We found that pre/early-pregnancy cannabis use increased in our study conducted before and after the legalization of recreational use in Canada. In this study, which followed over 2000 women, we found that those who used cannabis were younger and were more likely to drink alcohol, smoke tobacco, and use prescription pain medication. Reported use was lower in women in the highest income category but did not differ across other income categories. With respect to birth outcomes, we found that women who reported cannabis use in the 3 months prior to learning they were pregnant had infants born at lower average birth weights, and these infants had higher odds of being small for GA.

With respect to factors associated with cannabis use, our findings fall in line with those in comparable studies. For instance, Corsi and colleagues (2019) examined population-based data in Ontario over a similar time period and found that pregnant women who use cannabis are younger than their non-using counterparts and are more likely to drink alcohol and smoke tobacco [14].

The associations found between maternal cannabis use and birth outcomes are also consistent with other comparable studies. Corsi and colleagues (again examining population-based data in Ontario), similarly found that cannabis use reported at some point during obstetrical care is associated with infants born small for GA (as well as pre-term birth, placental abruption, and transfer to neonatal intensive care) [15].

Another Canadian study conducted in British Columbia also reported that women who reported using cannabis during their first prenatal visit had increased odds of having a small for GA infant, spontaneous preterm birth, and intrapartum stillbirth [16].

In terms of biological plausibility, previous literature has suggested that endocannabinoids can cross the placental barrier [17, 18]. Moreover, cannabinoid receptors and their endogenous ligands have been detected in the earliest stages of embryonic development, with the ECS appearing to play essential roles in these early stages for neuronal development and cell survival [6, 19, 20].Taken together, there exists biological plausibility that in-utero cannabis exposure may cause fetal growth abnormalities and influence birth outcomes. In fact, recent evidence has suggested that the ECS may play a role in placentation; altered placental ECS expression has been associated with spontaneous miscarriage [21]. Additional studies have also proposed that cannabis may affect glucose and insulin regulation, which could also influence fetal growth [22].

Our study extends the findings from existing literature through the homogeneity of our study population, which lends the advantage of better handling residual confounding. Even among our relatively healthy, high SES population exhibiting low rates of concurrent substance use, associations with adverse birth outcomes were found. Not only did the vast majority of the study population report high education attainment and household income, but the single centre hospital-based population effectively minimized confounding effects related to differences in quality of care. Notably, though the homogeneity of the population can be a major advantage, it also limits the generalizability of the study findings. For instance, 84.3% of women in our study had attained at least a Bachelor’s degree, compared to 50.4% across the Canadian population [23].

Additionally, all our covariate measures were captured on an individual basis and provided a high level of detail. Compared to similar studies using population-based registries, we did not estimate socioeconomic status, but instead had direct self-reports of income and education. We also captured detailed information on concurrent substance use including alcohol, tobacco, pain medication, and antidepressants. The data collection method likely provided more reliable covariate measures, especially concerning substance use. The aforementioned comparable studies collected these measures through antenatal care providers, wherein a patient may be more influenced by social desirability bias when asked to disclose substance use. Because OBS questionnaires are self-administered and not reported to a health care provider, we may have better minimized the risk of misclassification. The prevalence of pre/early-pregnancy use in our cohort (9.7%) was similar to the prevalence of use (10.0%) in women aged 25 to 44 years in Ontario in the 2012 Canadian Community Health Survey [24].

In terms of illicit substance use, only 50 women (< 5%) in our study reported use in the 3 months before pregnancy, which was not associated with cannabis use (p = 0.63). General smoking rates were also lower among our cannabis-using population compared to other studies (73.2% non-smokers among our cannabis users versus 29.7% among the Corsi et al. provincially representative population) [14]. These low rates of concurrent use helped further isolate the association in question. Our additional sensitivity analyses further strengthened our argument that concurrent tobacco use cannot explain all of the observed associations with cannabis.

Lastly, our study captured an early time point of cannabis use. As mentioned, previous studies have suggested that self-reported cannabis use is highest in the early first trimester, followed by substantial drop off as the pregnancy progresses [5]. Thus, our pre/early-pregnancy measure questioning participants about their cannabis use 3 months before learning their pregnancy status may be capturing a crucial window of high cannabis exposure, which other studies overlooked (of the 1778 women in the birth outcome models, only 10 reported using cannabis during pregnancy, 1510 reported no use, and 258 were missing a measure). It is possible that the associations we observed with the pre/early-pregnancy use period are due to under-reported use during pregnancy; the proportion of women reporting use during pregnancy in our study is lower than that reported in other studies [25].

There are a number of potential limitations that should be considered. Because our cannabis use measure was self-reported, its prevalence may be understated as some women who use cannabis may be misclassified as non-users. Non-disclosure rates may especially affect observations that were collected before recreational cannabis use was legalized, when stigma was higher. This reduction in stigma surrounding use may also contribute to our finding that cannabis use increased over time. However, this misclassification would bias our findings towards the null, suggesting that the true association may actually be larger than the reported results. Urinalysis screenings were not available, although it is important to consider that no gold standard measure exists. Differing rates of excretion, metabolism, cannabinoid potency, and half-life times affect urinalysis accuracy [26]. In fact, El Marroun and colleagues (2011) showed substantial agreement between urinalyses and self-reported measures [26].

Our measure of cannabis use also did not take frequency, amount, or mode of delivery into account. While deeper insight into women’s patterns of use could prove fruitful and help distinguish between light and heavy users, the variability in individual tolerance and cannabinoid concentrations poses significant barriers to the applicability of such measures. Regarding potential differences between different modes of consumption, data from the National Cannabis Study (2019) suggests that compared to men, Canadian women are much more likely to report any other mode of consumption other than smoking compared to men [27]. Deeper insights into the potential differences in risk between these different modes are warranted in light of growing trends in alternative modes of consumption, such as edibles and vaping.