Marijuana Use during Pregnancy and Preterm Birth: A Prospective Cohort Study

Abstract

Objective

The aim of this study is to estimate the association between marijuana use during pregnancy and total, spontaneous and indicated preterm birth.

Study Design

Prospective cohort study of women receiving antenatal care at The Ohio State University from 2010 to 2015. Marijuana use was assessed by questionnaire, record abstraction, and urine toxicology. Women were followed through the end of pregnancy. Relative risks were assessed with Poisson regression and time to delivery with proportional hazard models.

Results

Of 363 eligible women, 119 (33%) used marijuana in pregnancy by at least one measure. In this high-risk cohort, preterm birth occurred to 36.0% of users and 34.6% of nonusers (p = 0.81). The unadjusted relative risk of all preterm birth was 1.06 (95% confidence interval [CI]: 0.76–1.47); the adjusted relative risk was similar 1.04 (95% CI: 0.72–1.50). Spontaneous preterm birth was nonsignificantly elevated among users before 1.32 (95% CI: 0.89–1.96), and after 1.21 (95% CI: 0.76–1.94) adjustment. Indicated preterm birth was nonsignificantly reduced before 0.52 (95% CI: 0.22–1.23) and after 0.75 (95% CI: 0.29–2.15) adjustment. The unadjusted hazard ratio (HR) for time to preterm birth was 1.26 (95% CI: 0.84–2.00); the adjusted HR was 1.32 (95% CI: 0.80–2.07). Both unadjusted 1.77 (95% CI: 1.06–2.93) and adjusted 2.16 (95% CI: 1.16–4.02) HRs for spontaneous preterm birth were significantly elevated, primarily due to an increased risk of spontaneous birth <28 weeks among users. The unadjusted and adjusted HRs for time to indicated preterm birth were 0.69 (95% CI: 0.33–1.43) and 0.58 (95% CI: 0.23–1.46).

Conclusion

Marijuana use was not associated with total preterm birth in this cohort, suggesting that among women already at high risk of preterm birth, marijuana does not increase risk further. However, there was a suggestion that pregnant women who use marijuana may deliver earlier, particularly from spontaneous preterm birth, than women who do not use marijuana.

Marijuana is the most commonly used illicit substance in the United States, and use is increasing dramatically with state-level legalization and softening public attitudes. Despite guidelines from the American College of Obstetricians and Gynecologists,¹ many marijuana users do not cease during pregnancy. Past month use among pregnant women has doubled since 2002 to 7%, and daily/almost daily use has almost quadrupled to 3.4%.² Whether marijuana use in pregnancy impacts preterm (<37 completed weeks’ gestation) birth is uncertain. A recent systematic review and meta-analysis reported that use was associated with increased risk of preterm birth, but the risk was greatly diminished and no longer significant following control for concurrent tobacco smoking.³ A second review noted increased risks of several maternal and neonatal outcomes, but not preterm birth, but this review was unable to control for tobacco or other substance use.⁴ A recent publication, based on a large Canadian administrative database in which marijuana use was self-reported, found a significant increase in preterm birth among users.⁵ Another report, based on a cohort of young pregnant women in which use was assessed by self-report and urine toxicology, noted an increased risk of a composite neonatal outcome but no significant increase in spontaneous preterm birth.⁶

A limitation of many previous studies is reliance solely on self-report to assess marijuana use. We have previously shown that only 51% of women who used marijuana during pregnancy reported use, and misclassification may not be random: those who reported use tended to be less frequent users or users who have quit.⁷ Similar results have been reported more recently.^6,8 A second limitation is inconsistent control for other substance use, particularly tobacco or cocaine. A third limitation is that very few studies have differentiated indicated from spontaneous preterm birth.^6,9 In the present report, we address these concerns by employing multiple sources to assess substance use in pregnancy: self-report, clinical recognition in the prenatal record, and toxicological analysis of urine, and by subdividing preterm birth into spontaneous and indicated initiation.

Materials and Methods

This study includes women enrolled during pregnancy in the Ohio Perinatal Research Network Perinatal Research Repository (PRR).¹⁰ The PRR recruited women at several of the antenatal clinics at The Ohio State University Wexner Medical Center (OSUWMC) beginning in 2010. Many of the women were recruited from the prematurity clinic, which mainly received referrals of women who had a preterm infant in the past. The purpose of the repository was to serve as a research resource for affiliated investigators; studying the effect of substance use was not an explicit objective of the PRR at the time women were enrolled. Eligibility criteria for PRR recruitment in pregnancy included pregnancy in the first or second trimester, age 16 to 50 years, ability to communicate in English (with or without hospital translation services), and intention to deliver at OSUWMC. The PRR was approved by the institutional review board (IRB) at Nationwide Children’s Hospital, and a reliance agreement was executed with the IRB at OSUWMC. Eligible women were approached to provide written informed consent.

Women who consented completed an intake questionnaire covering a wide variety of medical, demographic, and socio-economic domains; the questionnaire included items on use of tobacco, marijuana, and other drugs of abuse to date during the current pregnancy. Women also provided urine and blood samples for unspecified future use at enrollment and approximately once in each subsequent trimester; the samples were archived at −80°C. They completed questionnaires regarding perceived stress,¹¹ depressive symptoms,¹² trait anxiety,¹³ sleep quality,¹⁴ and perceived everyday discrimination¹⁵ at enrollment and approximately once in each subsequent trimester. At the conclusion of the pregnancy, the obstetrical record was abstracted to a precoded form by an obstetrical research nurse; the form included specific items about recognized use of various illicit drugs, including marijuana. The neonate’s nursery record was also abstracted by trained personnel. The exact text of the marijuana question in the maternal questionnaire and record abstraction form are provided in Supplementary Material S1 (available online version only).

The PRR consent included an option indicating willingness to be contacted for unspecified future research. Therefore, the PRR served as the foundation for the ongoing Lifestyle and Early Achievement in Families (LEAF) study, the purpose of which is to evaluate associations between prenatal health and lifestyle factors, including substance use, and child development. Women in the PRR were eligible to participate in LEAF if they were willing to be contacted about future research and their child would be 42 to 95 months (3.5–7 years) during the anticipated course of LEAF. These LEAF-eligible women comprised the population for this report.

In preparation for LEAF, all archived urine samples from potentially eligible women were assayed for 11-nor-carboxy-Δ⁹-tetrahydrocannabinol (Δ9-THC-COOH)—the primary urine THC metabolite—by gas chromatography/tandem mass spectrometry following hydrolysis.^16–19 Women were considered to have used marijuana if they indicated use during pregnancy on their intake questionnaire, if use was noted on the obstetrical record abstraction, or if any urine specimen had a Δ9-THC-COOH concentration of >15 ng/mL; the concentration considered to represent active use when employing mass spectrometry.²⁰ During the time these women were pregnant, marijuana use was illegal in Ohio although since 1975 simple possession of up to 100 g was considered a “minor misdemeanor” with a maximum fine of $150.²¹ Further information on type of marijuana and method of administration was not collected.

The urine samples were also assayed for 16 additional substances, including several different opiates, benzoylecgo-nine (cocaine), and cotinine, by either liquid chromatography/mass spectrometry or liquid chromatography/tandem mass spectrometry. Assay results for these substances were reported only as positive or negative, with a cutoff of 2.5 ng/mL for positivity. We based tobacco use only on self-report and record abstraction because the binary cotinine assay, with a cutoff of 2.5 ng/mL, was too sensitive to reliably distinguish active from secondhand tobacco smoke exposure.^22,23 We have previously demonstrated that in prospective pregnancy cohorts not specifically focused on tobacco use, self-report is sufficiently accurate for classifying women as active tobacco users.^24,25 For cocaine, positivity was defined as self-report, notation in the obstetrical record, or at least one positive urine sample. Since opiates might be prescribed during the peripartum hospitalization, we did not consider as positive women whose only evidence of use was from a urine sample obtained no more than 2 days before delivery that contained an opiate that might be used clinically for short-term pain relief. Beyond that, our data collection could not distinguish legal opiate use by prescription from illicitly obtained use of legal drugs. LEAF and the toxicological analyses were approved as protocols separate from the PRR by the Nationwide Children’s Hospital IRB, and the latter was granted a waiver of consent.

Gestational age at delivery was based on last menstrual period, sonography, clinical estimate on admission to labor and delivery, and for in vitro fertilization, date of embryo transfer, utilizing the algorithm developed by the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network.²⁶ Preterm birth was defined as birth from 140 to 258 days’ (20 ^0/7–36 ^6/7 weeks) gestation. The initiating event of preterm birth was determined by the research nurse completing the abstraction. Preterm birth was spontaneous if the initiating event was labor or preterm PROM, regardless of the subsequent need for augmentation/induction of labor or delivery route; and indicated if the initiating event was induction or Cesarean in the absence of labor or preterm PROM.

Categorical variables were compared with the Chi-square test and continuous variables with the t-test. Relative risks were calculated by modified Poisson regression,²⁷ and time-to-delivery by proportional hazards (Cox) models. In the proportional hazards models, women delivering at term were censored at 37 completed weeks’ gestation; additionally, in the analysis of time-to-spontaneous preterm birth, women with indicated preterm birth were censored at the time of delivery, and in the analysis of indicated preterm birth, women with spontaneous preterm birth were censored at the time of delivery.²⁸ Since women who enrolled in the study after 20 weeks would not have been observed to be at risk of preterm birth for the entire time period, staggered entry was addressed by allowing women to enter the Cox model at the earliest date among the gestational age at completion of the questionnaire or the blood draw. The proportional hazards assumption was tested by including an interaction between marijuana use and log (gestational age), and found to be met in all models. Since women might have more than one pregnancy in the study and several of the questionnaires were repeated during each pregnancy, nonindependence of observations was addressed by nesting pregnancies within women utilizing generalized estimating equations for relative risks and robust methods for HRs. The depression, stress, anxiety, sleep, and discrimination questionnaires were not administered during the first year of PRR operation. In addition, clinic activities occasionally did not provide the women sufficient time to complete all questionnaires. Multiple imputation, by SAS PROC MI, was utilized to address missing data; imputed results were combined with PROC MIANALYZE.

Results

There were 497 pregnancies enrolled in the PRR with deliveries during the eligible range of dates such that the children would be 3.5 to 7 years old during the anticipated duration of the LEAF study; the mothers agreed to allow future contact for 389 of the children. Upon recontact for LEAF, one mother requested that her data and specimens be purged from the repository, leaving 388 children. Three children were born at <20 completed weeks’ gestation and were excluded from this analysis, as were 22 children who were twins. The 363 remaining singleton pregnancies comprise the population for this report; gestational age at delivery was unknown in 20 and was imputed as noted in the methods.

There were 119 (33%) users and 244 (67%) nonusers of marijuana. Table 1 provides the sources of positivity. If positivity is defined as recognition of use by any means, then the sensitivities for questionnaire, obstetrical record, and urine toxicology were 32% (38/119), 46% (55/119), and 85% (101/119), respectively. Of note, 39% of all users were identified only by urine toxicology.

Table 1.

Source of detection of marijuana use

Source	Number	Percent of all women (%)	Percent of all positives (%)
None	244	67	—
Any positive	119	33	100
Questionnaire only	6	2	5
Obstetrical record only	8	2	7
Urine only	46	13	39
Questionnaire and urine	12	3	10
Questionnaire and record	4	1	3
Record and urine	27	7	23
All three	16	4	13

Sensitivity of questionnaire = (6 + 12 + 4 + 16)/119 = 32%.

Sensitivity of obstetrical record = (8 + 4 + 27 + 16)/119 = 46%.

Sensitivity of urine = (46 + 12 + 27 + 16)/119 = 85%.

The associations between maternal characteristics and marijuana use are presented in Table 2. Marijuana use was more common among women who were currently unmarried, were African-American, had an unplanned pregnancy, and did not complete college. Use was more common among women who smoked tobacco, used cocaine, and drank alcohol although the latter was not significant; opiate exposure was similar between users and nonusers. Users reported higher perceived stress than nonusers. Users had urine sampled more times than nonusers. There were three women with four samples and one with five; this represents oversights as the protocol required a sample in each trimester; seven women (one user and six nonusers) had no urine samples. Users and nonusers enrolled at similar gestational ages. The number of pregnancies with missing data on demographic, medical, or substance use factors ranged from 0 (race, opiates, and cocaine) to 42 (homelessness); for the other questionnaires, which were not administered in the first year of repository operation, it ranged from 69 (everyday discrimination) to 94 (sleep).

Table 2.

Marijuana use by maternal characteristics

Characteristic	na (%)		p-Value
	Marijuana-positive	Marijuana-negative
Marital status			<0.001
Currently married	5 (4%)	61 (28%)
Never married	77 (72%)	102 (47%)
Other	25 (23%)	54 (25%)
Race			0.003
African-American	88 (74%)	141 (58%)
Non-African-American	31 (26%)	103 (42%)
Smoked tobacco during pregnancy			<0.001
No	42 (36%)	164 (68%)
Yes	49 (64%)	77 (32%)
Completed education			0.06
Less than high school	29 (27%)	46 (21%)
High school	40 (37%)	79 (35%)
Some college	36 (33%)	73 (33%)
College	3 (3%)	25 (11%)
Homeless within the past year			0.45
No	90 (87%)	194 (89%)
Yes	14 (13%)	23 (11%)
Physical abuse within the past year			0.31
No	93 (89%)	205 (93%)
Yes	11 (11%)	16 (7%)
Planned pregnancy			0.04
No	87 (84%)	159 (73%)
Yes	17 (16%)	58 (27%)
Alcohol use during pregnancy			0.11
No	86 (74%)	194 (82%)
Yes	30 (26%)	44 (18%)
Any opiate exposure			0.25
No	23 (77%)	221 (66%)
Yes	7 (23%)	112 (34%)
Any cocaine exposure			0.001
No	99 (83%)	229 (94%)
Yes	20 (17%)	15 (6%)
Parity			0.41
0	11 (9%)	29 (12%)
1+	107 (91%)	207 (88%)
Number of urine samples collected			0.01
1	32 (27%)	82 (34%)
2	45 (38%)	108 (45%)
3+	41 (35%)	48 (20%)
Mean (SD) gestation at entry (wk/d)	152/7 (44.1)	163/7 (51.2)	0.16
Mean (SD) height (cm)	163.1 (7.0)	163.1 (7.9)	0.95
Mean (SD) prepregnant weight (kg)	74.9 (22.6)	78.8 (22.8)	0.15
Mean (SD) age (y)	25.8 (5.3)	26.5 (5.3)	0.28
Mean total score (SD), CES-D	15.4 (10.9)	14.2 (10.8)	0.55b
Mean total score (SD), PSS	16.8 (7.4)	14.6 (7.8)	0.02b
Mean total score (SD), STAI	38.4 (11.0)	36.3 (10.1)	0.24b
Mean total score (SD), EDS	16.0 (8.9)	15.1 (8.6)	0.21b
Mean total score (SD), PSQI	7.5 (3.9)	7.7 (4.1)	0.74b

Abbreviations: CES-D, Center for Epidemiologic Studies-Depression Scale; EDS, Everyday Discrimination Scale; PSQI, Pittsburgh Sleep Quality Inventory; PSS, Perceived Stress Scale; STAI, Spielberger State Trait Inventory (only trait was asked).

^aNumbers in the table are unimputed.

^bp-Value from simple linear regression under generalized estimating equations to account for multiple assessments during the same pregnancy.

Mean (standard deviation) gestational ages at first, second, and third sample collection were 17 weeks, 1 day (56, n = 342), 24 weeks, 2 days (46, n = 242), and 29 weeks, 2 days (20, n = 90), respectively; 140 (39%) of the women had a sample in the first trimester. The percent of urine samples that were positive declined with advancing gestation; 37% of first, 26% of second, and 18% of third trimester samples were positive (p-value for trend < 0.001). These results were unchanged when women were stratified by the total number of samples they contributed, which means that the decrease in positive samples over pregnancy is not a simple artifact of nonusers having longer pregnancies and therefore more samples. In evaluating sequential samples from women, those who were positive at an earlier sample were positive in a subsequent sample approximately one-half to two-thirds of the time, while those who were negative were subsequently positive in <5% of samples. This further suggests that women tended to decrease use during pregnancy and were unlikely to start or resume use once they stopped.

Factors associated with preterm birth are presented in Table 3. The frequency of preterm birth was similar for both users and nonusers of marijuana (36.0 vs. 34.9%, p = 0.85). Although all preterm birth was similar between users and nonusers, a greater fraction of users than nonusers experienced spontaneous preterm birth (29.1 vs. 20.1%, p = 0.22), but users were less likely than nonusers to experience an indicated preterm birth (4.6 vs. 12.1%, p = 0.05); complete records were not always available to allow classification of the initiating event of preterm delivery. Interestingly, in 13 indicated preterm pregnancies to nonusers, the indication was a hypertensive disorder of pregnancy, while no pregnancies among the users had this indication for preterm delivery. Women delivering preterm enrolled in the study on average 18 days later than women delivering at term (p = 0.003). In addition, the Everyday Discrimination Scale was significantly lower, as was the number of samples collected per woman among women delivering preterm as compared with those delivering at term. The latter result suggests that shorter gestation afforded less opportunity to collect urine samples. Among tobacco nonsmokers preterm birth was somewhat more common among marijuana users than nonusers (43 vs. 32%, p = 0.21), while among tobacco smokers, preterm birth was somewhat less common among users than nonusers (32 vs. 41%, p = 0.27). However, the association of marijuana and preterm birth did not differ significantly by smoking status (p-value for interaction 0.09). Table 4 presents the unadjusted and adjusted associations between marijuana use and all preterm births as well as its subtypes spontaneous and indicated preterm birth. Risk ratios for total and spontaneous preterm birth were adjusted for all of the characteristics listed in Table 2 except that total number of samples was standardized to the total time the women were in the study to avoid the “reverse causation” of total time under study being a result, not cause of duration of pregnancy. In addition, gestation at registration could not be controlled in the models for total and spontaneous preterm birth because the models would not converge. We addressed this by running models for these outcomes restricted to women who enrolled at <20 weeks’ gestation, with results similar to models including all women regardless of gestation at enrollment. The relative risk for indicated preterm birth could not be adjusted for opiate use because the model would not converge. Overall, marijuana users were at only slightly, and not significantly increased risk of total preterm birth as compared with nonusers, and adjustment had minimal impact on the risk ratios. The unadjusted risk ratio was 1.06 (95% confidence interval [CI]: 0.76–1.47). Although users were at somewhat increased risk of spontaneous preterm birth (risk ratio: 1.32, 95% CI: 0.89–1.96) and reduced risk of indicated preterm birth (risk ratio: 0.52, 95% CI: 0.22–1.23) as compared with nonusers, none of the differences approached statistical significance, either before or after adjustment, and some estimates were imprecise. As a sensitivity analysis, we evaluated the unadjusted and adjusted associations for total preterm birth by marijuana use, after deleting, rather than imputing, missing gestational ages. Results were not substantially different (unadjusted risk ratio: 1.03, 0.76–1.39; adjusted risk ratio: 1.04; 95% CI: 0.71–1.50).

Table 3.

Characteristic’s associated with preterm (<37 weeks’ gestation) birth

Characteristic	na (%)		p-Value
	Term	Preterm
Marijuana use			0.85
No	149 (67%)	80 (66%)
Yes	73 (33%)	41 (34%)
Marital status			0.49
Currently married	43 (22%)	19 (18%)
Never married	110 (55%)	60 (55%)
Other	46 (23%)	29 (27%)
Race			0.95
African-American	144 (65%)	79 (65%)
Non-African-American	78 (35%)	42 (35%)
Smoked tobacco during pregnancy			0.75
No	130 (59%)	69 (57%)
Yes	91 (41%)	52 (43%)
Completed education			0.20
Less than high school	53 (26%)	20 (18%)
High school	64 (32%)	47 (42%)
Some college	68 (33%)	34 (31%)
College	18 (9%)	10 (9%)
Homeless within the past year			0.86
No	177 (89%)	95 (90%)
Yes	22 (11%)	11 (10%)
Physical abuse within the past year			0.98
No	183 (92%)	102 (92%)
Yes	16 (9%)	9 (8%)
Planned pregnancy			0.48
No	152 (78%)	80 (74%)
Yes	44 (22%)	28 (26%)
Alcohol use during pregnancy			0.68
No	174 (80%)	98 (82%)
Yes	44 (20%)	22 (18%)
Any opiate exposure			0.64
No	205 (92%)	110 (91%)
Yes	17 (8%)	11 (9%)
Any cocaine exposure			0.70
No	201 (91%)	108 (89%)
Yes	21 (9%)	13 (11%)
Parity			0.67
0	24 (11%)	15 (12%)
1+	195 (89%)	105 (88%)
Number of urine samples collected			0.001
1	54 (25%)	46 (38%)
2	95 (44%)	54 (45%)
3+	67 (31%)	20 (17%)
Mean (SD) gestation at entry (wk/d)	151/7 (42)	171/7 (58)	0.003
Mean (SD) height (cm)	163.0 (7.7)	163.2 (7.7)	0.87
Mean (SD) prepregnant weight (kg)	78.6 (23.4)	75.7 (21.8)	0.28
Mean (SD) age (y)	26.2 (5.3)	26.2 (5.2)	0.96
Mean total score (SD), CES-D	14.4 (11.1)	15.0 (9.7)	0.85b
Mean total score (SD), PSS	15.2 (7.7)	16.1 (7.6)	0.63b
Mean total score (SD), STAI	36.7 (10.9)	38.0 (9.7)	0.77b
Mean total score (SD), EDS	16.1 (9.4)	14.4 (7.4)	0.01b
Mean total score (SD), PSQI	7.4 (4.0)	8.1 (3.9)	0.20b

Abbreviations: CES-D, Center for Epidemiologic Studies-Depression Scale; EDS, Everyday Discrimination Scale; PSQI, Pittsburgh Sleep Quality Inventory; PSS, Perceived Stress Scale; SD, standard deviation; STAI, Spielberger State Trait Inventory (only trait was asked).

^aNumbers in the table are unimputed.

^bp-value from simple logistic regression under generalized estimating equations to account for multiple assessments during the same pregnancy.

Table 4.

Unadjusted and adjusted risk ratios for preterm birth and hazard ratios for time to delivery for marijuana users versus nonusers

Outcome	Unadjusted risk ratio	95% confidence limits	Adjusted risk ratio	95% confidence limits
All preterm births	1.06	(0.76–1.47)	1.04	(0.72–1.50)
Spontaneous preterm birth	1.32	(0.89–1.96)	1.21	(0.76–1.94)
Indicated preterm birth	0.52	(0.22–1.23)	0.75	(0.29–2.15)
	Unadjusted hazard ratioa	95% confidence limits	Adjusted hazard ratio	95% confidence limits
Time to all preterm delivery	1.26	(0.84–2.00)	1.32	(0.80–2.07)
Time to spontaneous preterm delivery	1.77	(1.06–2.93)	2.16	(1.16–4.02)
Time to indicated preterm delivery	0.69	(0.33–1.43)	0.58	(0.23–1.46)

^aHazard ratios <1 indicate longer time to delivery, and hazard ratios >1 indicate shorter time to delivery.

Table 4 also includes HRs for time to delivery. Ratios >1 indicate increased risk of delivery at any preterm gestational age, while ratios <1 indicate reduced risk. The time-to-delivery models supported inclusion of all the characteristics in Table 3. The HRs for total preterm delivery were modestly, but not significantly elevated among marijuana users as compared with nonusers either before or after adjustment. However, the unadjusted HR for spontaneous preterm birth was significantly elevated among users; adjustment for the characteristics in Table 3 had minimal impact on the HR. This was attributable to a nonsignificantly increased risk of spontaneous birth at <28 weeks among users versus nonusers (4.4 vs. 1.8%, p = 0.16). The unadjusted and adjusted HRs for indicated preterm birth were reduced among users as compared with nonusers although neither was statistically significant.

Discussion

Our study, which included multimodal assessment including urine toxicology, of marijuana use during pregnancy found that women who used marijuana were at neither clinically nor significantly increased risk of preterm birth in general. However, there was a suggestion that marijuana use was associated with increased risk of spontaneous and decreased risk of indicated preterm birth.

There have been numerous studies of marijuana use as a risk factor for preterm birth. The literature was summarized in two recent systematic reviews and meta-analyses. One reported a null association with preterm birth,⁴ while the other reported an increased risk of preterm birth that was substantially diminished and no longer statistically significant following adjustment for concurrent tobacco use.³ Since publication of those reviews, there have been several additional reports with variable results. One⁵ reported a substantially increased relative risk of preterm birth with self-reported use. A second²⁹ found no increased risk of preterm birth with marijuana use alone, but a relative risk of 2.56 for combined marijuana and tobacco use. A third, based on administrative data found nearly doubled risk of preterm birth in the presence of a diagnostic code for marijuana use.³⁰ None of these studies employed toxicology to identify use, and found that 1.4, 0.9, and 0.6% of women were users, respectively. This low prevalence of use suggests considerable underreporting, and indeed, studies that employed routine toxicological analysis have consistently noted that self-report understates actual use by a considerable amount.^6–8 Recent studies employing a biomarker for marijuana have not noted increased risk of preterm birth.^9,31 In addition to tobacco, concurrent abuse by other substances, such as cocaine and opiates, is commonly observed, but few studies have assessed this systematically nor controlled for it.

To the best of our knowledge, the shorter time to spontaneous preterm birth coupled with longer time to indicated preterm birth has not been reported previously. This result appeared to be due to an increase in very early deliveries among users, even though we observed little increase in total preterm birth. The mechanism for it is not obvious, and therefore, it requires confirmation in other studies.

Our study has numerous strengths: we evaluated a defined group of women, recruited into a general repository that was not focused on drug use, thereby diminishing the incentive to misreport or to decline participation among drug users. Marijuana use was assessed by questionnaire, record review, and universal urine toxicology. Marijuana use was very common in this population and was not associated with opioid use; although marijuana users were more likely than nonusers to use cocaine, the vast majority of users did not use cocaine. We employed the same multimodal assessment to evaluate use and abuse of other drugs. We collected detailed information on stress, depressive symptoms, anxiety, discrimination, and sleep, utilizing well-accepted instruments. We were thus able to control for a wide variety of potentially confounding factors.

Our study also has limitations. Although marijuana use in our population was common, as was preterm birth, the total number of women was relatively small, resulting in confidence limits for some of our results that are compatible with a clinically relevant adverse effect. Our assessment of marijuana and other substances allowed us to identify users and nonusers, but not to assess quantity or timing of use during pregnancy. In addition, we did not assay measures of urine concentration such as specific gravity and creatinine. However, only 30 of the 721 urine samples analyzed had Δ9-THC-COOH concentrations from 50 to 200% of the 15 ng/mL cutoff that defined use, suggesting that correction for urine concentration would not have impacted our definition of use substantially. We did not ask about nor did we assay for synthetic cannabinoids. Our IRB approval did not allow us to count or collect data about potentially eligible women, who were not approached or who declined to participate in the repository nor could we assay urine samples for women in the PRR who refused to allow future contact. Nevertheless, because our data and specimen collection was prospective with respect to pregnancy outcome, these omissions would have to be simultaneously associated with both marijuana use and pregnancy outcome to bias our results. Finally, although OSUWMC is a referral center and approximately 20% of births there during the study period occurred at <37 weeks, the rate of preterm birth in our study population is still elevated. This reflects the fact that many women were recruited in the prematurity clinic at OSUWMC. Therefore, our results may not be generalizable to more typical settings.

In conclusion, we found that among a prospective cohort of women in which substance use was assessed by questionnaire, record abstraction, and urine toxicological analysis, marijuana use was not associated with an increased risk of preterm birth, although there was a suggestion of an increased risk of early spontaneous preterm birth and a reduction in indicated preterm birth. Given the uncertainty of a possible mechanism for an increase in spontaneous and reduction in indicated preterm birth, the imprecision of our results, and the possibility for confounding, further research is needed to confirm our results. The clinical implication of our results is that among women already at high risk of preterm birth, marijuana use per se may not elevate the risk further. However, even if this conclusion is correct, it should not be assumed that marijuana use is harmless. In particular, use during pregnancy has been associated with subtle, but important long-term deficits in higher cognitive function among exposed fetuses.^32,33 Therefore, prudence dictates that ACOG’s recommendation that women who are pregnant or contemplating pregnancy be encouraged to discontinue use¹ should be followed.

Key Points.

• Marijuana was not associated with risk of all preterm birth.

• Marijuana was not associated with reduced time to delivery.

• However, users had reduced time to spontaneous preterm birth.