1 INTRODUCTION

Adverse maternal and paternal lifestyle habits during pregnancy may have lifelong consequences for cardio-metabolic health in offspring.1 Developmental adaptations in response to adverse exposures may increase the susceptibility of cardiovascular disease and metabolic diseases in later life.1, 2 Despite many public health campaigns, maternal tobacco smoking during pregnancy remains a commonly used and modifiable factor. The adverse effects of maternal tobacco use during pregnancy on foetal development are well known.3-5 In addition, the results from observational studies have suggested associations of foetal tobacco smoke exposure with obesity, cardiovascular disease and type 2 diabetes in adulthood.6-8 Studies in children focused on cardiovascular risk factors showed inconsistent results.9-13 These associations may be influenced by sex and ethnic differences.9, 14 Although less common than maternal tobacco smoking, maternal use of cannabis in pregnancy is increasing in western countries.15, 16 The prevalence of cannabis use in pregnant women is 7% in the United States − of which, 45% co-use tobacco.15, 16 Cannabis use has been associated with reduced foetal growth.17 Animal studies have shown that cannabis metabolites, for example, Δ9-tetrahydrocannabinol, may affect cardiovascular and metabolic development.18, 19 Whether maternal cannabis use in pregnancy also has adverse cardiovascular and metabolic consequences in human offspring is not known. Importantly, observational studies on the associations of maternal tobacco and cannabis use with offspring outcomes may be confounded by family-based social and lifestyle factors.20 Comparison of associations between maternal and paternal substance use may provide insight on direct foetal programming effects or confounding by family-based social and lifestyle factors.20 Stronger associations for maternal exposure with the outcomes would suggest direct foetal programming, whereas similar or stronger associations for paternal exposures with the outcomes may suggest confounding by family-based genetic, social and lifestyle factors.

Therefore, we examined the associations of maternal and paternal tobacco and cannabis use during pregnancy with child body mass index (BMI), body fat, blood pressure, and lipids, glucose and insulin concentrations at 10 years, in a population-based prospective cohort study. We also compared the associations between maternal and paternal exposure to disentangle whether any association is explained by direct foetal programming or confounded by family-based social and lifestyle factors.

2 METHODS 2.1 Study design

This study was embedded in the Generation R Study, a population-based prospective cohort study conducted in Rotterdam, the Netherlands.21 The study was approved by the Medical Ethics Committee of the Erasmus Medical Center, Rotterdam. The inclusion criteria for the pregnant women were (1) to be resident in the study area at their delivery date, (2) to have an expected date of delivery between April 2002 and January 2006, and (3) to give written informed consent.21 The aim was to enrol mothers in pregnancy, but enrolment was possible until the birth of their child. In total, 9778 mothers were enrolled in the study. Of these mothers, 91% (n = 8879) was enrolled in pregnancy, and 71% of all fathers were included.21 Of these, 8116 mothers had information on cannabis or tobacco and had singleton live-born children. Cardio-metabolic follow-up measurements at the age of 10 years were available for 4792 (60%) children (Figure 1).

Flowchart of the study population

2.2 Foetal tobacco and cannabis exposure

As previously described, questionnaires were collected in early pregnancy (median 12.9 weeks of gestation, 25th–75th percentiles 12.1–14.5), mid-pregnancy (median 20.4 weeks of gestation, 25th–75th percentiles 20.4–20.9), and late pregnancy (median 30.2 weeks of gestation, 25th–75th percentiles 29.9–30.8).22 In early pregnancy, mothers were asked whether they smoked during pregnancy. Then, in mid and late pregnancy, mothers were asked whether they smoked in the last 2 months.

Maternal information on cannabis use was collected using a combination of self-reports and urinalysis. In early pregnancy, mothers indicated whether they used cannabis before and/or during pregnancy, and whether they continued using cannabis after becoming aware of their pregnancy.23 Mothers also reported the frequency of use (daily, weekly or monthly). Urine samples were available in a subset of the cohort and were collected in three trimesters, and the first available sample was used for urinalysis of the cannabis metabolite, 11-nor-Δ9-THC-9-COOH, which were analysed using DRI® Cannabinoid Assay (Microgenics) with a cut-off value of 50 μg/L as recommended by the manufacturer and the Substance Abuse and Mental Health Security Agency. Agreement between the self-reported cannabis use and urinalysis was 0.77 (Yule's Y).23

Maternal smoking during pregnancy was categorized into three groups as follows, excluding mother's cannabis users: no, until pregnancy was known (first trimester only) and continued smoking. In addition, a second categorization was performed given that cannabis use is often used in combination with tobacco.16 We combined the information on maternal tobacco and cannabis use and categorized in four non-overlapping groups: no (included women that quit smoking tobacco until pregnancy was known), cannabis before pregnancy, cannabis during pregnancy (in combination with tobacco) and continued tobacco use during pregnancy (without cannabis).

Paternal information on tobacco and cannabis use during pregnancy was assessed by both maternal reports and self-reports during the first trimester of pregnancy, without specifying an exact period. The inter-rater agreement between maternal and self-report was high (Cohen's kappa cannabis use = 0.83, p < 0.001 and Cohen's kappa tobacco use = 0.86, p < 0.001). We used maternal reports because this information was available for more children as fewer fathers completed questionnaires (n = 4453).

Furthermore, the maternal and paternal frequency of smoking was also available and was categorized into three categories (no smoking, less than 5 per day, and more and equal than 5 per day).

2.3 Childhood body fat measurements and cardio-metabolic risk score

Information on child anthropometrics, body composition, and cardio-metabolic health was obtained at the median age of 9.7 years (95% range: 9.4–10.7). We measured the children's height and weight, without shoes and heavy clothing. We calculated BMI as total body weight in kilogram (kg) divided by height squared in meter (m2). BMI standard deviation score (SDS) was adjusted for sex and age according to Dutch reference growth curves (Growth Analyzer 4.0; Dutch Growth Research Foundation, Rotterdam, Netherlands).24 We also created a categorical variable for childhood BMI (underweight, normal weight, overweight and obesity) according to the International Obesity Task Force cut-offs.25 For the analysis, we combined the overweight and obesity groups, hereafter only referred to as the overweight group. Total, android, and gynoid body fat mass were measured using a dual-energy X-ray absorptiometry (DXA) scanner (iDXA; General electrics, Lunar, Madison, Wisconsin, USA). Then, we calculated the android/gynoid fat mass ratio.26 Childhood body fat mass is strongly influenced by the height of the child. We created index variables of body fat measurements independent of height, by using optimal adjustment estimated by log–log regression analysis.27 We calculated fat mass index (FMI) and fat-free mass index (FFMI) (total fat mass was divided by height at ‘4’ exponential, and fat-free mass by height at ‘2’ exponential).27

Systolic blood pressure and diastolic blood pressure were measured at the right brachial artery, four times with an interval of 1 min using the validated automatic sphygmomanometer Datascope Accutorr Plus.28 We calculated mean systolic and diastolic blood pressure values using the last three blood pressure measurements to reduce measurement error. Non-fasting venous blood samples were obtained to measure total cholesterol (mmol/L), high-density lipoprotein cholesterol (HDL) (mmol/L), triglycerides (mmol/L), glucose (mmol/L) concentrations using enzymatic methods (Cobas 8000, Roche, Almere, the Netherlands) and insulin (pmol/L) concentrations using electrochemiluminescence immunoassay on the E411 module (Roche, Almere, the Netherlands). For the clustering of cardio-metabolic risk factors, we defined whether there were any of three or more following components: android fat mass ≥ 75th percentile, systolic or diastolic blood pressure ≥ 75th percentile, triglycerides ≥75th percentile or HDL cholesterol ≤25th percentile and insulin ≥75th percentile.29 We used android fat mass as percentage of total body fat mass as a proxy for waist circumference because waist circumference was not available.

We additionally examined the association with a continuous composite cardio-metabolic score based on using standardized residuals (z-score) (details in Methods S1).30

2.4 Covariates

Potential covariates were selected based on previous literature and presented as a directed acyclic graphic (Figure S1).7, 8, 10, 15, 16, 31-33 Information on parental age, education, ethnicity, pre-pregnancy BMI and alcohol use during pregnancy was obtained from self-report questionnaires. Information on education and ethnicity were categorized according to the classification of Netherlands Statistics.34, 35 Maternal alcohol use was categorized as never drank, drank until pregnancy, and continued drinking during pregnancy. Like paternal smoking and cannabis use, paternal alcohol use was based on maternal report with a high inter-rater (Cohen's kappa alcohol use = 0.80, p < 0.001). Paternal anthropometric measurements were assessed at enrolment. Height and weight were measured without shoes and heavy clothing, and BMI was calculated. Maternal psychopathology score was assessed with the Brief Symptom Inventory (BSI), a validated self-reported measure of 53-items covering a spectrum of psychopathology symptoms.36 Child sex was extracted from medical records.

2.5 Statistical analysis

First, we showed descriptive statistics of the study population and performed non-response analyses by comparing children with and without follow-up measurements at years using chi-squared for categorical and Student's t-test or Mann–Whitney U tests for continuous variables. Second, we used linear regression models to analyse the associations of foetal tobacco and cannabis exposure with offspring body composition and cardio-metabolic outcomes at 10 years, and used logistic regression to analyse the associations with risks of overweight and clustering of cardio-metabolic risk in offspring. We used two models for the analysis. The basic model was adjusted for child sex and age. The confounder model was additionally adjusted for maternal age, education, ethnicity, alcohol use, psychopathology score, and pre-pregnancy BMI. We tested the statistical interaction terms between parental tobacco and cannabis with child sex and with maternal ethnicity to examine potential differential associations.9, 14 In this article, we presented analyses for the full group in the main tables. Also, in the supplementary information, we showed the results for boys and girls separately and for Dutch mothers only. Analyses among the other ethnic subgroups were not possible because of the low numbers of the various ethnic subgroups. In addition, in the paternal tobacco and cannabis use models, we adjusted for paternal variables (age, ethnicity, alcohol use and BMI) instead of maternal variables. Finally, we examined the associations of foetal tobacco and cannabis exposure with the continuous composite cardio-metabolic score in order to capture potential subtle differences cardio-metabolic health. Not normally distributed outcomes measures (android/gynoid fat mass ratio, and insulin and triglycerides concentrations) were log-natural transformed. To enable comparison of effect estimates, we constructed SDS of outcomes. Missing information on the covariates was between 0% and 10.2%, with the exception of maternal pre-pregnancy BMI (14%), psychopathology score (15.9%) and paternal BMI (20.9%). To avoid the bias of complete case analyses, we used multiple imputation to impute missing information of the covariates in 25 datasets, using the mice package.37 We repeated all analyses among complete cases only and observed similar associations (data not shown).

We applied Bonferroni correction to take multiple testing into account, so we divided the α = 0.05 by three categories of outcomes (body composition, blood pressure, metabolic outcomes), setting the statistical significance as two-sided p < 0.017. All statistical analyses were performed using R statistical software, version 3.6.3 (R Foundation for Statistical Computing).

3 RESULTS 3.1 Subject characteristics and non-response analysis

Table 1 shows the study population characteristics. Of all mothers, 24.1% and 2.5% used tobacco and cannabis during pregnancy, respectively. Of all fathers, 42.8% and 9.6% used tobacco and cannabis, respectively. Median child BMI was 17.0 kg/m2 (range 95% 14.0–24.9), with 18.7% being overweight. Tables S1 and S2 show the characteristics according to tobacco and cannabis categorization are provided. Non-response analyses showed that participating mothers were slightly older, more often had European origin, had a higher education, had a lower psychopathology score, and less often used tobacco and cannabis during pregnancy compared to non-participating mothers (Table S3).

TABLE 1. Subject characteristics (N = 4792) Maternal characteristic Age, years, mean (SD) 30.8 (4.9) Ethnicity Dutch (%) 56.9 Non-Dutch Non-Western (%) 31.1 Non-Dutch Western (%) 12.0 Educational level None/Primary (%) 8.0 Secondary (%) 42.9 Higher (%) 49.1 Pre-pregnancy body mass index, kg/m2, median (95% range) 22.6 (18.1–34.9) Psychopathology score, median (95% range) 0.15 (0–1.36) Maternal alcohol use Never drank in pregnancy (%) 43.5 Drank until pregnancy was known (%) 13.9 Continued drinking (%) 42.6 Maternal tobacco use Never smoked in pregnancy (%) 75.9 Smoked until pregnancy was known (%) 8.8 Continued smoking in pregnancy (%) 15.3 Maternal cannabis use No use (%) 94.9 Cannabis before pregnancy (%) 2.6 Cannabis during pregnancy (%) 2.5 Paternal characteristics Age, years, mean (SD) 33.4 (5.8) Ethnicity Dutch (%) 58.1 Non-Dutch Non-Western (%) 32.0 Non-Dutch Western (%) 9.9 Alcohol use, yes (%) 77.9 Tobacco use, yes (%) 42.8 Cannabis use, yes (%) 9.6 Body mass index, kg/m2, median (95% range) 25.1 (19.6–32.9) Child characteristics Female sex, yes (%) 50.5 Age, years, mean (SD) 9.8 (0.3) Weight, kilograms, median (95% range) 34 (25.2–53.8) Height, centimetres, mean (SD) 141.6 (6.7) Body mass index, kg/m2, median (95% range) 17.0 (14.0–24.9) Body composition Total fat mass, kg, median (95% range) 8.5 (4.5–22.2) Android/gynoid fat mass ratio, median (95% range) 0.24 (0.15–0.49) Fat-free mass, kg, median (95% range) 25.3 (19.1,33.9) Blood pressure Systolic, mmHg, mean (SD) 103.2 (7.9) Diastolic, mmHg, mean (SD) 58.6 (6.4) Lipid concentrations Total cholesterol, mmol/L, mean (SD) 4.31 (0.66) HDL cholesterol, mmol/L, mean (SD) 1.48 (0.34) Triglycerides, mmol/L, median (95% range) 0.98 (0.42–2.62) Insulin, pmol/L, median (95% range) 176.7 (35.7–646.4) Glucose, mmol/L, mean (SD) 5.2 (0.9) Overweight, yes (%) 18.7 Cardio-metabolic clustering risk, yes (%) 9.5 Note: Values are presented as means (SD), medians (95% range) or percentages. There were no missing data on these variables as they were imputed using multiple imputation methods. There were no missing data on these variables as they were imputed using multiple imputation methods. Abbreviations: HDL, high-density lipoprotein; SD, standard deviation. 3.2 Parental tobacco and cannabis exposure and childhood body fat outcomes at 10 years

Maternal tobacco smoking in the first trimester only was not associated with childhood BMI nor body composition (Table 2). Compared to non-exposed children, those exposed to maternal continued smoking during pregnancy had a higher android/gynoid fat mass ratio (difference 0.22 SDS, 95% confidence interval [CI] 0.13–0.30), a higher fat mass index (difference 0.20 SDS, 95% CI 0.12–0.28) and a higher risk of overweight (odds ratio [OR] 1.35, 95% CI 1.07–1.71). Dose–response association displayed the highest effect estimates in children whose mothers continued smoking ≥5 cigarettes per day (Table 2). Maternal cannabis use before pregnancy was not associated with childhood body fat outcomes. As compared to non-exposed children, those exposed to maternal cannabis use during pregnancy had a higher BMI (difference 0.26 SDS, 95% CI 0.08–0.44), a higher android/gynoid fat mass ratio (difference 0.21 SDS, 95% CI 0.04–0.39), and a higher fat-free mass index (difference 0.24 SDS, 95% CI 0.06–0.41) (Table 2). Dose–response analyses displayed the highest effect estimates in children whose mothers used daily cannabis (data not shown). No associations with fat mass index were observed (Table 2). We also observed largely similar associations of maternal and paternal tobacco/cannabis use with offspring outcomes (Table 2).

TABLE 2. Associations of maternal and paternal tobacco and cannabis use with childhood body fat outcomes at age 10 years Tobacco only Body mass index (SDS)a (n = 4252) Android-gynoid ratio (SDS)a (n = 4208) Fat mass index (SDS)a (n = 4199) Fat-free mass index (SDS)a (n = 4199) Overweightb (n = 4252) Difference (95% CI) Difference (95% CI) Difference (95% CI) Difference (95% CI) OR (95% CI) Maternal use No Reference Reference Reference Reference Reference First trimester only 0.02 (−0.09, 0.13) 0.11 (0.00, 0.21)* −0.02 (−0.12, 0.08) 0.03 (−0.07, 0.13) 1.16 (0.85, 1.60) Continued 0.16 (0.07, 0.25)** 0.22 (0.13, 0.30)** 0.20 (0.12, 0.28)** 0.07 (−0.01, 0.16) 1.35 (1.07, 1.71)** <5 per day 0.10 (−0.02, 0.23) 0.13 (0.00, 0.25)* 0.11 (0.00, 0.22) 0.04 (−0.08, 0.16) 1.13 (0.81, 1.60) ≥5 per day 0.21 (0.10, 0.33)** 0.29 (0.18, 0.40)** 0.28 (0.18, 0.38)** 0.10 (−0.01, 0.21) 1.54 (1.15, 2.07)** Paternal use No Reference Reference Reference Reference Reference Yes 0.13 (0.06, 0.19)** 0.17 (0.11, 0.23)** 0.15 (0.09, 0.20)** 0.03 (−0.03, 0.09) 1.30 (1.09, 1.55)** <5 per day 0.07 (−0.02, 0.15) 0.08 (0.00, 0.17) 0.06 (−0.02, 0.13) 0.02 (−0.06, 0.10) 1.10 (0.85, 1.44) ≥5 per day 0.16 (0.09, 0.24)** 0.22 (0.15, 0.29)** 0.21 (0.14, 0.27)** 0.03 (−0.03, 0.10) 1.40 (1.15, 1.72)** Cannabis Body mass index (SDS)a (n = 4781) Android-gynoid ratio (SDS)a (n = 4732) Fat mass index (SDS)a (n = 4721) Fat-free mass index (SDS)a (n = 4721) Overweightb (n = 4781) Difference (95% CI) Difference (95% CI) Difference (95% CI) Difference (95% CI) OR (95% CI) Maternal use No Reference Reference Reference Reference Reference Cannabis before 0.08 (−0.10, 0.25) 0.01 (−0.16, 0.18) 0.03 (−0.13, 0.19) 0.15 (−0.02, 0.32) 1.25 (0.75, 2.08) Cannabis during 0.26 (0.08, 0.44)** 0.21 (0.04, 0.39)** 0.14 (−0.02, 0.30) 0.24 (0.06, 0.41)** 1.33 (0.82, 2.16) Continued tobacco only 0.17 (0.08, 0.25)** 0.21 (0.13, 0.30)** 0.21 (0.13, 0.28)** 0.07 (−0.01, 0.15) 1.34 (1.07, 1.69)** Paternal use No Reference Reference Reference Reference Reference Cannabis use 0.12 (0.03, 0.22)** 0.12 (0.02, 0.22)** 0.08 (−0.01, 0.17) 0.17 (0.08, 0.27)** 1.29 (0.99, 1.69) Note: Confounder parental models were adjusted for maternal age, maternal education, maternal ethnicity, maternal alcohol use, maternal psychopathology score, pre-pregnancy body mass index (BMI), child sex and child age. *p-value <0.05, **p-value <0.017 (Bonferroni corrected values for multiple testing). Abbreviation: Android-gynoid ratio (Android/gynoid fat mass ratio). a Values represent regression coefficients (difference) and 95% confidence interval (95% CI) from linear regression models that reflects the differences in childhood outcomes standard deviation score (SDS) for maternal or paternal tobacco and/or cannabis use during pregnancy, compared to the reference group. b Values are odds ratio (OR) and 95% CI from logistic regression models that reflect the risk of childhood overweight and obesity for maternal and paternal tobacco and/or cannabis use during pregnancy, compared to the reference group. 3.3 Parental tobacco and cannabis exposure and childhood cardio-metabolic risk factors at 10 years

First-trimester maternal tobacco was not associated with childhood cardio-metabolic outcomes (Table 3). Compared to children of mothers who did not use tobacco during pregnancy, children exposed to maternal continued tobacco use during pregnancy had higher triglyceride concentrations (difference 0.15 SDS, 95% CI 0.04–0.26). Also, children exposed to continued smoking during pregnancy ≥5 cigarettes per day had a higher systolic blood pressure (difference 0.15 SDS, 95% CI 0.03–0.26) and a higher risk of cardio-metabolic clustering (OR 1.59, 95% CI 1.09, 2.32) (Table 3). We did not observe associations of maternal tobacco use with cholesterol and glucose outcomes (Table 3). The association of maternal continued tobacco use during pregnancy with diastolic blood pressure was explained by family-based social and lifestyle factors (data not shown). No associations of maternal cannabis use with childhood cardio-metabolic outcomes were observed (Table 3). We observed largely similar associations of maternal and paternal tobacco/cannabis use with offspring outcomes (Table 3).

TABLE 3. Associations of maternal and paternal tobacco and cannabis use with childhood cardio-metabolic risk factors at age 10 years Tobacco only Systolic blood pressure (SDS)a (n = 4109) Diastolic blood pressure (SDS)a (n = 4109) Total cholesterol (SDS)a (n = 2959) HDL cholesterol (SDS)a (n = 2960) Triglyceride (SDS)a (n = 2950) Glucose (SDS)a (n = 2958) Cardio-metabolic clusteringb (n = 2862) Difference (95% CI) Difference (95% CI) Difference (95% CI) Difference (95% CI) Difference (95% CI) Difference (95% CI) OR (95% CI) Maternal use No Reference Reference Reference Reference Reference Reference Reference