Strengths and limitations of this study

We used biochemical data to define THC use, which decreased the probability of under-reporting of marijuana use during pregnancy.

We controlled for important confounders that have limited previous research on this subject.

We excluded meconium drug screens with substances other than THC, eliminating the effect of polysubstance abuse.

We evaluated tobacco and alcohol use through self-report rather than through biochemical data.

Introduction

Marijuana is frequently used in pregnancy with increasing prevalence of use over the past 10 years.1 In the 2018 National Survey on Drug Use and Health, 4.7% of pregnant women aged 15–44 years and 9.8% of pregnant women aged 18–25 years used marijuana in the previous month.1 Complicating the issue is data suggesting that the self-report of marijuana use may underestimate the actual prevalence.2–4 Both the American Academy of Pediatrics (AAP) and the American College of Obstetrics and Gynecology (ACOG) have policy statements recommending against marijuana use during pregnancy.5 6 In addition, the Centers for Disease Control and Prevention (CDC) and the Surgeon General recommend not using marijuana during pregnancy.7 8 Despite public health campaigns, there remains a large proportion of pregnant women who perceive marijuana use as without risk.9 This discussion is particularly important with studies showing increased use of marijuana in states with legalisation.10

Previous literature examining the effect of marijuana on neonatal outcomes is varied.11 12 A 2016 metanalysis by Gunn et al11 found a decrease in birth weight and higher neonatal intensive care unit (NICU) admission rates in infants exposed to marijuana. One limitation of this metanalysis was many of the studies did not control for or exclude individuals with polysubstance use, including alcohol and tobacco, which limited the ability to examine the independent effect of marijuana.11 In addition, many of the studies relied on the self-report of marijuana rather than on biochemical samples.11 A separate metanalysis by Conner et al12 did control for tobacco and polysubstance drug use. In the unadjusted analysis, marijuana use was associated with lower birth weight and preterm birth.12 However, in the adjusted analysis, when controlling for concomitant tobacco use, marijuana use was not found to be associated with low birth weight or preterm birth.12 One of the limitations of this metanalysis was that 20 of the 31 included studies determined marijuana exposure by self-report alone.12 Meconium drug screens are an objective way to evaluate drug exposure and have traditionally been considered the gold standard for detection.13 Meconium screens are thought to primarily reflect second and third trimester drug exposure and are therefore most useful in assessing drug use in the later portion of pregnancy.13 14

With the background of this varied literature, the objective of this current study was to examine the effect of prenatal marijuana exposure on the following neonatal outcomes: gestational age, preterm birth (<37 weeks), birth weight, low birth weight (defined as <2.5 kg), length, head circumference, Apgar scores and admission to the NICU.

MethodsDesign, setting and participants

This was a retrospective cohort study using an electronic medical record with individual chart review from 1 January 2017 to 20 June 2019 for a complete hospital network in the Pacific Northwest. Recreational use of marijuana was legal during the entire study timeframe. Inclusion criteria included all cases with meconium drug screens recorded. Cases were excluded if any other drug (other than delta-9- tetrahydrocannabinol (THC)) was detected in meconium and in charts with significant missing data. Meconium drug screens evaluated for the presence of the following: methamphetamine, amphetamines, barbiturates, cocaine, opiates, oxycodone, phencyclidine, methadone, propoxyphene and benzodiazepines. Meconium drug screen tests used a homogeneous enzyme immunoassay method for analysis. Initial positive screens were reflexed to mass spectroscopy methodology. Test results were reported as positive if equal to or greater than threshold and negative if below threshold.15 This test was developed, and its analytical performance characteristics were determined by Quest Diagnostics Nichols Institute Chantilly, Virginia, USA.15 Validation was pursuant to the Clinical Laboratory Improvement Amendments regulations and the test is used for clinical purposes.15

Meconium drug screens are routinely obtained on infants within the hospital system based on the following criteria: no prenatal care, less than five prenatal visits, prenatal care initiated at 20 weeks or later, documented or admitted drug use by the mother or spouse within 2 years, mother in drug rehabilitation programme or infant exhibiting drug withdrawal. Alcohol and tobacco use was evaluated from maternal self-report through routine prenatal visit questionnaires. The timing and amount of exposure to alcohol and tobacco was not specifically evaluated. The study received exempt status from the hospital system’s institutional review board.

Outcomes

The primary predictor was prenatal exposure to marijuana as defined by a positive meconium test for THC. Covariates collected included maternal age, race/ethnicity, self-reported alcohol/tobacco use, cervical insufficiency, multiple gestation, maternal diabetes and hypertension. Outcomes included gestational age, preterm birth, NICU admission, low birth weight (defined as less than 2.5 kg), birth weight, length, head circumference and Apgar scores. To examine the bivariate association between rates of preterm birth and NICU admission with prenatal marijuana exposure, we performed χ2 tests. To determine if there was an adjusted difference in birth weight, height and head circumferences between those with versus without prenatal marijuana exposure, we used two-sample t-tests. To control for type I error, we calculated p values using the Benjamini and Hochberg false discovery rate correction. For outcomes with a significant (p<0.05) bivariate association with THC, we conducted multivariable regression analyses to control for important maternal and gestational factors, including tobacco use, alcohol use, diabetes, hypertension, cervical insufficiency and multiple gestation. For the dichotomous outcome of preterm birth, we utilised multivariable logistic regression. For the continuous outcomes of birth weight, length and head circumferences, we used multivariable linear regression.

Patient and public involvement

Patients or the public were not involved in the design, conduct or reporting of this study.

ResultsPopulation characteristics

There were 1804 patients for which a meconium sample was screened, 101 (5.6%) of which were excluded for significant missing data (figure 1). There were a total of 11 617 births in the hospital network during the study period; therefore, close to 15% of all newborns had a meconium drug screen obtained. For the primary analysis, we excluded patients whose sample contained any substances in addition to/other than THC (163, 9.6%) (online supplemental file 1), leading to a final sample size of 1540. THC was detected in 483 (31.3%) of meconium samples. Within this cohort, patients who tested positive for THC were more likely to be Caucasian, use tobacco and less likely to have diabetes (table 1).

Table 1

Patient characteristics, comorbidities and risk factors by THC status, n=1540

Figure 1

Flow chart for study cohort. Inclusion criteria included all cases with meconium drug screens recorded. Cases were excluded if any other drug (other than delta-9- tetrahydrocannabinol (THC)) was detected in meconium and in charts with significant missing data.

In unadjusted analyses, neonates who tested positive for THC had significantly lower birth weight, shorter length and smaller head circumference (p<0.003) (table 2).

Table 2

Unadjusted outcomes by THC status, n=1540

Adjusted analysis

In the adjusted analysis, neonates exposed to THC had significantly lower birth weight, shorter length and smaller head circumference (p<0.001) (table 3).

Table 3

Results from adjusted linear regression analyses, n=1539

Birth weight was on average 0.16 kg lower (95% CI 0.10 to 0.22, p<0.001) in those exposed to THC. Head circumference was on average 0.52 cm lower (95% CI 0.27 to 0.78, p<0.001) in those exposed to THC. Length was on average 0.71 cm lower (95% CI 0.39 to 1.03, p<0.001) in those exposed to THC. As compared with those unexposed to THC, those exposed had 1.9 times the odds (95% CI 1.3 to 2.7, p=0.001) of being defined as low birth weight in the adjusted analysis (table 4).

Table 4

Results from adjusted logistic regression analyses, n=1539

There were no significant association between THC exposure and preterm birth, NICU admission and Apgar scores. We were not able to include race in the multivariable analysis of dichotomous outcomes due to insufficient sample size. In preliminary adjusted logistic regression analyses, the p value for race was >0.60 for all dichotomous outcomes and thus was chosen for removal.

We used robust SEs for the analyses of head circumference and length due to evidence of heteroskedasticity of the residuals. All models were tested for multicollinearity, which was not present. All other regression diagnostics indicated good model fit.

Discussion

This study found that prenatal marijuana exposure was significantly associated with decreased birth weight, length and head circumference. In addition, infants exposed to marijuana were more likely to be defined as low birth weight compared with those unexposed.

Similar to many previous studies, our data showed a decreased birth weight in infants exposed to marijuana.3 11 16–19 On average, the birth weight in infants exposed to marijuana in our cohort was 160 g lower than in those unexposed and exposed infants were more likely to be classified as low birth weight. This finding is similar to previously published work demonstrating a higher incidence of low birth weight infants exposed to marijuana.18 20 21 These findings are particularly relevant in terms of newborn care as it relates to the increased need for blood work and testing.22 Increased newborn blood draws can be associated with breastfeeding disruption, hyperalgesia and parental anxiety, underscoring the importance in ameliorating factors such as THC use that may contribute to lower birth weight.23 24

Our study further demonstrated a decreased birth length in infants exposed to marijuana. Previous studies evaluating this outcome have been contradictory.3 6 19 25–27 In our cohort, infants exposed to marijuana were also more likely to have a decreased birth head circumference. Similarly, previous work evaluating this outcome has been conflicting.3 6 19 25–28 The finding of decreased head circumference in the exposed group is potentially multifactorial. Previous studies have linked maternal alcohol use with decreased head circumference.29 Given that alcohol use was self-reported, and potentially under reported, in our population, one could hypothesise that the decreased head circumference could be partially related to alcohol use.

Previous literature evaluating the effect of marijuana exposure on NICU admission is also inconsistent. Similar to previously reported data, our study did not show an increased risk of NICU admission in infants exposed to marijuana.3 12 20 This is in contrast to research demonstrating an increased risk of NICU admission in infants exposed to marijuana.16 18 21 Our study is also similar to previous literature that did not show an association with marijuana exposure and preterm birth.11 12 16 20 30 In contrast, other studies have shown an association with marijuana exposure and preterm birth.31–33 Lastly, our study did not show a significant difference in the 5 min Apgar scores for THC exposed infants, which is consistent with previously reported studies.3 17 26 27 34 35

Our study used a potentially higher risk initial population due to the inclusion criteria for obtaining a meconium drug screen. However, both the study group (THC positive meconium) and comparison group (THC negative meconium) were derived from this initial population of infants that had a meconium collected. Meconium drug screens that were positive for drugs other than THC were excluded from the analysis. Therefore, the comparison group consisted of infants with completely negative meconium drug screens. The authors intentionally did not derive a comparison group from infants who did not have meconium collected given the concern that this may have introduced significant bias between the study and comparison group.

The aetiology of discrepant findings of marijuana exposure on neonatal outcomes is likely multifaceted. Previous authors have hypothesised that the strong reliance on self-report of marijuana use could bias studies towards the null hypothesis by misclassifying marijuana users as non-users.3 To our knowledge, our study is one of the largest in the USA ever to examine the effects of marijuana on neonates using targeted drug screen results, rather than maternal self-report.11 12 As previously noted, in the metanalysis by Conner et al, 20 of the 31 studies included relied on maternal self-report of marijuana use.3 12 Unlike the Gunn et al11 metanalysis, which included many studies that did not control for tobacco use, our study rigorously controlled for potential confounders such as tobacco use, increasing the ability to evaluate for the independent effect of marijuana on neonatal outcomes. This ability to control for important confounders, large sample size, use of biochemical data to define THC use and the exclusion of polysubstance use may explain some of the differences found in our study compared with previous literature. Our findings underscore the importance in continued adherence to both AAP and ACOG guidelines that recommend counselling women against using marijuana during pregnancy. Our research adds to the growing literature demonstrating potential negative effects of marijuana use during pregnancy and highlights the need for continued national conversations regarding its widespread use.

This study has some limitations. The retrospective cohort design inherently limits the ability to determine causality. There was lack of racial diversity in the cohort, and we were unable to include race in the multivariable analysis, possibly limiting generalisability. We were unable to assess the precise reason for a meconium screen being obtained other the general category of reasons previously enumerated, which may have introduced unmeasured confounders. Both alcohol and tobacco use were self-reported, which may have resulted in the under-reporting of exposure. We may have introduced selection bias by only examining neonates who had meconium drug screens rather than using a cohort with universal testing. However, it could be hypothesised that if we had compared neonates without meconium drug screens, we may have found even greater differences. Future prospective studies could ameliorate this possible bias by studying cohorts that employ universal drug testing. As meconium screens primarily detect second and third trimester drug exposure, we did not evaluate early pregnancy drug use. There was no separation of maternal hypertensive disorders or maternal type of diabetes, and there was no exclusion of anomalous fetuses or those with genetic disorders that may have introduced confounding. We did not exclude mothers taking medications associated with low birth weight or exclude mothers with autoimmune conditions that may have also introduced confounding. Finally, we did not quantify marijuana exposure in our population that would have allowed for more granular interpretation and analysis.

Conclusions

To our knowledge, our study is one of the largest in the USA ever to examine the effects of marijuana on neonates using targeted drug testing results rather than maternal self-report. In our study, prenatal marijuana exposure was significantly associated with decreased birth weight, length, head circumference and risk of being low birth weight after controlling for important confounders. These findings highlight the need for continued education of pregnant women and adherence to both AAP and ACOG guidelines in avoiding marijuana use in pregnancy.

Data availability statement

Data are available on reasonable request.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Not applicable.