Sexually Transmitted Diseases:
Accuracy of Clinical Diagnosis of Bacterial Vaginosis by Human Immunodeficiency Virus Infection Status
Gallo, Maria F. PHD*; Jamieson, Denise J. MD, MPH*; Cu-Uvin, Susan MD†; Rompalo, Anne Msc, MD‡; Klein, Robert S. MD§; Sobel, Jack D. MD¶
From the *Division of Reproductive Health, Centers for Disease Control and Prevention, Atlanta, GA; †Brown University School of Medicine, Providence, RI; ‡Johns Hopkins University School of Medicine, Baltimore, MD; §Mount Sinai School of Medicine, New York, NY; and ¶Department of Medicine, Wayne State University School of Medicine, Detroit, MI
Supported by cooperative agreement Nos. U64/CCU106795, U64/CCU206798, U64/CCU306802, and U64/CCU506831 with the Centers for Disease Control and Prevention.
The HIV Epidemiology Research Study Group consists of the following: Robert S. Klein, MD, Ellie Schoenbaum, MD, Julia Arnsten, MD, MPH, Robert D. Burk, MD, Chee Jen Chang, PhD, Penelope Demas, PhD, and Andrea Howard, MD, MSc, from Montefiore Medical Center and the Albert Einstein College of Medicine; Paula Schuman, MD, and Jack Sobel, MD, from the Wayne State University School of Medicine; Anne Rompalo, MD, David Vlahov, PhD, and David Celentano, PhD, from the Johns Hopkins University School of Medicine; Charles Carpenter, MD, and Kenneth Mayer, MD, from the Brown University School of Medicine; Ann Duerr, MD, Lytt I. Gardner, PhD, Scott Holmberg, MD, Denise Jamieson, MD, MPH, Jan Moore, PhD, Ruby Phelps, Dawn Smith, MD, and Dora Warren, PhD, from the Centers for Disease Control and Prevention; and Katherine Davenny, PhD, from the National Institute of Drug Abuse.
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
Correspondence: Maria F. Gallo, PhD, Division of Reproductive Health, Centers for Disease Control and Prevention, 4770 Buford Highway, Mail Stop K-34, Atlanta, GA 30341–3724. E-mail: email@example.com.
Received for publication April 15, 2010, and accepted September 6, 2010.
Objective: To assess the accuracy of clinical diagnosis of bacterial vaginosis (BV) by using Amsel criteria, overall and by human immunodeficiency virus (HIV) infection status.
Methods: Women with HIV, or at risk for HIV, participated in the HIV Epidemiology Research Study, a prospective study conducted in 4 US sites. At enrollment and follow-up visits, scheduled at 6-month intervals for ≤5 years, participants received gynecologic examinations, had specimens collected, and underwent standardized interviews. We used McNemar test statistic to evaluate agreement between Amsel criteria and Nugent scoring. Using Nugent scoring as the reference standard, we calculated sensitivity and specificity for Amsel criteria and for 3 other classifications of clinical BV. Our results are based on data collected from 9140 study visits by 862 HIV-infected women and 421 HIV-uninfected women.
Results: Amsel criteria and Nugent scoring did not agree in the classification of BV cases (P < 0.01). Amsel criteria had poor sensitivity (60%; 95% confidence interval, 58%–61%) and specificity (90%; 95% confidence interval, 89%–91%) with wide differences in test properties by study site. We found no differences in diagnosing BV by HIV infection status.
Conclusions: The under- and overdiagnosing of BV clinically suggests that the accuracy of Amsel criteria for routine screening of asymptomatic women might be lower than previous estimates; that clinicians need more rigorous training to apply subjective Amsel criteria accurately; or that wide heterogeneity in cases might prevent agreement between clinical and laboratory diagnoses, with future research needed to better understand the criteria or morphotypes associated with specific adverse outcomes.
Bacterial vaginosis (BV) is a common, lower genital tract syndrome resulting when the normally dominant lactobacilli in the vagina are replaced with greater concentrations of anaerobic bacteria.1 BV has been associated with prevalent and incident infection with human immunodeficiency virus (HIV).2–6 It also has been linked to increased vaginal shedding of HIV-1 RNA in the female genital tract,7–9 and BV-associated microorganisms appear to enhance HIV-1 expression in vitro.10 Thus, BV could contribute to the spread of HIV by increasing susceptibility in HIV-negative women as well as by increasing infectiousness in HIV-positive women.
The control of BV could hold an important role in limiting the transmission of HIV. BV typically is diagnosed for patient care on the basis of its clinical manifestations (“Amsel criteria”).11 In contrast, for research purposes, BV often is diagnosed with a semiquantitative method of scoring bacterial morphotypes seen on Gram-stained vaginal smears (“Nugent scoring”).12 Although Nugent scoring is a more objective diagnostic method, it requires specialized laboratory training and more time to perform. Therefore, public health programs designed to control BV in HIV-endemic areas or among HIV-positive subgroups usually depend on Amsel criteria for detecting cases. In general, diagnoses that are based on Amsel criteria have been shown to have fairly good agreement with diagnoses based on Nugent scoring,13–18 although whether clinical diagnosing performs comparably among HIV-positive women has not been well established.
MATERIALS AND METHODS
The HIV Epidemiology Research Study was a prospective cohort study that enrolled 871 HIV-infected women and 439 HIV-uninfected women in 1993–1995 in 4 US sites (Baltimore, MD; Bronx, NY; Detroit, MI; and Providence, RI).19 Eligible women were 16 to 55 years of age, did not have acquired immune deficiency syndrome-defining clinical diagnosis, and were either injection-drug users or had high-risk sexual behaviors (defined as >5 sex partners in the prior 5 years, trading sex for money or drugs, or having sex with a man who was either an injection-drug user or who was known or suspected to be HIV infected). Only women who gave informed consent for participation were enrolled in the study. Institutional review boards of the participating sites as well as the Centers for Disease Control and Prevention approved the research.
After the enrollment visit, participants were asked to attend follow-up visits at six-month intervals, at which they were interviewed regarding demographics; psychosocial factors; medical, obstetric, and gynecologic histories; sexual and drug behaviors; and medication use. At study visits, participants also received a standardized limited physical examination, which included a pelvic examination, and had specimens collected for testing, including HIV, BV, candidiasis, trichomoniasis, yeast, gonorrhea, and chlamydial infection. Study visits were not scheduled or conducted for the purpose of evaluating or treating symptoms.
BV was diagnosed using both laboratory and clinical methods. Swabs of the posterior vaginal fornix were used to prepare Gram-stained slides, which were then air-dried, fixed in methanol, and shipped to a central laboratory (Detroit Medical Center University Laboratories, Detroit, MI). A single technician, who was masked to the clinical status of participants, used oil immersion with ×1000 magnification to quantify and score Lactobacillus morphotypes, Gardnerella and Bacteroides species morphotypes, and Mobiluncus species morphotypes. Following the method described by Nugent et al,12 scores for the first 2 morphotypes ranged from 0 to 4, whereas scores for Mobiluncus species morphotypes ranged from 0 to 2. Specimens with higher numbers of Gardnerella and Bacteroides species morphotypes and Mobiluncus species morphotypes identified were assigned higher scores. In contrast, scores for the lactobacillus morphotypes were inversely related to the number of organisms identified. Scores for the 3 morphotypes were summed for a final Nugent score, which had a possible range of 0 to 10. Nugent scores of 0 to 6 were defined as “normal,” and scores of 7 to 10 were defined as “BV.” An independent investigator read a random sample of Gram-stained slides (n = 10); the 2 readers had 100% agreement on the final Nugent scores.
Clinical BV was diagnosed when at least 3 of the following 4 modified Amsel criteria were met: abnormal vaginal discharge identified by a clinician, vaginal fluid pH >4.5, release of an amine (“fishy”) odor from vaginal fluid when mixed with 10% potassium hydroxide, and “clue cells” (cell borders obscured by a coating of bacteria) on wet microscopy of vaginal fluid.11 Women were classified as BV negative if they met fewer than 3 criteria. Those with insufficient criteria assessed were coded as missing BV status. Our criteria differed from those described by Amsel et al only in that the vaginal discharge did not specifically have to be “thin” and “homogeneous” appearing.
We calculated McNemar test statistic to evaluate the agreement between Amsel criteria and Nugent scoring for diagnosing BV. To assess the accuracy of clinical diagnosing of BV among HIV-positive women, we calculated sensitivity and specificity for BV diagnosed by the Amsel criteria stratified according to HIV status, using BV diagnosed by Nugent scoring as the reference standard. Note that the choice of diagnostic test as the reference standard is somewhat arbitrary. The prevalence of BV has been shown to be higher among HIV-infected women than HIV-uninfected women in the study population.20,21 Thus, we focused on sensitivity and specificity rather than measures of test performance, such as positive and negative predictive values, which are influenced by the prevalence of the condition in the population.22
In addition, we assessed the test properties of 3 alternate classifications for clinical BV. In many settings, women are assessed for BV only if vaginal discharge is present. To assess the effects of this “selective testing,” we created 2 new outcomes for clinical diagnosis of BV, in which women were classified by using Amsel criteria only if vaginal discharge was present as assessed (1) by either participant reports or clinician detection or (2) by participant reports alone. Participants without vaginal discharge were categorized as BV negative to simulate a lack of testing among asymptomatic women. Finally, we created a fourth outcome, “simplified” Amsel criteria, to evaluate whether diagnosing BV on the basis of meeting 2 or more of the Amsel criteria (instead of ≥3) would improve sensitivity without sacrificing test specificity, as suggested by others.14,15
We performed a sensitivity analysis by repeating the analyses after restricting the analysis population to visits without a diagnosis of trichomoniasis (n = 7843). The specificity was slightly improved in several of the analyses after cases of trichomoniasis were excluded; however, the interpretation of the study findings remained unchanged and we do not report these data.
A total of 871 HIV-infected and 439 HIV-uninfected women were enrolled in HIV Epidemiology Research Study. The current report excludes participants who HIV seroconverted during the study (n = 12) and is limited to the first 10 visits from each participant. Only visits with sufficient data for classifying BV according to both Amsel criteria and Nugent scoring were included. Results are based on 9140 visits by 862 HIV-infected and 421 HIV-uninfected participants. We used SAS 9.1.3 (SAS Institute, Cary, NC) for data analysis.
The median age of participants was 35 years, and 43% of participants had less than a high school education. The median annual income was $9000. In all, 58% of participants were black, and the remainder were white (24%), Hispanic (17%), or other (1%). BV was diagnosed with Amsel criteria at 28% of the visits, and with Nugent scoring at 37% of the visits (Table 1).
Overall, the clinical and laboratory methods did not agree in the classification of BV cases (P < 0.01) (Table 2). Applying Amsel criteria led to the diagnosing of 2590 episodes of BV, of which only 78% were identified as BV-positive by Nugent scoring. An additional 1375 women-visits were identified as positive for BV on the basis of Nugent scoring. With Amsel criteria, 6550 women-visits were classified as negative for BV, of which only 79% were negative by Nugent scoring. Nugent scoring identified an additional 572 women-visits as BV-negative. Similar patterns were seen when the results were stratified by HIV status.
The sensitivity of Amsel criteria in identifying overall BV episodes (with Nugent scoring as the reference standard) ranged from 23% (95% confidence interval [CI], 21%–27%) to 85% (95% CI, 83%–87%) between the sites and the specificity ranged from 78% (95% CI, 76%–80%) to 94% (95% CI, 92%–95%) (Table 3). Amsel criteria with selective testing of women with any vaginal discharge (participant-reported or clinician-detected) resulted in the identification of fewer overall episodes (sensitivity, 42%; 95% CI, 41%–44%) without improvement in specificity (92%; 95% CI, 91%–93%). Selective testing requiring participant report of vaginal discharge for testing to occur had even lower sensitivity (14%; 95% CI, 0.13–0.15) along with higher specificity (98%; 95% CI, 97%–98%). Simplified Amsel criteria had superior overall sensitivity (74%; 95% CI, 72%–75%) compared with the other clinical BV outcomes but had reduced specificity (80%; 95% CI, 79%–81%).
We found no differences in sensitivity and specificity between HIV-infected and HIV-uninfected women when using the 4 methods of clinically diagnosing BV (Table 3). The sensitivity and specificity of Amsel criteria differed by the study site (Table 3), but did not differ by study visit (data not shown).
We found lack of agreement between clinical and laboratory methods in diagnosing BV. Amsel criteria had sensitivity of 60% and specificity of 90% for identifying BV cases (with Nugent scoring used as the reference standard). With the exception of a study (N = 533) by Karani et al, which reported a sensitivity of 45%,13 prior studies have reported higher levels of sensitivity (69%–100%) when using Amsel criteria to identify BV cases.14–18 The specific methods of categorizing the 4 Amsel criteria often vary slightly between trials. Studies have required the discharge to be white, thin, homogeneous, and/or adherent,13–15,17,18 or the clue cells to constitute ≥20% of vaginal epithelial cells.14,15,17 However, this is an unlikely explanation for the lower sensitivity found in the current study, as we used less stringent definitions (i.e., any discharge and any clue cells) for these criteria. Consistent with previous research,17 we found differences in the accuracy of Amsel criteria by study site. Thus, the high degree of subjectivity involved in applying Amsel criteria, as well as variations in the proficiency of clinicians in identifying cases, are possible explanations for the lower sensitivity found in our study.
Alternatively, earlier studies may have overestimated the accuracy of clinical testing in identifying BV. Our study was much larger than previous ones. For example, the 2 studies reporting the highest sensitivities of 97%14 and 100%18 consisted of only 135 and 107 women, respectively. Thus, random error could be a factor in accounting for the higher levels of sensitivity previously observed.
Routine screening of asymptomatic women, except for high-risk pregnant women, currently is not recommended.23 Guidelines could change if treatment of BV were found to reduce the risk of HIV acquisition or transmission. However, the sole randomized controlled trial conducted on this topic did not show a decrease in HIV acquisition from community-level, routine treatment (regardless of test results or symptoms) for sexually transmitted infections, and reproductive tract infections, including BV.24 The lack of an effect, though, could be attributable to failure to adequately control BV; the trial used metronidazole, 2 g single-dose therapy, which is no longer recommended because of its low efficacy against BV.23 The effectiveness of therapy for BV, in general, is limited, and recurrence following treatment is common.25 Although presumptive treatment for BV could reduce the prevalence of BV26 and possibly, chlamydial infection27; therapy for BV could lead to increased risk of gonorrhea and vaginal candidiasis.25,27 Furthermore, metronidazole treatment for BV could have adverse effects,28 and routine treatment of asymptomatic women could contribute to emergence of resistant strains of bacteria. In the absence of new evidence, presumptive BV treatment for sexually transmitted infection/HIV prevention does not appear warranted.
Our findings are reassuring in that the sensitivity and specificity of Amsel criteria for diagnosing BV did not appear to differ by HIV status. However, assuming that Nugent scoring yields valid results, the clinical diagnosing of BV appeared to be much less sensitive overall than previously indicated.13–18 Furthermore, the differences in identification of cases by Amsel criteria among the study sites highlight the importance of rigorous clinical training. The study findings also suggest that in the typical patient care setting, which usually requires vaginal discharge to prompt testing for BV, the sensitivity of Amsel criteria would be even lower, without any gains in specificity. Relaxing the clinical diagnosing of BV to require a minimum of 2 (instead of 3) of the Amsel criteria, did improve sensitivity but with an accompanying decline in specificity. Thus, the choice of whether to use simplified or traditional Amsel criteria in a particular setting should be made after weighing the relative importance of minimizing false-positive versus false-negative results.
Nugent scoring is subjective. The identification and scoring of morphotypes on the Gram-stained slides could be influenced by differences in techniques used for specimen collection and slide preparation as well as by variability in readers.29 A strength of our study is that we used a single reader for all the specimens evaluated by Nugent scoring. The analysis also benefited from a large sample size, which permits precise estimates. The main limitation involves the assumption that Nugent scoring is more accurate in identifying BV cases than Amsel criteria. Because the methods used for clinical and laboratory diagnosing of BV are measuring different entities,30 determining the relative accuracy of the 2 tests would be difficult. Although BV cases are identified by clinical or laboratory methods, they encompass much heterogeneity. Future research should focus on identifying specific criteria or morphotypes associated with particular adverse events, which could then facilitate the development of better diagnostic methods and treatments for the identified condition.
1.Hillier S, Holmes KK. Bacterial vaginosis. In: Holmes KK, Sparling PF, Mardh P, et al, eds. Sexually Transmitted Diseases, 3rd ed. New York, NY: McGraw-Hill, 1999:563–586.
2.Atashili J, Poole C, Ndumbe PM, et al. Bacterial vaginosis and HIV acquisition: A meta-analysis of published studies. AIDS 2008; 22:1493–1501.
3.Bukusi EA, Cohen CR, Meier AS, et al. Bacterial vaginosis: Risk factors among Kenyan women and their male partners. Sex Transm Dis 2006; 33:361–367.
4.Myer L, Denny L, Telerant R, et al. Bacterial vaginosis and susceptibility to HIV infection in South African women: A nested case-control study. J Infect Dis 2005; 192:1372–1380.
5.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; 180:1863–1868.
6.Taha TE, Hoover DR, Dallabetta GA, et al. Bacterial vaginosis and disturbances of vaginal flora: Association with increased acquisition of HIV. AIDS 1998; 12:1699–1706.
7.Sha BE, Zariffard MR, Wang QJ, et al. Female genital-tract HIV load correlates inversely with Lactobacillus species but positively with bacterial vaginosis and Mycoplasma hominis. J Infect Dis 2005; 191:25–32.
8.Spinillo A, Debiaggi M, Zara F, et al. Factors associated with nucleic acids related to human immunodeficiency virus type 1 in cervico-vaginal secretions. BJOG 2001; 108:634–641.
9.Cu-Uvin S, Hogan JW, Caliendo AM, et al. Association between bacterial vaginosis and expression of human immunodeficiency virus type 1 RNA in the female genital tract. Clin Infect Dis 2001; 33:894–896.
10.Al-Harthi L, Roebuck KA, Olinger GG, et al. Bacterial vaginosis-associated microflora isolated from the female genital tract activates HIV-1 expression. J Acquir Immun Defic Syndr 1999; 21:194–202.
11.Amsel R, Totten PA, Spiegel CA, et al. Nonspecific vaginitis: diagnostic criteria and microbial and epidemiologic associations. Am J Med 1983; 74:14–22.
12.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.
13.Karani A, De Vuyst H, Luchters S, et al. The Pap smear for detection of bacterial vaginosis. Int J Gynaecol Obstet 2007; 98:20–23.
14.Simoes JA, Discacciati MG, Brolazo EM, et al. Clinical diagnosis of bacterial vaginosis. Int J Gynaecol Obstet 2006; 94:28–32.
15.Gutman RE, Peipert JF, Weitzen S, et al. Evaluation of clinical methods for diagnosing bacterial vaginosis. Obstet Gynecol 2005; 105:551–556.
16.Posner SF, Kerimova J, Aliyeva F, et al. Strategies for diagnosis of bacterial vaginosis in a resource-poor setting. Int J STD AIDS 2005; 16:52–55.
17.Schwebke JR, Hillier SL, Sobel JD, et al. Validity of the vaginal gram stain for the diagnosis of bacterial vaginosis. Obstet Gynecol 1996; 88:573–576.
18.Platz-Christensen JJ, Larsson PG, Sundström E, et al. Detection of bacterial vaginosis in wet mount, Papanicolaou stained vaginal smears and in Gram stained smears. Acta Obstet Gynecol Scand 1995; 74:67–70.
19.Smith DK, Warren DL, Vlahov D, et al. Design and baseline participant characteristics of the Human Immunodeficiency Virus Epidemiology Research (HER) Study: A prospective cohort study of human immunodeficiency virus infection in US women. Am J Epidemiol 1997; 146:459–469.
20.Jamieson DJ, Duerr A, Klein RS, et al. Longitudinal analysis of bacterial vaginosis: Findings from the HIV epidemiology research study. Obstet Gynecol 2001; 98:656–663.
21.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; 9:133–141.
22.Grimes DA, Schulz KF. Uses and abuses of screening tests. Lancet 2002; 359:881–884.
23.Centers for Disease Control and Prevention. Sexually Transmitted Diseases Treatment Guidelines, 2006. MMWR Recomm Rep 2006; 55:1–94.
24.Wawer MJ, Sewankambo NK, Serwadda D, et al; Rakai Project Study Group. Control of sexually transmitted diseases for AIDS prevention in Uganda: A randomized community trial. Lancet 1999; 353:525–535.
25.Sobel J, Ferris D, Schwebke JR, et al. Suppressive maintenance antibacterial therapy with 0.75% metronidazole vaginal gel to prevent recurrent bacterial vaginosis. Am J Obstet Gynecol 2006; 194:1283–1289.
26.McClelland RS, Richardson BA, Hassan WM, et al. Improvement of vaginal health for Kenyan women at risk for acquisition of human immunodeficiency virus type 1: Results of a randomized trial. J Infect Dis 2008; 197:1361–1368.
27.Schwebke JR, Desmond R. A randomized trial of metronidazole in asymptomatic bacterial vaginosis to prevent the acquisition of sexually transmitted diseases. Am J Obstet Gynecol 2007; 196:517.e1–517.e6.
28.Carey JC, Klebanoff MA; NICHD MFMU Network. Metronidazole treatment increased the risk of preterm birth in asymptomatic women with trichomonas. Am J Obstet Gynecol 2000; 182:13.
29.Forsum U, Larsson PG, Spiegel C. Scoring vaginal fluid smears for diagnosis of bacterial vaginosis: Need for quality specifications. APMIS 2008; 116:156–159.
30.Tohill BC, Heilig CM, Klein RS, et al. Vaginal flora morphotypic profiles and assessment of bacterial vaginosis in women at risk for HIV infection. Infect Dis Obstet Gynecol 2004; 12:121–126.
© Copyright 2011 American Sexually Transmitted Diseases Association