Tobacco cigarette and cannabis use among new mothers

Abstract

Background and Aims:

In the United States, the prevalence of cannabis use during pregnancy has increased whereas tobacco smoking has decreased. This study aimed to estimate the prevalence of tobacco cigarette smoking and cannabis use among new mothers, stratified by breastfeeding status. Additionally, trend analysis was used to examine changes in tobacco and cannabis use over time.

Design:

Cross‐sectional study based on the National Health and Nutrition Examination Survey (NHANES, 2001–18)

Setting:

United States

Participants:

Population‐based sample of new mothers (within 2 years of childbirth) 20–44 years of age (n = 1332)

Measurements:

For each NHANES data cycle, the prevalence of tobacco smoking measured by self-report or serum cotinine was estimated. Additionally, we estimated the prevalence of cannabis use measured by self-report. Using Joinpoint Regression, we tested for significant changes in trends and produced the average ‘data-cycle’ percent change (APC).

Findings:

From 2001–18, more than one in five new mothers were tobacco cigarette smokers. New mothers who were breastfeeding were less likely to smoke cigarettes (7.4 %; 95 % confidence interval [CI] = 4.9 %, 10.9 %) compared with mothers who were not breastfeeding (25.7 %; 95% CI = 22.3 %, 29.3 %). Results from Joinpoint regression indicated no robust temporal trends for self-reported tobacco smoking (APC = −1.3; 95 % CI = −8.6, 6.7; p = 0.7) or for cotinine levels ≥ 4.47 ng/mL (APC = 0.3; 95 % CI = −5.0, 5.9; p = 0.9), whereas cannabis use had increased among new mothers (APC = 23.7; 95% CI = 4.8, 46.0; p = 0.02).

Conclusions:

Consistent with the increase seen among other subgroups, cannabis use has approximately doubled among United States new mothers since 2005. There is no significant evidence of a change in tobacco smoking among US new mothers since 2001.

INTRODUCTION

The first few months of life is a critical period for human health and development. Certain modifiable maternal behaviors are associated with adverse health outcomes in infants (1). For example, maternal tobacco smoking has been implicated in sudden infant death syndrome and respiratory illnesses (2). Cannabis use prevalence has increased among reproductive age women in the United States (US) and there has been a growing concern about cannabis use among mothers, postnatally (3).

In this study, we present estimates for tobacco and cannabis use among new mothers participating in the US National Health and Nutrition Examination Surveys (NHANES), where tobacco use status can be determined by self-report or cotinine levels. In addition, we present estimates stratified by breastfeeding status. The American College of Obstetricians and Gynecologists and the American Academy of Pediatrics recommend against using cannabis while breastfeeding (4,5). Yet, the prevalence of cannabis use among breastfeeding mothers is largely unknown (6,7). In 2018, Crume et al. used the Pregnancy Risk Assessment Monitoring System and reported a prevalence of 5% among breastfeeding mothers in Colorado, a state where cannabis is legalized for recreational use (8). Here, NHANES provides a unique opportunity to derive these estimates based on a US national sample, rather than a single state sample.

We also examined trends in tobacco and cannabis use among new mothers over time. Previous trend analyses have focused on prenatal or later childhood exposures and reported an increase in cannabis use prevalence whereas tobacco smoking has decreased (9,10). Studying trends of these exposures among new mothers is needed, given that a significant number of women who quit smoking during pregnancy relapse shortly after childbirth (11).

In summary, this study aimed to a) estimate the prevalence of tobacco smoking and cannabis use stratified by breastfeeding status and b) examine trends in tobacco and cannabis use over time among US new mothers. Given the changes in cannabis legalization as well as policies to reduce tobacco smoking (e.g. Excise taxes) in the US over the last decade, Joinpoint regression was used to assess if and when changes in tobacco and cannabis use estimates have occurred (12,13).

METHODS

Sample

Each single NHANES year is designed to be nationally representative of US non-institutionalized civilian population. Data are released in a 2-year cycle due to sample size considerations and to decrease the possibility of disclosing participants’ identities. The protocol for NHANES was reviewed by the National Center for Health Statistics. Michigan State University institutional review board ruled NHANES analysis as “non-human subjects research” as it involves publically available de-identified data.

Details on NHANES methodology are published elsewhere (14). Briefly, NHANES protocol involves an interview at the participant residence where tobacco-use questions are administered. This interview is followed by a separate exam in a NHANES mobile examination center (MEC) where reproductive health and cannabis use questions are administered. Tobacco and reproductive health questions are interviewer-administered using the Computer-Assisted Personal Interview system. Cannabis questions are self-administered using Audio Computer-Assisted Self-Interview system. Blood was collected from participants by a phlebotomist at the MEC. Blood was processed and aliquoted in the MEC and vials were then refrigerated or frozen before transport to remote laboratories for processing.

Variables

Women (20–44 years old) whose age of last live birth is equal to “current age” or “current age-1” were identified as new mothers. Breastfeeding was defined based on an answer of ‘yes’ to the question “Are you now breast feeding a child?”. Cigarette smoking was defined by answer of “every day” or “some days” to the question “Do you now smoke cigarettes?” Additionally, tobacco use was defined with a serum cotinine cutoff of 4.47 ng/mL (15). Cotinine levels were measured in serum by liquid chromatography / tandem mass spectrometry. Serum cotinine is a metabolite of nicotine and is an objective measure of tobacco smoking, yet high levels can also indicate use of other nicotine products or heavy exposure to secondhand smoke. Data on cannabis use were available only for NHANES 2005–18. Cannabis users were defined as participants who used cannabis in the 30 days prior to NHANES MEC.

Analysis

Prevalence estimates for tobacco smoking were generated for each data cycle with adjustments for NHANES design using Stata v15 ‘SVY commands (StataCorp). Weights for NHANES MEC were applied in all analyses. Using Joinpoint Regression Program (version 4.7; National Cancer Institute), we tested for significant changes in trends using a log-linear model and permutation test approach, and produced the average ‘data-cycle’ percent change (APC). For the self-report tobacco smoking analysis (2001–18; 9 data points), a minimum of 0 and a maximum of 2 joinpoints were pre-specified. Models with 1 (p = 0.24) or 2 joinpoints (p = 0.22) were not statistically significant, thus model with zero joinpoint was selected. Similarly, for the cotinine analysis a minimum of 0 and a maximum of 2 joinpoints were pre-specified. Models with 1 (p = 0.45) or 2 joinpoints (p = 0.07) were not statistically significant, thus model with zero joinpoint was selected. For the cannabis use analysis (2005–18; 7 data points), a minimum of 0 and a maximum of 1 joinpoints were pre-specified. A model with 1 joinpoint was not statistically significant (p = 0.24), thus model with zero joinpoint was selected.

To adjust for potential confounding variables, we then used Poisson regression, which accounted for the preponderance of zeros, to regress tobacco use on NHANES data-cycle indicator (continuous variable) adjusting for age, race and education. We re-ran the analysis separately for serum cotinine and cannabis use. Among new mothers, 2 participants had missing information on smoking status. A total of 51 participants had missing information on serum cotinine levels. A total of 27 participants had missing information on cannabis use status. Participants with missing data were not included in the analyses. Hypotheses for this study were not pre-registered.

RESULTS

From 2001–18, 10917 women (20–44 years) were examined in NHANES MEC, among which 1332 (weighted % = 11.2) were identified as new mothers. Aggregation of samples across 2001–18 indicated that compared to non-smokers, new mothers who reported tobacco smoking were younger, and more likely to be non-Hispanic White and with high school education or less (Table 1).

Table 1.

Selected characteristics of the study population by tobacco smoking or cannabis use status. Data for the US National Health and Nutrition Examination Survey

Panel A: Self-reported tobacco smoking, 2001–18
Characteristics	n (weighted %) or mean (SE)		p valuea
	Non-smoker (n = 1096)	Active smoker (n = 236)
Age, years	29.3 (0.3)	27.4 (0.5)	<0.001
Race/ethnicity			<0.001
Non-Hispanic White	361 (53.5 %)	144 (75.8 %)
Non-Hispanic Black	219 (14.7 %)	45 (11.0 %)
All others	516 (31.8 %)	47 (13.2 %)
Education			<0.001
< High school	257 (15.8 %)	66 (23.4%)
High school	226 (19.7 %)	82 (37.1 %)
> High school	613 (64.5 %)	88 (39.5 %)
Panel B: Self-reported cannabis use, 2005–18b
	Non-user (n = 920)	Active user (n = 98)
Age, years	29.3 (0.3)	27.1 (0.9)	0.02
Race/ethnicity
Non-Hispanic White	331 (58.1 %)	39 (57.4 %)	0.003
Non-Hispanic Black	161 (11.7 %)	36 (23.7 %)
All others	428 (30.2 %)	23 (18.9 %)
Education
< High school	207 (15.8 %)	23 (21.7 %)	0.18
High school	208 (23.1 %)	29 (29.5 %)
> High school	505 (61.1 %)	46 (48.8 %)

^aAnalysis of variance for continuous variables and Pearson x2 tests for categorical variables.

^bData on cannabis use were only available for NHANES 2005–18

From 2001–18, 20.9% of new mothers were tobacco cigarette smokers. New mothers who were breastfeeding were less likely to smoke cigarettes (7.4%; 95% confidence interval [CI] = 4.9%, 10.9%) compared to mothers who were not breastfeeding (25.7%; 95% CI = 22.3%, 29.3%; Pearson’s χ test p <0.001). This association remained when serum cotinine cutoff of 4.7 ng/mL was used to define tobacco smoking (8.9%; 95% CI = 6.4%, 12.3% vs. 27.9%; 95% CI = 24.4%, 31.7%; Pearson’s χ test p <0.001).

The 2001–02 prevalence of cigarette smoking among new mothers was 22.9% (95% CI = 15.0%, 33.3%) and in 2017–18, cigarette smoking prevalence was 27.5% (95% CI = 19.1%, 37.9%). Results from Joinpoint regression indicated no significant evidence of a change (APC = −1.3; 95% CI = −8.6, 6.7; p = 0.7, Figure 1A). Similarly, there was no significant evidence of a change in the prevalence of cotinine levels ≥ 4.47 ng/mL over time among new mothers (APC = 0.3; 95% CI = −5.0, 5.9; p = 0.9, Figure 1B).

Figure 1:

Tobacco cigarette smoking and cannabis use among new mothers. Data for the US National Health and Nutrition Examination Survey.

X-axis represents NHANES data cycle and unweighted n of new mothers. Data on cannabis use were not available for NHANES 2001–04.

APC: Average ‘data cycle’ percent change

Aggregation of samples indicated that compared to non-users, new mothers who reported cannabis use were younger and more likely to be non-Hispanic Black (Table 1). From 2005–18, 10.6% of new mothers used cannabis at least once in the 30 days prior to assessment in NHANES MEC. New mothers who were breastfeeding were less likely to use cannabis (5.7%; 95% CI = 3.2%, 10.0%) compared to mothers who were not breastfeeding (12.2%; 95% CI = 9.3%, 15.9%; Pearson’s χ test p = 0.03). The 2005–06 prevalence of cannabis use was 9.0% (95% CI = 3.3%, 22.2%). By 2017–18, estimates rose to 19.5% (95% CI = 13.9%, 26.7%). Results from Joinpoint regression indicated increasing temporal trend (APC = 23.7; 95% CI = 4.8, 46.0; p = 0.02).

Adjustment for age, race and education level in Poisson regression models did not change our conclusion, that cannabis use has increased among new mothers (Prevalence ratio [PR] = 1.20; 95% CI = 1.05; 1.36), whereas there was no evidence of a change in self-reported tobacco smoking (PR = 1.00; 95% CI = 0.95, 1.05) or serum cotinine levels ≥ 4.47 ng/mL (PR = 1.02; 95% CI = 0.97, 1.07). Stratified by breastfeeding status, exploratory trend analyses revealed no evidence of a change over time in self-reported tobacco smoking or cotinine levels, whereas cannabis use has increased among new mothers who were not breastfeeding (PR = 1.17; 95% CI = 1.01, 1.36; p = 0.04, results are not shown in table/figure). However, results from stratified trend analyses need to be interpreted with caution because of the relatively small sample size, particularly for breastfeeding mothers.

DISCUSSION

Consistent with the increase seen among other US subgroups (16), cannabis use has approximately doubled in the recent years among new mothers whereas we did not observe evidence of a change in tobacco smoking. Overall, cannabis use prevalence estimates were lower among breastfeeding mothers, compared to mothers who were not breastfeeding. These results can guide clinician efforts to screen and counsel new mothers on cannabis use.

The increase in cannabis use prevalence among new mothers is alarming. In humans, cannabinoid-1 receptors were detectable in the developing brain starting 14-week gestation (17). Levels of delta-9-tetrahydrocannabinol (THC), the main psychoactive compound in cannabis, were detected in urine of children exposed to cannabis (18). It is possible that cannabis exposure in early life can affect long-term neurodevelopment. It is, however, difficult to separate the effects of prenatal cannabis exposure from postnatal exposure especially among breastfeeding mothers. Prior research suggested that most breastfeeding mothers who used cannabis were also cannabis users during pregnancy (19).

Strengths of this study include the nationally representative sample, the standardized assessment and the use of serum cotinine to verify tobacco use status. In addition, new mothers were defined in NHANES based on age of last live birth, compared to other national surveys where parenting status was derived from household roster and can include biological, step-, adopted, or foster children. Limitations include the observational cross-sectional nature of NHANES, the reliance on self-reported cannabis use with no toxicological assays and the relatively small sample size which might affect the precision of NHANES estimates.

Prior evidence suggests that the decision to breastfeed and breastfeeding duration are important factors influencing tobacco smoking relapse (20). This and the timing of NHANES assessment might have impacted the study findings. Although a large number of mothers initiates breastfeeding after birth, the majority discontinue by 6 months postpartum (21). Sensitivity analyses on the subsample of new mothers who had their baby in the 6 months prior to NHANES assessment revealed similar findings and did not change the conclusions of our study.

Another possible limitation is the lack of overlap between breastfeeding and tobacco or cannabis use. However, infrequent or irregular breastfeeding is unlikely. In addition, cotinine can be detected in the milk up to 72 hours and self-reported smoking was assessed as “every day or some days” (22). Also, THC can be detected in breast milk for up to 6 weeks (23). For all these reasons, the lack of overlap is unlikely to be an issue affecting the findings and conclusions of our study.

In summary, with the rise of cannabis use among new mothers in the US, studies on the effects of infant exposure to cannabis through breast milk or secondhand smoke are needed.