

^**p <.01

^*p <.05.

Note. Mothers’ BMI was grand-mean centered.

Table 3.

Longitudinal multilevel modeling predicting child BMI from birth to mid-childhood (N = 259)

Predictor	B (SE)	p	95% CI
Prenatal Substance Exposure
Tobacco Only Use (ref. = No-use)	−.31 (.30)	0.291	−.91, .27
Tobacco and Cannabis Co-use (ref. = No-use)	−.03 (.28)	0.913	−.58, .52
Mothers’ BMI	.00 (.02)	0.906	−.03, .03
Child Sex (1 = Male, 0 = Female)	.22 (.23)	0.337	−.23, .67
Child BMI slope
Child Age	.40 (.05)***	0.000	.30, .51
Tobacco Only Use x Child Age	.22 (.07)**	0.001	.09, .35
Tobacco and Cannabis Co-use x Child Age	.14 (.06)*	0.016	.03, .26
Mothers’ BMI x Child Age	.01 (.003)**	0.001	.004, .02
Child Sex x Child Age	−.08 (.05)	0.089	−.18, .01

^***p <.001

^**p <.01

^*p <.05.

Note. Mothers’ BMI was grand-mean centered.

Figure 1.

Child BMI by Prenatal Substance Exposure Status. Maternal age, years of education, race/ethnicity, parity, prenatal stress, depression, anger/hostility, breastfeeding status, and maternal BMI were covariates in model testing.

Table 4.

Longitudinal multilevel modeling testing sex difference in the association between prenatal substance exposure and child BMI trajectories (N = 259)

Predictor	B (SE)	p	95% CI
Prenatal Substance Exposure
Tobacco Only Use (ref. = Co-use)	−.34 (.41)	0.409	−1.13, .46
No-use (ref. = Co-use)	−.31 (.37)	0.411	−1.04, .43
Child Sex (1 = Male, 0 = Female)	−.03 (.35)	0.943	−.71, .66
Child Sex x Tobacco Only Use vs. Co-use	.12 (.54)	0.820	−.94, 1.19
Child Sex x No-use vs. Co-use	.71 (.55)	0.199	−.37, 1.79
Mothers’ BMI	.00 (.02)	0.904	−.03, .03
Child BMI slope
Child Age	.61 (.05)***	0.000	.52, .71
Tobacco Only Use vs. Co-use x Child Age	−.09 (.09)	0.322	−.26, .08
No-use vs. Co-use x Child Age	−.26 (.08)**	0.001	−.42, −.11
Child Sex x Tobacco Only Use vs. Co-use x Child Age	.31 (.12)**	0.008	.08, .54
Child Sex x No-use vs. Co-use x Child Age	.27 (.12)*	0.024	.04, .50
Child Sex x Child Age	−.24 (.07)**	0.001	−.38, −.10
Mothers’ BMI x Child Age	.01 (.003)***	0.000	.01, .02

^***p <.001

^**p <.01

^*p <.05.

Note. Mothers’ BMI was grand-mean centered.

Figure 2.

Child BMI by Child Sex and Prenatal Substance Exposure Group Status. Maternal age, years of education, race/ethnicity, parity, prenatal stress, depression, anger/hostility, breastfeeding status, and maternal BMI were covariates in model testing.

Association between prenatal substance exposure and childhood overweight and obesity risks

Based on BMI percentile at 24-months-old, kindergarten, and mid-childhood, logistic regression results demonstrate that children in the co-exposed group consistently had a significantly higher risk of having overweight at all time points (Supplemental Table 2). For example, the odds of having overweight at 24-mo for children in the co-exposed group were 3.6 times higher than for the no-use group (p <.05), and these odds increased to 4.7 times (p <.01) and 6.8 times (p <.01) at kindergarten and mid-childhood, respectively. Furthermore, there was a sex difference in the association between mothers’ co-exposed and her child’s risk of having overweight. Among children who were exposed to both tobacco and cannabis, females had 2.9 times higher odds (p <.05) of having overweight at kindergarten, and 2.8 times higher odds (p <.10) at mid-childhood compared with males. When examining the odds of having obesity, children in the co-exposed group had significantly higher odds [OR=11.6, (p <.01)] during mid-childhood compared with children in the no-use group, but the tobacco-only group was not significantly different from the no-use group (Supplemental Table 2).

Association between prenatal substance exposure and child body composition during mid-childhood

When testing associations between prenatal substance exposure and body fat composition during mid-childhood, children in the co-exposed group had greater fat mass (kg) (B=5.51, SE=2.29, p <.05) and fat mass % (B=5.89, SE=2.25, p <.05), compared with children in the no-use group (Supplemental Table 3). However, children in the tobacco-only group were not significantly different from those in the no-use group. There were no associations between trimester specific exposures and child BMI or body composition.

DISCUSSION

In this study, we found that children exposed to tobacco and cannabis or only tobacco experienced a comparable increase in BMI trajectory over time, and this was significantly greater than that of non-exposed children. Co-exposed children also had an OR of 11.6 for having obesity and displayed more fat mass relative to demographically similar controls during mid-childhood. But they were not significantly different from children in the tobacco-only group. Additionally, among co-exposed children, girls had 2.9 times and 2.8 times higher odds of having overweight at kindergarten and mid-childhood compared with boys of the same age, respectively.

Overall, there were few differences between the exposure groups. Results regarding BMI trajectories seem to be driven by effects of tobacco rather than co-exposure as the increase in BMI trajectory for the substance-exposed compared with the non-exposed children is similar to tobacco exposure effects noted previously.^31,32 Tobacco used during pregnancy is associated with a well-established pattern of low birth weight followed by subsequent catch-up growth that leads to childhood obesity later in life. Interestingly, others have demonstrated that prenatal cannabis exposure seems to follow a similar growth pattern in that the offspring typically has low birth weight,^33–35 followed by higher adiposity in childhood.^21,22 In term of prenatal tobacco exposure, mechanisms have been proposed that include nicotine effects on suppressed appetite during pregnancy followed by hyperphagia and rapid weight gain in the postnatal period or the “thrifty phenotype” hypotheses^36,37, as well as nicotine effects on the endocrine system resulting in increased adiposity.^36–40 However, the mechanisms for prenatal cannabis exposure on offspring obesity risk is largely unknown. Research suggests that cannabis use during pregnancy could cause fetal growth restriction and long-term neurological consequences.^41,42 In light of a lack of synergistic effects of tobacco and cannabis co-exposure in our cohort, we can only speculate as to why maternal use of cannabis during pregnancy in the context of maternal tobacco use had a negligible contribution on BMI trajectories. Perhaps unlike tobacco, exposure to cannabis while in the womb – both in frequency and amount – wasn’t enough to exert any detectable harm in the presence of tobacco. Research in larger cohorts of substance using pregnant women are needed to determine whether any adverse effects on child body weight might emerge with higher rates of usage, and/or if such effects are simply masked by effects of tobacco in instances of co-exposure.

Girls prenatally exposed to both cannabis and tobacco displayed a steeper BMI trajectory than male counterparts, though no sex differences were observed for children in the tobacco use and no-use groups. Research thus far did not examine the influence of prenatal co-exposure on offspring obesity risk, but there are data for maternal tobacco or cannabis only usage. For example, in the US Collaborative Perinatal Project (n=34,866) girls born to smoking mothers exhibited a higher chance of having overweight [1.30 (95% CI 1.08;1.56)] than boy counterparts [1.22 (95% CI 1.03;1.46)] at age 7.⁴³ By contrast, assessments of growth from birth to 9–10 years of age in Project Koshu (n=1,619) noted that boys were more likely than girls to have significant increases in BMI at each checkup if they were prenatally exposed to tobacco. However, some of the limitations of this study include small effect sizes, single item measures of maternal smoking, and no report on the use of biomarkers.⁴⁴ In addition, they did not account for demographic differences by design.⁴⁴ It is possible that some of the mixed findings are based on differences in measures and methods of ascertaining smoking and lack of controls for demographic differences. In term of cannabis only usage, Gillies and colleagues found that in rat model, female offspring exposed to Δ9-THC in utero were glucose intolerant, and experienced insulin insensitivity, but not male offspring at 5 months of age.⁴⁵ Altogether, these findings indicate that girls may be more vulnerable to health insults following exposure to both substances, but more studies are needed to examine potential underlying mechanisms.

We did not identify a dose-dependent relationship between prenatal tobacco or co-exposure to tobacco and cannabis and offspring BMI trajectory. This contrasts to previous findings in that higher fetal exposure to tobacco (based on meconium assays) and average number of cigarettes in the 2^nd and 3^rd trimesters were associated with higher conditional weight-for-length at 24 months of child age.⁴⁶ Thus, against this time-specific association, BMI trajectories over time were predicted by exposure alone but not by amount of exposure. One explanation is that the exposure variable (group status) was based on multiple methods of assessment (repeated maternal oral fluid assays across pregnancy, infant meconium for all substances, and maternal self-reports on the TLFB), while the dose response variable was based on maternal self-reports alone. In addition, we did not find any literature on a dose-dependent relationship between prenatal exposure to cannabis and obesity risk, although results from one meta-analysis showed a positive relationship between child overweight and number of cigarettes that mothers smoked during pregnancy when it ranged between 1–15 cigarettes daily (OR 1.03, 95% CI 1.02; 1.03 for each cigarette increase); beyond 15, no addition to obesity risk was observed.⁴⁷ Future studies with multiple measures of prenatal substance exposure with demographically similar non-using exposed families are needed to see if these results may be replicated and extended across a longer developmental period. Perhaps dose-response associations become more apparent with increased child age.

There were many strengths to our study. Controls were matched to our study participants on maternal age and education. Having this group is critical because earlier research often failed to use comparison groups that were similar in demographics, making it challenging to determine the influence of substance exposure vs. potential confounders on obesity risk. Our assessment of prenatal smoking was prospective, unlike most retrospective reports, which is important given that ~23% of pregnant smokers deny smoking during pregnancy and may be erroneously grouped with non-smokers.⁴⁸ Furthermore, we quantified participants’ amounts of fat mass and fat free mass to gain deeper insight into how prenatal tobacco and cannabis exposure modifies adiposity.^23,49 As to limitations, we did not include a comparison group of children born to mothers who only used cannabis during pregnancy, although this is a difficult group to recruit given co-use of cannabis with other substances. We also did not invite the full sample to participate in the mid-childhood assessment (given funding restrictions) and lost participants to follow-up from the planned assessment of 160 participants (80% of 160 were assessed) due to pandemic-related delays and closures at mid-childhood. This may have led to an underestimation of the effect size if such individuals happened to have overweight or obesity and loss of statistical power to examine smaller effects. In addition, we included children born prematurely (< 37 weeks’ gestation) in our cohort. Full-term vs. preterm infants often exhibit differential patterns in growth during the first couple years,^50,51 so keeping them in our analyses may slightly bias our findings. In addition, given that the major goal of this study was to examine associations between prenatal substance exposure and child outcomes, we oversampled pregnant women for substance use (with tobacco as the primary focus). Thus, the results are generalizable to mostly urban pregnant women who were using tobacco at the onset of pregnancy and demographically similar non-substance using pregnant women. Our analytic models were also focused on examining individual difference in change over time across our measurement ages and as a function of substance exposure status. Previous methodological papers have proposed multiple approaches to examining questions related to peak BMI and adiposity rebound⁶⁴, that were outside the scope of the current paper, but would address additional research question related to time varying associations between substance exposure and child obesity risk. Finally, these analyses were restricted to examining prenatal exposure associations with child obesity risks. It is possible that some of the associations may be due to postnatal variables not included in the current analyses, such as continued maternal substance use and children’s dietary intake. Future studies examining postnatal processes may be needed to understand potentially complex associations between modifiable maternal-child lifestyle variables (i.e., nutritional intake and physical activity participation) and substance exposure with children’s obesity risks.

In conclusion, our investigation supports previous findings on tobacco exposure effects on BMI trajectories and expands this literature to include the association between prenatal tobacco and cannabis co-exposure and childhood overweight and obesity, particularly for girls. Future studies that include groups of pregnant women using cannabis alone are needed to examine potential unique effects of cannabis on obesity risk. Results indicate that risk for obesity increases for tobacco-exposed children particularly between kindergarten age and mid-childhood. Future studies with longer follow-up may examine developmental changes in these trajectories and include potential postnatal risk and protective influences. Findings also highlight the importance of using multiple measures for assessing use of both substances during pregnancy and their influence on obesity risk. Results indicate stronger effects of tobacco and cannabis co-exposure on fat mass and percent of body fat in mid-childhood. Finally, results highlight increased risk for obesity among co-exposed girls than for boys. Overall, results point to the importance of examining polysubstance exposure risks for obesity, sex differences in these associations, and underscore the importance of targeting polysubstance use in pregnancy cessation efforts.

IMPLICATIONS.

A steeper BMI increase from birth to mid-childhood occurs in substance-exposed children. Prenatal tobacco and cannabis co-exposure produces higher odds of obesity than no exposure. Co-exposed children have higher fat mass in mid-childhood compared with no exposure. Co-exposed girls become more overweight by mid-childhood than co-exposed boys.

ABBREVIATIONS

THC

Tetrahydrocannabinol

TLFB

Timeline Follow-Back Interview