点击显示 收起
Infectious Diseases Epidemiology Unit, School of Public Health and Family Medicine
Department of Obstetrics and Gynaecology, University of Cape Town, Cape Town, South Africa
Department of Epidemiology, Mailman School of Public Health
Department of Pathology and Gertrude H. Sergievsky Center, College of Physicians and Surgeons, Columbia University, New York, New York
Background.
Bacterial vaginosis (BV) may increase women's susceptibility to HIV infection, but there are few prospective data.
Methods.
During follow-up for up to 36 months, 86 new HIV seroconverters (case patients) were identified among 5110 women enrolled in a cervical cancer screening trial. Nonseroconverting control subjects (n = 324) were frequency matched to case patients by age and duration of follow-up. At enrollment, case patients and control subjects were evaluated for clinical signs of BV, and Gram stains of vaginal fluid were scored using Nugent criteria.
Results.
BV was diagnosed on the basis of clinical criteria at enrollment in 20% of seroconverters and 16% of control subjects (summary odds ratio [OR], 1.31 [95% confidence interval {CI}, 0.712.41]). Nugent criteria for BV were met by 74% of seroconverters and 62% of control subjects. Diagnosis of BV on the basis of Nugent criteria was significantly associated with an increased risk of HIV seroconversion, after adjustment for demographic characteristics, other sexually transmitted infections, and sexual behaviors (adjusted OR, 2.01 [95% CI, 1.123.62]).
Conclusions.
BV may account for a substantial fraction of new HIV infections in this setting. Treatment of BV and other interventions to promote normal vaginal flora warrant attention for HIV prevention.
Bacterial vaginosis (BV), which is defined as an imbalance in the vaginal flora characterized by low levels of lactobacilli and an increased frequency of facultatively anaerobic bacteria, is among the most common reproductive tract infections in women worldwide [1]. Estimates of the prevalence of BV in African populations range from 20% to 50% [2, 3], with higher levels documented in female sex workers [46]. Although BV was originally thought to be of little long-term clinical significance, it has more recently been implicated in increasing the risks of preterm birth [7, 8] and the development of pelvic inflammatory disease [9, 10].
Several epidemiological studies have suggested that BV may increase women's risk of acquiring HIV infection. Most data come from cross-sectional studies demonstrating that HIV prevalence is higher in women with BV than in women with normal vaginal flora [11, 12]. In a population-based study of 4718 women in rural Uganda, women with any microbiological evidence of abnormal vaginal flora were significantly more likely than those with no such evidence to be infected with HIV (odds ratio [OR], 1.52 [95% confidence interval {CI}, 1.221.90]), and women with severe BV were >2 times more likely than women with normal vaginal flora to be infected with HIV (OR, 2.08 [95% CI, 1.482.94]) [3]. Similarly, in women attending antenatal clinics in Malawi, prevalent HIV infection was 3 times more likely in those with clinical signs of BV (OR, 3.0 [95% CI, 2.43.8]) [13]. However, these cross-sectional findings are unable to distinguish whether BV predisposes women to HIV infection or whether HIV-infected women are more likely to develop BV. Because there is evidence that BV is more prevalent and persistent in HIV-infected women [14, 15], this is a fundamental limitation of these studies.
There are few prospective data examining whether preexisting BV predisposes women to incident HIV infection. In a study of 657 Kenyan sex workers, the risk of HIV seroconversion was associated with microbiological evidence of BV (hazard ratio, 1.9 [95% CI, 1.13.1]) [16]. In addition, an increasing severity in clinical symptoms linked with BV was associated with HIV seroconversion in pregnant and postnatal women in Malawi [17]. However, not all studies have identified this association [18], and additional data are required to help establish whether BV truly increases women's risk of acquiring HIV infection. In addition, few studies have included older women, in whom the prevalence of BV may be higher than it is in younger women [19].
To complement existing epidemiological data, a number of laboratory studies have explored the potential for BV to increase susceptibility to HIV infection. The increased production of interleukin (IL)10 during BV may assist viral entry into target cells [20]. Proinflammatory cytokines with increased secretion during BV, including tumor necrosis factor and IL-1, may help to increase viral expression within infected cells, and this may facilitate the establishment of an initial infection within tissues in the reproductive tract [21]. Although these mechanisms have been investigated only in vitro, they suggest that the potential for BV to increase women's risk of acquiring HIV infection is biologically plausible.
Given the high prevalence of BV in many parts of sub-Saharan Africa, any causal association between this infection and women's risk of acquiring HIV infection may have significant implications for the prevention of HIV infection. In this light, we investigated associations between BV and HIV seroconversion in a cohort of women participating in a cervical cancer screening trial near Cape Town, South Africa.
SUBJECTS, MATERIALS, AND METHODS
Study design.
We conducted a case-control study nested within a randomized controlled trial evaluating cervical cancer screening modalities in Khayelitsha, which is outside Cape Town, South Africa. For the trial, 6555 volunteer women were randomized to 1 of 3 cervical cancer screening modalities between June 2000 and December 2002. Participants were recruited via community outreach advertising a health service to prevent cervical cancer. Women were eligible for the trial if they were 3565 years old, had no previous cervical neoplasia screening or treatment, did not have cervical cancer, and were suitable candidates for cryotherapy. All participants provided written, informed consent before being enrolled, and approval for the study was obtained from the Institutional Review Board of Columbia University and the Research Ethics Committee of the University of Cape Town.
Women participating in the trial were tested for the presence of HIV at enrollment and were scheduled for clinical follow-up and for repeat HIV testing 6 months later. In addition, approximately half of the cohort (n = 3199) was scheduled for long-term follow-up that included HIV testing at 12, 24, and 36 months after enrollment. All results of HIV testing were recorded using an anonymous, linked system, and separate voluntary counseling and testing services were provided for participants who wished to know their HIV serostatus.
For the nested case-control study described here, of the 5110 women who were seronegative at enrollment, we selected as case patients all women who had seroconverted by December 2003 (n = 86). Of the 5110 women, 89% participated in the 6-month follow-up visit; of the 3199 seronegative women scheduled for long-term follow-up, 73% participated in the 12-month follow-up visit. Follow-up visits at 24 and 36 months after enrollment were still being conducted when this nested case-control study was undertaken. For comparison, control subjects (n = 324) were selected at random from the cohort of women who had not seroconverted by the visit at which the index case patient had seroconverted (i.e., incidence-density sampling was used) and were frequency matched to case patients by age category (3539, 4049, or 5065 years old) to facilitate statistical adjustment for age [22, 23].
Methods.
During the enrollment visit, all women underwent gynecological examination by 1 of 4 trained study nurses. BV was assessed using Amsel criteria [24] and was considered to be present on the basis of at least 3 of the following 4 signs: moderate or severe vaginal discharge, a positive result for the whiff test after application of a 10% potassium hydroxide solution, clue cells found to be present by wet mount examination, and vaginal pH >4.5. Swabs of vaginal fluid were rolled on glass slides that were air dried and stored for later Gram staining, which was performed in accordance with standard procedures. For quality control, retraining and on-site supervision of examination procedures and diagnoses occurred during the course of the study. Slides were Gram stained and scored at the Department of Pathology, Columbia University, by use of the standard definitions developed by Nugent et al. [25]. The individuals who read the slides were blinded to all information, including case-control status, about the participants. This scoring system demonstrated excellent reliability (correlation coefficient for continuous Nugent scores, 0.9; Cohen's for Nugent criteria, 0.8).
HIV serostatus was determined using the Abbott HIV 1/2 g 0 Kit on the Abbott AXSYM system, and testing was performed in accordance with the manufacturer's instructions (sensitivity, 100%; specificity, 99.94%). Reactive samples were retested using the Vironosticka HIV Uniform 2 plus 0 Kit (Organon Teknika), and testing was also performed in accordance with the manufacturer's instructions (sensitivity, 100%; specificity, 99.90%); samples were classified as positive if they were reactive according to both tests. At enrollment, samples to be tested for the presence of Neisseria gonorrhoeae and Chlamydia trachomatis were collected using endocervical cone brushes and were tested using the Hybrid Capture GC/CT DNA Assay (Digene). Samples obtained by use of an endocervical cone brush were also tested for the presence of human papillomavirus (HPV) DNA by use of the Hybrid Capture II HPV DNA Assay (Digene). Wet mount examination of cervicovaginal secretions for the identification of clue cells and Trichomonas vaginalis was performed by trained study nurses. Women with cervicitis on clinical examination at enrollment or who tested positive for C. trachomatis and/or N. gonorrhoeae were treated with 200 mg of doxycycline daily for 7 days, a single 500-mg dose of ciproflaxin, and a single 2-g dose of metronidazole, and participants who were found to have trichomoniasis and/or clue cells present by wet mount examination were treated with a single 2-g dose of metronidazole.
Detailed questionnaires on sexual behaviors and biological risk factors were administered at the enrollment and 6-month follow-up visits. Additional data on intravaginal practices (including "dry sex" and douching) [26] were collected at the 6-month visit only.
Statistical analysis.
Data were analyzed using SAS (version 9.0; SAS Institute). Amsel (clinical) criteria for BV were analyzed as an ordinal variable scored 14 in accordance with the number of criteria met and as a dichotomous variable categorized as <3 and 3 criteria. Nugent (microbiological) scores for BV were categorized as normal vaginal flora (score, 03), intermediate flora (score, 46), moderate BV (score, 78), or severe BV (score, 910); the latter 2 categories were collapsed into the single category "any BV." We used separate analyses for case patients identified at the 6-month follow-up visit (the whole trial cohort) or later (the subset scheduled for long-term follow-up) to investigate the potential for bias due to differential losses to follow-up. Because there were no differences in results, we report the results from the full 36 months of observation. All crude analyses were stratified by age category, with Mantel-Haenszel summary odds ratios (ORs) and 95% confidence intervals (CIs). In multivariate analysis, conditional logistic regression was used to examine the association between different measures of BV and the relative odds of HIV seroconversion, taking into account control selection procedures; because of the incidence-density sampling of control subjects, the resulting ORs approximate incidence rate ratios [27]. Other covariates were included in multivariate models if they demonstrated an appreciable crude association with HIV seroconversion. Variables were removed from this analysis if they showed no adjusted association with HIV seroconversion and their removal did not alter appreciably the associations involving other covariates.
RESULTS
A total of 86 HIV seroconverters were identified during the trial (overall incidence rate, 2.1/100 person-years), and 324 matched control subjects were selected for comparison. The majority of seroconversions (n = 55; 64%) were identified at the 6-month follow-up visit (because of the study design), but the risk of HIV seroconversion remained constant throughout the 36-month follow-up period (data not shown). The demographic characteristics and sexual behaviors for case patients and control subjects are summarized in table 1. Because of the matched design of the study, the age distribution of case patients and control subjects was similar. Case patients who seroconverted were significantly more likely to be unmarried, to report having had >1 sex partner in the month before enrollment, and to report having a new sex partner at the 6-month follow-up visit.
Of case patients, 20% (n = 17) met Amsel criteria for BV at enrollment, compared with 16% of control subjects (n = 51), and this difference was not statistically significant (summary OR, 1.31 [95% CI, 0.712.41]). The distributions of participants who met various Amsel criteria are shown in table 2. Most participants had a vaginal pH >4.5 and vaginal discharge noted on the clinical examination. There was a trend toward increased severity of vaginal discharge associated with HIV seroconversion (P = .04).
Of the 410 participants, 383 (93%) had Gram-stained samples that were suitable for Nugent scoring. Of the HIV seroconverters with suitable samples, 74% (n = 59) had Nugent scores between 7 and 10, which is diagnostic of BV, compared with 62% of control subjects (n = 189). The prevalence of BV in control subjects increased slightly with age, from 60% in women 3539 years old to 67% in women >50 years old. The distribution of Nugent scores was strongly bimodal, with only 25 samples (6.5%) scoring between 4 and 6, which reflected intermediate vaginal flora (table 2). Compared with individuals with normal vaginal flora, HIV seroconverters were slightly more likely to have intermediate vaginal flora (summary OR, 1.44 [95% CI, 0.474.46]) and were significantly more likely to have BV (summary OR, 1.83 [95% CI, 1.003.85]) diagnosed on the basis of Nugent criteria.
The agreement in the diagnosis of BV when Amsel versus Nugent criteria were used was poor, with Amsel criteria being 21% sensitive and 91% specific, compared with Nugent criteria (overall percent agreement, 46%). The sensitivity of Nugent criteria in detecting BV was significantly lower in women who reported intravaginal practices than in those who did not (14% vs. 30%; P = .03), although there was no difference in the sensitivity of Amsel criteria in detecting BV by HIV seroconversion status. Women's intravaginal practicesprimarily washing inside the vagina with water, fingers, and/or a cloth [26]were inversely associated with the presence of clue cells on wet mount examination (24% of women who reported intravaginal practices had clue cells present, compared with 29% of women who reported no intravaginal practices; P = .005), but they were not associated with any other Amsel criteria for BV. In an analysis limited to control subjects, BV diagnosed on the basis of Nugent criteria was negatively associated with condom use (summary OR, 0.37 [95% CI, 0.150.96]) and was positively associated with trichomoniasis (summary OR, 5.93 [95% CI, 1.3426.24]) and chlamydia and/or gonorrhea (summary OR, 8.87 [95% CI, 1.2165.09]) (table 3).
The overall association between BV diagnosed on the basis of Nugent criteria and incident HIV seroconversion persisted in the multivariate analysis (table 4). After adjustment for demographic characteristics, other sexually transmitted infections (STIs) and sexual behaviors, women with BV diagnosed on the basis of Nugent criteria were significantly more likely to seroconvert than were women with normal vaginal flora (adjusted OR, 2.01 [95% CI, 1.123.62]). When the multivariate model was restricted to women 3539 years old (in whom the majority of seroconversions were observed), the association between microbiological evidence of BV and HIV seroconversion strengthened slightly (adjusted OR, 2.81 [95% CI, 1.206.62]). Other significant predictors of HIV seroconversion in the analysis of all age groups included being unmarried, reporting having had >1 sex partner in the month before enrollment, reporting having a new sex partner at the 6-month follow-up visit, being sexually active at the 6-month follow-up visit, and having ever been treated for an STI.
Women with BV diagnosed on the basis of Amsel or Nugent criteria were substantially more likely than women without such a diagnosis to be treated with metronidazole at enrollment. Eighty-six percent of women who met Amsel criteria for BV were treated with metronidazole, compared with only 10% of women who did not meet Amsel criteria. Similarly, 26% of women with microbiological evidence of BV were treated with metronidazole, compared with 13% of those with normal vaginal flora. Overall, there was no evidence for a protective effect of metronidazole on HIV seroconversion, because 27% of case patients and 22% of control subjects received metronidazole for reproductive tract infections identified at enrollment (n = 23 and n = 71, respectively), leading to an insignificant positive association between metronidazole treatment and subsequent HIV seroconversion (summary OR, 1.30 [95% CI, 0.752.24]) that did not change in multivariate models (data not shown). The magnitude of the adjusted association between BV diagnosed on the basis of Nugent criteria and the risk of HIV seroconversion was similar in women who received metronidazole (adjusted OR, 2.15 [95% CI, 0.4510.23]) and women who did not (adjusted OR, 1.74 [95% CI, 0.913.34]).
DISCUSSION
This study examined the association between BV and HIV seroconversion in a population of women who were somewhat older (3565 years old) than is usually included in studies of incident HIV infection, and it is also one of the few prospective studies of this association in a sample of women from the general population. After adjustment for demographic characteristics, other STIs, and sexual behaviors, we observed that BV diagnosed on the basis of Nugent criteria was significantly associated with HIV seroconversion. These findings add to the evidence suggesting that BV may increase women's susceptibility to HIV infection.
The association between BV and HIV seroconversion was observed with Nugent (microbiological) criteria for BV but not with Amsel (clinical) criteria. Related to this, there was poor agreement between the 2 measures in the diagnosis of BV, because a high proportion of women with BV diagnosed on the basis of Nugent criteria did not meet the Amsel definition. Some disagreement is to be expected, because the shift in the vaginal flora that is characteristic of microbiological BV does not necessarily lead to clinical signs of BV. Nevertheless, reasonable agreement is usually anticipated [28], although data on this point are few. In our study population, the prevalence of vaginal discharge and elevated vaginal pH was high, but the percentages of women with positive whiff test results and the presence of clue cells on wet mount examination were substantially lower. Part of this disagreement may be attributed to variability in the assessment of clinical symptoms of BV, despite the standardized training and supervision provided to the study nurses; such variability would likely reduce the association between the 2 measures in the diagnosis of BV. Interestingly, the sensitivity of clinical criteria to detect microbiologically confirmed BV was significantly lower in women who reported intravaginal practices. We speculate that women's intravaginal practices may have reduced certain clinical symptoms of BV without altering the underlying abnormal vaginal flora. The association between intravaginal practices and the components of BV require further investigation in different populations. If these findings are confirmed, then the interpretation of clinical signs in the assessment of BV, both in patient care and research, may need to be qualified by information on women's douching or other intravaginal practices.
In our study population, 62% of control subjects had microbiologically defined BV, which is a slightly higher prevalence than has been documented in other African populations. This increase may be attributable, in part, to the older age of our study population [19]. Several other factors associated with a high prevalence of BV were also common in our study population, including the nonuse of condoms [29, 30] and nonulcerative STIs [31]. Given the negative health outcomes associated with BV (other than the risk of acquiring HIV infection) [8, 10], the determinants of this high prevalence warrant further attention.
The interpretation of these data is subject to a number of important limitations. Women were assessed for BV once at enrollment, and HIV seroconversions were identified during the next 36 months. The dominant vaginal flora may shift over time [32, 33], and, as a result, a single assessment for BV may not accurately reflect the vaginal environment at the time of HIV exposure. However, most (64%) of the seroconversions detected occurred during the first 6 months of observation, and our findings for an association between BV and HIV seroconversion were similar in seroconverters at and after the 6-month follow-up visit. If the vaginal flora did fluctuate during follow-up, the use of a baseline assessment as a proxy for the vaginal flora at the time of sexual exposure to HIV later in time would introduce some degree of random measurement error into the assessment for BV. This type of nondifferential misclassification would most likely attenuate the observed association between BV and HIV seroconversion [34]; if this is the case, then the true association between BV and HIV seroconversion may be stronger than that identified here. Another limitation of the study is that we did not assess the presence of herpes simplex virus type 2 or other ulcerative STIs that may be independent risk factors for the acquisition of HIV infection. However, given the distinct etiologies and pathological disturbances in the vaginal floral and ulcerative STIs, these may not necessarily influence the observed association between BV diagnosed on the basis of Nugent criteria and HIV infection.
In contrast, unmeasured confounding may have inflated the association between BV and HIV seroconversion. Although we adjusted for a range of potential confounding variables, including sexual behaviors and different STIs (which may also be independent risk factors for abnormal vaginal flora), we cannot definitively rule out the possibility of residual confounding that may contribute to the observed adjusted associations.
If BV increases women's susceptibility to HIV infection, interventions to reduce the occurrence of BV may have an impact on the spread of HIV at a population level. The high prevalence of BV in our study population means that almost one-third of all new HIV infections in women in this setting might be prevented if all cases of BV could be cured (i.e., a population-attributable fraction of 31% [35]), paralleling previous estimates [17]. However, it is important to note that treatment of BV can be difficult, because treatment with metronidazole may fail in up to 50% of all cases, and recurrence even after successful treatment is common [36, 37]. In our study, there did not appear to be any attenuation of the association between BV and HIV seroconversion in women who were treated with metronidazole at enrollment. Only a single experimental study has attempted to treat BV to prevent HIV infection (as part of a mass treatment intervention for STIs), and this treatment led to a modest reduction in prevalent BV and had no effect on the incidence of HIV infection [38]. Although there is little evidence to suggest that the use of metronidazole to treat BV is effective in reducing the incidence of HIV infection, a number of new treatments for BV, including vaginal suppositories to promote lactobacilli growth and some vaginal microbicides [3941], are under development. If these new technologies prove to be more efficacious against BV over the long term, the possibility of treating BV to help prevent HIV infection will require renewed research attention.
In conclusion, this study adds to the existing evidence that suggests that BV diagnosed on the basis of Nugent criteria may increase women's susceptibility to HIV infection. Because of the relatively high prevalence of BV in this and many other settings where the risk of acquiring HIV infection is high, further research is required to evaluate interventions to reduce BV as a possible means of preventing HIV infections in women.
References
1. Hillier SL, Holmes KK. Bacterial vaginosis. In: Holmes KK, Sparling PF, Mardh PA, et al., eds. Sexually transmitted diseases. 3rd ed. New York: McGraw-Hill, 1999:56386. First citation in article
2. Sobel JD. Gynecologic infections in human immunodeficiency virusinfected women. Clin Infect Dis 2000; 31:122533. First citation in article
3. Sewankambo N, Gray RH, Wawer MJ, et al. HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 1997; 350:54650. First citation in article
4. Fonck K, Kaul R, Keli F, et al. Sexually transmitted infections and vaginal douching in a population of female sex workers in Nairobi, Kenya. Sex Transm Infect 2001; 77:2715. First citation in article
5. Riedner G, Rusizoka M, Hoffmann O, et al. Baseline survey of sexually transmitted infections in a cohort of female bar workers in Mbeya Region, Tanzania. Sex Transm Infect 2003; 79:3827. First citation in article
6. Ramjee G, Karim SS, Sturm AW. Sexually transmitted infections among sex workers in KwaZulu-Natal, South Africa. Sex Transm Dis 1998; 25:3469. First citation in article
7. Hillier SL, Nugent RP, Eschenbach DA, et al. Association between bacterial vaginosis and preterm delivery of a low-birth-weight infant. The Vaginal Infections and Prematurity Study Group. N Engl J Med 1995; 333:173742. First citation in article
8. Nelson DB, Macones G. Bacterial vaginosis in pregnancy: current findings and future directions. Epidemiol Rev 2002; 24:1028. First citation in article
9. Jossens MO, Eskenazi B, Schachter J, Sweet RL. Risk factors for pelvic inflammatory disease: a case control study. Sex Transm Dis 1996; 23:23947. First citation in article
10. Koumans EH, Kendrick JS, CDC Bacterial Vaginosis Working Group. Preventing adverse sequelae of bacterial vaginosis: a public health program and research agenda. Sex Transm Dis 2001; 28:2927. First citation in article
11. Royce RA, Thorp J, Granados JL, Savitz DA. Bacterial vaginosis associated with HIV infection in pregnant women from North Carolina. J Acquir Immune Defic Syndr Hum Retrovirol 1999; 20:3826. First citation in article
12. Cohen CR, Duerr A, Pruithithada N, et al. Bacterial vaginosis and HIV seroprevalence among female commercial sex workers in Chiang Mai, Thailand. AIDS 1995; 9:10937. First citation in article
13. Taha TE, Gray RH, Kumwenda NI, et al. HIV infection and disturbances of vaginal flora during pregnancy. J Acquir Immune Defic Syndr Hum Retrovirol 1999; 20:529. First citation in article
14. Jamieson DJ, Duerr A, Klein RS, et al. Longitudinal analysis of bacterial vaginosis: findings from the HIV epidemiology research study. Obstet Gynecol 2001; 98:65663. First citation in article
15. Warren D, Klein RS, Sobel J, et al. A multicenter study of bacterial vaginosis in women with or at risk for human immunodeficiency virus infection. Infect Dis Obstet Gynecol 2001; 9:13341. First citation in article
16. Martin HL, Richardson BA, Nyange PM, et al. Vaginal lactobacilli, microbial flora, and risk of human immunodeficiency virus type 1 and sexually transmitted disease acquisition. J Infect Dis 1999; 180:18638. First citation in article
17. Taha TE, Hoover DR, Dallabetta GA, et al. Bacterial vaginosis and disturbances of vaginal flora: association with increased acquisition of HIV. AIDS 1998; 12:1699706. First citation in article
18. Gray RH, Wawer MJ, Kiwanuka N, et al. Is bacterial vaginosis a cause or a consequence of HIV infection Studies from Rakai, Uganda (abstract 146). In: Program and abstracts of the 11th Conference on Retroviruses and Opportunistic Infections (San Francisco). Alexandria, VA: Foundation for Retrovirology and Human Health, 2004. First citation in article
19. Hillier SL. Normal vaginal flora. In: Holmes KK, Sparling PF, Mardh PA, et al., eds. Sexually transmitted diseases. 3rd ed. New York: McGraw-Hill, 1999:191203. First citation in article
20. Cohen CR, Plummer FA, Mugo N, et al. Increased interleukin-10 in the endocervical secretions of women with non-ulcerative sexually transmitted diseases: a mechanism for enhanced HIV-1 transmission AIDS 1999; 13:32732. First citation in article
21. Sturm-Ramirez K, Gaye-Diallo A, Eisen G, Mboup S, Kanki PJ. High levels of tumor necrosis factor- and interleukin-1 in bacterial vaginosis may increase susceptibility to human immunodeficiency virus. J Infect Dis 2000; 182:46773. First citation in article
22. Wacholder S, Gail M, Pee D. Selecting an efficient design for assessing exposure-disease relationships in an assembled cohort. Biometrics 1991; 47:6376. First citation in article
23. Clayton D, Hills M. Statistical models in epidemiology. New York: Oxford University Press, 1993. First citation in article
24. Amsel R, Totten PA, Spiegel CA, Chen KC, Eschenbach D, Holmes KK. Nonspecific vaginitis: diagnostic criteria and microbial and epidemiologic associations. Am J Med 1983; 74:1422. First citation in article
25. 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; 29:297301. First citation in article
26. Myer L, Denny L, De Souza M, Barone MA, Wright TC, Kuhn L. Intravaginal practices, HIV and other sexually transmitted diseases among South African women. Sex Transm Dis 2004; 31:1749. First citation in article
27. Hosmer DW, Lemeshow S. Applied survival analysis: regression modeling of time-to-event data. New York: Wiley & Sons, 1999. First citation in article
28. Schwebke JR, Hillier SL, Sobel JD, McGregor JA, Sweet RL. Validity of the vaginal Gram stain for the diagnosis of bacterial vaginosis. Obstet Gynecol 1996; 88:5736. First citation in article
29. Schwebke JR, Richey CM, Weiss HL. Correlation of behaviors with microbiological changes in vaginal flora. J Infect Dis 1999; 180:16326. First citation in article
30. Smart S, Singal A, Mindel A. Social and sexual risk factors for bacterial vaginosis. Sex Transm Infect 2004; 80:5862. First citation in article
31. Wiesenfeld HC, Hillier SL, Krohn MA, Landers DV, Sweet RL. Bacterial vaginosis is a strong predictor of Neisseria gonorrhoeae and Chlamydia trachomatis infection. Clin Infect Dis 2003; 36:6638. First citation in article
32. Hay PE, Ugwumadu A, Chowns J. Sex, thrush and bacterial vaginosis. Int J STD AIDS 1997; 8:6038. First citation in article
33. Klebanoff MA, Hauth JC, MacPherson CA, et al. Time course of the regression of asymptomatic bacterial vaginosis in pregnancy with and without treatment. Am J Obstet Gynecol 2004; 190:36370. First citation in article
34. Myer L, Morroni C, Link BG. Impact of measurement error in the study of sexually transmitted infections. Sex Transm Infect 2004; 80:31823. First citation in article
35. Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health 1998; 88:159. First citation in article
36. McCormack WM, Covino JM, Thomason JL, et al. Comparison of clindamycin phosphate vaginal cream with triple sulfonamide vaginal cream in the treatment of bacterial vaginosis. Sex Transm Dis 2001; 28:56975. First citation in article
37. Hanson JM, McGregor JA, Hillier SL, et al. Metronidazole for bacterial vaginosis: a comparison of vaginal gel vs. oral therapy. J Reprod Med 2000; 45:88996. First citation in article
38. Wawer MJ, Sewankambo NK, Serwadda D, et al. Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomised community trial. Rakai Project Study Group. Lancet 1999; 353:52535. First citation in article
39. Patton DL, Cosgrove Sweeney YT, Antonio MA, Rabe LK, Hillier SL. Lactobacillus crispatus capsules: single-use safety study in the Macaca nemestrina model. Sex Transm Dis 2003; 30:56870. First citation in article
40. Harwell JI, Moench T, Mayer KH, Chapman S, Rodriguez I, Cu-Uvin S. A pilot study of treatment of bacterial vaginosis with a buffering vaginal microbicide. J Womens Health (Larchmt) 2003; 12:2559. First citation in article
41. Gerli S, Rossetti D, Di Renzo GC. A new approach for the treatment of bacterial vaginosis: use of polyhexamethylene biguanide. A prospective, randomized study. Eur Rev Med Pharmacol Sci 2003; 7:12730. First citation in article