Factors Associated with Vaginal Lactobacillus Predominance Among African American Women Early in Pregnancy

Abstract

Background:

Vaginal Lactobacillus is considered protective of some adverse reproductive health outcomes, including preterm birth. However, factors that increase or decrease the likelihood of harboring Lactobacillus in the vaginal microbiome remain largely unknown. In this study, we sought to identify risk and protective factors associated with vaginal Lactobacillus predominance within a cohort of pregnant African American women.

Materials and Methods:

Vaginal microbiome samples were self-collected by African American women (N = 436) during their 8–14th week of pregnancy. Sociodemographic information and measures of health behaviors, including substance use, antibiotic exposure, sexual practices, frequency of vaginal intercourse, and the use of vaginal products, were collected through participant self-report. The V3–V4 region of the 16S rRNA gene was targeted for amplification and sequencing using Illumina HiSeq, with bacterial taxonomy assigned using the PECAN classifier. Univariate and a series of multivariate logistic regression models identified factors predictive of diverse vaginal microbiota or Lactobacillus predominance.

Results:

Participants who used marijuana in the past 30 days (aOR 1.80, 95% CI 1.08–2.98) were more likely to have diverse non-Lactobacillus-predominant vaginal microbiota, as were women not living with their partners (aOR 1.90, 95% CI 1.20–3.01). Cohabitating or marijuana usage were not associated with type of Lactobacillus (non-iners Lactobacillus vs. Lactobacillus iners) predominance (aOR 1.11, 95% CI 0.52–2.38 and aOR 0.56, 95% CI 0.21–1.47, respectively).

Conclusions:

Living with a partner is conducive to vaginal Lactobacillus predominance. As such, cohabitation may be in important covariate to consider in vaginal microbiome studies.

Introduction

Numerous studies report that more diverse, non-Lactobacillus-predominant vaginal microbiomes may be associated with higher risk for complications during pregnancy, including preterm birth and greater risk for sexually transmitted infections.^1–4 Specific Lactobacillus spp.,^1,5,6 notably Lactobacillus crispatus, assists in maintaining a balanced vaginal ecosystem by the production of lactic acid, antimicrobial peptides, and bactericidal compounds that are protective for reproductive health.^7,8 These microbial factors help to protect the reproductive tract from pathogenic microbes that may result in inflammation and increase reproductive health risks.^9,10 In contrast, having a diverse non-Lactobacillus-predominant vaginal microbiome has been associated with increased risk of bacterial vaginosis (BV), vaginal infections, and preterm birth among some cohorts of women.^1,9,11–13 While many studies have evaluated associations between vaginal microbiome composition and reproductive health outcomes, very few have examined factors that influence or are associated with the composition of the vaginal microbiome. For example, with the exception of research demonstrating that women who use products marketed for “vaginal hygiene” (e.g., douches, sexual lubricants, and antiseptic agents) are more likely to experience alterations in vaginal fluids resulting in increased vaginal pH, decreased levels of Lactobacillus spp., and overgrowth of pathogenic microbes,^14,15 very few other factors contributing to low levels of protective Lactobacillus predominance have been identified.

Understanding which factors hinder vaginal Lactobacillus is essential, given that preterm birth and other adverse reproductive health outcomes, such as vaginal infections, remain persistent among women with lower prevalence of vaginal Lactobacillus,^1,11–13 despite decades of research. Furthermore, some Lactobacillus species, particularly L. crispatus, may be more protective of reproductive health than other common vaginal Lactobacillus species like Lactobacillus iners.¹⁶ L. crispatus has been associated with greater stability of the vaginal microbiome, whereas L. iners has been associated with transitions between different vaginal microbiome communities.¹⁷ However, vaginal L. iners is more robust to changes to the vaginal environment, such as menses, than L. crispatus, suggesting that sensitivity to environmental influences varies across Lactobacillus species. To date, it is unclear which other host factors may contribute to differences in Lactobacillus spp. colonization among women. For example, specific social and behavioral factors that have the potential to alter the vaginal microbiome among women who remain underexplored include socioeconomic status, cohabitating with a partner, and sexual practices. Given that a decreased prevalence of any Lactobacillus early in pregnancy is associated with higher risk for preterm birth, it is essential that such relationships be explored to provide guidance for future intervention. Therefore, in this study, we sought to identify whether key social and behavioral factors may influence the vaginal microbiome, and contribute to higher diversity or decreased vaginal Lactobacillus early in pregnancy.³

Thus, the purpose of this study was to conduct a secondary analysis, within a population of African American women, to investigate whether vaginal microbiome Lactobacillus predominance in early pregnancy (8–14 weeks) was associated with measures of socioeconomic, behavioral, sexual, or vaginal hygiene factors. It is clear the vaginal microbial community composition and its homeostasis are complex. Few studies are inclusive of a significant number of African American women, and even fewer look within this population to identify significant risk and protective factors as a recommended first step by health disparity experts.¹⁸ Given the disproportionate occurrence of preterm birth and other birth complications among African American women, this first step is essential.

Materials and Methods

Participants included in this report (N = 436) are a subset of women participating in an ongoing longitudinal study being conducted, with Institutional Review Board approval (Protocol No. 68441), at Emory University, which evaluates biobehavioral factors associated with the microbiome and preterm birth among African American women.¹⁹ Women were recruited from one private and one public facility in metro Atlanta and provided written consent to participate in the study. Inclusion criteria included the following: (1) self-report of Black/African American and born in the United States; (2) singleton pregnancy with initial contact between 8 and 14 weeks gestation as per clinical record and/or ultrasound; (3) comprehension of written and spoken English; (4) between 18 and 40 years of age; and (5) no history of chronic illness or medication use, verified through medical record. Overall, 436 women from this ongoing study had vaginal microbiome samples collected between 8 and 14 weeks of gestation that pass quality control measures (detailed below) and were eligible for these analyses.

Sociodemographic survey

Detailed data collection procedures have been described elsewhere.¹⁹ In this study, we will briefly summarize the metrics used for these analyses, all of which were collected between 8 and 14 weeks of gestation. Maternal self-report and prenatal administrative record review were used to ascertain maternal age, cohabitating with their partner, years of education (collected as a four-level variable: less than high school, high school or general education development test [GED], some college, or college graduate), and prenatal health insurance type (categorized as Medicaid, pregnancy Medicaid, or private insurance). In Georgia, women are eligible for Medicaid during pregnancy if their total household income is <200% of the federal poverty level.

Health survey

Upon enrollment in the study, women completed a health survey to ascertain within the last month diagnoses, medications, sexual encounters (e.g., number and type of intercourse), hygiene self-care practices (i.e., douching and feminine sprays/wipes), and substance use (e.g., tobacco, marijuana, and alcohol). Items included within these analyses occurred within 1 month of sample collection, as per the participant's self-report. Medical chart abstraction was completed using a standardized chart abstraction tool to determine gestational age at sample collection, diagnoses of any infection, and/or prescription of systemic or oral antibiotics by comparing the date of these occurrences to the estimated date of confinement based on last menstrual period and/or ultrasound before 14 weeks of gestation according to standard clinical criteria. Chart abstraction was also used to ascertain obstetrical history, specifically whether the woman had any previous births.

Vaginal swabs

Participants were provided verbal and pictorial instructions to obtain self-collection of vaginal swabs using methods and protocols consistent with the Human Microbiome Project.²⁰ Samples were self-collected^21,22 by participants using a Sterile Catch-All™ Sample Collection Swab (Epicentre Biotechnologies, Madison, WI); after swabbing, the swab was immediately handed to the study coordinator and placed in MoBio bead tubes (MoBio Laboratories, Inc.) that were frozen upright on ice until transported to the laboratory, where they were stored at −80°C until DNA extraction. Studies support that vaginal self-collection swabs sample the same microbial diversity as physician-collected swabs of the mid-vagina and have high overall morphotype-specific validity compared with provider-collected swabs.²¹ Separate vaginal swabs, collected using a Bactiswab (Thermo Scientific, Waltham, MA) in the same manner as above, were transported to the Emory Clinical Microbiology laboratory for Nugent criteria scoring for evaluation of BV.

Microbiome data

16S rRNA gene amplification was used to sequence the V3–V4 region using Illumina HiSeq at the University of Maryland, details have been published previously.¹⁰ Data quality control and raw data processing were completed using a QIIME-dependent script and the DADA2 workflow.^23,24 Closed-reference operational taxonomic unit picking was used to classify amplicon sequence variants at the genus or species level using PECAN, University of Maryland's in-house custom database for classifying taxa in the vaginal microbiome.²⁵ Microbial community state types (CSTs) were assigned using hierarchical clustering with Jensen-Shannon divergence and Ward linkage.^5,26 CSTs have different signature microbiomes that are dominant in a given CST: CST I with L. crispatus, CST II with Lactobacillus gasseri, CST III with L. iners, CST IV is diverse with no specific dominant microbe, and CST V with Lactobacillus jensenii. For our population, non-iners Lactobacillus CSTs are not as common and all are thought to be somewhat protective of reproductive health; therefore, we collapsed non-iners Lactobacillus into one group (non-iners Lactobacillus). For analysis, we considered the following CST groupings: CST IV (Diverse) versus CST I, II, III, and V (Lactobacillus dominated) and CST III (L. iners dominated) versus CST I, II, and V (non-iners Lactobacillus dominated). Parameters evaluated included vaginal and receptive oral sex within 30 days, vaginal product use (any douche, spray, or cream within 30 days), vaginal infections, use of antibiotics before sampling, parity, frequency of vaginal sex, cohabitation, tobacco, marijuana, and alcohol use.

Statistical analyses

All analyses were completed using the R statistical computing environment.²⁷ Welch's t-tests and chi square tests were completed to identify baseline differences of categorical variables for comparisons between groups (e.g., Lactobacillus vs. Diverse non-Lactobacillus predominance and L. iners vs. non-iners Lactobacillus). Specifically, we evaluated differences in sociodemographic, sexual, and vaginal hygiene practices between women with Lactobacillus versus diverse non-Lactobacillus-predominant vaginal microbiota and then looked for differences among women with L. iners versus other Lactobacillus predominance, since previous works suggest L. crispatus, L. jensenii, and L. gasseri predominance may be associated with better reproductive health outcomes.^9,16

We then completed a series of logistic regression models to evaluate if any social factors were associated with different types of vaginal microbiome composition. We tested three models: (1) Diverse (reference/outcome) versus Lactobacillus predominance, (2) Diverse (reference/outcome) versus non-iners Lactobacillus predominance, and (3) non-iners Lactobacillus (reference/outcome) versus L. iners predominance. Covariates in the three models include maternal age, education level, insurance type, cohabitation status, parity, first prenatal body mass index category, gestational age at sampling, antibiotic exposure before sampling, tobacco use, marijuana use, receptive oral and vaginal sex, and vaginal product usage based on previous literature and differences in baseline characteristics.

Results

A total of 436 participants with vaginal microbiome samples that passed quality control measures were included in these analyses, and are described in Table 1. The mean age of participants was 25 years and most participants (78%) identified Medicaid as their prenatal insurance provider. Approximately half (47%) of the women in our sample were primiparas, with a similar percentage (49%) married to or cohabitating with their partner, and reported a high school education or less (54%). Most participants did not use tobacco (81%), marijuana (64%), or alcohol (90%). As expected, women with diverse microbiota, versus those with Lactobacillus predominance, had a higher total read count, as well as a higher pH, Nugent score. and Shannon diversity index (all p < 0.001, Table 2), while women with L. iners predominance were more likely to have Nugent scores classified as intermediate or BV than women with non-iners Lactobacillus predominance (p < 0.001).

Table 1.

Sociodemographic and Behavioral Factors by Lactobacillus Dominance

Characteristic	Diverse vs. any Lactobacillus			Lactobacillus iners vs. other Lactobacillus
	Diverse, N = 214	Any Lactobacillus, N = 222	p	L. iners, N = 151	Non-iners Lactobacillus, N = 71	p
Age, mean ± SD	24.4 ± 4.6	25.5 ± 5.0	0.016	25.1 ± 4.7	26.4 ± 5.4	0.107
Prior birth, n (%)	104 (49)	129 (58)	0.058	93 (62)	36 (51)	0.165
Not cohabiting, n (%)	129 (60)	94 (42)	<0.001	65 (43)	29 (41)	0.870
Pregnancy insurance, n (%)			<0.001			<0.001
Low-income Medicaid	78 (36)	80 (36)		66 (44)	14 (20)
Pregnancy Medicaid	105 (49)	78 (35)		51 (34)	27 (38)
Private	31 (15)	64 (29)		34 (23)	30 (42)
Education, n (%)			<0.001			<0.001
Less than high school	43 (20)	24 (11)		18 (12)	6 (8)
High school or GED	93 (43)	77 (35)		59 (39)	18 (25)
Some college	56 (26)	71 (32)		56 (37)	15 (21)
College graduate	22 (10)	50 (23)		18 (12)	32 (45)
First prenatal BMI, n (%)			0.734			0.100
Underweight	8 (4)	10 (5)		9 (6)	1 (1)
Healthy weight	82 (38)	93 (42)		60 (40)	33 (47)
Overweight	43 (20)	46 (21)		27 (18)	19 (27)
Obese	81 (38)	73 (33)		55 (36)	18 (25)
Current practices last 30 days only: yes n (%) vs. no/not reported
Tobacco use	48 (22)	34 (15)	0.063	28 (19)	6 (9)	0.077
Marijuana use	95 (44)	61 (28)	<0.001	48 (32)	13 (18)	0.040
Alcohol use	18 (8)	26 (12)	0.367	19 (13)	7 (10)	0.692
Vaginal sex	164 (77)	163 (73)	0.394	111 (74)	52 (73)	0.647
Receptive oral sex	86 (40)	77 (35)	0.258	52 (34)	25 (35)	0.990
Douche/sprays/creams	43 (20)	26 (12)	0.020	17 (11)	9 (13)	1
Parenteral or oral antibiotics	32 (15)	25 (11)	0.260	21 (14)	4 (6)	0.073
Parenteral or oral antibiotics (except nitrofurantoin)	24 (11)	21 (9)	0.637	17 (11)	4 (6)	0.225
Diagnosis of BV	15 (7)	11 (5)	0.421	8 (5)	3 (4)	1
Diagnosis of Chlamydia	9 (4)	9 (4)	1	7 (5)	2 (3)	0.722
Diagnosis of Gonorrhea	2 (1)	0 (0)	0.240	0	0	1
Diagnosis of Trichomoniasis	1 (<1)	4 (2)	0.373	4 (3)	0	0.309
Diagnosis of UTI	17 (8)	14 (6)	0.578	13 (9)	1 (1)	0.041

Bold values are statistically significant p < 0.05.

BMI, body mass index; BV, bacterial vaginosis; GED, general education development test; SD, standard deviation; UTI, urinary tract infection.

Table 2.

Vaginal Sample Characteristics by Lactobacillus Dominance

Characteristic	Diverse vs. any Lactobacillus			Lactobacillus iners vs. other Lactobacillus
	Diverse, N = 214	Lactobacillus, N = 222	p	iners, N = 151	Non-iners, N = 71	p
GA at sampling	11.0 ± 2.39	11.5 ± 2.44	0.055	11.5 ± 2.65	11.5 ± 1.92	0.998
Read count	40,385 ± 19,724	51,600 ± 26,972	<0.001	52,327 ± 26,624	50,053 ± 27,826	0.566
pH	4.75 ± 0.43	4.60 ± 0.34	<0.001	4.60 ± 0.33	4.61 ± 0.37	0.845
Nugent score	7.19 ± 2.49	2.50 ± 2.60	<0.001	2.85 ± 2.67	1.75 ± 2.29	0.002
Nugent categories, n (%)
Normal (1)	20 (9)	150 (68)	<0.001	93 (62)	57 (80)	0.010
Intermediate (2)	33 (15)	36 (16)		31 (21)	5 (7)
BV (3)	155 (72)	26 (12)		20 (13)	3 (8)
White blood cells, n (%)
None	102 (48)	106 (48)	0.816	66 (44)	40 (56)	0.020
Rare	54 (25)	53 (24)		41 (27)	12 (17)
Few	44 (21)	44 (20)		34 (23)	10 (14)
Moderate	5 (2)	8 (4)		3 (2)	5 (7)
Many	3 (1)	1 (<1)		0	1 (1)
CST, n (%)
I	0	50 (23)	<0.001	0	50 (70)	<0.001
II	0	8 (4)		0	8 (11)
III	0	151 (68)		151 (100)	0
IV	214 (100)	0		0	0
V	0	13 (6)		0	13 (18)
Chao	3.89 ± 0.46	3.89 ± 0.56	0.966	3.87 ± 0.53	3.95 ± 0.61	0.293
Shannon	1.79 ± 0.65	1.03 ± 0.79	<0.001	1.04 ± 0.80	1.02 ± 0.78	0.948

Bold values are statistically significant p < 0.05.

CST, community state type; GA, gestational age.

In bivariate analysis, women with Lactobacillus predominance were more likely to have private insurance and higher education, while women with diverse, non-Lactobacillus-predominant microbiota were more likely to not live with their partners, have smoked marijuana, and have used vaginal products in the past 30 days (Table 1). When contrasting women with L. iners versus non-iners Lactobacillus predominance in bivariate analysis, those with L. iners predominance were more likely to have smoked marijuana in the past 30 days and to have had a diagnosis of urinary tract infection.

When considering the influence of multiple factors using logistic regression models (Table 3), women who were not living with their partners were more likely to have diverse, non-Lactobacillus microbiota (aOR 1.90 95% CI 1.20–3.01) than any type of Lactobacillus predominance. Women who smoked marijuana were also more likely to have diverse, non-Lactobacillus-predominant vaginal microbiomes (aOR 1.80, 95% CI 1.08–2.98). This relationship was strongest when comparing only non-iners Lactobacillus predominance compared to diverse, non-Lactobacillus microbiota (aOR 2.50, 95% CI 1.06–5.90). Women who consumed alcohol within 30 days of sample collection were less likely to have diverse, non-Lactobacillus microbiota than non-iners Lactobacillus predominance. Women who were college graduates were less likely to have diverse, non-Lactobacillus microbiota (aOR 0.17, 95% CI 0.05–0.56) and more likely to have non-iners Lactobacillus versus L. iners predominance (aOR 6.32, 95% CI 1.91–20.35). Women with low-income Medicaid insurance were less likely to have non-iners Lactobacillus than L. iners predominance (aOR 0.38, 95% CI 0.26–0.90), although no significant difference by type of insurance was observed for non-Lactobacillus- versus Lactobacillus-predominant groups.

Table 3.

Logistic Regression Adjusted Odds Ratios of Binary Outcomes of Lactobacillus iners Predominance, Non-iners Lactobacillus Predominance, or Diverse Vaginal Microbiome Composition

Characteristic	Diverse (n = 214) vs. any Lactobacillus (n = 222)	Diverse (n = 214) vs. non-iners Lactobacillus (n = 71)	Non-iners Lactobacillus (n = 71) vs. L. iners (n = 151)
Maternal age	1.01 (0.95–1.06)	0.99 (0.91–1.07)	1.01 (0.93–1.10)
Education level
Less than high school	1.14 (0.58–2.25)	0.76 (0.24–2.36)	1.66 (0.48–5.73)
High school or GED	Ref.	Ref.	Ref.
Some college	0.75 (0.43–1.28)	0.75 (0.30–1.83)	1.04 (0.41–2.68)
College graduate	0.55 (0.25–1.25)	0.17 (0.05–0.56)	6.23 (1.91–20.35)
Insurance type
Low-income Medicaid	0.71 (0.43–1.61)	1.54 (0.66–3.62)	0.38 (0.26–0.90)
Pregnancy Medicaid	Ref.	Ref.	Ref.
Private	0.55 (0.28–1.10)	0.76 (0.29–2.03)	0.73 (0.26–2.04)
Parity	0.66 (0.41–1.07)	1.12 (0.53–2.35)	0.60 (0.28–1.30)
Gestational age at sampling	0.95 (0.87–1.04)	0.97 (0.85–1.12)	0.94 (0.81–1.09)
Not married or cohabiting	1.90 (1.20–3.01)	1.63 (0.79–3.36)	1.11 (0.52–2.38)
First prenatal body mass index
Underweight	0.43 (0.13–1.36)	0.80 (0.07–8.62)	0.57 (0.06–5.62)
Healthy weight	Ref.	Ref.	Ref.
Overweight	1.19 (0.66–2.15)	0.80 (0.33–1.90)	1.68 (0.69–4.10)
Obese	1.42 (0.85–2.38)	2.10 (0.92–4.80)	1.02 (0.44–2.36)
Tobacco use	1.20 (0.63–2.32)	2.15 (0.60–7.72)	0.64 (0.18–2.34)
Marijuana use	1.80 (1.08–2.98)	2.50 (1.06–5.90)	0.56 (0.21–1.47)
Alcohol use	0.47 (0.21–1.03)	0.23 (0.06–0.87)	0.99 (0.32–3.01)
Use of vaginal douche, sprays, or creams	1.54 (0.87–2.77)	1.31 (0.51–3.36)	2.03 (0.70–5.91)
Vaginal sex in last month	1.35 (0.76–2.39)	2.06 (0.89–4.80)	0.50 (0.20–1.27)
Oral sex (received) in last month	1.37 (0.85–2.22)	1.84 (0.85–4.01)	0.80 (0.36–1.77)
Parenteral or oral antibiotics in month before sampling	1.24 (0.64–2.39)	2.40 (0.70–8.23)	0.37 (0.0–1.36)

Bold values are statistically significant p < 0.05.

Discussion

To date, research evaluating the vaginal microbiome among pregnant women has largely focused on evaluating composition and taxa that may be associated with birth outcomes, particularly preterm birth. Various contradictory findings exist between studies, which may be correlated with differences in vaginal microbiome composition observed by race and geographic locations.^2,3,6 Moreover, previous works have reported that African American women are more likely to have a more diverse or L. iners-predominant vaginal microbiome profile that, when present in white women, is associated with preterm birth.^1,5,6 However, the link between a diverse non-Lactobacillus-predominant vaginal microbiome and preterm birth is not consistent among African American women since most women with a diverse non-Lactobacillus-predominant vaginal microbiome do not experience vaginal symptoms or poor pregnancy outcomes.^1,3,5 Conversely, Stout and others have reported that certain vaginal microbiome compositions early in pregnancy may initiate pathways that predispose African American women to preterm birth.^3,28 To try to identify factors that may contribute to these differences in microbiome composition, which others have reported among African American women, we evaluated factors that have not yet been reported in previous studies focused on the vaginal microbiome of pregnant African American women.

Roughly half of the women in our population (n = 222, 51%) harbored a microbiome dominated by various Lactobacillus spp., with L. iners being most common (n = 151, or 35%) among the entire cohort. Within this cohort, comprising entirely African American women, the factors most consistently associated with vaginal microbiome composition were socioeconomic status (i.e., education and type of insurance), cohabitating with a partner, and marijuana usage. Our previous work identified associations with socioeconomic status and more diverse vaginal microbiome composition, yet we had not previously assessed the influence of marijuana usage or cohabitation.²⁹ To date, research evaluating the influence of marijuana on the human microbiome is scant. One study did identify significant differences in the gut microbiome associated with marijuana use; however, the authors were unable to determine how much of the effect was due to differences in diet between those who used marijuana compared to those who did not.

Although the association between a diverse vaginal microbiome and marijuana usage in this study is intriguing, the underlying mechanisms for the association remain unclear. Research generally indicates primarily anti-inflammatory actions associated with the use of cannabis.³⁰ However, the immune effects of cannabinoids may vary depending on concentration, and in some cases, have been reported to increase proinflammatory cytokine production.³¹ Moreover, women using marijuana were more likely to smoke [χ² (1, N = 436) = 65.78, p < 0.001] and drink alcohol [χ²² (1, N = 436) = 4.64, p = 0.031], perhaps indicating broader associations among these variables and increased likelihood of having diverse vaginal microbiota. Although women with L. iners are more likely to smoke marijuana than those with other protective Lactobacillus spp., the relationship was not maintained in multivariate analysis, suggesting other factors may more strongly influence why women harbor one type of Lactobacillus over another. Given the legalization of marijuana in many regions across the United States, increased evaluation of the influence of marijuana on health outcomes may warrant additional study.

Post hoc analysis revealed that women living with a partner were more likely to engage in vaginal sex than those not living with a partner, χ²² (1, N = 436) = 10.02, p = 0.002, and significantly less likely to use vaginal hygiene products, χ² (1, N = 436) = 5.56, p = 0.018. Few of the previous studies have reported data regarding sexual practices, and none, to our knowledge, has reported on cohabitation status.^{2,3,5,32–35} Moreover, while several previous studies evaluating the vaginal microbiome have considered vaginal sex or number of sexual partners as contributors to vaginal dysbiosis,^1,26,36,37 none of these studies included evaluation of the microbiome in pregnant women in light of these variables, or whether women were cohabitating with one partner. Among nonpregnant women, vaginal sex has been reported to disrupt the baseline vaginal microbiome composition, although the disturbance is smaller in magnitude than changes observed during menses and changes in hormone levels.²⁶ Given that the hormonal milieu is different between pregnant and nonpregnant women, it is surprising that the influence of sexual practices during pregnancy has not been reported previously. Previous works have noted, however, that the vaginal microbiome during pregnancy tends to be less diverse and more stable than among nonpregnant women,² which may help prevent perturbation in the vaginal microbiome composition from vaginal sex.

Although this study was one of the first to specifically assess the influence of social factors on the vaginal microbiome among pregnant women, instead of focusing on birth outcomes, there are a number of limitations. Our study was not designed to identify mechanisms associated with variations in the type of microbiome pregnant women harbor. Additional work evaluating inflammatory pathways, ancestry, and stress may offer additional insights as to why vaginal microbiome differs among women. Psychological and physiological stress may be of particular interest due to their association with reproductive outcomes, such as preterm birth.^38,39 Future studies assessing the interaction of stress and variations in vaginal microbiome composition may elucidate additional associations among social factors and Lactobacillus predominance. Proinflammatory cytokines have been associated with diverse microbiome composition,^9,10 and if psychological stress can influence vaginal cytokine levels, these changes could influence the microbes within the vaginal microbiome. Future studies investigating psychological and biological markers associated with stress could provide additional insight into why some women are more likely to harbor Lactobacillus.

In summary, in this study, we identified within a cohort of socially and economically diverse African American women in the United States, that the most significant factors influencing the composition of the vaginal microbiome early in pregnancy were cohabitation with a partner, marijuana usage, and socioeconomic factors. Moreover, cohabitation was itself associated with reports of increased vaginal sex and decreased use of vaginal hygiene products, including douches and deodorant sprays. This suggests that either the actual presence of and/or regular intercourse with a cohabitating partner may be an important factor related to microbial colonization and stability that has previously been underappreciated in vaginal microbiome studies during pregnancy. Previous research has shown that couples living in the same household have more similar gut, oral, and skin microbiomes than to those outside the household.⁴⁰ Although not previously considered a factor for stability of the vaginal microbiome, we hypothesize that environmental factors or even day-to-day physical contact associated with cohabitation that influence other microbiome sites may also influence the vaginal microbiome. Or, as suggested by Dill-McFarland et al. in regard to the gut microbiome, the sustained, close contact among couples living together might influence the microbiome either directly through frequent physical contact or through less clear affective benefits, such as increased social support.⁴¹ Future works evaluating diet, chemicals in the home, and work environments, and comparing multiple microbiome sites among partners may elucidate additional factors that influence the composition and stability of the vaginal microbiome.

Authors' Contributions

A.L.D. and E.J.C. conceived the study design and oversaw data collection. M.L.W. drafted the article and performed statistical analysis. M.L.W., A.L.D., A.B.D., E.F.W., R.M.M., and E.J.C. contributed to interpretation of results and the article. All authors discussed results and provided feedback that contributed to the final development of the article.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This study was supported by the National Institutes of Health, National Institute of Nursing Research [R01NR014800 and K01NR017903], National Institute on Minority Health and Health Disparities [R01MD009064], National Institute of Environmental Health Sciences [R24ES029490], and the Office of the Director [UG3OD023318/UH3OD023318]. This study was also supported, in part, by the Emory Integrated Genomics Core, which is subsidized by the Emory University School of Medicine and is one of the Emory Integrated Core Facilities. Additional support was provided by the National Center for Advancing Translational Sciences of the National Institutes of Health under award number UL1TR000424 and UL1TR000454. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the article. Emily F. Wissel is supported by the National Science Foundation Graduate Research Fellowship under grant number 1937971.