Objective
Recent data suggest vitamin D deficiency (VDD) is associated with bacterial vaginosis (BV) during pregnancy. We hypothesized that VDD is a risk factor for BV in nonpregnant women.
Study Design
Using National Health and Nutrition Examination Survey data, we conducted multivariable logistic regression analyses stratified by pregnancy.
Results
VDD was associated with BV only in pregnant women (adjusted odds ratio [AOR], 2.87; 95% confidence interval [CI], 1.13–7.28). Among nonpregnant women, douching (AOR, 1.72; 95% CI, 1.25–2.37), smoking (AOR, 1.66; 95% CI, 1.23–2.24), and black race (AOR, 2.41; 95% CI, 1.67–3.47) were associated with BV; oral contraceptive use was inversely associated with BV (AOR, 0.60; 95% CI, 0.40–0.90). VDD moderated the association between smoking and BV in nonpregnant women.
Conclusion
Risk factors for BV differ by pregnancy status. VDD was a modifiable risk factor for BV among pregnant women; evaluation of vitamin D supplementation for prevention or adjunct therapy of BV in pregnancy is warranted.
Bacterial vaginosis (BV) is exceedingly prevalent, affecting nearly 30% of women of child-bearing age in the United States. Although the dramatic changes in vaginal microflora that characterize BV often occur in the absence of symptoms, they are associated with numerous adverse sequelae including an increased risk of sexually transmitted infections, preterm labor, and postpartum endometritis. Although eradication of BV is possible with appropriate antimicrobial therapy, recurrent disease remains a formidable challenge. Treatment of pregnant women with BV is particularly problematic because large clinical trials have failed to demonstrate a reduction in adverse pregnancy outcomes following antibiotic therapy. The identification of potentially modifiable risk factors represents a unique opportunity to reduce the burden of BV and its associated morbidities.
Recently Bodnar et al identified vitamin D deficiency (VDD) as an independent risk factor for BV in pregnant women. This cross-sectional analysis revealed a significant association between serum concentrations of 25-hydroxyvitamin D (25D), the major circulating vitamin D metabolite, and the presence of BV in the first trimester. These findings have enormous public health implications in light of the high prevalence of VDD in the United States, estimated at 78% among nonpregnant women and ranging from 83% in the first trimester to 47% in the third trimester among pregnant women. Because VDD is modifiable through supplementation at very low cost, it provides a possible point of intervention in reducing the burden of BV.
It is now well recognized that vitamin D is an important regulator of host immune responses, and VDD has been associated with increased susceptibility to numerous infectious diseases. Binding of 1, 25-dihydroxyvitamin D to its receptor ultimately results in the transcription of hundreds of genes, including integral components of the innate immune system. Therefore, vitamin D may locally regulate host immune signaling. Altered immunity in the vaginal microenvironment provides a potential mechanistic explanation for the observed association between VDD and the development of BV.
The relationship between VDD and BV in nonpregnant women has not yet been explored. Furthermore, it is unknown whether other risk factors for BV differ by pregnancy status. We assessed whether BV risk factors differ between pregnant and nonpregnant women with a special focus on VDD across subgroups using a large, nationally representative data set.
Materials and Methods
The National Health and Nutrition Examination Survey (NHANES) is an ongoing national survey conducted by the National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention, that assesses disease and risk factors among the civilian, noninstitutionalized population of the United States. NHANES uses a complex, multistage, probability sample design and oversampled African Americans, Mexican Americans, low-income persons, and adolescents aged 12-19 years during 2001-2004 to obtain sample sizes large enough for stable estimates among these groups. Health examinations and some interviews, including sexual behavior and reproductive health, were conducted at a mobile examination center (MEC).
Inclusion criteria for this analysis were women aged 14-49 years who participated in the MEC components of NHANES. Exclusion criteria included no reported Nugent score for BV and/or no reported 25D serum concentration.
Participants aged 12 years or older provided consent for examination and interview; parents of participants younger than 18 years also provided consent for their children’s participation. The National Center for Health Statistics Research Ethics Review Board approved all protocols. The Columbia University Medical Center Institutional Review Board provided exemption for this analysis.
Race/ethnicity was self-reported and classified as non-Hispanic white, non-Hispanic black, Mexican American, and other, including multiracial persons and persons of non-Mexican Hispanic descent. Poverty index was calculated by comparing the family’s self-reported income to the family’s appropriate poverty threshold. An index less than 1 indicated income below the poverty threshold. Educational attainment, marital status, current oral contraception use, and douching frequency in the past 6 months were self-reported. Sexual behaviors, including ever having sex, number of lifetime sexual partners (including both male and female sexual partners), ever having a female sexual partner, age at first sex, and unprotected sex in the past 30 days, were self-reported in an audio computer-assisted self-interview.
Participants were instructed to consider oral, anal, and vaginal sex as “sex” in their responses. Only women who reported multiple sex partners in the past year were asked about condom use in the preceding 30 days. Therefore, the classification of having “unprotected sex” is limited to women who had more than 1 sexual partner in the past year and who reported sex without a condom at least once in the preceding 30 days. Women with only 1 sexual partner in the past year were categorized as not having “unprotected sex” because they were not asked about condom use. This analysis includes sexual behavior data only for women 20 years old or older.
Pregnancy status was confirmed by urine testing. Body mass index (BMI) was calculated from height and weight measurements and classified into the following categories: normal, BMI ≥18.5 kg/m 2 and ≤24.9 kg/m 2 ; underweight, BMI <18.5 kg/m 2 ; overweight, BMI ≥25.0 kg/m 2 and ≤29.9 kg/m 2 ; obese, BMI ≥30.0 kg/m 2 . Vitamin D and cotinine (a serum metabolite of nicotine) levels were determined using serum collected via venipuncture by certified phlebotomists at the MEC. Analysis methods are described elsewhere. In this analysis, VDD was classified as less than 30 ng/mL based on current recommendations. Cotinine levels greater than 3 ng/mL were indicative of active smoking. Vaginal swabs used for determining BV were self-collected in the MEC. The MEC staff then rolled the swabs onto glass slides, which were subsequently shipped to Magee Women’s Hospital (Pittsburgh, PA) for Gram staining and Nugent scoring. A Nugent score of 7-10 indicates BV.
Statistical analysis was conducted using survey procedures in SAS (version 9.2; SAS Institute, Cary, NC), which account for the unequal weighting of persons in the complex NHANES sample design, and missing data were excluded from analysis. Weights reported by NCHS for the 2001-2002 and 2003-2004 data cycles were used to reflect the unequal probability of selection and to adjust for nonresponse among sample persons. Taylor series linearization was used for variance estimation to produce unbiased estimates.
In bivariable analyses, unadjusted logistic regression was used to compute crude odds ratios with 95% confidence intervals. Associated P values were calculated using Wald χ 2 tests. Variables with P ≤ .10 were included as covariates in the multivariable logistic regression analysis as were age, which was included out of convention, and pregnancy status, which was of interest based on the literature.
A second analysis stratified the study population by pregnancy status. Bivariable analyses for pregnant and nonpregnant women were conducted as described earlier. Multivariable logistic regression analyses were conducted separately for pregnant and nonpregnant women and included the same covariates used in the unstratified multivariable logistic regression analysis, excepting pregnancy status.
Finally, a multivariable logistic regression analysis was stratified by vitamin D status. Results of other multivariable logistic regression models in this larger analysis suggested the association between cotinine levels and BV may be dependent on vitamin D status. Stratifying by vitamin D status allowed an examination of this hypothesized moderating effect. Covariates included in this model were the same as those used in the unstratified multivariable logistic regression analysis, excepting vitamin D status.
Results
Of 3527 women who satisfied the inclusion and exclusion criteria, 4 women who reported never having sex had positive urine pregnancy tests at time of examination. These women were excluded from the analysis resulting in an n = 3523 for the sample. BV prevalence was 29%.
Similar to previous findings, several demographic, behavioral, and clinical characteristics were significantly associated with BV in the bivariable analysis, including black race ( P < .01), Mexican American/other race/ethnicity ( P < .01), less than a high school education ( P < .01), ever having sex ( P = .05), sexual debut 14 years old or younger ( P < .01) or at ages 15 or 16 years ( P = .03), number of lifetime sexual partners (5-8; 9 or more; P < .01 for each), douching at least once in the past 6 months ( P < .01), and being overweight ( P = .03) or obese ( P < .01).
VDD ( P < .01), having unprotected sex in the last 30 days ( P < .01), and cotinine levels of 3 ng/mL or greater ( P < .01) were also significantly associated with BV. Income above the poverty level ( P < .01), being married/living as married ( P < .01), and current oral contraceptive (OC) use ( P < .01) were significantly inversely associated with BV. Median serum concentration of 25D was 23.00 ng/mL (interquartile range [IQR], 17.00–30.00).
In the bivariable analysis stratified by pregnancy status, among nonpregnant women, VDD ( P < .01), black race ( P < .01), Mexican American/other race/ethnicity ( P < .01), less than a high school education ( P = .02), ever having sex ( P = .05), sexual debut at age 14 years or younger ( P < .01) or at age 15 or 16 years ( P = .02), number of lifetime sexual partners (5-8; ≥9; P < .01 for each), ever having had a female sexual partner ( P < .01), unprotected sex in the past 30 days ( P < .01), douching at least once in the past 6 months ( P < .01), cotinine levels of 3 ng/mL or greater ( P < .01), and being overweight ( P = .03) or obese ( P < .01) were all significantly associated with BV. Income above the poverty level ( P < .01), being married/living as married ( P < .01), and current OC use ( P < .01) were significantly inversely associated with BV among nonpregnant women.
Among pregnant women, VDD ( P < .01), black race ( P = .02), and ever having a female sexual partner ( P < .01) were significantly associated with BV in the bivariable analysis; marital status ( P < .01) was significantly inversely associated with BV. Median serum concentrations of 25D were 23.00 ng/mL (IQR, 17.00–30.00) among nonpregnant women and 25.00 ng/mL (IQR, 17.00–31.00) among pregnant women.
Several characteristics remained associated with BV in the multivariable logistic regression analysis ( Table 1 ). Black race (adjusted odds ratio [AOR], 2.41; 95% confidence interval [CI], 1.64–3.55), Mexican American/other race/ethnicity (AOR, 1.54; 95% CI, 1.00–2.37), sexual debut at age 14 years or younger (AOR, 3.04; 95% CI, 1.12–8.24), ever having a female sexual partner (AOR, 1.76; 95% CI, 1.07–2.91), douching at least once in the past 6 months (AOR, 1.68; 95% CI, 1.21–2.33), and serum cotinine level of 3 ng/mL or greater (AOR, 1.63; 95% CI, 1.22–2.16) were significantly associated with BV. Current OC use (AOR, 0.60; 95% CI, 0.40–0.90) was significantly inversely associated with BV.
| Characteristic | Overall (n = 3523) | |
|---|---|---|
| AOR (95% CI) | P value a | |
| Vitamin D b | ||
| Sufficient | 1.00 (reference) | |
| Deficient | 1.03 (0.75–1.44) | .84 |
| Age, y | 1.00 (0.98–1.02) | .95 |
| Race/ethnicity | ||
| White | 1.00 (reference) | |
| Black | 2.41 (1.64–3.55) | < .01 |
| Mexican American/other | 1.54 (1.00–2.37) | .04 |
| Education | ||
| High school graduate/GED or more | 1.00 (reference) | |
| Less than high school | 1.12 (0.74–1.70) | .58 |
| Poverty index | ||
| At or below poverty level | 1.00 (reference) | |
| Above poverty level | 0.78 (0.60–1.00) | .05 |
| Marital status | ||
| Not married | 1.00 (reference) | |
| Married/living as married | 0.89 (0.62–1.27) | .52 |
| Age at first sex, y c | ||
| Never had sex | 1.00 (reference) | |
| ≤14 | 3.04 (1.12–8.24) | .03 |
| 15 or 16 | 2.55 (0.94–6.94) | .07 |
| 17 or 18 | 2.24 (0.79–6.33) | .13 |
| ≥19 | 2.16 (0.85–5.49) | .10 |
| Number lifetime partners c | ||
| 0 or 1 | 1.00 (reference) | |
| 2-4 | 0.78 (0.53–1.17) | .23 |
| 5-8 | 0.94 (0.61–1.47) | .80 |
| ≥9 | 0.89 (0.62–1.27) | .51 |
| Ever had female sex partner c | ||
| No | 1.00 (reference) | |
| Yes | 1.76 (1.07–2.91) | .03 |
| Unprotected sex c , d | ||
| No | 1.00 (reference) | |
| Yes | 1.13 (0.78–1.64) | .51 |
| Pregnancy status | ||
| Not pregnant | 1.00 (reference) | |
| Pregnant | 0.59 (0.32–1.07) | .08 |
| Current oral contraception use | ||
| No | 1.00 (reference) | |
| Yes | 0.60 (0.40–0.90) | .01 |
| Douching frequency last 6 mo | ||
| Never | 1.00 (reference) | |
| At least once | 1.68 (1.21–2.33) | < .01 |
| Cotinine level | ||
| <3 ng/mL | 1.00 (reference) | |
| ≥3 ng/mL | 1.63 (1.22–2.16) | < .01 |
| BMI e | ||
| Normal | 1.00 (reference) | |
| Underweight | 0.82 (0.38–1.77) | .62 |
| Overweight | 1.29 (0.88–1.90) | .19 |
| Obese | 1.26 (0.87–1.81) | .22 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
