Our systematic review included 7,920,383 women and found that cannabis consumption was associated with increased risks of low birth weight, small for gestational age, major anomalies, decreased head circumference, decreased neonatal weight, decreased birth length, decreased gestational age, NICU admission, perinatal mortality, and preterm delivery; however, it was associated with decreased risk of gestational diabetes. This constitutes the largest meta-analysis on this subject to date, and hopefully will add strong evidence to the argument that cannabis use in pregnancy is associated with poor neonatal outcomes. As stated below, however, many questions still remain unanswered as far as if these findings apply to all methods of ingesting cannabis, and if these results remain relevant when controlling for tobacco smoking, environmental exposures, and alcohol use in pregnancy. As for the unexpected finding of an association between cannabis use and reduced gestational diabetes risk, our researchers speculate that this may be the result of the common practice of using cannabis to alleviate chronic joint pain from morbid obesity. Many of these individuals likely have already been diagnosed with Type II diabetes prior to pregnancy, thus making it impossible for them to receive a diagnosis of gestational diabetes, and giving the misleading impression that cannabis may protect against the same. This hypothesis requires further investigation due to limited data on pregestational diabetes prevalence.

We acknowledge the significant heterogeneity observed across most outcomes, which is not unexpected given the inclusion of 51 studies spanning diverse populations, methodologies, and exposure definitions. Potential sources of this heterogeneity include variations in the frequency, quantity, and recency of cannabis use, which our binary classification (users vs. non-users) may not fully capture. For instance, heavy or frequent use might amplify adverse outcomes compared to occasional use, while recency, such as use concentrated in the first trimester versus throughout pregnancy, could influence fetal development differently due to critical windows of organogenesis. Additionally, the method of assessing cannabis exposure varied across studies, with some relying on self-reports and others using biological validation (e.g., urine toxicology or meconium screening), as detailed in Table 1. These differences could contribute to heterogeneity by affecting the accuracy and consistency of exposure classification. For example, self-reports may underestimate use due to social desirability bias, whereas biological measures might detect use that participants did not disclose.

Many systematic reviews and meta-analyses have supported the effect of cannabis consumption in increasing risks of neonatal adverse effects, especially preterm delivery, NICU admission, low birth weight, and smaller head circumference, as was seen in our findings [16, 18, 77].

Our meta-analysis, encompassing 51 studies and 7,920,383 women, aligns with and extends findings from prior meta-analyses by Conner 2016 et al. [15], Gunn 2016 et al. [16], Lo 2023 et al. [17], and Marchand 2022 et al. [18]. Like Gunn 2016 and Marchand 2022, we found significant associations between prenatal cannabis use and increased risks of low birth weight (RR = 1.69, 95% CI = 1.34–2.14 vs. Gunn’s OR = 1.77 and Marchand’s OR = 1.87), preterm delivery (RR = 1.39, 95% CI = 1.23–1.56 vs. Gunn’s OR = 1.43, Lo’s elevated risk, and Marchand’s OR = 1.42), SGA (RR = 1.79, 95% CI = 1.52–2.1, consistent with Lo and Marchand), and NICU admission (RR = 1.55, 95% CI = 1.36–1.78, echoing Gunn’s OR = 2.02 and Marchand’s findings). However, our results diverge from Conner 2016, which reported no independent cannabis effect after adjusting for tobacco (OR = 1.43 for low birth weight reduced post-adjustment), suggesting our broader, unadjusted associations may partly reflect confounding. Unlike Gunn’s unique finding of maternal anemia (OR = 1.36), we found no significant maternal outcomes except a decreased gestational diabetes risk (RR = 0.64, 95% CI = 0.55–0.75), potentially a spurious signal. Compared to Lo 2023, which found no clear cannabis-only mortality link, our increased perinatal mortality (RR = 1.72, 95% CI = 1.09–2.71) suggests newer studies may amplify this signal, though with borderline significance. Our inclusion of 35 additional studies beyond Marchand 2022’s 16 reinforces these associations, adding novel outcomes like major anomalies (RR = 1.81, 95% CI = 1.48–2.23) and decreased head circumference (MD = −0.34, 95% CI = −0.57 to −0.11), not emphasized in earlier works. This expanded scope, current to March 2024, suggests a consistent pattern of neonatal risk, though confounding remains a challenge, aligning with all four prior reviews’ cautions.

Associated smoking with cannabis consumption could be an important confounding factor that can be responsible for this association as found in Conner 2016 et al. who found that there was no significant difference between cannabis users and non-users regarding neonatal outcomes after controlling confounders like tobacco smoking [15] which was supported also by English 1997 et al. [78] who included only studies which adjusted the tobacco use. This effect results from the larger percentage of cannabis smokers also smoking cigarettes during pregnancy than non-users [79]. While our large sample size (over 7 million women) suggests robustness, uncontrolled tobacco use remains a potential confounder, as noted in prior studies [15, 78]. Further evidence for this has been presented in the 2017 cross-sectional analysis by Haight et al. [80], which found high frequency cannabis use was related to lower birth weights regardless of cigarette use. To further explore this, we reviewed the 51 included studies and found that approximately 20 (39%) explicitly reported adjusting for smoking status in their statistical analyses (e.g., Conner et al., 2015; Metz et al., 2017; Avalos et al., 2023), as noted in their respective methodologies or results sections [15, 21, 37]. The remaining studies either did not adjust for smoking or did not clearly report such adjustments, often due to reliance on self-reported data or lack of detailed covariate control. This variability likely contributes to the observed heterogeneity across outcomes. While we considered stratifying our analysis by adjustment status, the inconsistent reporting of adjustment methods and the lack of standardized data on smoking adjustment across studies precluded a meaningful meta-analytic separation. Instead, we relied on the random-effects model to account for this variability, ensuring our pooled estimates reflect the real-world diversity of study designs and confounder handling.

Another potential source of heterogeneity could be the timing of cannabis exposure during pregnancy, which our study did not stratify due to limited data granularity in the included studies. Early exposure during the first trimester, a period of rapid fetal organogenesis, might pose different risks compared to use later in gestation, potentially affecting outcomes like congenital anomalies or preterm delivery differently. While some studies in our review (e.g., Dodge et al., 2023; El Marroun et al., 2009) explored timing-specific effects, the majority provided only aggregate exposure data, precluding a meta-analytic stratification by trimester [29, 32]. This limitation is inherent to the retrospective nature of our source material, but it highlights an important avenue for future research.

Almost all recent systematic reviews have agreed with cannabis increasing the risk of poor neonatal outcomes, especially weight outcomes [18, 77, 81], preterm delivery [18, 77, 81], and NICU admission [18, 81]. However, secondary to the large number of included studies, this analysis was able to include many other neonatal outcomes that have not been thoroughly addressed in previous analyses. These outcomes included fetal anomalies, neonatal mortality, birth length, head circumference, and decreased gestational age. This is considered a strength of our review. Moreover, we found maternal cannabis use was associated with an increased risk of infant death during the first year of life, with an adjusted risk ratio of 1.72 compared to non-users. This findings is consistent with a 2023 retrospective study, Bandoli et al. [82], that specifically analyzed this outcome and further found that the specifically increased causes of mortality were sudden unexpected death and death attributable to perinatal conditions.

Many recent studies have also supported the association of cannabis consumption with anomalies affecting many systems like gastrointestinal, neuronal, nephrological, cardiovascular and musculoskeletal, although there is no consensus as to what the mechanism of this damage truly is [83–86]. Some authors have hypothesized that this may be secondary to cannabis’s role in the methylation of fetal DNA, which may increase the risk of birth defects and other anomalies [87]. Others have postulated that it could be cannabis’s role in glucose and insulin regulation that affects fetal growth and may explain its teratogenicity [32]. As the endocannabinoid system is important in the early stages of cell survival and formation of the neuronal system [88], other authors have suspected that disruption of this system may be the cause of birth defects and other adverse neonatal outcomes associated with cannabis [89]. Lastly, other authors have speculated that cannabis damages placental endocrine function by enhancing ESR1 and CYP19 AI transcription, which may increase estradiol production, causing disruption [90].

Besides neonatal outcomes, the association between cannabis use and maternal complications is also controversial. Many studies have found pregnant cannabis users were found to have higher risks of less studied outcomes not included in this study, including alcohol consumption, anemia, depression, and anxiety [16, 50, 91]. However, when focusing on the most commonly studied outcomes, such as placental abruption, antepartum or postpartum bleeding, and gestational hypertension, most [50, 62, 91], but not all [92] studies showed no significant association with cannabis use. Lastly, we found an unexpected result compared to the previous literature on the decreased risk of gestational diabetes mellitus in cannabis users compared to non-users. Most previous studies have found no association [50, 62, 91], and one study, Porr et al. [93], actually found that cannabis use was associated with increased HbA1c in diabetes mellitus. Another study, Ayonrinde et al. [94], also found that cannabis use increased the caloric intake, weight, and percentage of fatty liver during pregnancy, which in turn increased insulin resistance. Pan et al. [95] in 2023 found that preconceptional cannabis use was associated with increased gestational diabetes risk in pregnant women who never used tobacco; however, among those on current or previous using tobacco, no significant results were observed. Consistent with these studies and as stated above, we believe the protective association we have seen against gestational diabetes is most likely not a true signal, and is secondary to the likely higher percentage of pregestational diabetics in the cannabis use group, making it impossible for these women to receive a diagnosis of gestational diabetes during pregnancy. Unfortunately we do not have the specific data as to the percentages of pre-existing diabetics in both groups that would be necessary to test this hypothesis.