Article Text

Download PDFPDF

Prospective study of correlates of vaginal Lactobacillus colonisation among high-risk HIV-1 seronegative women
  1. J M Baeten1,2,
  2. W M Hassan3,4,
  3. V Chohan4,
  4. B A Richardson5,
  5. K Mandaliya6,
  6. J O Ndinya-Achola4,
  7. W Jaoko4,
  8. R S McClelland2,3,4
  1. 1
    Department of Global Health, University of Washington, Seattle, Washington, USA
  2. 2
    Department of Medicine, University of Washington, Seattle, Washington, USA
  3. 3
    Department of Epidemiology, University of Washington, Seattle, Washington, USA
  4. 4
    Department of Medical Microbiology, University of Nairobi, Nairobi, Kenya
  5. 5
    Department of Biostatistics, University of Washington, Seattle, Washington, USA
  6. 6
    Coast Provincial General Hospital, Mombasa, Kenya
  1. Dr J Baeten, Department of Global Health, University of Washington, 901 Boren Ave, Suite 1300, Seattle, WA 98104, USA; jbaeten{at}


Objective: Vaginal colonisation with Lactobacillus species is characteristic of normal vaginal ecology. The absence of vaginal lactobacilli, particularly hydrogen peroxide (H2O2)-producing isolates, has been associated with symptomatic bacterial vaginosis (BV) and increased risk for HIV-1 acquisition. Identification of factors associated with vaginal Lactobacillus colonisation may suggest interventions to improve vaginal health.

Methods: We conducted a prospective cohort study of correlates of vaginal Lactobacillus colonisation among Kenyan HIV-1 seronegative female sex workers. At monthly follow-up visits, vaginal Lactobacillus cultures were obtained. Generalised estimating equations were used to examine demographic, behavioural and medical correlates of Lactobacillus isolation, including isolation of H2O2-producing strains.

Results: Lactobacillus cultures were obtained from 1020 women who completed a total of 8896 follow-up visits. Vaginal washing, typically with water alone or with soap and water, was associated with an approximately 40% decreased likelihood of Lactobacillus isolation, including isolation of H2O2-producing strains. Recent antibiotic use, excluding metronidazole and treatments for vaginal candidiasis, reduced Lactobacillus isolation by ∼30%. H2O2-producing lactobacilli were significantly less common among women with Trichomonas vaginalis infection and those who were seropositive for herpes simplex virus type 2. In contrast, H2O2-producing lactobacilli were significantly more common among women with concurrent vaginal candidiasis.

Conclusions: Modifiable biological and behavioural factors are associated with Lactobacillus colonisation in African women. Our results suggest intervention strategies to improve vaginal health in women at high risk for HIV-1.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The normal vaginal microflora in women of reproductive age is composed predominantly of Lactobacillus species.1 These Gram-positive rods create an inhospitable environment for infectious pathogens by producing antimicrobial substances including hydrogen peroxide (H2O2) and lactic acid, which maintains a low vaginal pH.2

There is increasing recognition that vaginal lactobacilli may mediate HIV-1 risk. In vitro, H2O2-producing lactobacilli are virucidal to HIV-1.3 One longitudinal study found a twofold increased HIV-1 risk among women without vaginal lactobacilli.4 Moreover, in several prospective studies, bacterial vaginosis (BV), a vaginal infection characterised by decreased Lactobacillus colonisation, has been associated with elevated HIV-1 risk.57 Epidemiological studies also found an association between a lack of lactobacilli and the presence of sexually transmitted infections (STIs), such as Neisseria gonorrhoeae4 and Trichomonas vaginalis,8 which are risk factors for HIV-1 acquisition and could further amplify HIV-1 risk.

Understanding factors that influence vaginal Lactobacillus colonisation may offer an insight into novel vaginal health strategies to prevent HIV-1 transmission to women. We used data from a prospective observational cohort study to examine correlates of vaginal Lactobacillus isolation among African women at high risk for HIV-1.


Study population

In 1993, a prospective open cohort study was initiated of risk factors for HIV-1 and STI acquisition among women attending a municipal sex-worker clinic in Mombasa, Kenya.9 At monthly follow-up, data on recent sexual behaviour and contraceptive use were recorded, and genital and blood samples were collected. STIs were treated according to Kenyan national guidelines, either syndromically at the time of examination or on the basis of laboratory testing at a scheduled results visit after 1 week. At each visit, risk reduction counselling and free condoms were provided, as well as primary care medical services. The study protocol was approved by the Kenyatta National Hospital/University of Nairobi Ethical Review Committee and the Human Subjects Review Committee of the University of Washington. Participants provided informed consent.

Laboratory methods

HIV-1 serology was by ELISA (Detect HIV, Biochem ImmunoSystems, Montreal, confirmed by Recombigen, Cambridge Biotech, Worcester, Massachusetts). Serological testing for herpes simplex virus type 2 (HSV-2) was performed on archived samples using a type-specific HSV-2 gG based ELISA (HerpeSelect 2, Focus Diagnostics, Cypress, California), as described previously.10

Endocervical secretions were cultured on modified Thayer–Martin media for N gonorrhoeae and tested for Chlamydia trachomatis antigen by EIA (MicroTrak II, Syva, San Jose, California). Testing for C trachomatis was discontinued in June 1999 because of a low incidence in the cohort (which may have related to low sensitivity of the EIA-based testing). Cervicitis was defined by a mean of >30 polymorphonuclear cells in three oil-immersion fields of Gram-stained cervical secretions.

Samples of vaginal secretions were examined microscopically using normal saline and 10% potassium hydroxide (KOH) for the presence of motile T vaginalis and yeast. We defined vaginal candidiasis (VC) as the presence of yeast buds and/or pseudohyphae. BV was diagnosed by Nugent criteria as a score of ⩾7 on microscopic evaluation of Gram-stained vaginal fluid.11

Beginning in 1995, Lactobacillus species were isolated by culture of vaginal secretions collected at each monthly visit. Samples were directly inoculated onto Rogosa agar using Dacron swabs. Plates were incubated in anaerobic jars at 37°C for 72 h. Colonies were identified as Lactobacillus species on the basis of morphology and Gram stain appearance. Three Lactobacillus colonies from each plate were selected and stored in skim milk at −20°C for batch testing for H2O2 production, which was done by subculturing isolates on tetramethylbenzidine (TMB) agar containing horseradish peroxidase. After 72 h of anaerobic incubation, subculture plates were exposed to air and H2O2-producing isolates were identified by visualisation of a blue coloration resulting from oxidation of TMB.

Data analysis

Data were analysed using SPSS version 15.0 (SPSS, Chicago) and S-PLUS 2000 (MathSoft, Seattle, Washington). All visits between May 1995, when Lactobacillus cultures were begun, and May 2003, when a randomised trial of periodic presumptive treatment of vaginal infections was initiated in the cohort,12 were selected. Visits that took place as part of two studies of the vaginal microbicide nonoxynol-9 were excluded from the analysis.13 14 Women who acquired HIV-1 during follow-up (n = 120) were censored at their last HIV-1 seronegative visit.

To account for multiple measurements from individual women during prospective follow-up, we used generalised estimating equations with a logit link, exchangeable correlation structure and robust variance estimates to examine correlates of Lactobacillus isolation. All models adjusted for the total number of visits contributed by each woman. Two outcomes were defined: (1) the presence of any lactobacilli and (2) the presence of H2O2-producing lactobacilli. Univariate analyses were performed for a number of characteristics including demographic factors, contraceptive use, sexual behaviour, concurrently diagnosed genital tract infections and recent antibiotic use. Multivariate models were then constructed using those variables demonstrating statistically significant associations (at p⩽0.05) in a univariate analysis.

Demographic and behavioural correlates included age, education, parity, tobacco use, alcohol use and intravaginal washing practices. Of these, age was assessed as a time-dependent variable, while the remainder were from study enrolment data.

Contraceptive use was assessed in a time-dependent fashion. A 70-day window of persistent effect was estimated for women who changed their contraceptive method during follow-up, as we have done previously.10 Sexual behaviour correlates included number of sex partners and the occurrence of any unprotected sex, assessed as time-dependent variables measured during the week prior to each clinic visit. Sex during menses and whether lubrication was used for sex, both measured only at enrolment, were also assessed. The relationship between Lactobacillus isolation and the presence of concurrent genital tract infections (measured as time-dependent variables) was assessed.

Recent antibiotic use (as a time-dependent factor) was classified into three categories: (1) metronidazole, (2) treatment of VC (usually intravaginal clotrimazole or nystatin pessaries) and (3) any other antibiotics. Agents used for treatment of tuberculosis or helminths were excluded. Antibiotics were considered to have an effect for 60 days after administration, as we have done previously.15


Between May 1995 and May 2003, 1020 HIV-1 seronegative women contributed 8896 follow-up visits at which vaginal Lactobacillus cultures were performed. The median number of Lactobacillus cultures per woman was 5 (interquartile range (IQR) 2 to 11, range 1 to 84), and the median time between visits was 36 days (IQR 28 to 67). Most women worked as barmaids (table 1) and supplemented their income with sex work.

Table 1 Enrolment characteristics of study population (N = 1020)

Lactobacillus cultures were collected at enrolment for 775 women (ie, those who enrolled after May 1995). Of these, 169 (22%) had any lactobacilli, and 75 (10%) had H2O2-producing lactobacilli. During follow-up, Lactobacillus species were isolated at 1694 (19.0%) follow-up visits, of which 731 (8.2% of total follow-up visits) had H2O2-producing lactobacilli isolated. Five hundred and thirty women (52.0%) had lactobacilli detected at least once during follow-up, and 309 (30.3%) had H2O2-producing lactobacilli detected at least once.

Recent antibiotic exposure was present at 559 (6.3%), 214 (2.4%) and 639 (7.2%) follow-up visits for metronidazole, agents for treatment of VC and other antibiotics, respectively.

Correlates of detection of Lactobacillus species

In univariate analysis, older age, alcohol use, vaginal cleansing, current bacterial vaginosis, HSV-2 seropositivity, recent use of metronidazole and recent use of other antibiotics (except those for treatment of VC) were associated with decreased Lactobacillus isolation (table 2). Tobacco use and a higher number of sexual partners were associated with increased detection.

Table 2 Univariate and multivariate correlates of overall Lactobacillus isolation among HIV-1 seronegative women

In the multivariate analyses, two important factors were significantly associated with decreased Lactobacillus isolation: (1) vaginal washing, which decreased detection by approximately 40% both among women who used water alone and among those who used soap and (2) recent use of antibiotics other than metronidazole or preparations to treat VC, which decreased detection by approximately 30%. There was a stepwise decrease in the likelihood of Lactobacillus isolation with older age. A higher number of sexual partners (>1 per week) was associated with a 30% greater likelihood of Lactobacillus detection.

Correlates of detection of H2O2-producing Lactobacillus species

A number of factors associated with overall vaginal Lactobacillus colonisation were also associated with vaginal H2O2-producing-Lactobacillus colonisation. In univariate analysis, older age, parity, alcohol use, vaginal cleansing, current BV, T vaginalis infection, HSV-2 seropositivity and recent use of other antibiotics were associated with decreased Lactobacillus isolation (table 3). Tobacco use, cervicitis, VC and recent treatment for VC were associated with increased detection.

Table 3 Univariate and multivariate correlates of H2O2-producing Lactobacillus isolation among HIV-1 seronegative women

In multivariate modelling, vaginal washing and recent use of antibiotics other than metronidazole or preparations to treat VC were associated with significantly decreased detection of H2O2-producing lactobacilli, with magnitudes of effect similar to those seen in multivariate analysis for overall Lactobacillus isolation. Women older than 40 had an approximately 50% decreased likelihood of isolation of H2O2-producing lactobacilli. Alcohol use was not consistently associated with H2O2-producing Lactobacillus colonisation, although an intake of 1–7 drinks per week was associated with an approximately 40% decrease in detection. HSV-2 seropositivity and T vaginalis infection were associated with decreased detection of H2O2-producing lactobacilli, whereas the identification of these Lactobacillus strains was increased when concurrent cervicitis and VC were detected.

Bacterial vaginosis and vaginal Lactobacillus

In both multivariate models, bacterial vaginosis was associated with an ∼40% reduced likelihood of Lactobacillus detection. Because decreased Lactobacillus colonisation may be reflective of bacterial vaginosis pathogenesis, we performed additional multivariate models, excluding the bacterial vaginosis variable—results for other variables in those models were essentially unchanged (results not shown).


In this large, prospective study of HIV-1-uninfected Kenyan women, we found that biological and behavioural factors were associated with isolation of vaginal Lactobacillus species, a key component of healthy vaginal ecology. The identification of modifiable biological factors that influence vaginal Lactobacillus colonisation may have important implications for the design of vaginal health strategies to prevent HIV-1 acquisition in women.

Intravaginal practices such as douching or wiping inside the vagina with water or antiseptic preparations have been associated with increased risk for HIV-1 acquisition.6 We previously reported elevated HIV-1 risk among women who performed vaginal washing: adjusted hazard ratio (aHR) 2.64 (95% CI 1.00 to 6.97) for those who used water and aHR 3.84 (95% CI 1.51 to 9.77) for those who used soap, compared with those who did not vaginally cleanse.16 An unadjusted preliminary analysis demonstrated decreased Lactobacillus colonisation associated with vaginal washing.17 Another report from our cohort found that Lactobacillus colonisation was associated with decreased HIV-1 incidence.4 The results of the present study bridge and reinforce these earlier findings, suggesting that decreased Lactobacillus colonisation as a result of vaginal washing might be an important factor mediating the relationship between intravaginal practices and HIV-1 risk in this population. One limitation of our study was that vaginal washing was measured at enrolment, and practices may have changed during follow-up; however, our previous report, associating vaginal washing with HIV-1 risk, used this same measure.16 Our findings parallel the results of studies in US women, which have also found douching to be associated with BV and decreased vaginal colonisation with H2O2-producing lactobacilli.18 19

Antibiotic use has been reported as a risk factor for loss of H2O2-producing lactobacilli,20 and our results confirm this association. We also found decreased Lactobacillus isolation after metronidazole use and increased H2O2-producing Lactobacillus isolation after antifungal treatment, although these did not achieve statistical significance in multivariate analysis. These latter associations may reflect confounding by indication, as metronidazole was very frequently used to treat BV and T vaginalis infections in the cohort, while the antifungals were used to treat VC. These vaginal conditions were associated with decreased (BV and T vaginalis) and increased (VC) identification of Lactobacillus, and perturbations of the vaginal flora may persist or recur after treatment.

The VC results are consistent with a recent analysis of women from this cohort who participated in a clinical trial of oral periodic presumptive treatment of vaginal infections between 2003 and 2006.12 Among women in the control arm, concurrent colonisation with H2O2-producing Lactobacillus was associated with increased risk of symptomatic VC (adjusted OR 3.31, 95% CI 1.09 to 10.08), while prior metronidazole treatment was associated with a trend towards increased risk of symptomatic VC (adjusted OR 6.41, 95% CI 0.68 to 60.25).21 Given that BV, trichomoniasis and VC have all been associated with increased HIV-1 risk,22 new approaches to treating vaginal pathogens need to investigate methods to support Lactobacillus colonisation without trading one vaginal infection for another.

Key messages

  • In this prospective longitudinal cohort study among Kenyan women, modifiable biological and behavioural factors were associated with Lactobacillus colonisation.

  • Vaginal washing, with water alone or with soap and water, was associated with an approximately 40% decreased likelihood of Lactobacillus isolation.

  • Recent antibiotic use reduced Lactobacillus isolation by ∼30%.

  • H2O2-producing lactobacilli were less common among women with concurrent Trichomonas vaginalis infection and among those who were seropositive for herpes simplex virus type 2.

HSV-2 seropositivity was associated with an ∼30% decrease in isolation of H2O2-producing lactobacilli in this cohort, consistent with other studies.23 24 Subclinical HSV-2 reactivation is frequent in seropositive persons (∼20% of days) and leads to long-lasting immunological changes to the genital mucosa.25 HSV-2 seropositivity has been associated with more than a threefold increased risk of HIV-1 acquisition,26 including in this cohort.10 While the HSV-2/HIV-1 association has generally been hypothesised to be mediated through changes to mucosal integrity and mucosal immune cell populations, it is interesting to consider that an HSV-2 effect might be mediated in part through changes in vaginal ecology.

Older age was associated with a stepwise decrease in Lactobacillus isolation, potentially reflecting lower estrogen levels among older women. However, hormonal contraceptive use was not associated with Lactobacillus isolation. We and others have previously shown that hormonal contraception decreases BV risk.23 27 Taken together, these data suggest that hormonal contraception may reduce the level of disturbances of the vaginal flora, without promoting Lactobacillus colonisation.

We were surprised to find that a higher number of sex partners was associated with detection of Lactobacillus (though not with H2O2-producing lactobacilli). Previous studies have found lactobacilli to be less common and BV more common among women with a higher number of sexual partners.19 23 The reason for the association observed in our study is not clear but could relate to incompletely controlled differences in sexual risk (eg, reduced partner number when women had symptomatic vaginal discharge).

There are several strengths to this study. The large number of women and total number of visits make this the largest study of correlates of vaginal Lactobacillus colonisation to date. Longitudinal and frequent follow-up permitted careful analysis of time-dependent factors like age, antibiotic use, contraception and sexual behaviour.

We found a low colonisation rate of vaginal Lactobacillus, especially H2O2-producing lactobacilli. The low isolation rates may be attributed to our population (Kenya vs US, where most studies of vaginal flora have been conducted) or isolation methods (culture versus PCR28). Some Lactobacillus species, like L iners (which is common among African women28), do not grow on Rogosa agar, and our colonisation rates may be underestimates. The high prevalence of vaginal washing and high incidence of BV in our study population also likely contributed to low Lactobacillus isolation rates. Finally, the results of this study may not be generalizable to all populations—they may be most directly relevant to women at high risk for HIV-1; HIV-1 incidence in this cohort was >8% per year during the period included in this analysis.10

Our findings suggest that modifiable biological and behavioural correlates—intravaginal practices, antimicrobial use and genital tract infections—influence vaginal colonisation with Lactobacillus among African women at high risk for HIV-1. Increasing attention to vaginal health interventions as an HIV-1 prevention strategy12 requires continued study of risk factors for vaginal pathogens and correlates of healthy vaginal flora. New interventions aimed at improving vaginal Lactobacillus colonisation have the potential to alter HIV-1 susceptibility.


Authorship contributions: study concept—JMB, WMH, RSM; clinical and laboratory data collection—JMB, WMH, VC, KM, JON-A, WJ, RSM; data analysis—JMB, BAR; first draft of manuscript—JMB, WMH; critical review and approval of final draft of manuscript—JMB, WMH, VC, BAR, KM, JON-A, WJ, RSM.

We thank the clinical, laboratory and administrative staff in Mombasa and Seattle for their valuable contributions to this work, the Mombasa Municipal Council for allowing use of clinical facilities and Coast Provincial General Hospital for providing laboratory space, and the study participants for their time and dedication.



  • Funding: This work was supported by US National Institutes of Health (NIH) through grants R01-AI33873, R01-AI52480, R01-AI38518, D43-TW00007 and K23-AI52480, and by Family Health International (subcontract N01-A1-35173-119).

  • Competing interests: None.

  • Ethics approval: Ethics approval was provided by the institutional review boards of the University of Washington and the University of Nairobi.

  • Patient consent: Obtained.