The Association between Trichomonas vaginalis Infection and Level of Vaginal Lactobacilli, in Nonpregnant Women

Trichomoniasis is the most common curable sexually transmitted infection (STI) worldwide [1]. The prevalence of Trichomonas vaginalis infection in women in the United States is 6%–54%, depending on the population studied [2]. Trichomoniasis can lead to health problems, including pelvic inflammatory disease and preterm delivery, and is a risk factor for HIV infection [3]. Until recently, trichomoniasis has received little clinical and research attention, despite its public health importance.

Lactobacillus species are the most prevalent organisms in normal human vaginal flora and may protect the vagina against infection [4]. The addition of lactobacilli to in vitro T. vaginalis culture slows growth of the protozoa, and T. vaginalis phagocytoses lactobacilli [5]. However, epidemiologic evidence is inconsistent [6–9], and the association between T. vaginalis and lactobacilli is not understood.

We conducted a case-control study to assess the association between level of vaginal lactobacilli and T. vaginalis among nonpregnant women. Although this topic has been examined previously, this is the first study of nonpregnant women to have an adequate sample size, use the diagnostic reference standard, and comprehensively evaluate the effect that covariables have on the association between the level of vaginal lactobacilli and T. vaginalis. Clarifying this relationship could improve our understanding of T. vaginalis and may have implications for the treatment of trichomoniasis.

Participants, materials, and methods. We conducted a prospective, case-control study in 3 health-department sexually transmitted disease clinics, located in Durham and Raleigh in North Carolina and Birmingham in Alabama, from November 2001 to July 2003. For every 3 women with positive wet mount- microscopy results for T. vaginalis whom we enrolled, we recruited 1 woman with negative wet mount-microscopy results; this case:control ratio was selected as part of a larger study that focused on T. vaginalis-infection concordance between infected women and their male partners [10]. Female patients who were ⩾18 years old, spoke English, and had not previously been prescreened or enrolled in the study were eligible for prescreening. Women were prescreened on the basis of wet mount- microscopy results and T. vaginalis cultures. Prescreened patients with positive wet mount-microscopy results for T. vaginalis were recruited immediately and completed questionnaires at enrollment; prescreened patients with negative wet mount-microscopy results for T. vaginalis but with positive culture results for it were contacted and asked to return for treatment and/or counseling; during the follow-up visit, they were recruited for the study and completed questionnaires. The unmatched control group was selected from prescreened women whom wet-mount microscopy had found to be uninfected with T. vaginalis, but control subjects whom T. vaginalis cultures subsequently identified as having trichomoniasis were reclassified as case patients.

Prescreened women who had a history of vaginal sex during the preceding 60 days and had not used metronidazole during the 4 weeks before T. vaginalis cultures were collected were eligible for enrollment in the study. All enrolled participants provided written informed consent, and the study was approved by the University of North Carolina at Chapel Hill and the University of Alabama at Birmingham Biomedical Institutional Review Boards.

Vaginal swabs for Gram staining and for T. vaginalis culturing were obtained from prescreened women. Vaginal wet-mount microscopy and routine testing for Neisseria gonorrheae were performed. Routine testing for Chlamydia trachomatis was also performed, although in 1 of the 3 clinics it was limited to women ⩽25 years old. All participants underwent routine physical and bimanual pelvic examinations. Clinical, demographic, and behavioral information was obtained from medical records and a detailed questionnaire.

At each of the 3 clinics, wet-mount microscopy was performed according to routine clinic procedures performed by trained laboratory personnel. T. vaginalis culture was performed by use of the InPouch culture system (Biomed). Cultures were read daily for up to 5 days after inoculation. Vaginal smears were sent to the University of Alabama at Birmingham for Gram-stain evaluation and were quantified by use of criteria of Nugent et al. [11]. Bacterial vaginosis was diagnosed on the basis of the Nugent et al. score for the vaginal Gram stain [11]. Testing for N. gonorrheae was performed by use of culturing or DNA hybridization.

Data were analyzed by use of SAS software (version 8). Women with positive wet mount-microscopy results or positive T. vaginalis culture results were classified as having trichomoniasis. Women with negative wet mount-microscopy results and negative T. vaginalis culture results were defined as not having T. vaginalis infection. We conducted descriptive analyses of the study population by examining the frequency distribution, measures of central tendency, and variability for covariables.

The level of vaginal lactobacilli was measured by use of the Lactobacillus-species vaginal Gram–stain score based on the categories described by Nugent et al. [11]. Bivariable analyses were performed in parallel, using both dichotomous variables (based on data on the distribution of the vaginal lactobacilli) and 4- level indicator variables for vaginal lactobacilli. Because both coding schemes yielded similar results, we dichotomized the data on vaginal lactobacilli to facilitate interpretation of the study.Women who had ⩾30 lactobacillus morphotypes per oilimmersion field (i.e., a Lactobacillus-species vaginal Gram-stain score of 4+) were classified as having “high” levels of vaginal lactobacilli, whereas women who had ⩾30 lactobacillus morphotypes per oil-immersion field (i.e., a Lactobacillus-species vaginal Gram-stain score of 0 to 3+) were classified as having “low” levels of vaginal lactobacilli.

We used bivariable analyses to evaluate both potential confounding and intersubgroup differences in the association between the level of vaginal lactobacilli and T. vaginalis. Differences between the subgroups were assessed by use of the Breslow-Day test of homogeneity. Variables that resulted in were assessed in multivariable analyses of intersubgroup P < .20 differences. Variables that were associated with both the level of vaginal lactobacilli and T. vaginalis were considered to potentially confound the association between them; we evaluated these variables by using a stratified tabular approach [12]. Unconditional multiple logistic regression was used to estimate an adjusted odds ratio (OR) and 95% confidence interval (CI) for the association between the level of vaginal lactobacilli and the presence of trichomoniasis.

We used a backward-elimination model-building strategy. Interaction terms were assessed by comparing the likelihood ratios (LRs) for models with and without the interaction term. Variables resulting in P ⩽ .10 for an LR were retained in the model. Confounding of the association between the level of vaginal lactobacilli and T. vaginalis was assessed by comparing the fully adjusted OR to the adjusted OR after potentially confounding variables had been removed until the cumulative change in the OR of the exposure was >10%.

Results. Of the 3836 women prescreened, 790 were diagnosed as having T. vaginalis infection. Of these, 540 (68.4%) infected women participated in the study and 521 (65.9%) were included in the present analyses (19 were excluded because of missing data). Of the 203 women negative for T. vaginalis recruited as controls for the study, 177 (87.2%) agreed to participate. One uninfected woman was excluded because of missing data.

The median age of participants was 27 years (SD, 9 years), and most (91%) were non-Hispanic black (table 1). Half of the participants reported having had a household income of ⩾$12,000 during the previous year, and 35% had education/ training beyond high school or the General Educational Developmental test. Almost half (48%) of the participants had douched during the preceding 2 months, and most (78.6%) of those who had done so had used a commercial douche preparation.

Women with T. vaginalis tended to have lower levels of vaginal lactobacilli, be non-Hispanic black, and have a lower level of education and of household income than did women without T. vaginalis (table 1). The proportion of women who reported having douched during the preceding 2 months and at least once during the preceding week was higher in those with T. vaginalis than in those without T. vaginalis.

Demographic variables such as non-Hispanic black race (OR, 2.1 [95% CI, 1.2–3.7]), lower educational level (OR, 2.2 [95% CI, 1.5–3.2]), and lower household income (OR, 1.5 [95% CI, 1.0–2.3]) were positively associated with low levels of vaginal lactobacilli. Oral-contraceptive use and condom use were inversely associated with low levels of vaginal lactobacilli (OR, 0.4 [95% CI, 0.2–0.6] and 0.6 [95% CI, 0.4–0.8], respectively). Douching during the preceding 2 months was associated with lower levels of vaginal lactobacilli (OR, 1.8 [95% CI, 1.2–2.6]), as was douching more than once during an average week (OR, 2.0 [95% CI, 1.0–3.9]). Neither sexual behavior nor coinfection with either N. gonorrheae or C. trachomatis was associated with the level of vaginal lactobacilli.

In bivariable analyses, low levels of vaginal lactobacilli were positively associated with T. vaginalis (OR, 2.2 [95% CI, 1.4–3.2]) and remained so after adjustment for age, race, education, concurrent bacterial vaginosis, concurrent gonorrhea, number of sex partners during the preceding 2 months, number of times the woman had sex during an average week, condom use, and oral contraceptive use (adjusted OR, 2.8 [95% CI, 1.6–5.0]).

The results of multivariable analyses suggested that the association between the level of vaginal lactobacilli and T. vaginalis differs by educational level ( for the LR). Among P = .02 women with more than a high school education, low levels of vaginal lactobacilli were positively associated with T. vaginalis (OR, 3.3 [95% CI, 1.8–5.9]); this association was not observed among women with a high school education or less (OR, 1.1 [95% CI, 0.6–2.0]). Among women with more than a high school education, low levels of vaginal lactobacilli remained positively associated with T. vaginalis after adjustment for the same cofactors that had been adjusted for in the bivariable analyses (adjusted OR, 4.6 [95% CI, 2.2–9.5]). Among women with a high school education or less, T. vaginalis was not independently associated with low levels of lactobacilli after adjustment for potentially confounding variables (adjusted OR, 1.6 [95% CI, 0.7–3.4]). These associations, which are presented in table 2, were not substantially altered after the final model was adjusted for C. trachomatis coinfection and for previous T. vaginalis infection.

Discussion. Trichomoniasis is the most common curable STI and an apparent contributor to HIV acquisition [1]. The relationship between vaginal lactobacilli and T. vaginalis remains unclear, although it could affect important aspects of infection; for example, the presence of abnormal vaginal flora after cervicovaginal conditions and antimicrobial treatment of these conditions may promote relapse of infection [13]. A clearer understanding of the role that lactobacilli play in the natural history of T. vaginalis may help refine this therapy and possibly could improve the treatment of T. vaginalis.

In the present study, lower levels of vaginal lactobacilli were found to be positively associated with prevalent T. vaginalis. The association between the level of vaginal lactobacilli and T. vaginalis varied between women with different levels of education: among women with a high school education or less, T. vaginalis was not independently associated with low levels of vaginal lactobacilli; however, such an association was found among women with more than a high school education, after adjustment for covariables that could potentially confound this relationship.

The reason for the observed education-dependent effect between these groups of women is unclear; given the relative imprecision of the education-specific estimates, this observation may have been due to chance. However, education may also be a proxy for another factor, which we either did not measure or did not measure well. We explored whether, in multivariable models, education could be a proxy for other study variables, such as race, douching, and geographic region; however, an education-correlated difference in adjusted OR was not observed for them, and, furthermore, they were not strongly associated with education, as a proxy would be. The associations between the level of vaginal lactobacilli and T. vaginalis when multilevel indicator variables were used for the level of vaginal lactobacilli and educational level were also consistent with those observed when dichotomous coding of these variables was used. The education-correlated difference that was observed is probably due not to educational level itself but to another factor associated with education, a factor that was not captured in this study.

We were unable to explore whether the level of vaginal lactobacilli and T. vaginalis infection change over time, and prospective studies are needed to clarify whether low levels of vaginal lactobacilli increase the risk for T. vaginalis infection or whether T. vaginalis reduces the level of vaginal lactobacilli. Furthermore, vaginal flora are dynamic, and the level of lactobacilli measured in the present study may not reflect the level of lactobacilli present when the participant was infected with T. vaginalis; however, a study of vaginal flora has shown that the level of lactobacilli does not significantly change during the menstrual cycle [14].

The results of the present study suggest that lower levels of vaginal lactobacilli are associated with the presence of T. vaginalis in women with higher educational levels (or a factor related to higher educational levels) and that this relationship is not explained by other demographic, clinical, or behavioral factors measured in the present study. Elucidation of this relationship may improve our understanding of the pathogenesis, acquisition, and management of trichomoniasis.

• Received January 12, 2007.
• Accepted April 29, 2007.

• © 2007 by the Infectious Diseases Society of America