Bacterial vaginosis (BV) is the most common vaginal infection among women of reproductive age, causing considerable physical and psychosocial discomfort.1 BV is additionally associated with an increased risk of multiple adverse outcomes, including preterm birth, pelvic inflammatory disease, endometritis, and transmission and acquisition of human immunodeficiency virus type 1 (HIV-1) and sexually transmitted infections (STIs).1
Metronidazole has been the standard treatment for BV for more than 25 years,2 with 4-week cure rates as high as 85%.3 However, 58% of successfully treated cases recur within 1 year.4 Additionally, evidence that existing BV treatments prevent adverse sequelae is limited. Randomized trials of BV treatment have shown a reduction in risk of vaginal cuff infections after hysterectomy and reduced risk of preterm delivery among high-risk pregnant women, but have failed to reduce the incidence of preterm birth among asymptomatic women of heterogeneous risk.5 Efforts to improve BV treatment include antimicrobials with a lower dosing burden,6 antimicrobials plus a probiotic adjuvant to reduce recurrence rates via recolonization of a healthy vaginal microbiome, and partner treatment.1
An understanding of the global burden of BV and associated costs is needed to highlight the potential benefits of improved prevention and treatment of BV and its sequelae and to prioritize distribution of such treatment. A previous systematic review described prevalence among diverse populations of women, including both low- and high-risk subsets of the general population, reporting widely heterogeneous estimates of prevalence globally, ranging from 5% among pregnant European women to 58% among South African women living in informal housing.7 However, approximately one fifth of included estimates were obtained from samples of women selected on the basis of symptomatology or a BV-associated condition,123s–140s thereby inducing considerable risk of bias in estiamtes and introducing variability in estimates that would preclude meaningful pooled prevalence estimates.
Prevalence among intraregional subpopulations also varies, with higher prevalence of BV among women who have sex with women (WSW)8 and black and Hispanic women in the United States.9 Estimates also vary widely among pregnant women, a group of particular interest given the high number of BV-associated adverse obstetric outcomes, ranging from 8%10 to 51%.11
The wide range of prevalence estimates in general and pregnant populations indicates a similarly wide range of potential cost burdens of BV, with the cost of BV treatment alone likely amplified by the costs incurred for treatment of BV-associated sequelae. A comprehensive inventory of the costs of treatment of BV and its sequelae is necessary to assess the potential benefits of improved treatment and prevention of BV and the possible subsequent prevention of adverse outcomes.
In this systematic review and meta-analysis, we describe the prevalence of BV among reproductive-aged women of the general population in global regions using pooled estimates, and evaluate race- and ethnicity-specific prevalence within the United States. We furthermore estimate the direct cost burden of treating BV and its sequelae.
MATERIALS AND METHODS
Search Strategy and Study Selection Criteria
Our review protocol is registered in the International Prospective Register of Systematic Reviews database with registration number CRD42017073625.
We searched the PubMed and Embase databases for studies of BV prevalence, incidence, and sequelae, and costs of treatment for BV and its sequelae. A full description of our search terms is included in Text S1 in Supplemental Digital Content (SDC) 1, http://links.lww.com/OLQ/A335. We did not impose date restrictions in the search, and included studies in English, Spanish, Italian, French, Portuguese, and Chinese. A single reviewer screened all titles and abstracts (or title alone if the abstract was unavailable) for identification of potentially relevant articles. A randomly selected 10% of these articles were then assessed by 2 reviewers for inclusion decisions and data extraction. A single reviewer made inclusion decisions and extracted data of remaining articles.
We included studies in which participants are of reproductive age, defined as ages 15 to 49 years or indication that participants are postmenarchal and premenopausal. To limit the potential for selection bias, we excluded studies in which participants were selected on the basis of symptomatology or a BV-associated condition (including pelvic inflammatory disease, female infertility, women with or at risk of preterm birth or miscarriage, preterm premature rupture of membranes, premature rupture of membranes, diabetes, women engaging in intravaginal practices, and women with an intrauterine device). We also excluded estimates from study populations with BV risk that differs from the general population, including sex workers and STI clinic attendees, and from study populations that are not representative of the general population, such as incarcerated women. In countries with high HIV-1 prevalence (defined as >5%), we excluded studies of exclusively HIV-1-negative women; given the association between BV and HIV-1,12 the distribution of BV among HIV-1-negative women likely differs from the general population.
We included measures of prevalence and incidence in which estimates were obtained by Nugent score.13 Due to the wide range in published estimates of sensitivity and specificity of Amsel's criteria relative to Nugent score,14,15 we include estimates obtained by Amsel's criteria only for sub-populations for which Nugent score estimates were unavailable.
Estimation of BV Prevalence and Incidence
We estimated BV prevalence separately for each World Bank region16 and study population. Estimates from the United States were excluded from these region-specific estimates if their study populations differed meaningfully in racial and ethnic composition from the general US population (>10% difference). However, these studies were included in race- and ethnicity-specific BV prevalence estimates in the United States.
We estimated pooled prevalence with a random effects model with inverse variance weights calculated with the Tukey-Freeman double arcsine transformation.142s We adjusted estimates of prevalence obtained by Amsel's criteria using the formula and estimates of sensitivity and specificity relative to Nugent score.14 Differences in population and region across studies reporting BV incidence preclude estimation of a pooled incidence rate. We therefore present each incidence estimate separately.
Estimation of Cost Burden of BV
We converted all currencies to US dollars and adjusted costs to 2017 US dollars. We present each cost estimate separately with a description of the components included in the cost estimate. We selected the most comprehensive estimate of direct costs for standard treatment in each region for use in subsequent cost analyses. The cost estimates for the regions of North America100s and Europe and Central Asia101s include the cost of the diagnostic test, medication, and an office visit, whereas the cost estimate used for remaining regions106s includes the cost of the office visit and medication, under the assumption that the syndromic management commonly employed in these regions would preclude the cost of a diagnostic test.
We estimated the annual global economic burden of treated BV cases as the product of the region-specific cost of treatment per BV case 100s,101s,106s; Table S6 in SDC 3, http://links.lww.com/OLQ/A337), the annual number of incident cases per region, and the proportion of cases that are symptomatic (as a proxy for healthcare seeking behavior that would result in treatment). We estimated the number of incident cases per region as the product of region-specific prevalence, the size of region-specific female population ages 15 to 49 years,17 and the proportion of cases that are expected to recur within 1 year after treatment.4 We used a random-effects model with inverse variance weights calculated with the Tukey-Freeman double arcsine transformation to obtain a pooled estimate of the proportion of cases that are symptomatic from population-based samples of reproductive-aged women. Using the race-specific number of incident cases,18,19 we also estimated the race-specific annual economic burden of BV in the United States. We repeated both the global- and US-specific economic analyses, irrespective of symptomatology, to estimate the cost of treating all BV cases.
We estimated the US economic burden of 2 BV-associated sequelae for which there is consistent evidence of an independent relationship: preterm birth and HIV acquisition.12,20 Due to limitations in the availability of detailed data in all regions, we limited estimates of the economic burden of BV-associated sequelae to the United States. We estimated the race-specific annual cost of BV-associated preterm births in the United States as the products of 2015 race-specific number of BV-associated preterm births21,22 and the cost per low and normal birthweight preterm birth107s, weighted by the proportion of preterm births that are low birthweight.23 We also estimated the race-specific annual costs resulting from BV-associated prevalent HIV cases in 2016 as the product of the number of race- and age-specific BV-associated HIV cases24 and the annual cost of HIV care.110s A detailed description of parameters and formulas used in these estimates is included in SDC 2, http://links.lww.com/OLQ/A336.
We identified 2361 potentially relevant records and include data from 122 publications (Fig. 1, Tables S1, S5, S6, S7, and S10 in SDC 3, http://links.lww.com/OLQ/A337).
Prevalence of BV
The majority of prevalence estimates among women of the general population were obtained from clinic-based populations or other convenience samples (21 of 36 and 7 of 36, respectively), whereas only 8 studies provided population-based estimates of BV prevalence (Table S1 in SDC 3, http://links.lww.com/OLQ/A337). Similarly, 34 of 50 prevalence estimates among pregnant women were obtained from antenatal care attendees, with an additional 11 estimates from other clinic-based settings, 4 convenience samples from other settings, and a single population-based estimate (Table S1 in SDC 3, http://links.lww.com/OLQ/A337). Among the 8 population-based samples of the general population, 5 provide an estimate of the proportion of women with symptomatic BV (Table S2 in SDC 3, http://links.lww.com/OLQ/A337).
Reported BV prevalence was high globally among reproductive-aged women of the general population, ranging from 23% to 29% (Europe and Central Asia: 22.8%, n = 5; East Asia and Pacific: 24.2%, n = 4; Latin America and Caribbean: 24.2%, n = 4; sub-Saharan Africa: 24.6%, n = 9; Middle East and North Africa: 25.1%, n = 2; North America: 27.4%, n = 3; and South Asia: 28.7%, n = 8) (Figs. 2 and 3). Within sub-Saharan Africa, BV prevalence was lower in Western and Central Africa (20.6%; 95% confidence interval [CI], 6.1–40.6) than in Southern and Eastern Africa (33.3%; 95% CI, 17.4–51.5), although this difference was not statistically significant (P = 0.33). Prevalence estimates across sampling strategies (population-based, clinic-based, or other convenience sampling method) followed a similar distribution (Fig. S1 in SDC 4, http://links.lww.com/OLQ/A338). Across 5 population-based samples, 34.9% (95% CI, 17.7–54.4) of those with BV were symptomatic (Table S2 in SDC 3, http://links.lww.com/OLQ/A337).
Within North America, black and Hispanic women of the general population had significantly higher (33.2% and 30.7%, respectively) BV prevalence than other racial and ethnic groups (white: 22.7%; Asian: 11.1%) (P = 0.001; Fig. 3). This disparity persisted when study populations were classified as majority (>50%) black women or majority non-black women, with approximately 2-fold higher prevalence among majority black study populations (46.5%; 95% CI, 37.5–55.6 vs. 21.3%; 95% CI, 16.7–26.3; P < 0.001).
Among pregnant women, BV prevalence ranged from 11.7% in South Asia (95% CI, 9.0–14.7) to 33.2% in Latin America and the Caribbean (95% CI, 14.8–54.7) (Table S3 in SDC 3, http://links.lww.com/OLQ/A337; Fig. S2 in SDC 4, http://links.lww.com/OLQ/A338). Within the United States, the prevalence of BV in pregnancy was highest among black (49.0%; 95% CI, 40.2–57.8) and Hispanic women (42.7%; 95% CI, 36.4–49.1) and lowest among Asian (20.3%; 95% CI, 5.4–41.2) and white women (19.9%; 95% CI, 8.0–35.5) (Fig. 4).
Six studies reported BV prevalence among women living with HIV in South Asia (n = 2) and sub-Saharan Africa (n = 4). Relative to women of the general population, BV prevalence was higher among women living with HIV (35.6%; 95% CI, 25.7–46.2 vs. 25.6; 95% CI, 22.6–28.7; P = 0.054) (Table S4 in SDC 3, http://links.lww.com/OLQ/A337).
The BV prevalence among WSW was not assessed by Nugent score in any studies, and was assessed by Amsel's criteria in a single study meeting our inclusion criteria. Prevalence was approximately 20% higher (33.5%, 95% CI: 30.5, 40.7) in WSW than in women of the general population (P = 0.007).
Incidence of BV
Of reviewed studies, only 5 provided estimates of BV incidence (Table S5 in SDC 3, http://links.lww.com/OLQ/A337), ranging from 1.6 per 100 woman-years among young Australian women96s to 74.2 per 100 woman-years among pregnant women in Anhui Province, China.48s
Cost Burden of BV and Associated Sequelae
The direct medical cost per treated case of BV is highest in North America ($90.47),100s followed by Europe and Central Asia ($27.13)101s, and lowest in sub-Saharan Africa ($11.73)106s (Table S6 in SDC 3, http://links.lww.com/OLQ/A337). The cost associated with BV sequelae varied by region and sequelae. Reproductive sequelae of BV were relatively high-cost, ranging from US $4146.56 per premature rupture of membranes in North America107s to US $87,664.42 per preterm birth of a low birthweight infant in Europe and Central Asia108s (Table S7 in SDC 3, http://links.lww.com/OLQ/A337). The costs per case treated for curable STIs were comparatively low, and lower in low-resource settings (US $8.75–40.28 in sub-Saharan Africa109s) than high-resource settings (US $85.78–104.14 in Europe and Central Asia101s) (Table S7 in SDC 3, http://links.lww.com/OLQ/A337).
The global annual cost burden of BV treatment is high, at an estimated US $4.8 billion to treat symptomatic BV cases (Table 1), with more than half of costs due to recurrent BV cases. The total annual cost of BV treatment is highest in North America (US $1.3 billion for symptomatic cases), due to both high prevalence and a high cost per treated case. In both South Asia and East Asia and Pacific regions, the relatively low cost per case treated nonetheless translates to a large cost burden (US $914 million and US $1.0 billion for symptomatic cases, respectively) due to high prevalence coupled with a large population size.
Within the United States, the cost burden of BV-associated sequelae is also high, with black women bearing a disproportionate burden of these costs, accounting for 39% of the US $2.7 billion annual burden of BV-associated preterm births and 83% of the US $916 million annual costs of BV-associated HIV cases among black and white women in the United States, despite making up only 15% of this same population (Tables S8 and S9, respectively, in SDC 3, http://links.lww.com/OLQ/A337).
In this systematic review and meta-analysis, we estimate that BV prevalence is high globally, with approximately a quarter of the women of the general population meeting the Nugent score criteria for BV. This high burden of BV carries a concomitantly high economic burden, despite the relatively low direct costs per case treated. Additionally, there are marked racial disparities in the prevalence of BV within the United States, with an accompanying disproportionate cost burden borne by black women.
In contrast to the region-specific variation in BV prevalence identified in a previous systematic review,7 our meta-analysis showed similar BV prevalence across global regions. Kenyon et al. identified regions of low, moderate, and high BV prevalence according to the proportion of included studies with prevalence of >30%. By this metric, prevalence only in sub-Saharan Africa was classified as high, given that half of studies reported prevalence in excess of 30%. We also found that approximately half of estimates among reproductive-aged general population women in sub-Saharan Africa exceeded 30%. However, a pooled estimate of prevalence across these studies was similar to other global regions, indicating that categorization using the threshold indicator may reflect intraregional heterogeneity rather than a region-specific high prevalence. Previous meta-analyses of BV prevalence in sub-Saharan Africa among pregnant women11 and HIV-negative women25 have produced estimates higher than those presented here (30%25–51%11). However, both of these meta-analyses focused on populations other than the general population of women, and included in their estimates groups of women that may have higher BV prevalence than the general population (e.g., women living with HIV11 and women at high risk of HIV acquisition25). Our estimates of BV prevalence in the general population are additionally strengthened by the exclusion of estimates of prevalence that were obtained from study populations selected on the basis of symptomatology or the presence of BV-associated conditions, limiting the potential for bias in our estimates. Nevertheless, population-based estimates of BV prevalence are needed to clarify the burden of BV in sub-Saharan Africa and to understand intraregional heterogeneity. Indeed, even among samples composed only of women of the general population, our meta-analysis results indicated considerable heterogeneity in BV prevalence within global regions (range I2 79–99%; Fig. 2) that, given the similarity in distribution of prevalence estimates grouped by sampling strategy (Fig. S1 in SDC 4, http://links.lww.com/OLQ/A338), did not appear to be driven by methodological variation in sample recruitment. Further research is needed to elucidate the drivers of such intraregional variation.
Our review confirms previous findings of racial and ethnic disparities in BV prevalence within the United States, and additionally highlights the resulting higher financial burden of treatment for BV and its sequelae among black women. Previous attempts to isolate race as an independent predictor of BV risk have had mixed conclusions,26 with investigators noting that observed disparities may be the result of both structural factors, such as socioeconomic status, and biological factors, such as lower vaginal concentrations of protective Lactobacillus species among black women.27 Further research is needed to elucidate the mechanisms underlying intraregional disparities in BV prevalence to most effectively target treatment and prevention.
In population-based estimates, the proportion of women with BV who were symptomatic varied considerably, from less than 10% in sub-Saharan Africa72s to greater than one third in North2s and South America32s (Table S2 in SDC 3, http://links.lww.com/OLQ/A337). Should symptomatic women be more likely to seek treatment for BV, treatment rates may be higher in some regions than others, and may differ among subpopulations within regions, as well. However, given that annual recurrence rates are as high as 58% with current treatments,4 differences in care seeking are unlikely to result in considerable variation in prevalence estimates.
The global economic burden of treating BV cases is high, and this burden likely increases when including the cost of BV-associated sequelae. Indeed, in the United States, the cost burden of BV-associated sequelae is 3-fold the cost of treating symptomatic cases. Although we limited the estimates of the cost of BV-associated sequelae to preterm births and HIV cases in the United States, BV is associated with additional sequelae, so the total cost of BV-associated sequelae may be higher than the estimates provided here. The high cost of BV-associated preterm births and HIV treatment highlights the need for research to understand potential causal linkages between BV and adverse health outcomes, and to determine whether treatment and prevention of BV can prevent these outcomes.141s This work also highlights disparities in the cost burden of BV among regional and sub-regional populations. Although our analysis considers only direct medical costs, disparities in prevalence in the United States would also result in a higher societal cost to black women incurred through the time burden and costs of transportation and childcare necessary to attend clinic visits for BV treatment.
Despite decades-long recognition of BV, the etiology of BV remains undefined.122s Etiological theories propose that incident BV is triggered by intravaginal practices and consequent disruption of vaginal microbiota, or that BV is a sexually transmitted infection in which Gardnerella vaginalis may operate synergistically with microorganisms normally present in low concentrations in the vagina to produce the dysbiosis characteristic of BV.122s Given the high burden of BV described here and high rates of recurrence with existing therapies,4 research to determine the etiology of BV and corresponding prevention and sustainable, affordable treatment strategies are urgently needed to reduce the burden of BV among women.
There are several limitations in the work presented here. First, estimates of prevalence obtained from samples of general population women are more likely to exclude those who seek BV treatment shortly after the onset of BV symptoms; such length-biased sampling may result in lower prevalence estimates than would be observed otherwise. Additionally, the majority of prevalence estimates were obtained from clinic-based populations, with few studies providing population-based BV estimates, potentially limiting generalizability to populations of care-seeking women. Second, our estimates of the cost burden of BV-associated sequelae assume a causal relationship between BV and preterm birth and HIV acquisition. However, both causality and the impact of treatment to reduce these outcomes remain uncertain, and our estimates should be viewed in light of this uncertainty. Third, although previous studies have established declining BV prevalence as pregnancy progresses,28–30 gestational age in studies included here was variable, and differences in how gestational age was reported across studies precluded assessment of how the distribution of gestational age may explain the regional variation in BV prevalence we observed among pregnant women. Finally, despite limiting included estimates to those of comparable study populations of reproductive-aged women of the general population, there is substantial unexplained heterogeneity across study estimates (all I2 > 75%). Nevertheless, the mean prevalence estimate produced by our random-effects meta-analysis provides an important summary of BV prevalence globally among this population.
In conclusion, BV prevalence is high globally, with a consequently high economic burden. Within the United States, there are striking racial disparities in BV prevalence. Treatment that results in effective and sustained resolution of BV is needed to reduce the burden of BV and potentially its associated sequelae.
1. Bradshaw CS, Brotman RM. Making inroads into improving treatment of bacterial vaginosis—striving for long-term cure. BMC Infect Dis 2015; 15:292.
2. Sobel R, Sobel JD. Metronidazole for the treatment of vaginal infections. Expert Opin Pharmacother 2015; 16:1109–1115.
3. Oduyebo OO, Anorlu RI, Ogunsola FT. The effects of antimicrobial therapy on bacterial vaginosis in non-pregnant women. Cochrane Database Syst Rev 2009; Cd006055.
4. 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; 193:1478–1486.
5. Koumans EH, Markowitz LE, Hogan V. Indications for therapy and treatment recommendations for bacterial vaginosis in nonpregnant and pregnant women: A synthesis of data. Clin Infect Dis 2002; 35(Suppl 2):S152–S172.
6. Nyirjesy P, Schwebke JR. Secnidazole: Next-generation antimicrobial agent for bacterial vaginosis treatment. Future Microbiol 2018; 13:507–524.
7. Kenyon C, Colebunders R, Crucitti T. The global epidemiology of bacterial vaginosis: A systematic review. Am J Obstet Gynecol 2013; 209:505–523.
8. Evans AL, Scally AJ, Wellard SJ, et al. Prevalence of bacterial vaginosis in lesbians and heterosexual women in a community setting. Sex Transm Infect 2007; 83:470–475.
9. Koumans EH, Sternberg M, Bruce C, et al. The prevalence of bacterial vaginosis in the United States, 2001-2004; associations with symptoms, sexual behaviors, and reproductive health. Sex Transm Dis 2007; 34:864–869.
10. Wennerholm UB, Holm B, Mattsby-Baltzer I, et al. Fetal fibronectin, endotoxin, bacterial vaginosis and cervical length as predictors of preterm birth and neonatal morbidity in twin pregnancies. Br J Obstet Gynaecol 1997; 104:1398–1404.
11. Chico RM, Mayaud P, Ariti C, et al. A systematic review and meta-analysis of malaria and sexually transmitted and reproductive tract infections in pregnancy in sub-Saharan Africa: Opportunities for antenatal intervention. Am J Trop Med Hyg 2012; 87:288.
12. Low N, Chersich MF, Schmidlin K, et al. Intravaginal practices, bacterial vaginosis, and HIV infection in women: Individual participant data meta-analysis. PLoS Med 2011; 8:e1000416.
13. Nugent RP, Krohn MA, Hillier SL. Reliability of diagnosing bacterial vaginosis is improved by a standardized method of gram stain interpretation. J Clin Microbiol 1991; 29:297–301.
14. Schwebke JR, Hillier SL, Sobel JD, et al. Validity of the vaginal gram stain for the diagnosis of bacterial vaginosis. Obstet Gynecol 1996; 88(4 Pt 1):573–576.
15. Myziuk L, Romanowski B, Johnson SC. BVBlue test for diagnosis of bacterial vaginosis. J Clin Microbiol 2003; 41:1925–1928.
17. The World Bank Group2018; Pages. Accessed at https://data.worldbank.org/indicator/SP.POP.TOTL.FE.IN
. Accessed 6 February 2018.
18. United States Census Bureau; Pages. Accessed at https://www.census.gov/quickfacts/fact/table/US/PST045216
. Accessed 6 February 2018.
20. Leitich H, Bodner-Adler B, Brunbauer M, et al. Bacterial vaginosis as a risk factor for preterm delivery: A meta-analysis. Am J Obstet Gynecol 2003; 189:139–147.
22. Ferre C, Callaghan W, Olson C, et al. Effects of maternal age and age-specific preterm birth rates on overall preterm birth rates—United States, 2007 and 2014. MMWR Morb Mortal Wkly Rep 2016; 65:1181–1184.
23. Centers for Disease Control and Prevention 2017;Pages. Accessed at https://www.cdc.gov/nchs/fastats/birthweight.htm
. Accessed 4 May 2018.
24. Centers for Disease Control and Prevention 2018;Pages. Accessed at https://www.cdc.gov/hiv/group/gender/women/index.html
. Accessed 12 March 2018.
25. Torrone EA, Morrison CS, Chen PL, et al. Prevalence of sexually transmitted infections and bacterial vaginosis among women in sub-Saharan Africa: An individual participant data meta-analysis of 18 HIV prevention studies. PLoS Med 2018; 15:e1002511.
26. Ness RB, Hillier S, Richter HE, et al. Can known risk factors explain racial differences in the occurrence of bacterial vaginosis? J Natl Med Assoc 2003; 95:201–212.
27. Beamer MA, Austin MN, Avolia HA, et al. Bacterial species colonizing the vagina of healthy women are not associated with race. Anaerobe 2017; 45:40–43.
28. Redelinghuys MJ, Ehlers MM, Dreyer AW, et al. A cross-sectional study on the relationship of age, gestational age and HIV infection to bacterial vaginosis and genital mycoplasma infection. BMJ Open 2015; 5:e008530.
29. Hay PE, Morgan DJ, Ison CA, et al. A longitudinal study of bacterial vaginosis during pregnancy. Br J Obstet Gynaecol 1994; 101:1048–1053.
30. Waters TP, Denney JM, Mathew L, et al. Longitudinal trajectory of bacterial vaginosis during pregnancy. Am J Obstet Gynecol 2008; 199:431.e1–431.e5.