TABLE 3.

Relative risk of bacterial vaginosis^* among women who regularly practiced vaginal douching as compared with women who never douched, Longitudinal Study of Vaginal Flora (n = 3,620), Birmingham, Alabama, 1999–2003

	Adjusted relative risk	95% confidence interval
Marginal structural model†
Never douching	Referent
Douching in all study intervals	1.21	1.08, 1.38
Conventional model (univariable)‡
Never douching	Referent
Douching in all study intervals	1.81	1.60, 2.04
Conventional model (adjusted)§
Never douching	Referent
Douching in all study intervals	1.13	1.05, 1.22

^*Bacterial vaginosis was defined as a Nugent Gram stain criteria score of 7 or higher.

^†Estimated from weighted logistic regression models, with adjustment for age, race/ethnicity, high school education, and exposures reported at the previous visit: douching, bacterial vaginosis, cigarette smoking, alcohol drinking, season, sexual behavior (number of sex partners, a new sex partner, receptive oral sex, condom use, use of an insertive sex toy), menstrual regularity, time since last menstrual period, vaginal symptoms (odor, discharge, itching, irritation), vaginitis treatment (metronidazole, clindamycin, antifungal medication), contraception (oral contraceptives, tubal ligation, depo-medroxyprogesterone acetate), sexually transmitted infections (Trichomonas vaginalis, Chlamydia trachomatis, Neisseria gonorrhoeae), and vulvovaginal candidiasis.

^‡Univariable model from modified Poisson regression.

^§Multivariable model from modified Poisson regression, with adjustment for age, race/ethnicity, high school education, and history of confounding factors as listed above for the marginal structural model.