Birth outcomes following intrauterine exposure to electronic nicotine delivery systems: a systematic review and meta-analysis

Abstract

Introduction

Despite electronic nicotine delivery systems (ENDS) being used for smoking cessation during pregnancy, the relative consequences are still unclear. Here, we systematically reviewed and meta-analyzed birth outcomes associated with ENDS use.

Methods

This review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement. OVID (MEDLINE, EMBASE, PsycINFO) and PubMed databases were searched from inception until August 2025, for studies comparing outcomes from in utero exclusive ENDS exposure to non-exposed, exclusive nicotine replacement therapy–exposed, or exclusive combustible cigarette (CC)–exposed neonates.

Results

We included k = 22 studies with k = 10 independent samples. In ENDS-exposed vs non-exposed births, there were significantly greater odds of preterm birth (k = 3; moderate confidence) but not small for gestational age (SGA; k = 3; low confidence); these findings were less robust to sensitivity analysis. Additionally, there was no significant difference in SGA (k = 3; moderate confidence) or preterm birth (k = 3; high confidence) between CC-exposed and ENDS-exposed neonates; these findings were robust to sensitivity analysis. Analysis of continuous outcomes found lower birthweight in ENDS- vs non-exposed neonates (k = 4; moderate confidence) and lower birthweight in CC- vs ENDS-exposed neonates (k = 4; low confidence), and no difference in gestational age in either comparison (low confidence). Narrative synthesis suggests quitting ENDS was associated with better birth outcomes, and associations between ENDS and delayed development of infant motor maturity and reflexes.

Conclusion

ENDS exposure may confer comparable risk as cigarette smoking to preterm birth. Further research is needed given low to moderate confidence in findings and risk of bias conferred by limited classification of exclusive ENDS use.

Implications

With respect to adverse birth outcomes, ENDS may constitute a risk compared to no-nicotine exposure, suggesting the importance of investigating vaping cessation interventions, as well as greater research into alternative smoking cessation interventions for people who are pregnant or want to become pregnant.

Introduction

A wealth of literature supports that combustible cigarette (CC) smoking during pregnancy has deleterious effects on birth outcomes and childhood neurodevelopmental and behavioral outcomes.^1-4 Public health messaging, including the Surgeon General report, has been unequivocal about the physiological and psychological risks associated with intrauterine exposure to CCs.⁵ Accordingly, there has been substantial decreases in CC use during pregnancy in the United States, with recent estimates indicating a prevalence of 4.6% in 2021.⁶ Comparatively, research and public health recommendations related to intrauterine exposure to parental use of electronic nicotine delivery systems (ENDSs) have not kept pace with the rapid rise in ENDS use.⁷ Indeed, a 2018 report by the US National Academies of Sciences (NASEM) stated that there was no available evidence regarding ENDS and pregnancy outcomes and that the evidence on fetal development was insufficient.⁸ Public perceptions of ENDS as risk-free or as less risky than CCs may have contributed to greater ENDS use during pregnancy, with recent estimates ranging up to 7.0%.⁹ Indeed, several studies have found that pregnant women endorsed believing there was lower prenatal risk of ENDS compared to smoking.^10-12 Thus, calls have been made for research comparing the effects of prenatal ENDS exposure to CC exposure and no nicotine exposure.

Earlier (2020–2021) systematic reviews related to non-combustible nicotine use included 1–7 studies reporting pregnancy outcomes associated with ENDS.^13-15 They concluded that there are potential for harms from ENDS use and that there is insufficient data to support ENDS as a harm reduction tool and also described challenges in the literature with respect to categorization of “exclusive” ENDS use, including a lack of inclusion of biomarkers.^13-15 Given the rapidly evolving nature of ENDS research, with several studies published since 2021, we systematically review and meta-analyze the updated evidence base on birth and extended the search to infant neurodevelopmental outcomes associated with intrauterine exclusive ENDS exposure, relative to exclusive CC exposure, no exposure, and exclusive exposure to nicotine replacement therapy (NRT) to triangulate the effects of nicotine exposure and provide a comprehensive understanding of risks associated with in utero ENDS exposure. Note that groups were classified based on study-defined exclusivity, which is detailed in Supplemental Table S1.

Notably, two meta-analyses on this topic were published while the present meta-analysis was underway and concluded there was evidence for adverse neonatal outcomes associated with ENDS exposure and minimal concerns for certainty of evidence.^16,17 One meta-analysis compared intrauterine exposure to ENDS with CC or no exposure. They found greater odds of preterm birth, low birthweight, and small for gestational age (SGA) in ENDS relative to no nicotine use and lower odds of preterm birth (but not SGA or LBW) in ENDS relative to smoking.¹⁷ Of eight studies, six were from overlapping samples, and sensitivity analyses of estimated effects were not conducted. The other study only compared intrauterine exposure to ENDS with non-exposure and found significantly greater odds of adverse neonatal outcomes. The authors also included overlapping samples (ie, 10/17); a sensitivity analysis excluded overlapping samples, although prioritization decisions for specific studies in each outcome are unclear.¹⁶ Thus, the present meta-analysis builds on the extant literature by improving the accuracy of estimates with our handling of overlapping samples and detailed sensitivity analyses,¹⁷ and separate evaluation of multiple neonatal outcomes against CCs and no nicotine use.

Effects of combustible cigarette smoking during pregnancy

CC smoking during pregnancy results in exposure to nicotine, the primary psychoactive component of cigarettes, as well as numerous other reproductive toxins and carcinogens, including carbon monoxide (CO), carbon dioxide, and nitric oxide, each of which may affect birth and infant outcomes.¹⁸ Notably, CC exposure may contribute to persistent epigenetic alterations in DNA methylation that underlie adverse birth outcomes and subsequent developmental delays.¹⁹ Indeed, prenatal CC exposure has been associated with significantly lower birthweights and preterm birth; these effects have been found to be dose-dependent,¹ and quitting smoking during pregnancy is associated with significantly better birth outcomes than continued smoking.²⁰ In addition, significantly reduced brain sizes and neurological dysfunction have been found in children exposed to CC smoke compared to those who were not exposed.^21,22 Critically, long-term consequences may arise from adverse birth outcomes associated with exposure. For example, LBW is associated with subsequent reduced cognitive performance compared to children born at normal weights.²³ This effect was attenuated but still significant after controlling for psychosocial factors (eg, socioeconomic status).²³ In conjunction with genetic and socioeconomic factors, exposure may confer risk to other psychological and behavioral outcomes, including conduct disorder, cognitive impairment, the impulsivity dimension of attention-deficit hyperactivity disorder, and poor academic outcomes.²⁴ However, given the considerable reproductive toxin exposure from CC smoking, this evidence does not inform the degree to which nicotine specifically contributes to these outcomes.

Current evidence of effects of intrauterine nicotine exposure

While the literature indicates multiple reproductive toxins present in CCs, existing evidence suggests that nicotine exposure alone negatively affects fetal development, implicating ENDS as a teratogenic risk. Nicotine readily crosses the placenta and can independently affect the development of fetal brain tissue¹ through a variety of mechanisms. In particular, nicotine activates nicotinic acetylcholine receptors within the placenta, potentially impacting cellular reproductive processes, placental development, and growth.^18,25 Studies in animal models have found that subcutaneous nicotine exposure in utero resulted in significantly lower fetus weights and fetal brain weights²⁶ and lower body weights,²⁷ in addition to structural and neurotransmitter abnormalities.²⁶ Furthermore, nicotine acts as a vasoconstrictor, reducing delivery of oxygen and nutrition to the fetus in humans.¹⁸ Vasoconstriction was previously considered a mechanism of intrauterine CC harms, as animal models exposed to short-term high doses of nicotine had significantly lower fetal weight.²⁸ However, some studies have suggested that in humans, uteroplacental oxygen insufficiency did not predict low birthweights, except for when blood flow was severely compromised.²⁹ This literature has been used to support prescription of nicotine replacement therapy (NRTs) during pregnancy, under close medical supervision, if behavioral attempts to quit smoking fail, with researchers purporting that nicotine was not the main driver of intrauterine CC harms.^29,30

Animal models of intrauterine electronic nicotine delivery system exposure

Harms from ENDS exposure may exceed those from delivery methods used in NRT. Notably, inhaled nicotine has a rapid cellular uptake^31,32 and overall nicotine intake from ENDS may be comparable to that from cigarettes.³³ In addition, ENDS use also results in intrauterine exposure to several other ENDS vapor constituents, including heavy metals, formaldehydes, volatile organic compounds, and tobacco-specific nitrosamines,³⁴ albeit typically to a lower degree than found in CCs.^35-37 Other evidence suggests that ENDS emit nicotine, as well as a range of potential toxicants, the degree to which can vary widely based on use patterns, vape generation, and settings.^38,39 Studies on animal models have found that intrauterine e-cigarette exposure (ie, including other vapor constituents) resulted in reduced birthweight and placental weight, as well as dose-dependent reductions in uterine artery functioning.^40,41 Other studies have found e-cigarette aerosol exposure resulted in significant impairments in cardiovascular functioning that persisted long term, in both nicotine and non-nicotine containing aerosols⁴² and significantly lower birthweight of female offspring, although not in male offspring.⁴³

Investigating the effects of ENDS on birth outcomes will be critical to understanding potential long-term outcomes in cognitive, psychological, and behavioral domains. Indeed, animal models suggest nicotine exposure in utero is associated with increased motor activity, hyperactivity, and impairments in attention, learning, and memory in offspring,²⁴ as well as impaired baseline startle reactivity, enhanced sensory motor gating, and increases in anxiety-associated behaviors in adolescent offspring⁴⁴ and deficits in long-term memory in adult offspring.⁴⁵ These findings support a direct association with nicotine exposure on neurodevelopmental outcomes; thus, we aimed to additionally examine the available evidence in humans addressing neurodevelopmental outcomes associated with ENDS exposure.

Present study

While the evidence from animal models is fairly conclusive, it may not necessarily generalize to humans, who do not undergo regimented short-term high-dose exposures to e-cigarette vapor. Similarly, the evidence from CC smoking is useful in predicting potential risks of vaping but cannot be assumed to generalize to ENDS, as ENDS differ from cigarettes in the associated exposure to reproductive toxins. Notably, since NRTs may be used as a smoking cessation aid, it is further prudent to compare ENDS exposure to NRT exposure in determining the utility of ENDS as harm reduction in pregnancy. Therefore, to clarify the risks specific to ENDS use during pregnancy, we aim to meta-analyze the effects of intrauterine ENDS exposure on birth and infant/toddler outcomes compared to CC exposure, NRT, or no nicotine exposure.

Materials and methods

This systematic review and meta-analysis approach was pre-registered with PROSPERO (CRD42024584628) and follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA).⁴⁶ The study-inclusion PRISMA diagram is included in Figure 1.

Figure 1.

PRISMA diagram of study eligibility determination.

Search strategy

We searched PubMed, OVID MEDLINE, EMBASE, and PsycINFO in September 2024, March 2025, and August 2025. Search terms were terms related to ENDS (eg, “e-cigarettes,” “vaping”) and pregnancy (eg, “intrauterine,” “prenatal”). The original search strategy and revised search strategy following peer review are included in Supplemental Material Section 1. Three raters independently screened studies for eligibility (S.J., S.C., A.S.) based on title and abstract content followed by a full-text review of each text by S.C. and A.S.; any disagreements were resolved by discussion with S.J.

Eligibility criteria

Eligibility criteria for this study follow PECOS,⁴⁷ as is standard practice per PRISMA. Manuscripts were included if they were original, peer-reviewed, studies assessing (P) newborns, infants, or toddlers (0 days to 3 years), (E) who were exposed in utero to ENDS exclusively, (C) compared to newborns or infants/toddlers not exposed to nicotine, exposed in utero to CCs exclusively, or exposed in utero to nicotine replacement therapy exclusively, (O) on validated measures of birth or infant neurodevelopmental outcomes, assessed by (S) cross-sectional, retrospective, and prospective longitudinal study designs.

Studies were excluded if they were animal models, meta-analyses, or systematic reviews. Given the focus on birth outcomes, we planned to exclude studies with children only over the age of 3 years old (however, no studies were identified with older children). One study⁴⁸ was excluded because there were no outcome events in the small (n = 18) ENDS group and the effect size was reported as not calculable.

Data extraction

We extracted birth outcomes (birthweight, preterm birth, etc.) and infant/toddler neurodevelopmental outcomes (cognitive functioning, psychological/behavioral outcomes). Infant/toddler outcomes that are outside the scope of neurological or psychological functioning (ie, respiratory health) are beyond the scope of this review. Where possible, we extracted most-adjusted odds ratio (OR) and confidence intervals for comparisons of interest (ENDS vs CC or NRT or non-use). Otherwise, we extracted means and standard deviations (or Cohen’s d or t-values) for continuous outcomes and frequencies for binary outcomes. Other extracted data included study year, country, demographics of sample and control group, eligibility criteria, frequency of nicotine exposure, ENDS nicotine concentration, methods of determining exposure, and study limitations. Authors were contacted for data if needed.

Risk of bias

The Risk Of Bias In Non-randomized Studies–of Exposure (ROBINS-E) tool was used to assess study bias.⁴⁹ The ROBINS-E provides a structured approach to assess the strength of evidence of a potential effect of an exposure on an outcome by evaluating the risk of bias and whether the bias is sufficiently high to threaten the conclusions under the domains of: confounding, measurement of exposure, selection of participants, post-exposure interventions, missing data, measurement of outcomes, and selection of outcomes. With respect to evaluating studies on domain 1 (controlling for confounding variables), relevant covariates were specified a priori: maternal age, maternal race/ethnicity, prenatal care/healthcare, socioeconomic status, and alcohol or cannabis use. One intervention study was identified and evaluated Risk of Bias In Non-randomized Studies–of Intervention (ROBINS-I),⁵⁰ which is intended to assess studies that were not blinded. Bias domains include randomization, deviation from intended intervention, missing data, outcome measurement, and selection of reported results. After reaching 100% agreement on two initial article assessments, bias assessments were conducted independently by A.S. and S.C.; all assessments were reviewed by S.J.

Statistical analysis

To characterize the risk of ENDS exposure, we aimed to compare outcomes in exclusively ENDS-exposed births to no-nicotine exposure, NRT, and to CC exposure. For a given outcome to be meta-analyzed, at least three included studies had to assess the outcome. For binary outcome variables (eg, SGA), a random-effects meta-analysis with the most adjusted ORs was conducted. If another ratio statistic was reported, it was converted to an OR. For continuous outcome variables, a random-effects meta-analysis on the standardized mean difference between ENDS exposure vs comparator exposure was conducted. Through study eligibility determination, we identified numerous studies using the Pregnancy Risk Assessment Monitoring Study (PRAMS), a population-based retrospective cross-sectional survey surveillance project with the intention to monitor changes in maternal and child health.⁵¹ One PRAMS study⁵² assessed adolescent mothers only; findings were excluded from the meta-analysis, given that data were wholly overlapping with other PRAMS samples that did not exclude adolescents and variability would likely be explained by age. Handling of PRAMS samples was not pre-registered as the amount of overlapping data was not anticipated. Prior to analysis, we decided to prioritize the Ammar et al.⁵³ PRAMS study in the meta-analysis given that it had the largest sample size and longest years of coverage (2016–2020) and conducted a sensitivity analysis by replacing it with each of the other PRAMS sample studies.

Variation in effect sizes due to heterogeneity (rather than chance) was assessed with I².⁵⁴ Trim-and-fill methods assessed publication bias and the potential effect of missing studies on confidence intervals.⁵⁵ Influential case analyses assessed study outliers through externally standardized residuals, “difference in fit” values, Cook’s distance, covariance, leave-out heterogeneity, hat values, and weights. If a study was determined to be influential by any assessment, it was identified. Likelihood log-ratio plots for all models determined that all parameters were identifiable. All analyses were conducted in R-Studio v4.3 with the package “metafor.”⁵⁶

Results

Study characteristics

We included k = 22 studies (see Supplemental Table S1). We found k = 4 studies from Europe, and the remainder were from the United States. Of the studies from the United States, k = 2 utilized data from the Population Assessment for Tobacco and Health (PATH), which is a population-based longitudinal survey. In addition, k = 10 utilized data from PRAMS. We identified k = 12 retrospective cross-sectional cohort studies (k = 10 PRAMS), k = 5 longitudinal cohort studies, k = 1 longitudinal case–control study, k = 2 cross-sectional studies, and k = 1 randomized clinical trial (comparing ENDS to NRT for smoking cessation reported in two manuscripts, outcome data were only extracted from participants who were determined to be abstinent from smoking). All studies reported birth outcomes from either birth certificates or medical records. Only one study also assessed infants at 1 month old; therefore, we could not meta-analyze neurodevelopmental outcomes.

The median number of pregnant women reporting exclusive ENDS use was 125.5 (range 6 to a weighted 61 173) and exclusive cigarette smoking was 719 (range 29 to a weighted 773 586). There was a high degree of variability among studies with respect to classification of exposure, with k = 8 PRAMS studies defining exclusive use in the last 3 months of gestation and the remaining PRAMS studies defining exclusive use in the 3 months prior to and last 3 months during gestation, and k = 2 assessing use in the past 30 days at survey completion. Comparatively, non-epidemiological studies tended to have bioverification of self-report to support exposure classification. Indeed, k = 5 sample-level studies used biological markers of exposure data; critically, one study that employed biomarkers of nicotine exposure with hair nicotine, salivary cotinine, and exhaled CO found substantial non-reporting of ENDS and smoking exposure (24% and 12%, respectively⁵⁷), suggesting comparison of nicotine product groups through only self-report may be affected by self-reporting bias impacting classification.

Notably, where reported, most studies included a range of ENDS exposure frequency (ie, from a few times a month to multiple times a day), whereas cigarette exposure tended to be daily, indicating variability in exposure. Furthermore, all studies investigating both cigarettes and ENDS reported a higher prevalence of smoking during pregnancy versus ENDS use. In one PRAMS sample, ENDS use increased among pregnant adolescents from 0.8% to 4.1% between 2016 and 2021.⁵²

Risk of bias

Most of the included studies were determined to have some concerns or high risk of bias by the ROBINS-E and ROB-2 and one study as a very high risk of bias. Studies were considered to have a high risk of bias if they did not exclude or control for other substance use, as well as other predetermined covariates (eg, socioeconomic status). In addition, studies were considered to have some risk of bias in classification of exposure when there was limited coverage of assessing exposure across the gestational period. Finally, Hajek et al.⁵⁸ were considered to have some risk of bias due to deviation from intended intervention (ie, NRT or ENDS for smoking cessation). Domain ratings for risk of bias assessments can be found in Supplemental Material Section 2.

Meta-analytic results

Categorical outcomes

When comparing exclusive ENDS exposure during pregnancy to no-nicotine exposure, we included k = 8 studies that assessed preterm birth (k = 3 independent samples and k = 5 other studies that assessed PRAMS data) and k = 8 studies that assessed SGA (k = 3 independent samples and k = 5 others that assessed PRAMS data). Only one included study assessed NRT as a comparator to ENDS and therefore could not be meta-analyzed.

Among the k = 3 independent samples with preterm birth as an outcome,^59-61 the risk of preterm birth was significantly greater in ENDS-exposed (n = 334) vs non-exposed (n = 56 571) births (OR = 1.25, 95% CI [1.01, 1.56]; see Figure 2) and the index of between-study heterogeneity was not significant (I² = 0%, P = .55). Opondo et al.⁵⁹ was considered influential. Trim-and-fill analysis indicated two missing studies, when the 95% CI was re-estimated it remained significant [1.18, 1.98]. When substituting the Ammar et al. PRAMS study with the other PRAMS sample studies, the risk of preterm birth remained significantly higher in two of the other models (⁵³OR = 1.50 [1.16, 1.94]⁶²; OR = 1.51 [1.07, 2.12]) but not with the remaining three models (⁶³OR = 1.59 [0.81, 3.11]⁶⁴; OR = 1.07, [0.50, 2.26])⁶⁵; OR = 1.13 [0.63, 2.02].

Figure 2.

Odds of preterm birth (upper panel) and small for gestational age (SGA; lower panel) outcomes in electronic nicotine delivery system–exposed relative to non-exposed births. Note: ^*indicates that the study uses the PRAMS sample; each were substituted for Ammar et al. as a sensitivity analysis.

Among k = 3 independent samples with SGA as an outcome,^53,57,61 the odds of SGA were not significantly greater for ENDS-exposed (n = 988) compared to non-exposed (n = 175 247) neonates (OR = 3.48 [0.68, 17.76]; see Figure 2), and the test for heterogeneity was not significant (I² = 61%, P = .06). Ammar et al.⁵³ and Nanninga et al.⁶¹ were considered influential. Trim-and-fill analysis indicated two missing studies, and the 95% CI remained non-significant [0.17, 6.50]. When substituting with the other PRAMS sample studies, the risk of SGA was significantly higher in two models (⁶³OR = 4.02, [1.10, 14.61]⁶⁴; OR = 4.49 [1.53, 13.18]) but not with the other models (⁶⁰OR = 3.55 [0.72, 17.53]⁶²; OR = 3.35 [0.59, 18.92]⁶⁵; OR = 3.69, [0.78, 17.53]⁶⁶; OR = 3.76, [0.91, 15.60]).

With respect to CC smoking, we included k = 8 studies that assessed preterm birth (k = 3 independent samples and k = 5 other studies that assessed PRAMS data) and k = 8 studies that assessed SGA (k = 3 independent samples and k = 5 others that assessed PRAMS data).

Among the k = 3 independent samples with preterm birth as an outcome,^53,59,61 the odds of preterm birth between ENDS-exposed (n = 1010) and CC-exposed (n = 15 061) neonates was not significant (OR = 0.96, 95% CI [0.82, 1.12]; see Figure 3). Between-study heterogeneity was not significant (I² = 0%, P = .87). Ammar et al.⁵³ was considered influential. Trim-and-fill analysis indicated zero missing studies. An additional sensitivity analysis was conducted substituting with the other PRAMS sample studies and the difference in odds of preterm birth remained non-significant.^62,65,67

Figure 3.

Odds of preterm birth (upper panel) and small for gestational age (SGA; lower panel) in electronic nicotine delivery system exposed relative to combustible cigarette–exposed births. Note: ^*indicates that the study uses the PRAMS sample; each were substituted for Ammar et al. as a sensitivity analysis.

Among the k = 3 independent samples with SGA as an outcome,^53,57,61 there was also no significant difference in odds of SGA between ENDS (n = 988) compared to CC (n = 14 932) exposed (OR = 1.32, 95% CI [0.37, 4.69]; see Figure 3). However, there was significant between-study heterogeneity (I² = 67.2, P = .03). Ammar et al.⁵³ and Nanninga et al.⁶¹ were considered influential. Trim-and-fill analysis indicated two missing studies, but the 95% CI remained non-significant [0.16, 2.24]. When substituting with four of the other PRAMS sample studies, the difference in odds of SGA remained non-significant.^62-64,67 When substituting with one of the other PRAMS studies, the odds of SGA were significantly higher in ENDS exposed versus CC-exposed neonates (⁶⁵OR = 1.60 [1.10, 2.34]). Notably, this study included a smaller sample of ENDS and CC-exposed neonates.

Continuous outcomes

We found k = 5 studies^{59,62,68–70} that assessed gestational age and found there was no significant difference between ENDS-exposed (n = 597) vs non-exposed (n = 54 460) neonates (SMD = 0.8, [−0.30, 0.47], P = .65; see Figure 4). The test for heterogeneity was not significant (I² = 0.0, P = .25). Trim-and-fill analysis indicated one missing study; the estimated confidence interval remained non-significant [−0.26, 0.49]. Froggatt et al.⁶⁸ and Hawkins et al.⁶² were considered influential. Note that k = 3 studies only included term births. In addition, we found k = 4 studies^59,68,69,71 that assessed birthweight in ENDS-exposed (n = 726) vs non-exposed (n = 29 390) neonates. There was significantly lower weight in grams at birth among ENDS-exposed neonates (SMD = −94.2, 95% CI [−132.2, −4.79], P < .001; see Figure 4). The test for heterogeneity was not significant (I² = 0.0, p = .25). Azar et al.⁷¹ was considered influential. Trim-and-fill analysis indicated two missing studies; the confidence interval remained significant after estimating their effects [−136.0, −60.2].

Figure 4.

Mean difference in gestational age and birthweight between electronic nicotine delivery system exposed relative to non-exposed births.

In k = 3 studies^59,68,69 that assessed gestational age, there was no significant difference between ENDS-exposed (n = 232) and CC–exposed (n = 271) neonates (SMD = .39, [−0.21, 1.00], P = .20; see Figure 5); the test for heterogeneity was not significant (I² = 58.9, P = .08). Froggatt et al.⁶⁸ and McDonnell et al.⁶⁹ were considered influential. Trim-and-fill analysis indicated two missing studies; the confidence interval remained non-significant [−0.63, 0.63]. In k = 4 studies^59,68,69,71 that assessed birthweight in ENDS-exposed (n = 726) vs CC-exposed (n = 2405) neonates, ENDS-exposed neonates were significantly heavier at birth (SMD = 223.2, [29.6, 416.8], P = .20; see Figure 5). The test for heterogeneity was significant (I² = 83.5, P < .001). Azar et al.⁷¹ was considered influential. Trim-and-fill analysis indicated one missing study, and the confidence interval was no longer significant after estimating the effect [−3.90, 347.6].

Figure 5.

Mean difference in gestational age and birthweight between electronic nicotine delivery system–exposed relative to combustible cigarette–exposed births.

Narrative synthesis

Additional data from included studies that could not be meta-analyzed evaluated other potential risks of intrauterine ENDS exposure (see Supplemental Table S2). Four studies (k=3 PRAMS samples) found that ENDS-exposed neonates had significantly greater odds of LBW relative to non-exposed neonates and no difference in odds of LBW between ENDS-exposed versus CC-exposed neonates.^53,63,67,71 One study found ENDS-exposed neonates were significantly shorter with a smaller head circumference compared to non-exposed neonates.⁷⁰ Two PATH studies covering the same years of data collection (2013–2018) assessed adverse birth outcomes, which included preterm birth, LBW, birth defects, placenta previa, placenta abruption, and pre-eclampsia and found no significant differences between ENDS exposed and non-exposed or CC-exposed neonates.^72,73

In addition, four studies (k = 2 PRAMS samples) reported birth outcomes related to cessation and product switching behavior. Three studies (k = 2 PRAMS) reported heavier birthweights and higher gestational age among neonates born to women who quit exclusive ENDS use compared to those who continued to use,^53,67,74 whereas one study found that quitting ENDS was not associated with better outcomes.⁷³ Notably, one study reported no significant difference in outcomes based on switching from CCs to ENDS,⁷⁴ whereas another study reported that switching from cigarettes to ENDS use resulted in birth outcomes similar to non-exposed neonates.⁶⁶ However, initiating dual-use during pregnancy resulted in the greatest risk of SGA.⁶⁶ In addition, one treatment study^58,75 compared ENDS to NRT for smoking cessation during pregnancy and found greater smoking cessation in participants randomized to ENDS; intention-to-treat (ITT) analysis revealed greater odds of LBW in neonates of NRT-treated mothers compared to ENDS-treated mothers.

Finally, while nearly all studies collapsed findings across ENDS nicotine concentration, frequency of use, and flavors, one study reported greater risk of adverse outcomes in neonates exposed to mint/menthol-flavor ENDS products compared to other flavors among exclusive and dual ENDS exposure,⁷³ and one study that differentiated birth outcomes by frequency of ENDS use found significantly greater adverse birth outcomes (relative to non-exposed) only in those reporting daily use.⁶⁰

Only one study assessed neurodevelopmental outcomes after birth. Froggatt et al.⁶⁸ found that compared to non-exposed, infants/toddlers who were exposed to ENDS and CCs had significantly lower developmental attainment in motor maturity and reflexes. Notably, there were no significant differences on motor maturity, reflex, and self-regulation subscales between ENDS- and CC–exposed.

Certainty assessment: GRADE

GRADE ratings can be found in Table 1. All domains had some or high risk of bias; SGA analyses were downgraded due to heterogeneity and not meeting the optimal information size criterion. Analyses against non-exposed neonates were further downgraded due to lack of robustness against some sensitivity analyses. Final ratings for comparisons to non-exposed are low confidence for SGA and moderate confidence for preterm birth; comparisons to CC-exposed are moderate confidence for SGA and high-confidence for preterm birth. Final ratings for comparisons to non-nicotine exposed are moderate confidence for birthweight and low confidence for gestational age; comparisons to CC-exposed are low confidence for birthweight and gestational age.

Table 1.

GRADE certainty assessment of evidence for birth outcomes.

	Bias across studies		Heterogeneity		Indirectness		Imprecision of results		Publication bias		Other
SGA (vs non-use) Rating: Low confidence	Most some-concerns	−1	Low	0	Evidence is sufficiently direct	0	1/3 studies met OIS criterion	−1	Bias affected findings	−1	Large effect size, not robust to some sensitivity analysis	−1
SGA (vs CC) Rating: Moderate confidence	Most some-concerns	−1	High	−1	Evidence is sufficiently direct	0	1/3 studies met OIS criterion	−1	No bias or bis did not affect findings	0	Very small effect size, robust	+2
Preterm (vs non-use) Rating: Low confidence	Most some-concerns	−1	Low	0	Evidence is sufficiently direct	0	2/3 studies met OIS criterion	0	No bias or bis did not affect findings	0	Not robust to some sensitivity analyses	−1
Preterm (vs CC) Rating: High confidence	Most some-concerns	−1	High	−1	Evidence is sufficiently direct	0	2/3 studies met OIS criterion	0	No bias or bis did not affect findings	0	Very small effect size, robust	+2
Birthweight (vs non-use) Rating: Moderate confidence	Most some-concerns	−1	Low	0	Evidence is sufficiently direct	0	2/4 studies met OIS criterion	0	No bias or bis did not affect findings	0
Gestational age (vs non-use) Rating: Low confidence	Most some-concerns	−1	Low	0	Evidence is sufficiently direct	0	2/5 studies met OIS criterion	−1	No bias or bis did not affect findings	0
Birthweight (vs CC use) Rating: Low confidence	Most some-concerns	−1	High	−1	Evidence is sufficiently direct	0	2/4 studies met OIS criterion	0	Bias affected findings	−1
Gestational age (vs CC use) Rating: Low confidence	Most some-concerns	−1	Low	0	Evidence is sufficiently direct	0	2/5 studies met OIS criterion	−1	No bias or bis did not affect findings	0

Note: Imprecision of results (OIS criterion) estimated for 80% power to detect an odds ratio of 1.95 based on difference in odds between CC and no nicotine [92] for OR; a small effect size between CC and no nicotine or weeks between CC and no nicotine [62]

Discussion

Prevalence of ENDS use during pregnancy has been increasing,⁹ potentially in part due to limited knowledge of outcomes associated with intrauterine ENDS exposure. This study sought to systematically review and meta-analyze birth and neurodevelopmental outcomes associated with ENDS use. With the initial meta-analyses utilizing the largest PRAMS sample, we found significantly greater odds of preterm birth in ENDS-exposed neonates compared to no nicotine exposure. However, the difference in odds of SGA was not significant. Notably, we found inconsistency in findings when substituting with other PRAMS studies; however, these had less coverage in terms of years of data collection and participants. Comparatively, meta-analysis of continuous outcomes suggested significantly lower birthweight in ENDS-exposed vs non-exposed neonates, although there was no significant difference in gestational age; these findings were more robust to sensitivity analysis. Further research in this area is needed before drawing strong conclusions. Comparatively, relative to CC use, there was not a significant difference in odds of preterm birth and SGA, and these findings were more robust with respect to sensitivity analysis. In addition, a meta-analysis of continuous outcomes found significantly lower birthweight of CC-exposed vs ENDS-exposed neonates, although this did not survive correcting for trim-and-fill analysis. There was no significant difference in gestational age. Critically, our review suggests continued limitations in the literature regarding ENDS exposure in utero, with comparisons against no exposure only demonstrating low to moderate confidence, differing from conclusions of other meta-analyses on this topic.^16,17

Major concerns regarding ENDS use characterization may be contributing to inconsistency in the present findings. Specifically, there is a lack of consensus on how to best quantify ENDS exposure (ie, in a metric that is comparable to CPD), as well as a lack of assessment and reporting of frequency and duration of ENDS use, lack of reporting of nicotine concentration, and limited information on the effect of flavoring, which may contribute uniquely to harms.⁷³ With respect to quantification, only five studies used bioverification of nicotine exposure. Furthermore, most studies assessing ENDS use included participants with use ranging from less than once a week to once a day and compared against cigarette smoking at a rate of 1–41 CPD, limiting the comparability of exposure between ENDS and smoking, further affecting ascertainment of the relative harms of ENDS versus smoking to no nicotine exposure (ie, the harms of daily ENDS use are unclear). However, one PRAMS study that restricted analysis to daily use found greater effect sizes for odds of preterm birth and LBW in ENDS compared to non-exposed neonates than when the analysis was restricted to daily use.⁶⁰ More rigorous characterization of ENDS exposure, including assessment of cotinine, is critical for advancing our understanding of the impact of ENDS use during pregnancy.

Given that most of the included studies were rated as having some or high concerns for risk of bias primarily due to confounding and exposure assessment, and the variability in sensitivity analyses in outcomes associated with ENDS compared to no nicotine exposure, the confidence in the estimated effects is moderate. However, the present findings suggest that odds of preterm birth and SGA associated with ENDS exposure are not significantly less likely when compared to CC exposure and that ENDS exposure may be associated with significantly lower birthweight relative to no nicotine use. Two independent studies^53,74 found that quitting ENDS use was associated with better birth outcomes when compared to continued ENDS use through pregnancy (although one did not⁷³), further suggesting risks associated with intrauterine ENDS exposure. Notably, birthweight is considered one of the most important birth outcomes given that it is associated with social and health-related outcomes in childhood and adulthood, including morbidity and mortality.^76,77 For example, birthweight is predictive of childhood cognitive ability,⁷⁸ which is further associated with occupational prestige, long-term sickness, and educational attainment.^79-81 Thus, significantly lower birthweight in ENDS-exposed vs non-exposed neonates and similar odds of preterm birth and SGA between ENDS and CC-exposed neonates are striking. Our findings indicate a need for vaping cessation research among those who use ENDS and improvements in public health messaging regarding ENDS use during pregnancy. However, it is still notable that CC-exposed neonates had significantly lower birthweights than ENDS-exposed neonates and may indicate that exposure to cigarettes in utero continues to constitute the greatest risk.

Notably, delayed or impaired development may be associated with lower birthweight, SGA, and preterm birth.⁸² However, the ENDS literature remains in its infancy; in the present review, only one article (with a small sample size and limited assessment of exposure) investigated infant/toddler neurodevelopmental outcomes. Findings suggest significantly worse motor maturity and reflexes and possibly greater difficulty with emotion regulation, which may reflect delays in the prefrontal cortex, midbrain, and brainstem regions,⁸³ which may affect future neurodevelopment.⁸⁴ Notably, early neurodevelopment milestones are integral to developing behavioral flexibility, as well as perceptual, cognitive, and social abilities.⁸⁵ In addition, models of developmental psychopathology suggest that failure at earlier stages of neurodevelopment may confer risk to externalizing disorders.⁸⁶

Smoking during pregnancy continues to be a major public health issue. Given the interest in ENDS as a harm reduction tool, the comparison of exposure harms relative to other established cessation treatments is relevant in terms of triangulating the utility of ENDS in harm reduction. In a randomized clinical trial reported by Hajek et al.⁵⁸ and Pesola et al.,⁷⁵ birth outcomes associated with ENDS vs NRT treatment were assessed and there was no significant difference in high-risk birth or fetal death between ITT groups, or compared to those who were completely abstinent from nicotine. However, individuals randomized to NRT had significantly lower birthweight than those randomized to ENDS. Notably, since NRT was associated with lower odds of cessation, these findings may be explained by greater cigarette exposure throughout pregnancy. Relatedly, other studies in this review found that switching from cigarettes to ENDS was not associated with better outcomes while transitioning to dual use, a major concern when introducing ENDS for smoking cessation, was associated with greater odds of adverse birth outcomes.⁶⁶ Therefore, in the context of findings from the present meta-analysis, it may be prudent to investigate alternative smoking cessation methods. However, it should also be noted that findings related to NRT and switching behavior could not be meta-analyzed and are limited by small sample sizes; we urge caution in interpretation of these results. Studies with larger samples of people who switch from CCs to ENDS may inform the utility of ENDS as harm reduction among people who are pregnant.

Limitations and future research

There are several limitations of this meta-analysis. First, while some studies employed large sample sizes, others were quite small, and the sample of ENDS-exposed participants was substantially smaller than non-exposed or cigarette-exposed groups, which may have biased results. Prospective designs using enriched samples are required to ascertain associations between birth and future developmental outcomes of intrauterine ENDS exposure. Second, ENDS exposure was typically poorly quantified, with most studies not reporting frequency or quantity of use, nor e-liquid nicotine concentration. Comparatively, most studies included participants who used cigarettes daily. Therefore, the birth outcomes of daily ENDS use (likely indicative of dependence) to no nicotine exposure and to daily cigarette smoking may be more severe than what is presently reported. Third, classification of exposure was limited as most data were based on self-report. Additionally, ENDS and CC exposure were grouped based on classifications from included studies, most typically the product used in the last 3 months of pregnancy, either with or without consideration of use in the last 3 months before pregnancy. Critically, his method may not capture product switching throughout gestation. Future studies should employ multiple objective biochemical indicators (eg, cotinine and CO) repeatedly across pregnancy, along with detailed self-report around quantity and frequency of ENDS use and product switching behaviors.⁸⁷

Fourth, although we extracted the models adjusted for the most potentially confounding variables (eg, demographics, other substance use socioeconomic status, prenatal care, and previous smoking), when ORs for the comparisons of interest were not reported, we utilized cell counts, which did not address confounding variables.

Fifth, rates of co-use of cannabis, alcohol, and other substances among people who use nicotine during pregnancy is higher than among those who do not use nicotine.^88,89 In the present review, no studies investigated effects of co-exposure to alcohol or cannabis. Critically, co-exposure of nicotine and alcohol or cannabis may worsen birth and neurodevelopmental outcomes,^90-92 and thus should be investigated. Finally, one study suggested that flavoring chemicals may have greater negative effects, specifically menthol.⁷³ Future research should investigate differences in birth and neurodevelopmental outcomes based on menthol exposure, as well as other flavors.

Conclusion

The findings of this meta-analysis suggest the odds of preterm birth and SGA in ENDS-exposed neonates may be comparable to those associated with cigarette exposure. Comparisons against non-exposed neonates relative to ENDS-exposed were mixed, stronger study designs and larger sample sizes are needed. Furthermore, while CC–exposed neonates were significantly lighter than ENDS-exposed neonates; ENDS-exposed neonates were significantly lighter than non-exposed neonates. Narrative synthesis suggests improvements in birth outcomes associated with quitting ENDS. Additionally, infant/toddler neurocognitive development, specifically motor maturity and reflexes, may be affected by ENDS exposure. Therefore, findings suggest the importance of vaping cessation interventions—and alternative smoking cessation interventions—for people who are pregnant or want to become pregnant.

Funding

This manuscript was in part funded by the National Institutes of Drug Abuse (3R01DA054276-03S1) to L.H. and the Canadian Institutes of Health Research Doctoral Foreign Study Award (202310DFD-513256-DRA-86383) to S.J. L.W.H. has received funding from National Institutes of Drug Abuse (R01 DA054276-01A1), National Cancer Institute (R01 CA266170-01A1), and National Center for Advancing Translational Sciences (UG3 TR004797-02).

Declarations of interest

The authors have no conflicts of interest to report.

Author contributions

Samantha Johnstone (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Methodology [equal], Visualization [equal], Writing—original draft [equal]), Ashley Schenkel (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Writing—review & editing [equal]), Samantha Canazzi (Conceptualization [equal], Data curation [equal], Writing—review & editing [equal]), Rina Eiden (Methodology [equal], Supervision [equal], Writing—review & editing [equal]), and Larry Hawk Jr. (Conceptualization [equal], Supervision [equal], Writing—review & editing [equal])

Data availability

No new data were generated or analyzed in support of this research. Meta-analysis code is available upon request.