This study was conducted between March 2004 and December 2008 among HIV-1-seropositive women between 18 and 45 years old in Mombasa, Kenya. The participants were recruited from within a larger cohort of female sex workers in Mombasa . Women eligible for ART initiated a first-line regimen of stavudine or zidovudine, lamivudine, and nevirapine, as recommended by the World Health Organization (WHO) and the Kenyan Ministry of Health Guidelines at the time . All participants gave written informed consent. The study was approved by the ethical review committees of the Kenya Medical Research Institute, the University of Washington, and the Fred Hutchinson Cancer Research Center.
Participants in the ART cohort were asked to return for monthly follow-up visits. During each visit, study nurses completed a standardized interview covering medical, gynecological, and sexual history. The study physician performed a general physical examination and pelvic speculum examination with collection of specimens for laboratory diagnosis of genital tract infections. A Dacron swab was used to collect vaginal secretions for HIV-1 RNA quantitation by placing the swab firmly on the vaginal wall and rolling 3 times between the fingertips. The swab was then placed into a cryovial with freezing medium (70% RPMI, 20% fetal calf serum, and 10% dimethyl sulfoxide with added penicillin, streptomycin, and amphotericin B). Genital samples collected for HIV-1 RNA viral load were stored at -70°C until they were shipped on dry ice to the Fred Hutchinson Cancer Research Center in Seattle for testing. Blood samples were collected every three months for CD4 lymphocyte count.
Treatment of vaginal infections
Women with symptomatic vaginal conditions (vaginal discharge, itching, or burning) at an examination visit were treated syndromically according to WHO and Kenyan Ministry of Health Guidelines (oral metronidazole 2 g as a single dose combined with clotrimazole 100 mg intravaginal pessaries nightly at bedtime for 7 nights) . All genital specimens were obtained prior to provision of treatment. Patients were asked to return one week after each examination visit for laboratory results. At the results visit, additional treatment was provided if indicated based upon laboratory findings. Those with T. vaginalis infection received treatment with single dose oral metronidazole (2 g) only if metronidazole was not dispensed at the examination visit. Samples collected one month post-treatment were used to confirm cure. Women with persistent T. vaginalis infection were retreated, and then re-evaluated at the next monthly visit. Participants whose T. vaginalis infection resolved after treatment (including retreatment) were included in this study.
HIV-1 serostatus was determined by ELISA (Detect HIV1/2, BioChem Immunosystems, Montreal, Canada). Positive tests were confirmed using a second ELISA (Recombigen, Cambridge Biotech, Worcester, Massachusetts, USA). Quantitation of CD4 lymphocytes was performed using a manual system (Cytosphere, Coulter, Hialeah, Florida, USA) from March 2004 until October 2004, and thereafter by an automated method (FACS Count, Becton Dickinson, Forest Lakes, New Jersey, USA). HIV-1 RNA quantitation was performed using a Gen-Probe viral load assay on vaginal samples (Gen-Probe, San Diego, California, USA) .
Vaginal saline wet mounts were examined microscopically at 40X power for the presence of motile T. vaginalis parasites and yeast cells or hyphae. Vaginal Gram-stained slides were evaluated for bacterial vaginosis (BV) using Nugent's criteria . Bacterial vaginosis was defined as a Nugent's score of 7-10. Culture for Neisseria gonorrhoeae was performed on modified Thayer-Martin media (Difco BD Diagnostics, Oxford, UK). Endocervical samples were also tested for the presence of N. gonorrhoeae and Chlamydia trachomatis by transcription mediated amplification using the Aptima GC/CT Detection System (Gen-Probe, San Diego, California, USA). The presence of sperm in the genital tract was determined by microscopic examination of Gram stained slides of cervical and vaginal secretions.
Visits before, during and after successful treatment of T. vaginalis infection were included in this analysis. Women with concurrent cervical infections were excluded because previous studies demonstrate that cervical infections may influence cervicovaginal HIV-1 shedding [16, 17]. We also excluded episodes in which co-infection with both yeast and T. vaginalis were present. Cases with concurrent BV were not excluded, as BV commonly coexists with T. vaginalis, and our prior prospective studies have not shown a decrease in HIV-1 RNA shedding when women are treated for BV . For this analysis, only treated episodes of T. vaginalis infection that were followed by a visit with cure were included. For women with more than 1 consecutive visit with T. vaginalis infection, only the first visit with trichomoniasis was included in the analysis.
Our a priori sample size calculation was based on the assumption that the majority of women with a genital infection would have quantifiable genital HIV-1 RNA levels. Based on this assumption, 31 cases would allow 90% power to detect a 0.6 log10 copies/swab difference in vaginal HIV-1 RNA levels using a two-sided hypothesis test at α = 0.05. We have observed reductions in vaginal HIV-1 RNA greater than 0.6 log10 copies/swab with successful treatment of T. vaginalis in studies of ART-naïve women .
The lower limit for linear quantitation for this assay was 100 copies/ml, which corresponds to 100 copies/swab (swabs were placed in 1 ml of freezing media). Because the majority of vaginal HIV-1 RNA levels were below 100 copies/ml even at the time of T. vaginalis infection, we followed our pre-specified alternative analysis plan, which dichotomized the outcome (vaginal HIV-1 RNA) as detectable versus undetectable (< 100 copies/ml). Analyses were performed using SPSS 13.0 (SPSS Inc., Chicago, Illinois, USA) and STATA 9 (StataCorp, College Station, Texas, USA). The primary analysis utilized logistic regression with generalized estimating equations with an exchangeable correlation matrix to compare the presence of detectable HIV-1 RNA in vaginal secretions across the three visits, using the pre-infection visit as the baseline.
Our analysis plan included adjustment for number of months since ART initiation in all models, because increasing duration since ART initiation would be expected to be associated with greater levels of viral suppression, particularly during the first six months of treatment. We then considered adjustment for other potential confounding factors. Adjustment variables considered were adherence to ART (measured by visual analogue scale [VAS], pill count, or pharmacy refill timing), hormonal contraceptive use (use of oral contraceptives, Norplant or depot medroxyprogesterone acetate), presence of concurrent BV, vaginal washing (none, water only, water and soap), and week of the menstrual cycle (week 1-3, week 4 and 5, amenorrheic). We used two approaches to limit the number of potential confounding factors to include in the final model. First, we assessed the change across the three visits for each potential confounding variable, and second, we assessed the association between each variable and the outcome. In the final adjusted model, we only included variables that differed significantly across the three visits or were significantly related to the outcome (p = 0.1).