JOURNAL OF WOMENS HEALTH

Volume 18, Number 12, 2009

Mary Ann Liebert, Inc.
DOI: 10.1089=jwh.2008.1088

Treatment Considerations for Bacterial Vaginosis

and the Risk of Recurrence
Judy Y. Chen, M.D., MSHS,1 Haijun Tian, Ph.D.,1 and Richard H. Beigi, M.D., M.Sc.2

Abstract

Background: Recommended regimens for the treatment of bacterial vaginosis (BV) have similar efficacy; thus,
the choice of treatment should consider additional factors such as risk of BV recurrence and side effect profile.
The purpose of this study was to investigate BV recurrence rates and rates of acquiring vulvovaginal candidiasis
(VVC) after different BV treatments in a commercially insured population.
Methods: Private administrative insurance claims from 2004 to 2006 were used. Study subjects were continuously enrolled females 1250 years of age who filled prescriptions for BV treatment (n 32,268). The four BV
treatments (single-dose clindamycin vaginal cream (2%), multiple-dose clindamycin vaginal regimens, vaginal
metronidazole, and oral metronidazole) were compared for rates of recurrent BV and VVC after treatment using
multivariate analyses. Covariates included sociodemographic and clinical characteristics.
Results: Overall, the rate of BV recurrence (2.7%), and VVC posttreatment (2.9%) were low. Women who were
treated with single-dose clindamycin vaginal cream (2%) showed no significant difference from women treated
with oral metronidazole in the likelihood of BV recurrence. However, women who received other vaginal
treatments were significantly more likely to experience BV recurrence compared with women who received oral
metronidazole ( p < 0.01). Moreover, women who were treated with single-dose clindamycin vaginal cream (2%)
and vaginal metronidazole were significantly less likely to have VVC compared with those treated with oral
metronidazole ( p < 0.01).
Conclusions: This study suggests that single-dose clindamycin vaginal cream (2%) may be a good alternative to
oral metronidazole for the treatment of BV, given the low rates of recurrence and subsequent VVC demonstrated
in this analysis.

Introduction

acterial vaginosis (BV) is a highly prevalent condition

affecting almost one third of reproductive women in the
United States at some point in their lifetime.1 It is the most
common cause of vaginal irritation, more common than either
vulvovaginal candidiasis (VVC) or trichomoniasis.1,2 BV has
been associated with many untoward reproductive sequelae,
including spontaneous abortion, preterm labor and birth,
postpartum endometritis, acquisition and transmission of
numerous sexually transmitted diseases (STDs), including
human immunodeficiency virus (HIV) and pelvic inflammatory disease (PID), infections after gynecological surgeries,
and urinary tract infections (UTIs).1,2 In addition, a recent
prospective study demonstrated that women treated for
asymptomatic BV were less likely to acquire STDs.3 Thus,
effective treatment regimens that minimize the risk for re-

1Health

currence and are easily tolerated, with fewer side effects, are
critically important.
Recommended treatments for BV include both intravaginal
and oral formulations of the antibiotics clindamycin and
metronidazole.2 As current BV treatment regimens are reported to have similar initial clinical cure rates,48 the choice of
treatment should also consider such factors as patients preference of the route of administration, frequency of administration, and side effect profile. The recurrence rate of BV in the
year after treatment has been reported to be as high as 58%.9
As the risk of recurrence may be related to the antimicrobial
activity of the drug, this is another important determinant in
the selection of BV treatment.
Intravaginal antimicrobial treatments for BV may be more
advantageous than oral therapies because of less gastrointestinal side effects, such as nausea, vomiting, and taste aversion.7,8 However, many women may dislike the inconvenience

Benchmarks, Inc., IMS Health, Woodland Hills, California.

Hospital of the University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania.

2Magee-Womens

1997

1998
of repeatedly administering intravaginal therapy, leading to a
negative impact on treatment compliance.4 Thus, an equally
effective single-dose clindamycin vaginal cream may be
beneficial for women who prefer vaginal treatments, as it has
been demonstrated to improve compliance and tolerability.4
Little is known about the BV recurrence rate using different
antimicrobial therapies. In addition, although antimicrobial
treatment is a well-known risk factor for acquiring VVC,10,11
no study to date has examined the risks of acquiring VVC
after various antimicrobial treatments for BV. The purpose of
this study was (1) to investigate the rates of BV recurrence
after treatment using different antimicrobial therapies and (2)
to assess the differential risk of acquiring VVC after use of
these same standard BV antimicrobial therapies in a commercially insured population.
Materials and Methods

CHEN ET AL.
ronidazole (n 1,545), or multiple-dose vaginal clindamycin
treatments (n 320) and patients who filled more than a 1-day
supply of single-dose vaginal clindamycin (2%) (n 290) were
excluded from the analyses. We excluded the women who
filled more days of medication than indicated to minimize the
possibility that taking more than the usual indicated course of
the medication was a factor in preventing BV recurrence and
to enable a fair comparison of the single-dose vaginal clindamycin (2%) with other therapies. After application of these
inclusion and exclusion criteria, the study sample to assess BV
treatment failure (i.e., treatment for BV in the first 3 weeks
postindex date) was 32,268 women. To assess recurrent BV
(i.e., treatment for BV in the 428 weeks postindex date),
women who failed BV treatment in the first 3 weeks postindex
were excluded from the analysis (n 140), and the final
sample size was 32,128. To assess the likelihood of acquiring
VVC after BV treatment, women who were diagnosed or
treated for VVC 07 days prior to the index date were excluded from the analyses (n 2,611); the sample size for this
cohort was 29,657.

Administrative claims data were used from private insurers and large employers encompassing both fee-for-service
and managed care health plans, covering approximately
8 million member lives across the southern, western, and
midwestern United States between January 1, 2004, and December 31, 2006. Membership and health plan eligibility information were linked to claims from inpatient, outpatient,
professional, pharmacy, emergency department, and ancillary sources. Data elements drawn from these databases included member demographics (age, gender, residence region,
and enrollment), service dates, setting of the care episode
(outpatient, inpatient, professional, or emergency care), diagnosis codes, procedure codes, and prescription medications
dispensed or administered (date filled, days supplied, dosage,
drug name, and therapeutic class).

The outcomes of interest in this study were BV treatment

failure, recurrent BV, and occurrence of VVC at 128 weeks
after the index date. BV treatment failure and recurrent BV
were defined as filling a vaginal or oral prescription for BV
therapy in the first 3 weeks postindex date and 428 weeks
postindex date, respectively.13 Again, for the prescription of
oral metronidazole to count as treatment of BV, the diagnosis
of vaginitis (ICD-9 code 616.10) had to be present in the 07
days preceding the prescription. The occurrence of VVC postBV treatment was assessed by ICD-9 diagnosis code 112.1.

Subject selection

Main independent variables

Female health plan participants aged 1250 years who filled one of the following prescriptions for treatment of BV: (1)
single-dose clindamycin vaginal cream (2%), (2) multipledose vaginal clindamycin, (3) vaginal metronidazole, and (4)
oral metronidazole were considered eligible as study subjects.
Because oral metronidazole can be used for other reasons
besides treatment of BV, a filled prescription of oral metronidazole was paired with the diagnosis of vaginitis and vulvovaginitis (ICD-9 CM code 616.10) in the 07 days preceding
the prescription. Of note, oral clindamycin was not included
in this analysis because of its common use for treatment of
nonvaginitis conditions and the infrequency of its use for BV
(<1%).12 The index date was defined as the date the prescriptions were filled.
To accurately assess the baseline clinical characteristics and
outcome of interest (e.g., recurrent BV) and ensure the completeness of the data, the study included only patients who
were continuously enrolled in the health plan for both medical
and pharmacy benefits for 52 weeks before and 28 weeks after
the index date. The study excluded patients who had a diagnosis of or treatment for HIV or acquired immune deficiency syndrome (AIDS); had other infectious causes of
vulvovaginitis, including trichomonal vaginitis, gonorrhorea,
or Chlamydia, at any time during the 028 weeks after the
index date; and were pregnant during the 52 weeks prior to
the index date. In addition, patients who filled more than a
7-day supply of oral metronidazole (n 1,342), vaginal met-

The main independent variables of interest were four BV

treatments: (1) single-dose clindamycin vaginal cream (2%),
(2) multiple-dose vaginal clindamycin, (3) vaginal metronidazole, and (4) oral metronidazole. National drug codes
(NDC) for these drug therapies were used to identify prescriptions in pharmacy claims.

Dependent variables

Covariates
General covariates used in this study included age, median
household income obtained by matching ZIP code to 2000
U.S. Census data, racial=ethnic composition of ZIP code obtained by matching with 2000 U.S. Census data (i.e., percent of
whites living in the ZIP code), health plan region (south vs.
other), Charlson Index, medication count, receipt of Papanicolaou test (Pap smear) in the baseline year, number of outpatient visits in the baseline year, and physician specialty (i.e.,
obstetrics=gynecology, other, or unknown specialty). The
Charlson Index was calculated by a proprietary version of the
Charlson algorithm specifically adapted for administrative
datasets that has been shown to predict a variety of patient
outcomes, including mortality, postoperative complications,
length of stay, and hospital charges. A Charlson Index score of
0 indicates no comorbid diseases, and a higher score on the
Charlson Index indicates a greater burden of comorbid diseases.14,15 Medication count was measured by the total
number of unique medication types (excluding study thera-

TREATMENT OPTION FOR BACTERIAL VAGINOSIS

1999

pies) that each patient received during the baseline year. The
receipt of Pap smear and number of outpatient visits in the
baseline year were used as proxy measures for the likelihood of patients healthcare use. Of note, the income
measured is not an individual level variable. It is the median income of the community defined by the patients ZIP
code obtained using data published by the U.S. Census.
Researchers often use area-based measure to estimate the
effects of socioeconomic position and find it useful as a
measure of community-wide characteristics and a strong
predictor of health outcomes.1619
The BV-specific covariates included history of recurrent
BV, defined as having two or more episodes of BV in the year
prior to the index date prescription. History of recurrent BV
was controlled for because it was found to be highly significant in previous literature to predict the recurrence of BV.9
VVC-specific covariates include history of two or more episodes of VVC in the year prior to the index date, diabetes, any
oral antibiotic use 1 month before index date, estrogen or
estrogen-containing contraceptive use on the index date, and
oral steroid use 2 months prior to the index date. Diabetes
was defined among individuals with two or more claims of
diabetes (ICD-9 diagnosis codes 250.xx, 357.2x, 362.0x,
366.41, 648.0x and DRG 294, 295) during the study period.
Medications, such as antibiotics, contraceptives, and oral
steroids, were captured using appropriate NDC codes. All

the covariates were selected a priori based on literature review and expert opinions for their potential to predict recurrent BV or VVC.
Statistical analysis
Univariate descriptive statistics were calculated for variables of interest. Bivariate statistical tests (e.g., chi-square,
ANOVA) were conducted between drug cohorts and outcomes. Multiple logistic regression models were used to examine the impact of drug therapies on BV treatment failure
and BV recurrence while controlling for history of recurrent
BV, age, community demographic and socioeconomic characteristics (i.e., median income, and ethnic composition),
health plan region, comorbidity, healthcare use, medication
burden, and physician specialty. A third multivariate logistic
regression model was used to assess the impact of drug
therapies on the occurrence of VVC after BV pharmacological
therapy while controlling for history of recurrent VVC, age,
community demographic and socioeconomic characteristics,
health plan region, comorbidity, healthcare use, medication
count, physician specialty, history of diabetes, antibiotic use 1
month before index date, estrogen or estrogen-containing
contraceptive use on the index date, and oral steroid use 2
months prior to or on the index date. The results are presented
as odds ratios (ORs) and p values.

Table 1. Sociodemographic and Clinical Characteristics of Women Treated for Bacterial Vaginosis (BV)

Characteristic
Demographic characteristics
Mean age (SD)**
Median household
incomea $ (SD)**
Percentage of whites (SD)a,**
Region: South (%)**
Healthcare utilization
Pap smear (%)**
Mean number of
outpatient visits (SD)**
Mean medication countb (SD)**
Physician specialty**
Obstetrics=gynecology (%)
Other specialty (%)
Missing (%)
Clinical characteristics
Charlson Comorbidity
Index > 0 (%)*
History of recurrent BVc (%)**
History of recurrent VVCd (%)**
Diabetes (%)
Antibiotic use (%)**
Contraceptives (%)
Steroid use (%)

Total
(n 32,268)
(100%)

Single-dose
vaginal
clindamycin
(n 2,825)
(8.8%)

Multiple-dose
vaginal
clindamycin
(n 4,871)
(15.1%)

Vaginal
metronidazole
(n 15,566)
(48.2%)

Oral
metronidazole
(n 9,006)
(27.9%)

34.5 (9.6)
$45,595
(15,457)
64.4 (27.2)
72.8

34.3 (9.4)
$45,579
(18,060)
62.9 (26.3)
94.1

35.7 (9.5)
$47,912
(16,231)
62.4 (28.7)
63.4

35.1 (9.5)
$45,091
(14,929)
65.7 (26.5)
75.1

33.1 (9.8)
$45,195
(15,043)
63.5 (27.8)
66.5

85.7
4.4 (4.4)

89.1
5.0 (4.6)

86.6
4.6 (4.7)

85.4
4.1 (4.1)

84.8
4.6 (4.8)

5.8 (4.2)

6.5 (4.3)

5.9 (4.4)

5.5 (4.0)

6.1 (4.3)

35.5
49.9
14.6

44.3
40.5
15.2

38.0
46.7
15.3

33.5
49.9
16.6

35.1
54.7
10.2

11.1

10.7

11.8

10.6

11.6

7.2
1.4
2.6
13.6
23.9
2.4

9.1
1.4
2.2
12.9
24.7
2.8

10.4
1.4
2.4
15.9
24.1
2.3

6.7
1.1
2.6
13.2
23.8
2.3

5.6
1.7
2.8
13.2
23.5
2.5

*p < 0.05; **p < 0.01 using Pearsons chi-square test for categorical variables and ANOVA for continuous variables.
a
These were obtained by matching ZIP code with U.S. Census data; 8,631 members have missing ZIP code values.
b
Medication count was measured by the total number of unique medication types (excluding study therapies) that each patient received
during the baseline year.
c
History of two or more episodes of BV in a year.
d
History of two or more episodes of vulvovaginal candidiasis (VVC) in a year.

2000

CHEN ET AL.

Sensitivity analyses were conducted to confirm the robustness of the results. First, patients with VVC on the index
date were included and then excluded in the study period to
examine the impact of concurrent drug therapies for VVC on
the recurrent BV in the study period of interest. Second, a Cox
proportional hazard analysis was used to model the time to
first event (i.e., recurrence of BV or occurrence of VVC after
BV treatment). Third, to investigate the possibility that coding
variation could contribute to the results, we defined BV
treatment failure and recurrence using three other methods:
(1) diagnosis of BV (i.e., ICD-9 diagnosis code 616.10) in a faceto-face visit in the outpatient setting, (2) diagnosis of BV in a
face-to-face visit in the outpatient setting or the receipt of BV
treatments, and (3) diagnosis of BV in a face-to-face visit in the
outpatient setting, followed by receipt of BV treatment in the
07 days after the diagnosis. The results of all sensitivity analyses were consistent with the primary results presented
herein. SAS proprietary software, release 9.1 (SAS Institute
Inc., Cary, NC) and STATA (Statacorp, 2003, College Station,
TX) were used for all statistical analyses.
Results
A total of 32,268 women were included in the claims
study, and bivariate analyses revealed significant variation

in almost all sociodemographic and clinical characteristics

by drug cohorts (Table 1) and outcomes (Table 2). More
women were prescribed metronidazole (vaginal, 48%; oral,
28%) for their index prescription compared with clindamycin (single-dose clindamycin vaginal cream (2%), 9%;
multiple-dose vaginal clindamycin products, 15%) (Table 1).
The mean age was 35. The median household income by ZIP
code was $45,595 for the cohort, and women who were
prescribed multiple-dose clindamycin had higher incomes
on average ($47,912) than women who filled other BV
treatments. Similarly, women who had BV treatment failure
had higher incomes ($46,719) than women who did not
(Table 2). Although 7% of women had two or more episodes
of BV in the year previous to the index date, only 1% of
women had two or more episodes of VVC in the same period
(Table 1). Women who filled multiple-dose vaginal clindamycin prescriptions had the highest history of recurrent
BV (10%) (Table 1).
Overall, the rates of BV treatment failure (<1%), BV recurrence (2.7%), and VVC post-BV treatment (2.9%) were low
(Table 3). Bivariate analyses showed significant variation with
regard to BV treatment failure, BV recurrence, and VVC rates
by treatment. Women who were treated with multiple-dose
vaginal clindamycin products had the highest rate of treatment failure (0.6%) and BV recurrence (3.3%). Women treated

Table 2. Characteristics of Women with and without Bacterial Vaginosis (BV)

Treatment Failure or Recurrence

Characteristic
Antimicrobials received for BV treatment**
Single-dose vaginal clindamycin cream (%)
Multiple-dose vaginal clindamycin treatments (%)
Vaginal metronidazole (%)
Oral metronidazole (%)
Demographic characteristics
Mean age (SD)**
Median household income $ (SD)a,**
Percentage of whites (SD)a,**
Region: South (%)
Health care utilization
Pap smear (%)*
Mean number of outpatient visits (SD)**
Mean medication countb (SD)**
Physician specialty
Obstetrics=gynecology (%)
Other specialty (%)
Missing (%)
Clinical characteristics
Charlson Comorbidity Index > 0 (%)
History of recurrent BVc (%)**
History of recurrent VVCd (%)*
Diabetes (%)
Antibiotic use (%)*
Contraceptives (%)
Steroid use (%)

BV treatment
failure
(n 140) (0.43%)

BV recurrence
(n 873) (2.71%)

No BV treatment failure
or recurrence
(n 31,255) (96.86%)

10.0
22.9
41.4
25.7

9.1
18.3
50.2
22.5

8.7
15.0
48.2
28.1

34.5 (9.8)
46719 (17123)
70.9 (24.3)
70.7

33.6 (9.0)
46963 (16331)
60.4 (27.5)
74.7

34.6 (9.6)
45556 (15426)
64.5 (27.2)
72.7

78.6
5.2 (4.9)
7.5 (4.9)

85.3
4.8 (4.8)
6.6 (4.8)

85.8
4.4 (4.4)
5.8 (4.2)

32.9
52.9
14.3

38.4
49.1
12.5

35.5
49.9
14.6

13.6
19.3
1.4
1.4
17.9
25.7
4.3

10.7
21.3
2.4
2.6
16.0
25.4
2.2

11.1
6.7
1.3
2.6
13.5
23.8
2.4

*p < 0.05; **p < 0.01 using Pearsons chi-square test for categorical variables and ANOVA for continuous variables.
a
These were obtained by matching ZIP code with U.S. Census data; 8,631 members have missing ZIP code values.
b
Medication count was measured by the total number of unique medication types (excluding study therapies) that each patient received
during the baseline year.
c
History of two or more episodes of BV in a year.
d
History of two or more episodes of vulvovaginal candidiasis (VVC) in a year.

TREATMENT OPTION FOR BACTERIAL VAGINOSIS

2001

Table 3. Bacterial Vaginosis (BV) Treatment Outcomes by Antimicrobial Treatment

Total
BV treatment failurea,* (n 32,268) 0.43%
Recurrent BVb,**
(n 32,128) 2.72%
VVCc,**
(n 29,657) 2.94%

Single-dose
Multiple-dose vaginal
vaginal clindamycin
clindamycin
(n 2,825) 0.50%
(n 2,811) 2.81%
(n 2,620) 2.37%

(n 4,871) 0.66%
(n 4,839) 3.31%
(n 4,521) 3.07%

Vaginal
metronidazole

Oral
metronidazole

(n 15,566) 0.37% (n 9,006) 0.40%

(n 15,508) 2.82% (n 8,970) 2.19%
(n 14,437) 2.65% (n 8,079) 3.58%

*p < 0.05; **p < 0.01 using Pearsons chi-square test.

a
BV recurrence in the 13 weeks post-BV treatment.
b
BV recurrence in the 428 weeks post-BV treatment. Women with BV recurrence in the 13 weeks post-BV treatment were excluded in this
analysis.
c
Vulvovaginal candidiasis (VVC) in the 028 weeks post-BV treatment (i.e., index date). Of note, women who were diagnosed or treated for
VVC 07 days prior to index date were excluded from this analysis.

with oral metronidazole had the highest rate of VVC post-BV

treatment (3.6%).
Multivariate analyses, using oral metronidazole as the
reference drug, showed that there were no significant differences in the probability for treatment failure by BV therapies
(Table 4). Significant predictors of BV treatment failure include prior history of recurrent BV. There were, however,
differences observed in the probability for recurrent BV according to treatment therapies. Women who were treated
with single-dose clindamycin vaginal cream (2%) showed no
difference from women treated with oral metronidazole in the
likelihood of BV recurrence. However, women who received
multiple-dose vaginal clindamycin (OR 1.47, 95% CI 1.191.82, p < 0.001) or vaginal metronidazole (OR 1.36, 95 CI 1.141.61, p < 0.001) were significantly more likely to experience
BV recurrence. Other significant predictors of BV recurrence
include age, race area, sociodemographic measures, medication count, and prior history of recurrent BV. Of interest,
middle-aged women (2140 years) appear to have recurrent
BV more after than younger or older women, and women
who live in ZIP codes with higher percentages of minorities
were significantly more likely to have recurrent BV.
Differences were also observed in the likelihood of VVC
occurrence according to treatment therapies. Women who
were treated with single-dose clindamycin vaginal cream
(2%) (OR 0.66, 95% CI 0.50-0.88, p < 0.01) and vaginal metronidazole (OR 0.79, 95% CI 0.68-0.93, p < 0.01) were significantly less likely to have VVC post-BV treatment compared
with those treated with oral metronidazole, when controlling
for covariates. The odds of VVC occurrence in women treated
with multiple-dose vaginal clindamycin products did not
differ significantly from those of women treated with oral
metronidazole. Other significant predictors of VVC occurrence include race area, sociodemographic measure, region,
utilization measures, and prior history of recurrent BV and
VVC. Not surprisingly, although atypical and few in number
(<1.4% of total study population), women with a history of
two or more episodes of VVC in the past year were significantly more likely to experience VVC post-BV treatment.
Discussion
The similar likelihood for treatment failure across the
treatment therapies observed in this study indirectly supports
published findings that current BV treatments have similar
initial clinical cure rates. However, the significant variations
noted among the four treatment groups in the probability for
recurrent BV and VVC occurrence at 428 weeks postindex

treatment suggest that therapies may be differentiated from

each other. This finding is relevant to both clinicians and patients when selecting which drug to prescribe and use in the
management of BV.
This analysis demonstrates that women using single-dose
clindamycin vaginal cream (2%) have a similar risk for BV
recurrence to that of women treated with oral metronidazole
even after controlling for important confounding factors, such
as prior history of BV infection. Women treated with vaginal
metronidazole or multiple-dose vaginal clindamycin were
statistically more likely to have recurrent BV compared with
women treated with oral metronidazole. This finding is supported by a recently published study, which showed singledose vaginal clindamycin is more effective in reducing vaginal
mobiluncus morphotypes than is vaginal metronidazle in patients with BV.5 In addition, women treated with single-dose
clindamycin vaginal cream (2%) and vaginal metronidazole
were less likely to have VVC in the 28 weeks posttreatment
than women treated with oral metronidazole. Given the longlasting effects of single-dose clindamycin, our data suggest that
this new drug delivery system may have longer lasting beneficial effects on the vaginal flora. Further study is needed to
investigate the microbiological effects of single-dose clindamycin on yeast colonization and infection rates.
Prior studies have suggested that single-dose clindamycin
vaginal cream (2%) may cause fewer gastrointestinal side effects (i.e., nausea, vomiting, and taste perversion) and greater
satisfaction when compared with oral antibiotic therapy7,8,20
and improved compliance when compared with longer vaginal treatments.21 This investigation demonstrates that in
addition to these benefits, single-dose clindamycin vaginal
cream (2%) may be a good treatment option not only because
of its similarity to oral metronidazole in the risk of BV recurrence (and lower risk than the other regimens) but also for
its lower risk of posttreatment VVC when compared with oral
metronidazole.
Notably, although in this study we could not measure the
direct effect of race=ethnicity on BV recurrence, our findings,
consistent with prior studies, suggest that minority populations were significantly more likely to have recurrent BV.1 We
also found that women who were more likely to use healthcare were also more likely to have complaints of VVC. This
may be because VVC may be self-diagnosed and treated with
over-the-counter (OTC) products by the patient. Thus, patients with VVC may be undercounted in this study.
It is acknowledged that the BV recurrence rate found in this
study (2.72%) is substantially lower than the rates reported in
studies that defined recurrence based on examination of the

2002

CHEN ET AL.
Table 4. Multivariate Logistic Regression Results of Bacterial Vaginosis (BV)
and Vulvovaginal Candidiasis (VVC) Treatment Outcomes
BV treatment
failure OR (95% CI)
(n 32,268)

Antimicrobial treatments (reference: oral metronidazole)

Single-dose vaginal clindamycin
1.26 (0.67-2.37)
Multiple-dose vaginal clindamycin
1.60 (0.99-2.59)
Vaginal metronidazole
0.97 (0.64-1.48)
Demographic and clinical characteristics and healthcare utilization
Age (reference 1220 years)
2130 years
1.13 (0.59-2.13)
3140 years
1.14 (0.62-2.13)
4150 years
0.88 (0.47-1.65)
Median income of ZIP codec (reference: > $48,930)
<$36,930
0.92 (0.55-1.52)
$36,930$48,930
0.90 (0.56-1.46)
% Whites in ZIP codec (reference: > 83%)
< 60%
0.66 (0.40-1.10)
60%83%
0.71 (0.44-1.15)
Missing ZIP code
0.82 (0.48-1.42)
Region: South
0.86 (0.57-1.31)
Receipt of Pap smear
0.59 (0.39-0.91)**
Number of outpatient visits in 1 year (reference: 01)
25
1.01 (0.62-1.63)
>5
1.00 (0.57-1.75)
Number of distinct medications in 1 year (reference: 01)
25
1.17 (0.55-2.51)
>5
1.85 (0.84-4.09)
Treating physician specialty (reference: other specialty)
Obstetrics=gynecology
0.94 (0.64-1.37)
Missing specialty
0.93 (0.55-1.56)
Comorbidity
1.11 (0.67-1.84)
Recurrent BV
2.66 (1.71-4.12)**
Recurrent VVC

Diabetes

Antibiotic use

Contraceptives

Steroid use

BV recurrencea
OR (95% CI)
(n 32,128)

VVCb OR (95% CI)

(n 29,657)

1.19 (0.91-1.56)
1.47 (1.19-1.82)**
1.36 (1.14-1.61)**

0.66 (0.50-0.88)**
0.85 (0.69-1.04)
0.79 (0.68-0.93)**

1.52 (1.15-2.01)**
1.31 (1.00-1.73)*
0.94 (0.71-1.25)

1.15 (0.89-1.48)
0.94 (0.73-1.21)
0.89 (0.69-1.15)

0.82 (066-1.02)
1.07 (0.88-1.30)

0.95 (0.78-1.16)
0.83 (0.68-1.01)

1.54
1.29
1.63
0.96
0.91

1.27
1.04
1.33
0.75
1.48

(1.24-1.90)**
(1.04-1.59)*
(1.28-2.08)**
(0.80-1.15)
(0.74-1.12)

(1.04-1.56)*
(0.85-1.28)
(1.05-1.68)*
(0.63-0.88)**
(1.16-1.88)**

0.84 (0.70-1.01)
0.96 (0.77-1.20)

1.14 (0.93-1.38)
1.47 (1.17-1.85)**

1.46 (1.08-1.98)*
1.64 (1.18-2.26)**

0.85 (0.66-1.10)
0.92 (0.70-1.22)

1.08
0.85
0.91
3.37

1.11
1.16
0.91
1.56
4.69
0.92
1.09
0.96
0.93

(0.93-1.25)
(0.68-1.07)
(0.73-1.15)
(2.83-4.02)**

(0.96-1.30)
(0.94-1.44)
(0.72-1.16)
(1.25-1.94)**
(3.36-6.63)**
(0.56-1.49)
(0.90-1.32)
(0.82-1.14)
(0.59-1.45)

*p < 0.05; **p < 0.01.

a
Women with BV recurrence in the 13 weeks post-BV treatment were excluded in this analysis.
b
VVC in the 028 weeks post-BV treatment (i.e., index date). Of note, women who were diagnosed or treated for VVC 07 days prior to
index date were excluded from this analysis.
c
Income and percent whites were obtained by matching ZIP code with U.S. Census data and categorized into tertiles.

entire cohort of women using laboratory parameters.9,22,23

Because in this study, BV recurrence is established by filled
prescriptions for BV treatment, there may be some women
who had BV recurrence but did not seek BV treatment. It is
noteworthy, however, that despite the overall low rates based
on the specific definition applied in this study, BV recurrence
was documented and significant differences persisted across
treatment therapies even after controlling for relevant confounding factors.
The etiology of recurrent BV is poorly understood, and
there are many unproven theories. It is not clear whether BV
recurs based on antibiotic failure to eradicate organisms,
immune-mediated changes that predispose to recurrent BV,
sexual partner reintroduction of pathogenic organisms, or
failure of protective lactobacilli to recolonize. It is likely that
all four of these possible explanations contribute to a predisposition for recurrence to varying degrees in different patients. It may also be hypothesized that diverse antimicrobial

properties of the various agents used and the different

mechanisms of therapeutic delivery (oral vs. vaginal dosing,
prolonged release vs. immediate release of drug) may interface with the bacteria and inflammatory mediators in ways
not completely understood to produce the differences in recurrence rates suggested herein. Nevertheless, BV recurrence
is a challenging and highly understudied entity seen in clinical practice, and gaining an understanding of potential ways
to minimize recurrence is an important undertaking. Thus, the
findings in this investigation add to the small body of literature on this topic and deserve notice.
This study has some limitations. First, this study can only
be generalized to women in the United States who have
continuous commercial insurance coverage. Second, this is a
claims-based analysis with certain potential biases, including
coding variation between providers, imprecision of timing
of symptoms, and missing data. However, to investigate
the possibility that coding variation could contribute to the

TREATMENT OPTION FOR BACTERIAL VAGINOSIS

results, we defined BV treatment failure and recurrence using
other methods, including BV diagnosis, BV diagnosis or filled
prescription, and BV diagnosis code followed by filled prescription. We found that our results are consistent in our
sensitivity analyses with these alternate definitions. Thus, we
believe that the results presented are robust to coding variations. Third, there were important factors (e.g., race, education, smoking, sexual history, adherence to antimicrobials,
self-treatment with OTC medication) that predict BV recurrence that are not available in administrative claims data and,
thus, could not be controlled for in this study. Fourth, although we excluded women who filled more days of medication than indicated to minimize the possibility that taking
more than the usual indicated course of the medication was a
factor in preventing BV recurrence, this may be a source of
bias. Lastly, as laboratory data confirming the diagnosis of BV
or VVC were not available, it is possible that patients were
misdiagnosed or miscoded. Nevertheless, administrative
claims data have been used to effectively examine and document patterns of healthcare use,24,25 detect opportunities to
improve quality of care,26 estimate incidence of disease,2729
and even assess outcomes of pharmaceutical,30 radiological,31
and surgical procedures,32 and thus provide useful and important insights.
In conclusion, women significantly differed in the risk of
BV recurrence and VVC by BV treatments. In the discussion of
BV treatment options with patients, healthcare providers
should address not only the patients preferred method of
drug administration (oral vs. vaginal) and tolerance of side
effects (e.g., gastrointestinal symptoms, posttreatment VVC)
but also the risk of BV recurrence. This study demonstrates
that single-dose clindamycin vaginal cream (2%) is similar to
oral metronidazole and is superior to the other agents in terms
of prevention of BV recurrence and, therefore, may be a good
alternative to oral metronidazole, given its favorable side
effect profile and lower risk of posttreatment VVC.
Acknowledgments
This study was funded by the Ther-Ex Corporation.

2003

6.

7.

8.

9.

10.

11.
12.

13.

14.

15.

16.

17.
18.

Disclosure Statement
The authors have no conflicts of interest to report.
References
1. Allsworth JE, Peipert JF. Prevalence of bacterial vaginosis:
20012004 National Health and Nutrition Examination Survey data. Obstet Gynecol 2007;109:114120.
2. ACOG Practice Bulletin. Clinical management guidelines for
obstetrician-gynecologists, Number 72, May 2006: Vaginitis.
Obstet Gynecol 2006;107:11951206.
3. Bradshaw CS, Morton AN, Garland SM, et al. Higher-risk
behavioral practices associated with bacterial vaginosis
compared with vaginal candidiasis. Obstet Gynecol 2005;
106:105114.
4. Faro S, Skokos CK. The efficacy and safety of a single dose of
Clindesse vaginal cream versus a seven-dose regimen of
Cleocin vaginal cream in patients with bacterial vaginosis.
Infect Dis Obstet Gynecol 2005;13:155160.
5. Nyirjesy P, McIntosh MJ, Steinmetz JI, et al. The effects of
intravaginal clindamycin and metronidazole therapy on

19.

20.

21.

22.

23.

24.

vaginal mobiluncus morphotypes in patients with bacterial

vaginosis. Sex Transm Dis 2007;34:197202.
Nyirjesy P, McIntosh MJ, Gattermeir DJ, et al. The effects of
intravaginal clindamycin and metronidazole therapy on
vaginal lactobacilli in patients with bacterial vaginosis. Am J
Obstet Gynecol 2006;194:12771282.
Fischbach F, Petersen EE, Weissenbacher ER, et al. Efficacy
of clindamycin vaginal cream versus oral metronidazole in
the treatment of bacterial vaginosis. Obstet Gynecol 1993;
82:405410.
Paavonen J, Mangioni C, Martin MA, et al. Vaginal clindamycin and oral metronidazole for bacterial vaginosis: A
randomized trial. Obstet Gynecol 2000;96:256260.
Bradshaw CS, Morton AN, Hocking J, et al. High recurrence
rates of bacterial vaginosis over the course of 12 months after
oral metronidazole therapy and factors associated with recurrence. J Infect Dis 2006;193:14781486.
Eckert LO, Hawes SE, Stevens CE, et al. Vulvovaginal candidiasis: Clinical manifestations, risk factors, management
algorithm. Obstet Gynecol 1998;92:757765.
Maccato ML, Kaufman RH. Fungal vulvovaginitis. Curr
Opin Obstet Gynecol 1991;3:849852.
Russo CA, Andrews RM. The national hospital bill: The
most expensive conditions, by payer, 2004. HCUP statistical
brief. Rockville, MD: Agency for Healthcare Research and
Quality, 2006.
Sobel JD, Ferris D, Schwebke J, et al. Suppressive antibacterial therapy with 0.75% metronidazole vaginal gel to prevent recurrent bacterial vaginosis. Am J Obstet Gynecol
2006;194:12831289.
Klabunde CN, Potosky AL, Legler JM, et al. Development of
a comorbidity index using physician claims data. J Clin
Epidemiol 2000;53:12581267.
Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613619.
Robert SA. Neighborhood socioeconomic context and adult
health. The mediating role of individual health behaviors
and psychosocial factors. Ann NY Acad Sci 1999;896:465468.
Robert SA. Community-level socioeconomic status effects on
adult health. J Health Soc Behav 1998;39:1837.
Thomas AJ, Eberly LE, Neaton JD, et al. Census-based and
personally reported income measures as long-term riskadjusted mortality predictors. Public Health 2007;121:898901.
Thomas AJ, Eberly LE, Davey SG, et al. ZIP-code-based
versus tract-based income measures as long-term risk-adjusted
mortality predictors. Am J Epidemiol 2006;164:586590.
Ferris DG, Litaker MS, Woodward L, et al. Treatment of
bacterial vaginosis: A comparison of oral metronidazole,
metronidazole vaginal gel, and clindamycin vaginal cream.
J Fam Pract 1995;41:443449.
Richter A, Anton SE, Koch P, et al. The impact of reducing
dose frequency on health outcomes. Clin Ther 2003;25:2307
2335.
Brotman RM, Erbelding EJ, Jamshidi RM, et al. Findings
associated with recurrence of bacterial vaginosis among
adolescents attending sexually transmitted diseases clinics.
J Pediatr Adolesc Gynecol 2007;20:225231.
Myer L, Kuhn L, Denny L, et al. Recurrence of symptomatic
bacterial vaginosis 12 months after oral metronidazole
therapy in HIV-positive and -negative women. J Infect Dis
2006;194:17971799.
Gilmore AS, Zhao Y, Kang N, et al. Patient outcomes and
evidence-based medicine in a preferred provider organization

2004

25.

26.
27.

28.

29.

30.

setting: A six-year evaluation of a physician pay-forperformance program. Health Serv Res 2007;42:21402159.
Wennberg JE, Freeman JL, Shelton RM, et al. Hospital use
and mortality among Medicare beneficiaries in Boston and
New Haven. N Engl J Med 1989;321:11681173.
Lohr KN. Use of insurance claims data in measuring quality
of care. Int J Technol Assess Health Care 1990;6:263271.
McBean AM, Warren JL, Babish JD. Measuring the incidence
of cancer in elderly Americans using Medicare claims data.
Cancer 1994;73:24172425.
McBean AM, Babish JD, Warren JL. Determination of lung
cancer incidence in the elderly using Medicare claims data.
Am J Epidemiol 1993;137:226234.
Nattinger AB, Laud PW, Bajorunaite R, et al. An algorithm
for the use of Medicare claims data to identify women with
incident breast cancer. Health Serv Res 2004;39:17331749.
OConnor RD, Rosenzweig JR, Stanford RH, et al. Asthmarelated exacerbations, therapy switching, and therapy
discontinuation: A comparison of 3 commonly used con-

CHEN ET AL.
troller regimens. Ann Allergy Asthma Immunol 2005;95:
535540.
31. Min JK, Kang N, Shaw LJ, et al. Cost and clinical outcomes
after multidetector coronary computed tomographic angiography in patients without known coronary artery disease:
A comparison to myocardial perfusion SPECT. Radiology
2008;249:6270.
32. Katz JN, Barrett J, Mahomed NN, et al. Association between
hospital and surgeon procedure volume and the outcomes
of total knee replacement. J Bone Joint Surg Am 2004;86:
19091916.

Address correspondence to:

Judy Y. Chen, M.D., MSHS
Health Benchmarks, Inc., IMS Health
21650 Oxnard Street
Woodland Hills, CA 91367
E-mail: judy.chen@us.imshealth.com

Copyright of Journal of Women's Health (15409996) is the property of Mary Ann Liebert, Inc. and its content
may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express
written permission. However, users may print, download, or email articles for individual use.