The study population was comprised of women (≥18 years) who responded to recruitment through newspaper advertisements, posted fliers, and referral from the Johns Hopkins Bayview Department of Obstetrics and Gynecology between December 2005 and March 2007. Inclusion criteria were report of regular menstrual cycle and any douching in the two months before screening.
Because of the theoretical risk associated with douching and no known health benefit, we chose a cessation design for the preliminary study. A 4-week douching observation interval (phase I), in which participants were expected to douche at least once, was initiated on day one of the menstrual cycle and was followed by a 12-week douching-cessation interval (phase II). During the douching cessation phase, women were asked to discontinue the use of all feminine hygiene products including douches, wipes, creams, sprays, vaginal films and powders. Participants were encouraged to use unscented soaps and to only perform external genital washing. In the final four weeks of the study, participants chose to resume douching or continue douching cessation (phase III). At the end of the study (week 20), participants were contacted by phone and surveyed on their current douching practice. Participants were instructed to douche at their preferred frequency in phase I and phase III and to choose products that they normally used. All were asked to return for a visit at week 4 for the douching cessation educational intervention and a final visit at week 16.
Participants used daily diaries, in the form of a yes/no check-off list, to report menstrual bleeding, vaginal douching, sexual activity (including vaginal intercourse, receptive oral sex, digital penetration, rectal sex, sex toys, condoms, spermicides, lubricants), thong undergarment, medications, or use of a diaphragm, sanitary napkin and/or tampons. Trained female staff conducted 15-minute questionnaires at baseline and at the end of phases II and III. At baseline, women were surveyed on demographic factors, including age, race, educational attainment, mother’s education, and contraceptive use. In addition, baseline and follow-up surveys included a collection of information on sexual activity and douching practices. A douche product was described to participants as “washing or cleansing inside your vagina.” At all visits, women were asked about their frequency, and with multiple selections, reasons for douching and product types used.
Self-collected vaginal samples were obtained twice weekly in phases I and II and a final sample in week 20. Self-collected swabs demonstrate high overall and morphotype-specific validity compared to provider-collected swabs.28
Participants were instructed to insert the vaginal swab 1–2 inches into the vagina, twist the swab to collect material on all sides of the cotton tip, wipe in several full circles on the vaginal wall, keep in the vagina for 20 seconds, and then roll each swab across a slide and allow the material to air-dry. The slides were labeled only with a unique study identifier and date of sample. Participants placed the slide into a storage container and mailed the slides and behavioral diaries to our laboratory weekly. Study staff logged incoming postal mail, and if a participant failed to submit her weekly samples, she was immediately contacted and reminded to send in the samples. The slides were heat-fixed and Gram-stained, read in random order and blinded to phase of study. A microscopy score of 0 to 10 was assigned using the standardized method described by Nugent29
. A Nugent Gram stain score of 0–3 was normal, 4–6 intermediate and 7–10 categorized BV. One experienced microbiologist scored all slides.
At baseline and final visits, we conducted beta human chorionic gonadotropin (hCG) tests on urine and nucleic acid amplification tests (NAATs) for Chlamydia trachomatis (CT) and Neisseria gonorrhoeae (GC) (Becton Dickinson, Sparks, Maryland, BD ProbeTec ET) on self-collected swabs. Women who tested positive for GC or CT were referred to the Baltimore City STD clinic for treatment and partner notification procedures and were permitted study enrollment at a later date. Participants were excluded if they were pregnant. Women not reporting the use of intrauterine device or hormonal contraception or a history of tubal ligation were asked to return for hCG pregnancy testing if the daily diaries demonstrated more than 35 days between menstrual cycles. One participant tested positive for CT at the screening visit, was treated successfully, but she subsequently was lost to follow-up before she began study sampling. One participant was not eligible due to pregnancy at the screening visit.
Participants were instructed to contact the study investigator if they experienced any genital or pelvic symptoms during the study. Symptomatic women were offered an appointment with the Johns Hopkins Center for Reproductive Tract Infections, or were referred to the Baltimore City Health Department free STD clinics or their personal clinician. Medications were noted by the participant on the daily diary forms.
Conditional logistic regression (CLR) was used in this crossover design to evaluate the risk of BV in a participant’s douching cessation phase compared to her douching observation (phase II vs phase I).30
Each woman acts as her own control, and therefore the effect of variation (and confounding) that occurs between participants is eliminated. The matched strata were made up of observations from each study phase and the matching factor was participant identification number. CLR accounts for within subject correlations. Time-varying factors occurring in the three days before sampling that had been identified on the basis of previous literature, biologic plausibility and preliminary univariable analyses as possible confounders were included in the multivariable models. Tests for multiplicative interaction were performed by examining stratum-specific results. One participant reported douching during the final week of the cessation phase; however, data were analyzed according to the intent to treat principle. 95% CIs are based on robust estimations of the standard errors. We also evaluated the data using random effects statistical models and the results were similar (data not shown). Data were analyzed using STATA/SE 9.2 for Windows (Stata Corporation, College Station, Texas).
The protocol was approved by the Institutional Review Board of the Johns Hopkins University School of Medicine. All participants provided written informed consent.