Bacterial vaginosis (BV) is a condition of unknown etiology in which there is a global disturbance of vaginal microbial ecology described as the replacement of the usual lactic acid-producing lactobacillus predominant flora with an overgrowth of Gardnerella vaginalis and mixed anaerobic organisms.1 It is the most common cause of symptomatic vaginal discharge in reproductive age women, and has been associated with a variety of conditions including postsurgical infection, pelvic inflammatory disease, preterm birth, and HIV acquisition.2 Observational studies have consistently demonstrated increased occurrence of BV among women who douche.3–5 However, while associations between BV and other personal hygienic behaviors have been studied extensively among women in Africa,6–11 hygiene behaviors other than douching have not been well studied among women in the United States. In one US study BV was slightly, although not significantly more prevalent among women who bathed frequently but less prevalent among women who showered frequently.3 This same study reported tampon use to be associated with reduced BV prevalence,3 but no association between BV and menstrual protection products was found in another study.12 In a study of adolescents, feminine deodorant spray use was increased among women with recently diagnosed sexually transmitted infection.13
Whether douching causes BV or subtle BV symptoms prompt a woman to douche is controversial. Study of other hygienic behaviors can provide insight into this question. If douching were merely a response to BV symptoms, then one would expect other genital hygienic behaviors such as use of deodorant sprays, wipes, etc, also to be more common among women with BV, even after controlling for douching. However, if douching were the only behavior associated with BV, this finding might provide evidence against douching being a response to BV symptoms. To address this question, we studied the association between personal hygienic practices and BV prevalence among women enrolled in the Longitudinal Study of Vaginal Flora.
MATERIALS AND METHODS
Data for this study are from the Longitudinal Study of Vaginal Flora, described previously.14 This cohort study was intended to evaluate the natural history of BV. Nonpregnant women age 15 to 44 were recruited from August 1999 to February 2002 upon presentation for a routine health care visit at 1 of 12 clinics in the Birmingham, AL area. Women with significant medical or gynecological conditions were ineligible, as were women with conditions precluding informed consent and those planning to move from the area in the next 12 months. All participants provided written informed consent; the study was approved by the Institutional Review Boards of the University of Alabama at Birmingham, the Jefferson County Department of Health, and the National Institute of Child Health and Human Development.
Participating women were seen at a research clinic for an initial visit and then for 4 quarterly visits, for a total of up to 1 year of observation. At each visit a pelvic examination was conducted at which time specimens were obtained for culture of Neisseria gonorrhoeae and Trichomonas vaginalis, for ligase chain reaction for Chlamydia trachomatis, for vaginal pH, and for a wet prep evaluating clue cells, trichomonads and yeast. Vaginal gram stains were also obtained and scored according to the method of Nugent15; 10% of gram stains were reread in a different laboratory, with a κ coefficient of 0.81 for the diagnosis of BV. BV was defined as a gram stain score of 7 to 10.
At each visit women underwent an extensive face-to-face private interview with a female interviewer, covering among other things demographic factors, personal hygiene, sexual behaviors, contraceptive use, and vaginal symptoms. The initial questionnaire referred to behaviors over the previous 6 months; subsequent questionnaires referred to behaviors since the previous visit. Specifically, the questionnaire at each visit asked about type of birth control used; frequency of bathing and of showering (for each behavior response options were daily, several times/week, about once/week, less than once/week, never); frequency of use of hygienic products including talcum or baby power on one's private parts, feminine towlettes or baby wipes, feminine hygiene spray, and vaginal suppositories (for each product–never, less than once/month, at least once/month, at least once/week, daily, twice or more daily); and frequency of douching (daily, several times/week, about once/week, less than once/week, and never). Presence of menstrual periods was ascertained at each visit, although specific menstrual protection products (pads only, tampons only, pads and tampons, other) were ascertained only at baseline. Use of sanitary pads or panty liners when not menstruating, as well as type of underwear most commonly worn (cotton, nylon, crotchless or none, panty hose only) were also ascertained only at baseline. For analysis, categories were collapsed based on available numbers as noted in Table 2.
The initial analysis compared the prevalence of studied behaviors among women who reported douching or not douching at that study visit, and the prevalence of studied behaviors among women with and without BV at each visit. Statistical significance of these associations was determined by logistic regression under Generalized Estimating Equations16 to account for the nonindependence of visits by the same woman.
Next, the association between hygienic behaviors and BV was assessed. Prevalence ratios were determined from Poisson regression with robust (empirical) standard errors17 under Generalized Estimating Equations.16 In several analyses, log-binomial regression,18 usually employed to estimate the prevalence ratio, failed to converge, so Poisson regression was always used to maintain comparability across different analyses. The first model presents the unadjusted prevalence ratio for BV among women practicing that behavior. The second model adjusts the association for other demographic and behavioral factors: race, age (continuous variable), body mass index (restricted cubic spline), marital status, years of education, smoking (cigarettes/day, restricted cubic spline, although presented categorically in Table 1), alcohol drinking, number of male sex partners during the interval under study, acquisition of a new sex partner, vaginal and receptive oral sex frequencies during the time period of the questionnaire, and current use of hormonal contraception or receipt of tubal ligation.19 Oral contraceptives and progestin-only injectable/implantable contraceptives were considered together because they manifested similar associations with BV in the Flora study.19 Race was considered as black versus other due to the small number of women who where neither black nor white, and the similarity of their BV prevalence to that of white women. The final model adjusted each hygienic behavior for these same demographic and behavioral factors, plus all the other hygienic behaviors. P-values were determined from the score statistic of a Poisson regression model with hygienic behaviors considered as categorical variables, as noted in the Tables. We evaluated the interaction between ethnicity and all hygienic behaviors simultaneously, and the interaction as between use of hormonal contraception and all hygienic behaviors simultaneously using the “chunk” test20; we also evaluated the interaction between showering and bathing. P-values for these tests used the score statistic. None of these tests was statistically significant (P = 0.71, 0.91 and 0.62, respectively); no other interactions were evaluated. Calculations utilized SAS software, version 9.1.3 (Cary, NC).
The study recruited 3620 women, who completed 13,587 study visits; BV status was missing for 70 visits (due to incomplete visits, broken slides or inadequate sample), leaving 13,517 eligible visits. Characteristics of the women at the time of the initial study visit are presented in Table 1. In addition to the characteristics described in the table, women with BV were on average older than women without BV (mean[SD] age, 25.1 [6.7] vs. 24.6 [6.7] years, respectively, P = 0.02). Women with BV also had higher median (interquartile range) body mass index, (29.1 [10.7] vs. 27.7 [11.0] kg/m2, P <0.001). BV was present at 5467 (40.4%) of the 13,517 visits with available data.
Associations between each studied hygienic behavior and douching, and between each behavior and BV, are presented in Table 2. Most genital hygienic measures were more common among women who douched, as were tampon use, use of pads and panty liners when not menstruating, and usually wearing nylon underwear. Neither frequency of showering nor of bathing was statistically significantly associated with douching, although black women were more likely than women of other races to bathe versus shower and also more likely to douche (data not shown). Associations between hygienic behaviors and BV were more variable. BV was positively associated with bathing frequency, use of powder and feminine hygiene spray, and usual type of underwear. BV was less common among women experiencing amenorrhea, but type of menstrual protection was not associated with BV.
After controlling for only the demographic and nonhygiene behavioral factors noted in the Methods, less than weekly use of towlettes was associated with an increased prevalence of BV, but more frequent use was not. Use of vaginal suppositories, feminine deodorant spray, or powder on the genitals was not associated with BV (Table 3). Daily bathing was associated with increased BV prevalence. Underwear material was not associated with BV. Although BV was less common among amenorrheic women, type of menstrual protection was not associated with BV among women experiencing menstrual periods (prevalence ratio for tampons only vs. pads only: 1.04, 0.95–1.12; prevalence ratio for pads and tampons vs. pads only: 1.00, 0.92–1.07). As has been reported previously from this study,5 douching was associated with a modest, but statistically significant increase in BV prevalence.
The fully adjusted model was based on 12,900 visits (95.4% of visits with data on BV). After further adjustment of the hygienic behaviors for all other hygienic behaviors, in addition to the demographic and nonhygiene behavioral factors, neither powder, towlettes, feminine hygiene spray nor vaginal suppositories had overall statistically significant associations with BV, although less than weekly use of towlettes was associated with increased BV prevalence (Table 3). Daily bathing, compared with less than daily, was also associated with increased BV prevalence. Amenorrheic women continued to have reduced BV prevalence, and type of menstrual protection continued to be unassociated with BV. The association between douching and BV was minimally changed: compared with women who did not douche, the prevalence ratio for BV among women who douched less than weekly was 1.14 (1.08–1.20), and the prevalence ratio for BV among women who douched weekly or more often was 1.17 (1.09–1.26). When no douching, douching less than weekly and douching weekly or more were considered as a continuous variable with values of 0, 1, and 2, the trend for increased douching to be associated with increased BV prevalence was highly statistically significant (P <0.0001). None of the other hygienic behaviors similarly categorized (powder, towlettes, and spray) even approached statistical significance.
We found that after adjustment for demographic, sexual behavior and other behavioral factors, nondouching personal hygienic behaviors which might be practiced in response to vaginal symptoms were by and large not strongly associated with prevalent BV. We also found that hygienic behaviors less likely to be a direct response to vaginal symptoms (i.e., type of underwear and menstrual protection) were not strongly associated with BV. In contrast, douching remained associated with BV in a dose-response manner, both before and after adjustment for other hygienic behaviors. If women douched in response to subtle BV symptoms then one would expect other behaviors done to prevent or reduce vaginal odor or irritation also to be more common among women with BV. Therefore, the lack of consistent association observed between nondouching hygienic behavior and BV in this study provides evidence against the hypothesis that the consistent observational association between douching and BV is due to women douching in response to subtle symptoms of BV.
Although a variety of vaginal practices, such as type of menstrual protection10 insertion of herbs and astringents into the vagina,21 vaginal “finger cleansing,” external genital wiping,6 as well as bathing have been extensively studied among African women, relatively few studies of nondouching genital hygiene behavior have been conducted in the United States. Holzman et al reported increased BV prevalence among women who bathed rather than showered, but the difference was not statistically significant after adjustment.3 They also reported nonsignificantly lower BV prevalence among women who used tampons, although Schwebke et al found tampon use to be unassociated with BV.12 Misra found that women who douched were also more likely to use feminine spray, wash or towlettes, as well as to use powder on their genitals, but the relationship between these behaviors and BV was not explored.22
Strengths of our study included large sample size, repeated measurements and inclusion of an extensive array of behaviors in our questionnaire. The study also has limitations. Type of underwear and use of pads or panty liners were assessed only at study entry; and although the presence of menstrual periods was assessed at every visit we collected data on menstrual protection only at baseline. Therefore, the association between these behaviors and BV may have been misclassified. If misclassification is random with regard to BV that may have developed after baseline, then associations between BV and underwear, menstrual protection and nonmenstrual use of pads or panty liners would be biased toward finding no association. However, if women changed their use of these products specifically in response to symptoms of BV (such as changing from nylon to cotton underwear), then the direction of bias may be less predictable. We did not collect data on brand of pads or tampons, and thus cannot address how absorbency or chemical composition of these products might influence the vaginal environment. Our question on “vaginal suppositories” was nonspecific, and we do not know what, if any medications they might have contained (although use of suppositories was rare).
Approximately half of the women used some type of hormonal contraception. Although all injectable/implantable contraceptives contain only progestin, we did not collect data on the specific composition of the oral contraceptives used. Thus we cannot explore the role of specific hormones in BV. The interaction between use of hormonal contraception and hygienic behavior on BV was not statistically significant (P = 0.91); thus we found no evidence that the association between hygienic behaviors and BV was different in users of hormonal contraceptives than among nonusers. Nevertheless our results might not apply to other populations where hormonal contraception was less common.
Approximately 600 visits were excluded from the adjusted analyses due to missing data on demographic and/or behavioral variables (4.6% of all visits with data on BV). Although this might be addressed by complex methods such as multiple imputation,23 simple deletion of observations with missing data is a reasonable option when these deletions comprise less than 5% of otherwise eligible observations.24 Although our study collected data longitudinally, this analysis was cross-sectional. Since many previous reports of the association between hygienic behaviors and BV were also cross-sectional, we used a cross-sectional analysis to maintain comparability to previous reports. In addition, our intention was to compare the cross-sectional association between douching and BV to the cross-sectional associations between other hygienic behaviors and BV, thus we do not believe this to be a substantial limitation. However, one should not infer that any statistically significant associations found are causal.
In summary, we found that the associations between nondouching hygienic behaviors and BV were not strong and for the most part not statistically significant. In contrast, douching was associated with increased BV prevalence. These results do not support the premise that the increased prevalence of BV among women who douche is because BV prompts the douching.
1.Hill GB. The microbiology of bacterial vaginosis. Am J Obstet Gynecol 1993; 169:450–454.
2.Nyirjesy P. Vulvovaginal candidiasis and bacterial vaginosis. Infect Dis Clin North Am 2008; 22:637–652.
3.Holzman C, Leventhal JM, Qiu H, et al. Factors linked to bacterial vaginosis in nonpregnant women. Am J Public Health 2001; 91:1664–1670.
4.Hutchinson KB, Kip KE, Ness RB, et al. Vaginal douching and development of bacterial vaginosis among women with normal and abnormal vaginal microflora. Sex Transm Dis 2007; 34:671–675.
5.Brotman RM, Klebanoff MA, Nansel TR, et al. A longitudinal study of vaginal douching and bacterial vaginosis–A marginal structural model analysis. Am J Epidemiol 2008; 168:188–196.
6.Myer L, Denny L, De Souza M, et al. Intravaginal practices, HIV and other sexually transmitted diseases among South African women. Sex Transm Dis 2004; 3:174–179.
7.McClelland RS, Richardson BA, Graham SM, et al. A prospective study of risk factors for bacterial vaginosis in HIV-1 seronegative African women. Sex Transm Dis 2008; 35:617–623.
8.van De Wijgert JH, Mason PR, Gwanzura L, et al. Intravaginal practices, vaginal flora disturbances, and acquisition of sexually transmitted diseases in Zimbabwean women. J Infect Dis 2000; 181:587–594.
9.Mbizvo ME, Musya SE, Stray-Pedersen B, et al. Bacterial vaginosis and intravaginal practices: Association with HIV. Cent Afr J Med 2004; 50:41–46.
10.Morison L, Ekpo G, West B, et al. Bacterial vaginosis in relation to menstrual cycle, menstrual protection method, and sexual intercourse in rural Gambian women. Sex Transm Infect 2005; 81:242–247.
11.Hassan WM, Lavreys L, Chohan V, et al. Associations between intravaginal practices and bacterial vaginosis in Kenyan female sex workers without symptoms of vaginal infections. Sex Transm Dis 2007; 34:384–388.
12.Schwebke JR, Richey CM, Weiss HL. Correlation of behaviors with microbiological changes in vaginal flora. J Infect Dis 1999; 180:1632–1636.
13.Ott M, Ofner S, Fortenberry JD. Beyond douching: Use of feminine hygiene products and STI risk among young women. J Sex Med 2009; 6:1335–1340.
14.Klebanoff MA, Schwebke JR, Zhang J, et al. Vulvovaginal symptoms in women with bacterial vaginosis. Obstet Gynecol 2004; 104:267–272.
15.Nugent RP, Krohn MA, Hillier SL. Reliability of diagnosing bacterial vaginosis is improved by a standardized method of gram stain interpretation. J Clin Microbiol 1991; 29:297–301.
16.Diggle P, Heagerty P, Liang K, et al. Analysis of Longitudinal Data. 2nd ed. New York, NY: Oxford University Press, 2002.
17.Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004; 159:702–706.
18.Skov T, Deddens J, Peterson MR, et al. Prevalence proportion ratios: Estimation and hypothesis testing. Int J Epidemiol 1998; 27:91–95.
19.Riggs M, Klebanoff M, Nansel T, et al. Longitudinal association between hormonal contraceptives and bacterial vaginosis in women of reproductive age. Sex Transm Dis 2007; 34:954–959.
20.Kleinbaum DG, Klein M. Logistic Regression: A Self-Learning Text. 2nd ed. New York, NY: Springer Science and Business Media Inc, 2002: 195–198.
21.Myer L, Denny L, de Souza M, et al. Distinguishing the temporal association between women's intravaginal practices and risk of human immunodeficiency virus infection: A prospective study of South African Women. Am J Epidemiol 2006; 163:552–560.
22.Misra DP, Trabert B, Atherly-Trim S. Variation and predictors of vaginal douching behavior. Womens Health Issues 2006; 16:275–282.
23.Donders AR, van der Heijden GJ, Stijnen T, et al. Review: A gentle introduction to imputation of missing values. J Clin Epidemiol 2006; 59:1087–1091.
24.Harrell FE. Regression Modeling Strategies. New York, NY: Springer-Verlag, 2001: 49.