Abstract
Background. Bacterial vaginosis (BV) is the most common cause of vaginitis worldwide. Currently recommended treatments have poor efficacy and are associated with high rates of BV recurrence. We examined whether a longer duration of treatment with metronidazole or combination therapy with metronidazole and azithromycin would enhance the cure rates for BV. In addition, we examined factors other than drug therapy associated with cure.
Methods. Women with symptomatic BV (defined by a modified Amsel criteria) were enrolled in a 4-arm study that compared metronidazole for 7 days versus 14 days, plus or minus azithromycin on days 1 and 3. Data regarding interim behaviors were also obtained, as were vaginal specimens for Gram staining.
Results. At the first follow-up visit (7 days after the completion of therapy), there was a significant difference in cure rates among patients who received 7 days of metronidazole therapy, compared with those who received 14 days of therapy, combined across azithromycin therapy (P = .0003). There was no effect associated with azithromycin therapy. There were no differences in cure rates between any of the treatment groups at 21 days after completion of therapy. Abstinence or protected sex, refraining from douching, and a lower baseline Nugent score for the vaginal Gram stain were all significantly associated with cure.
Conclusions. Cure rates for BV were significantly improved by 14 days of metronidazole treatment (compared with 7 days of treatment), but the effects were not sustained, suggesting that relapse or reinfection occurred. Combination therapy with the addition of azithromycin had no benefit. Lower baseline Nugent scores—presumably reflecting less complex vaginal flora—were significantly associated with cure, as was refraining from unprotected sex and from douching.
Bacterial vaginosis (BV), the most common cause of vaginitis, is associated with public health problems, including preterm birth and acquisition and/or transmission of sexually transmitted diseases (STDs), including HIV infection [1, 2]. Although debate exists as to the causative pathogen(s), BV is characterized by increased numbers of Gardnerella vaginalis and anaerobes, along with a decrease in lactobacilli [3]. There appears to be a continuum of microbiological abnormalities, as suggested by the findings of vaginal Gram stains [4], ranging from a Gardnerella-predominant flora to a more complex flora associated with increased concentrations of anaerobes, such as Mobiluncus and Atopobium species [5]. A recent evaluation of bacterial communities in vaginal biopsy specimens revealed that BV consists of a dense biofilm in which G. vaginalis predominates [6]. Treatment of BV has been traditionally aimed at the eradication of anaerobes with use of clindamycin or metronidazole. However, cure rates, although comparable for these 2 agents, are unacceptably low (50%–80%) [7]. Mycoplasmas and Mobiluncus curtisii, which may play a role in the etiology of BV, have in vitro resistance to metronidazole [8] but are susceptible to macrolides. Failure to adequately treat these organisms could conceivably contribute to treatment failure. Alternatively, the duration of therapy may be a factor, in that a longer duration of therapy may be required for adequate treatment of a biofilm community of organisms. Some experts have recommended treatment courses lasting 10–14 days for women with recurrent disease, but there are currently no data to support this approach [9]. We conducted a randomized, double-blind, placebo-controlled trial that examined the duration of therapy with metronidazole, as well as combination therapy with metronidazole and azithromycin, to determine whether either of these approaches would result in higher cure rates for BV.
Methods
Design. A 2 × 2 factorial design was use to test for the main effects of duration of metronidazole therapy and combination therapy with azithromycin on cure rates of BV. Women were randomized to receive either (1) extended-release metronidazole (750 mg orally once per day for 7 days), followed by metronidazole placebo (once per day for 7 days), plus azithromycin placebo (1 g orally on days 1 and 3); (2) metronidazole for 7 days, followed by placebo for 7 days, plus azithromycin on days 1 and 3; (3) metronidazole for 14 days plus azithromycin placebo on days 1 and 3; or (4) metronidazole for 14 days plus azithromycin on days 1 and 3. Because of its unique pharmacokinetics, azithromycin is usually given as single-dose therapy for STDs. However, our aim was to achieve higher peak levels of azithromycin for longer duration. Although a single 2-g dose would have achieved this, the percentage of subjects experiencing nausea and vomiting with this dosing regimen is high [10]. Therefore, we chose to repeat the 1-g dose at 48 h to achieve the same peak levels and areas under the curve desired without the aforementioned adverse effects [11].
Study population. Women who attended the Jefferson County Department of Health STD Clinic (Birmingham, AL) with symptomatic BV (as determined using a modified Amsel criteria [12]) were invited to participate in the study. Subjects were required to have a vaginal pH of >4.5, a positive “whiff” test result, and clue cells noted by microscopic evaluation to be eligible. Women were excluded from the study if they were pregnant or breast-feeding, allergic to metronidazole or azithromycin, had an STD that required treatment, or had HIV infection or other chronic disease.
Randomization. A blocked randomization scheme (block size, 4.8) was used to generate a listing for packaging and labeling of study drugs. Each eligible subject who provided consent was assigned the subsequent randomization number.
Data and sample collection. At enrollment, subjects were administered questionnaires and a pelvic examination. Subjects were screened for gonorrhea, chlamydia, syphilis, and trichomonas. Presence of BV was determined by Amsel criteria [12]; however, the subject was required to have an elevated vaginal pH, positive “whiff” test result, and clue cells to be eligible. Vaginal samples were obtained for Gram staining. Women with a screening test result positive for gonorrhea or chlamydia were notified, treated, and deemed ineligible. Subjects were asked to avoid sexual intercourse and douching or, at a minimum, to use condoms during intercourse.
Follow-up visits were conducted 7 days after the completion of therapy (on day 21), 21 days after the completion of therapy (days 35–40), and then monthly for an additional 2 visits. Questionnaires addressing interim behaviors and symptoms were administered along with a pelvic examination, and additional specimens were collected for diagnosis of BV. Women were classified as cured if the vaginal pH and whiff test results normalized and if clue cells were absent. Improvement was defined as normalization of 2 of the 3 criteria, and failure was defined as normalization of 0 or 1 of the criteria. Women for whom initial therapy failed or who had symptomatic recurrences were treated with 7 days of metronidazole and dropped from the study. Women with intercurrent vaginal candidiasis were treated with fluconazole and continued in the study.
Laboratory methods. Amsel criteria were described elsewhere [13]. Vaginal smears were obtained for Gram stains, and the findings were interpreted in accordance with the method of Nugent et al. [4] by technicians blinded to the clinical diagnosis. Neisseria gonorrhoeae and Chlamydia trachomatis infections were diagnosed using ligase chain reaction (Abbott Laboratories). Cultures for Trichomonas vaginalis were performed using In Pouch TV test (BioMed Diagnostics) [14].
Sample size and power. Sample size calculation was guided by assumptions of an expected cure rate in the 7-day metronidazole arm of 70%, compared with a 90% cure rate in the 14-day metronidazole arm. We also hypothesized that, among women who received azithromycin, the initial cure rate would be 90% in the 7-day metronidazole-plus-azithromycin arm and 95% in the 14-day metronidazole-plus-azithromycin arm. On the basis of data for a total of 400 women, we could detect the difference in proportion of cure rates of metronidazole averaged across women who received and did not receive azithromycin. The sample size allowed us to detect a similar amount of difference in initial cure rates between women who received azithromycin and those who did not receive azithromycin pooled across 2 different durations of metronidazole administration. Our goal was to enroll 600 women to account for an expected 20% dropout rate for women whose screening test results would be positive for an STD and an anticipated 15% dropout rate for patients who dropped out during the study period because of other factors and for general loss to follow-up.
Statistical analyses. To test the main effect of metronidazole (7 vs. 14 days), we compared regimens that included women who did not receive azithromycin with regimens that included women who did receive azithromycin. The Breslow-Day statistic was evaluated to confirm homogeneity of effect of duration of metronidazole on cure rates of women who received and those did not receive azithromycin [15]. A nonsignificant test for homogeneity, which allowed us to test the main effect of duration of metronidazole therapy, was estimated by pooling these separate comparisons using the Mantel-Haenszel χ2 test. A similar procedure was performed to test the main effect of azithromycin pooled across duration of metronidazole. All randomized patients were included in analyses based on an intent-to-treat model. For sensitivity analyses, patients with a <80% rate of compliance with therapy were excluded from the analyses.
The significance level for the final analysis was .04907, calculated on the basis of the final accrual and 1 interim analysis [16, 17]. The proposed sample size did not allow for enough power for testing of a statistically significant interaction, although we computed the estimate and P value from logistic regression modeling. Multivariate logistic regression was used to compare initial cure rates between the durations of metronidazole and azithromycin therapy, with adjustment for confounding variables, such as new sexual partner in the past 30 days, abstinence or protected sex, and baseline Nugent scores. Model-building procedures used stepwise selection, and the fit of the models was evaluated with Akaike Information Criteria. As described above, the P value for the treatment effect was .049, although the P value for other confounders in the multivariate model was set at .05.
Temporal associations. A cumulative repeated-measures model was fit to the clinical outcome, as defined by Amsel criteria (ordered cure, improvement, and failure), to predict the lowest category (cure). Models included the duration of treatment with metronidazole, receipt of azithromycin, and other covariates, including baseline Nugent score, new sex partner since the prior visit, abstinence or consistent condom use, and current douching. ORs and 95% CIs were computed with the Wald statistic, using the Proc Genmod procedure in SAS software, version 9.0 (SAS Institute), with a repeated statement and multinomial distribution.
Safety. Continuous variables were compared using the Kruskal-Wallis test or F test, and categorical variables were compared using the χ2 test or Fisher's exact test. Drug safety was evaluated on the basis of the frequency of reported adverse events and number of patients exhibiting adverse events.
Results
Descriptive baseline characteristics. The first patient was recruited on 10 July 2002, and a total of 568 patients were randomized. The trial was completed 31 August 2005. There were no significant differences between groups with respect to demographic characteristics or behaviors (table 1). The distribution of baseline Nugent scores at study entry was as follows: 63.2% of subjects had a score ⩽8, and 36.8% had a score of 9 or 10.
Demographic and behavioral characteristics of subjects, by randomized treatment group.
Primary outcomes. At the first follow-up visit (7 days after the completion of therapy), there was a significant difference in cure rates (as measured by Amsel criteria) among patients who received 7 versus 14 days of metronidazole therapy pooled across azithromycin therapy (P = .0003) (table 2). The cure rate for 7 days of metronidazole was 44.8%, compared with 63.4% for those who were randomized to receive 14 days of therapy. There was no effect of azithromycin therapy pooled across duration of treatment (P = .57). Review of the data indicated that all women categorized as improved were categorized as such only because of a persistently elevated pH. Thus, if these women are included in the cure category, the cure rates for 7 and 14 days of therapy were 65.8% and 79.5%, respectively (P = .0002 for duration effect [14 vs. 7 days of metronidazole] at 7 days of follow-up). At the second follow-up visit, there were no significant differences between treatment groups. Table 3 lists evaluation of patients stratified by Nugent score at the first and second follow-up visits, categorized as negative (<7) or positive (⩾ 7). There was a significant effect of duration of therapy (14 vs. 7 days) pooled across both azithromycin groups (P = .029). There was no effect for azithromycin therapy (P = .49). At the second follow-up visit, there were no significant differences across treatments.
Primary outcome, as defined by modified Amsel criteria, 7 and 21 days after the completion of therapy, by randomized treatment group.
Primary outcome, as defined by Nugent criteria, 7 and 21 days after the completion of therapy, by randomized treatment group.
Among patients who were cured or improved at the first follow-up visit, the Nugent score at that visit was highly predictive of outcome at the second follow-up visit. The odds of cure for women with Nugent scores of 4–6 (intermediate flora) were ∼30% lower than the odds for those with normal scores of 0–3 (P < .0001).
Sensitivity analysis based on treatment compliance. We repeated the analyses described above, including only those patients who complied with at least 80% of the treatment regimen, as based on tablet counts at follow-up. Fifty patients were classified as noncompliant. This proportion did not differ significantly by treatment arm (7 days of metronidazole, 10.8%; 14 days of metronidazole, 9.2%; 7 days of metronidazole plus azithromycin, 8.7%; and 14 days of metronidazole plus azithromycin, 11.9%; P = .84). The duration effect persisted for patients at the first follow-up visit, whether considering the Amsel criteria definition of cure (P = .001) or BV determined by Nugent scoring (P = .076).
Multivariate models. Table 4 shows the results of the cumulative repeated measures logit model. The model includes data from the second, third, and fourth visits. Baseline Nugent score was highly associated with successful treatment of BV (P = .03). The adjusted risk of being classified as having improvement or cure was 43% greater for patients with a baseline Nugent score of 5–8, compared with those with a baseline score of 9–10 (relative risk [RR], 1.43; 95% CI, 1.02–2.02). Subjects who abstained from sex or consistently used condoms were 50% more likely to be classified as improved or cured, compared with subjects who engaged in unprotected sex (RR, 1.50; 95% CI, 1.10–2.05). Subjects who refrained from douching during the study were also significantly more likely to be classified as improved or cured, compared with subjects who reported douching 1–2 times per month (RR, 1.82; 95% CI, 1.13–2.92). There were no significant effects of having a new sexual partner in the past 30 days. Duration of treatment was not found to be statistically significant when controlling for other factors over the entire course of the study. Of note, nearly 70% of women admitted to having unprotected sex during the course of the study.
Repeated measures modeling of risk factors for cure, defined by modified Amsel criteria, controlling for treatment status
Reasons for study termination. Of 568 women enrolled and randomized in the study, 420 (73.9%) completed the study through the first follow-up visit, and 321 (56.5%) completed through at the second follow-up visit. For the third follow-up visit, data were available for 248 patients (43.7%), and only 192 patients (33.8%) completed the fourth follow-up visit. Reasons for discontinuation included loss to follow-up, pregnancy, and screening tests positive for an STD at baseline (table 5). The distribution of losses across all visits did not differ by treatment arm (P = .78).
Reason for termination from the study, by treatment group, over the duration of the study.
Drug safety. The frequency of treatment-related adverse effects did not differ by treatment group (P = .28) (table 6). The incidence of vaginal candidiasis was similar among all groups, with a mean proportion of 24.4%.
Frequency of treatment-emergent adverse events, with probable relationship, for all randomized patients.
Discussion
BV is a common vaginal infection that is associated with preterm birth and acquisition and transmission of STDs and HIV [18–20]. There is a limited understanding of its pathogenesis, including the causative organism(s) and mode of acquisition. This lack of knowledge has greatly hampered our ability to effectively treat and manage BV.
Little has changed in our therapeutic approach to BV over the past 20 years. The efficacies of the currently recommended therapies are all suboptimal, with cure rates of ⩽80% [7]. In addition, many women have frequent recurrences of BV [7]. Theories regarding the lack of efficacy of current treatments include “resistance” to currently used drugs and reinfection from a sexual partner. Duration of therapy has been suggested as a factor in the successful treatment of BV. Some experts recommend longer courses of therapy, although there have been no data to support this approach. In the original dose-ranging study of metronidazole for treatment of BV, Eschenbach et al. [21] found superior cure rates for 7 days of therapy, compared with shorter courses. However, they did not extend the studies to determine whether durations >7 days improved cure rates.
Our study was designed to test the hypotheses that combination therapy with metronidazole and a macrolide (in this case, azithromycin) was superior to single-drug treatment and that metronidazole therapy with a duration >7 days was superior to the 7-day regimen. We found that the addition of azithromycin had no effect on cure rates. Extending the duration of metronidazole therapy to 14 days resulted in a significantly greater cure rate at the short-term follow-up visit (i.e., 7 days after the completion of therapy and 21 days after enrollment in the study). The women randomized to receive 14 days of metronidazole had an overall cure rate (defined by modified Amsel criteria) of 63.4%, compared with a cure rate of 44.8% among those randomized to receive 7 days of metronidazole and 7 days of placebo (P = .0002). However, by the second follow-up visit (21 days after the completion of therapy and 35 days after enrollment), the cure rates were the same among all groups, suggesting that either relapse or reinfection occurred.
Regarding predictors of cure, we found that baseline Nugent score was significantly correlated with clinical cure of BV. Women with baseline scores of 9–10 were significantly less likely to achieve cure than were women with scores <9. The Nugent score is related to the complexity of the changes of the flora [4]. Generally, scores ⩽8 represent a predominance of morphotypes consistent with Gardnerella species, whereas scores of 9–10 indicate the addition of morphotypes consistent with Mobiluncus species. Scores of 9–10 are believed to represent vaginal flora that is more complex and more likely to have a strong anaerobic component. Anecdotally, practitioners have described a subset of patients who are recalcitrant to standard therapy and more likely to have recurrence once treated. However, up until now, there have been no published data to support this observation. Our data strongly suggests that more-complicated BV flora—perhaps one with a significant anaerobic component—is more difficult to eradicate.
Behavioral factors associated with cure of BV in our study included abstinence and/or consistent condom use and refraining from douching. Women who abstained from sex or consistently used condoms were 50% more likely to be cured or to improve than were women who had unprotected sex. This is consistent with data from other studies. In a prospective study of incident BV, use of condoms with occasional partners was protective against acquisition of BV [22]. A study of treatment options for BV found that relapse between the first and second follow-up visits was significantly associated with an interim behavior of unprotected sex [23]. An Australian study found that women with BV were significantly less likely than the control group to have used condoms consistently [24]. Yet another study found that sexual exposure to a regular partner was significantly associated with recurrent BV [25]. These studies suggest that BV may be sexually transmitted and that reexposure to the causative agent is associated with relapse. Thus, one hypothesis for the failure of the treatment benefit of 14 days of metronidazole to be sustained beyond the first follow-up visit in our study is that 70% of women participated in unprotected sex during the course of the study and were reexposed to a causative agent.
Limitations to our study include retention difficulties, which are inherent in studies involving high-risk populations, as well as the extent of high-risk intercurrent behavior, such as unprotected sex.
In summary, we have shown a significant increase in short-term cure rates for BV by extending duration of therapy with metronidazole to 14 days. However, the benefit of prolonged duration of therapy was not sustained beyond 7 days after completion of therapy, perhaps as a result of intercurrent behaviors. Baseline Nugent scores ⩽8 were significantly associated with a more favorable response to therapy than were scores of 9–10. Consistent use of condoms and refraining from douching were also significantly associated with cure.
Acknowledgments
Pfizer kindly supplied azithromycin and metronidazole.
Financial support. National Institutes of Health (grant Therapy and Prevention of Bacterial Vaginosis R01 AI048044).
Potential conflicts of interest. All authors: no conflicts.
Comments