Analysis of Vaginal Acetic Acid in Patients Undergoing Treatment for Bacterial Vaginosis

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, Nov. 2004, p Vol. 42, No /04/$ DOI: /JCM Copyright 2004, American Society for Microbiology. All Rights Reserved. Analysis of Vaginal Acetic Acid in Patients Undergoing Treatment for Bacterial Vaginosis Amjad N. Chaudry, 1 Paul J. Travers, 2 Jeffrey Yuenger, 3 Lorraine Colletta, 4 Phillip Evans, 2 Jonathan M. Zenilman, 3 * and Andrew Tummon 2 Osmetech plc, Crewe, United Kingdom 1 ; Osmetech Inc., Roswell, Georgia 2 ; and Division of Infectious Diseases, Department of Medicine, Johns Hopkins University School of Medicine, 3 and Baltimore City Health Department, 4 Baltimore, Maryland Received 29 April 2004/Returned for modification 14 June 2004/Accepted 19 July 2004 A gold standard method for the diagnosis of bacterial vaginosis (BV) is lacking. The clinical criteria described by the Amsel technique are subjective and difficult to quantify. Alternatively, the reading of Gramstained vaginal smears by scoring techniques such as those that use the Nugent or Hay-Ison scoring systems is again subjective, requires expert personnel to perform the reading, and is infrequently used clinically. Recently, a new diagnostic device, the Osmetech Microbial Analyzer Bacterial Vaginosis (OMA-BV), which determines a patient s BV status on the basis of measurement of the amount of acetic acid present in a vaginal swab specimen, was approved by the Food and Drug Administration. The present study uses the conducting polymer gas-sensing technology of OMA-BV to measure the concentration of acetic acid in the headspace above vaginal swab specimens from patients undergoing treatment for BV with metronidazole. In 97.8% of the cases the level of acetic acid detected fell sharply during the treatment period, crossing from above to below the diagnostic threshold of 900 ppm. The diagnosis obtained on the basis of the level of vaginal acetic acid was compared with the diagnoses obtained by use of the Amsel criteria and the Nugent scoring system both at the time of initial entry into the study and at the repeat samplings on days 7 and 14. The results obtained with OMA-BV showed overall agreements compared with the results of the Amsel and Nugent tests of 98 and 94%, respectively, for the 34 patients monitored through the treatment process. This provides further evidence that the measurement of vaginal acetic acid by headspace analysis with conducting polymer sensors is a valid alternative to present tests for the diagnosis of BV. Bacterial vaginosis (BV) is the most common cause of vaginitis symptoms among women of childbearing age. The condition affects 10 to 15% of women (13) in the general female population, but an incidence as high as 40% has been reported among women attending sexually transmitted clinics (16). BV is associated with an increased risk for a host of obstetric, gynecological, and neonatal complications, including postoperative infection following hysterectomy (22), miscarriage (15), preterm birth (17), postabortion pelvic inflammatory disease (8), plasma cell endometritis (21), and human immunodeficiency virus infection (28). BV is characterized by a change or an imbalance in the vaginal ecosystem, whereby the number of Lactobacillus species decreases and there is an overgrowth of organisms such as Gardnerella vaginalis and anaerobic organisms such as Mobiluncus spp. (31). The change in flora is accompanied by biochemical changes in the vaginal fluid, including increases in the concentrations of diamines (6) as well as those of polyamines and volatile organic acids (29, 35). These biochemical markers were used for the detection of BV in this study. Treatment is directed at reducing the numbers of these bacteria by several of the different treatment regimens recommended, including treatment with metronidazole and clindamycin, which have * Corresponding author. Mailing address: Division of Infectious Diseases, Department of Medicine, Johns Hopkins University School of Medicine, 4940 Eastern Ave., Baltimore, MD Phone (410) Fax: (410) jzenilma@jhmi.edu. reported treatment failure rates of 4 and 6%, respectively (3, 9, 11). The diagnosis of BV is typically made by using the criteria of Amsel et al. (2) (referred to here as the Amsel criteria), which require three of the following: homogeneous discharge, vaginal fluid ph 4.5, positive amine whiff test, and the presence of 20% clue cells among the total epithelial cells. Individually, each of these tests varies in its sensitivity and specificity for the diagnosis of BV (7, 14, 34). Gram staining of vaginal secretions and grading of the various bacterial morphotypes present to produce a score or level of BV is another method used primarily in research settings (30). These methods include the Nugent scoring system (25, 27) and the Hay-Ison criteria (18). These are sensitive for the diagnosis of BV but require expert personnel to perform the quantification of morphotypes. None of the methods for the diagnosis of BV discussed above are considered to be a gold standard. The Osmetech Microbial Analyzer Bacterial Vaginosis (OMA-BV; Osmetech plc, Crewe, United Kingdom) is an automated in vitro diagnostic device that can be used as a clinical aid for the diagnosis of BV. The technology is based on the gas-phase detection of volatile metabolites produced by the bacteria in clinical samples. The analysis of chemicals in the gas phase is often referred to as headspace analysis. The metabolites are detected with an array of conducting polymer gas sensors which undergo changes in resistance on exposure to specific gases (5, 10, 20, 24, 26). Work conducted by Chandiok et al. (4) first demonstrated the possibility of using conducting polymer gas sensor arrays for the diagnosis of BV. 5170

2 VOL. 42, 2004 TRACKING VAGINAL ACETIC ACID 5171 Subsequent work with prototype OMA systems (12) demonstrated that acetic acid was the primary volatile compound produced from vaginal swabs. Comparative studies have shown that the results obtained with OMA-BV correlate well with those obtained by use of both the Amsel criteria and Gram staining methods (Food and Drug Administration approval no. for OMA-BV, k023677). Supporting studies by gas chromatography (GC)-mass spectrometry (MS) (P. Evans et al., submitted for publication) demonstrated that the discrimination achievable with OMA-BV was primarily due to the different levels of acetic acid present on vaginal swab specimens from patients positive and negative for BV. By using these data, a threshold concentration of 900 ppm was established; i.e., readings greater than 900 ppm indicate the presence of BV. This threshold concentration is expressed as the headspace concentration of acetic acid produced by the vaginal swab, which is equivalent to that produced by a 900-mg dm 3 (i.e., 900 ppm, by mass) solution of acetic acid in 0.01 M HCl. Other volatile compounds were also detected, including ammonia, but there was little or no correlation with BV status. MATERIALS AND METHODS Study design. The study was a single-center BV patient follow-up study carried out at the Baltimore City Health Department and Johns Hopkins University, Baltimore, Md. The objective was to determine whether successful treatment of BV-positive patients affected the concentration of acetic acid, the principal chemical marker used to identify BV. The primary efficacy criterion was a drop in the vaginal acetic acid concentration from above the 900-ppm threshold to below this concentration after treatment for BV. The secondary efficacy criteria were changes in the subject symptoms, clinical diagnosis, the Amsel criteria, and Nugent scores from the time of recruitment through the subsequent follow-up visits after treatment. The research protocol was approved by the Institutional Review Board at the Baltimore City Health Department and the Western Institutional Review Board for Johns Hopkins University. This was an observational study for a 2-week period after treatment. The inclusion criteria were female sex, age 18 years or older, a confirmed diagnosis of BV, and positivity for all four Amsel criteria at the time of the initial visit. Subjects were excluded on the basis of the following: the last menstrual period was 2 weeks prior to the initial visit (this increased the likelihood that the subject would have had menses on the follow-up visits), pregnancy, breast-feeding, known primary or secondary immunodeficiency, antifungal or antimicrobial therapy in the previous 14 days, and activities contraindicated during metronidazole treatment (e.g., active alcohol use). On the follow-up visits, the subjects were excluded if they had menses or a repeat physical examination was contraindicated or refused. After the subjects provided informed consent, they were asked to refrain from using any intravaginal products for the duration of the study (e.g., douches, feminine deodorant products, spermicides, lubricated condoms, and diaphragms) and to use a nonlubricated condom when engaging in vaginal intercourse for the duration of the study. The subjects were prescribed the standard oral metronidazole regimen of 500 mg twice a day for 7 days and were asked to return to the clinic 7 and 14 days after the initial visit. If symptoms persisted on day 7, a repeat course of oral metronidazole was prescribed. Clinical sampling and screening for sexually transmitted diseases. At enrollment, diagnostic tests were performed according to standard clinical practice, including screening for BV (by use of the Amsel criteria), Candida spp. (wet mount), Trichomonas vaginalis (wet mount), Neisseria gonorrhoeae (culture), Chlamydia trachomatis (PCR), herpes (culture, if indicated), syphilis (rapid plasma reagin card test), and human immunodeficiency virus (serologic testing by enzyme-linked immunosorbent assay). Additionally, two vaginal swab specimens, from the lateral wall and the posterior fornix of the vagina, were collected for testing with the OMA-BV and scoring with the Nugent system by use of a Gram stain smear. Demographic information, final diagnoses, concurrent medications, douching habits, the time since the last sexual intercourse, contraceptive use, and menstrual history were obtained. FIG. 1. Acetic acid calibration curve used to predict the amount of acetic acid in samples from BV patients analyzed with OMA. At the follow-up visits on days 7 and 14, the tests for the Amsel criteria were repeated, and two vaginal swab specimens were taken for scoring with the Nugent system and analysis with the OMA system. Instrumentation and operation. OMA maintains a carousel that contains both sample and calibration vials at a constant temperature of 40 C. The automatic sampling of each calibration and sample vial involves insertion of a coaxial needle through the vial septum and sweeping of the headspace gas out of the vial with a controlled flow of humidified nitrogen. The headspace gas is delivered to a 48-element sensor array which consists of 12 replicates of four different polymer types. The system is computer controlled, and data are captured to files on a local personal computer for further analysis. The measurement process compares the sensor response for a clinical sample with that for a series of calibration vials to determine the concentration of acetic acid in the vaginal swab. Clinical samples were prepared by transferring the tip of a vaginal swab specimen from a patient into 22-ml headspace vials and sealing the vials with an aluminum cap containing a polytetrafluoroethylene-lined silicone septum. Calibration vials were prepared by pipetting 1 ml of the standards (200, 500, 900, 2,000, and 5,000 ppm acetic acid in 0.01 M HCl and 10 ppm ammonia in 0.01 M NaOH) into 22-ml headspace vials and then sealing the vials, as described above. The inclusion of an ammonia standard in the calibration sequence arises as a result of the observation in previous studies that a small number of vaginal swab specimens produce significant levels of ammonia in the headspace. Data analysis. The resistance of each of the polymer sensors was measured during the sampling period, and the percent change from the initial value was calculated at 1-s intervals. An analysis time period was selected to optimize the analysis of acetic acid levels. The average response for each of the four polymer types was calculated at this point to produce a four-element response matrix for each calibration and sample vial in the carousel. The responses for the acetic acid and ammonia calibrants were transformed by principal component (PC) analysis into a set of scores and loading vectors (19, 36). This had the effect of separating out the acetic acid and ammonia responses into orthogonal vectors, which simplified the subsequent analysis. The PC1 score correlated with the acetic acid concentration, and the PC1 values for the calibrants were used to generate a calibration curve. The loadings from the PC analysis showed how the sensors related to one another and were used to calculate PC scores for the four-element response patterns generated from the patient samples. The PC1 score from a patient vial was converted into an acetic acid concentration by using the calibration curve illustrated in Fig. 1. Note that patient acetic acid levels were always calculated with calibration vials run on the same carousel. This minimized any potential impact of sensor response drift. To validate this method for determination of the acetic acid concentration, a study was carried out prior to the clinical trial in which aqueous solutions of acetic acid of known concentration were used as pseudosamples. Some of these test solutions had been spiked with 5 ppm ammonia. Tests were carried out with two modified OMA systems and a total of 340 randomized pseudosamples. The results for these test solutions are summarized in Table 1. The percent deviation columns in Table 1 record how far the mean values calculated for the repeat samples of each sample type were from the prepared

3 5172 CHAUDRY ET AL. J. CLIN. MICROBIOL. TABLE 1. Summary of results from an in-house study conducted to determine the acetic acid concentrations in pseudosamples with known concentrations Instrument 1 (142 test samples) Instrument 2 (198 test samples) Sample type Mean (ppm) SD (ppm) Relative SD (ppm) % Deviation of mean from real value Mean (ppm) SD (ppm) Relative SD (ppm) % Deviation of mean from real value 100 ppm ppm ppm ppm ,000 ppm 2, , ,000 ppm 5, , n n ,000 n 1, , concentrations. Note that the significantly larger percent errors for the test solutions with 100 ppm are a result of the fact that these values were calculated by extrapolation rather than interpolation. For the clinical study, accuracy below 200 ppm was not considered critical. The performance of OMA with the test samples spiked with 5 ppm ammonia (i.e., 200n, 900n, and 2,000n [where n denotes the presence of 5 ppm ammonia; e.g., 200n indicates a solution of 200 ppm acetic acid and 5 ppm ammonia] in Table 1) was not significantly different from the performance of OMA with the samples that contained only acetic acid. This suggested that the presence of significant amounts of ammonia in a clinical sample would not affect the accuracy of the calculated acetic acid concentration. When the data for 100 ppm are ignored, the relative standard deviations for the calculated acetic acid concentrations ranged from 4.8 to 10.8%, with an average of 7.8%. Therefore, to a first approximation, the 95% confidence intervals for a particular value can be determined by calculating a relative standard deviation of 15.6%. For example, assume that the system returns a calculated acetic acid concentration of 1,000 ppm. The estimated 95% confidence interval for this measurement is thus 840 to 1,160 ppm. RESULTS A total of 66 subjects with a clinical diagnosis of BV on the initial visit, as confirmed by positivity for all four Amsel criteria, were initially enrolled in the study. Of these, 32 subjects did not meet the inclusion criteria for the study and were excluded from the final analysis. The reasons for exclusion were as follows; 27 subjects did not return for either of the scheduled follow-up visits 7 or 14 days after the initial visit (classified as loss to follow-up), which we speculate occurred because of a lack of adequate incentive once the condition had been treated. Four subjects did not meet the entry criteria (one did not have a full record of results according to the Amsel criteria, one was pregnant, and two were using an antifungal or antimicrobial medication), and the sample from one subject was not analyzed by OMA due to an instrument failure. Of the remaining 34 subjects, 24 completed both of the follow-up visits in the prescribed time frame. The data for 10 subjects who were able to return for at least one of the follow-up visits were also included. All of these patients had completed the course of treatment. The majority of subjects examined were African-American (94%) and had an average age of 29 years (maximum age, 45 years; minimum age, 18 years). The results for the subjects included in the study are listed in Table 2. It shows how the level of acetic acid for each subject varied during the treatment phase (i.e., between days 1 and 7) and the recovery phase (between days 7 and 14) of the study. Table 2 also shows patient status for the clinical diagnosis of BV (on the basis of the Amsel criteria) and the Nugent scores for each patient visit, together with any concurrent infections. Note that the values presented for acetic acid concentrations are single-point measurements because only one patient swab was available for analysis with OMA. The estimated 95% confidence intervals for each patient can be determined by calculating the value 15.6%. A box-whisker plot of the acetic acid response for each sample analyzed is shown in Fig. 2. The box has horizontal lines at the lower quartile, median, and upper quartile values, which were calculated for the day 1, day 7, and day 14 data sets. The whiskers are lines extending from each end of the box to show the 2.5 and 97.5 centiles. For the day 7 and the day 14 data sets, a number of points plotted outside the whiskers. Typically, for a normally distributed data set these would be described as outliers, but in this instance they reflect the fact that the distributions of acetic acid are skewed. A previous study (Evans et al., submitted for publication) has shown that the distribution of acetic acid in non-bv patients approximates the lognormal distribution. It should be noted that three of the sample points from Table 2 were excluded from the box-whisker plot (i.e., patient 60 for day 7, patient 11 for day 14, and patient 54 for day 14) because for these three patients the clinician determined by use of the Amsel criteria that the BV had not resolved, which agreed with the results obtained with OMA. It is evident from the box-whisker plots that, after treatment with metronidazole, on day 1 BV-positive swab samples show much lower levels of acetic acid than swabs taken from the same patient before treatment. Furthermore, the results demonstrate that the acetic acid marker threshold (i.e., 900 ppm) used by the OMA-BV device to determine whether a sample is BV positive or negative holds true in this study. By using the calculated acetic acid concentrations, the diagnosis of BV status was determined with OMA for each sample. The agreements between the results obtained with OMA and

4 VOL. 42, 2004 TRACKING VAGINAL ACETIC ACID 5173 TABLE 2. Acetic concentrations, clinical diagnosis, and Nugent scores for the patients included in the study a Patient no. Acetic acid concn (ppm) Diagnosis b Nugent score Day 1 Day 7 Day 14 Day 1 Day 7 Day 14 Day 1 Day 7 Day , BV Res Res , BV, Can Res Res , BV Res Res , BV Res Res , BV Res Res 7 URS 0 9 6, BV Res Res ,300 1,600 BV BV , BV Res Res, Can , BV Res Res BV, CT Res, CT Res , BV Res , BV Res Res , BV Res Res, Can, Crohn s , BV, CT Res, CT Res , BV, PID Res, TV , BV Res Res , BV Res , BV Res, Tinea capitis Res , BV Res, Can Res BV, MPC, Syphilis Res Res , BV, CT Res, Can , BV, CT Res, Can Res 8 URS ,400 0 BV Res , BV, NGU Res, Can Res ,500 0 BV Res , BV Res Res , ,000 BV Res, Can BV , BV, NG Res, NG Res, Can , BV, TV Res Res ,100 2, BV BV Res ,600 0 BV Res , BV Res Res ,900 0 BV, Fibroids Res, NG , BV Res 8 0 a Abbreviations and symbols: Res, BV resolved; TV, Trichomonas vaginalis; Can, Candida; PID, pelvic inflammatory disease; URS, unreadable slide; NG, gonorrhoea; CT, chlamydia; NGU, contact nongonorrheal urethritis; MPC, mucopurulent cervicitis;, sample not taken. b The clinical diagnosis of BV was based on the Amstel criteria. those obtained by use of the Amsel criteria and the Nugent system are summarized in Table 3. There was excellent agreement between the results obtained by use of the Amsel criteria and those obtained with OMA, with disagreements between the tests for only 2 of the 92 samples with which both tests were performed (i.e., 98% agreement). Typically, a Nugent score of 7 or greater is defined as being consistent with BV and a Nugent score of 3 or less is indicative of normal flora. A score between 4 and 6 is indicative of an intermediate flora. In this study the samples with intermediate results according to the Nugent scoring system were classified as BV negative to reflect the common belief that a Nugent score of 6 or less does not necessarily correlate with BV (23, 33) and should not be interpreted to be consistent with incipient BV. Note that none of the samples from the patients scored intermediate on the initial visit, while the samples from three patients scored intermediate on days 7 and 14 (see Table 2 for details). This simplifies comparison with the binary outcome of the test with OMA. The overall correlation between the test with OMA and the Nugent scoring system was also very good, with disagreement of the results for only 5 of the 90 samples with which both tests were performed (i.e., 94% agreement). DISCUSSION Many of the organisms associated with BV (e.g., G. vaginalis and anaerobic organisms such as Mobiluncus and Prevotella spp.) are known to produce acetic acid as a metabolic byproduct (32). Al-Mushrif et al. (1) have shown that the fatty acids produced by these organisms impair chemotaxis. They further suggest that this slowing of the immune response allows further proliferation of the infecting organisms. It is therefore logical to expect elevated vaginal acetic acid concentrations in patients with BV. The results presented in this study provide further evidence that the vaginal acetic acid level can be a useful diagnostic for BV. This assertion is supported by the results obtained by tracking patients in the treatment and recovery phases of an oral metronidazole treatment regimen. In the vast majority of cases the level of acetic acid detected fell sharply during the treatment period, crossing from above to below the 900-ppm threshold.

5 5174 CHAUDRY ET AL. J. CLIN. MICROBIOL. FIG. 2. Box-whisker plot of study data, including the 900-ppm acetic acid concentration threshold for the diagnosis of BV. In this study, BV status was tracked by using the two established clinical methods for diagnosis together with the measurement of vaginal acetic acid concentrations. Comparison of the disagreements between the three different methods shows that statistically there was no significant difference in the performance of each of the methods. The disagreements between the three methods were tested by the chi-square test ( ). Because multiple comparisons were carried out, the Bonferroni correction was applied and gave a nonsignificant (P 0.2) outcome for both the standard and the corrected tests. This supports the assertion that there is no significant difference between the three methods, indicating that the use of vaginal acetic acid levels to diagnose BV is as accurate as present techniques. Existing methods for the clinical diagnosis of BV have important limitations. In this study, only patients positive for all four of the Amsel criteria at the outset were included in order to minimize the inclusion of patients with false-positive diagnoses. This was stipulated to try to avoid the inclusion of patients who were not truly BV positive. The diagnosis of BV by measuring vaginal acetic acid concentrations for vaginal swab specimens by headspace analysis has a number of advantages over the other tests. Namely, it is objective and does not rely on subjective observations of the user, no specialized knowledge or expertise is required to use the test, and it provides a straightforward classification of the patient as BV positive or negative. This work confirms the findings from previous studies demonstrating the suitability of using Osmetech s conducting polymer gas-sensing technology to determine the acetic acid concentrations on vaginal swab specimens. An advantage of this technique is its relative simplicity and its potential for translation to an office-based system. Other techniques for the measurement of acetic acid concentrations exist, but none readily lend themselves to this application. High-performance liquid chromatography (HPLC) and GC may both be used to analyze swab specimens for acetic acid concentrations. Most methodologies for these techniques rely on derivatization of the acid to an ester to simplify the analysis. This is time-consuming and relatively complex and has the disadvantage of variable recovery efficiencies, imposing the need for the use of an internal control to compensate for the variability. Advances in HPLC and GC column technologies have produced phases that can tolerate the highly polar nature of the analyte, but the initial sample weight must still be determined to be able to quantify the acetic acid present in the sample. In the case of GC-based techniques, high temperatures are normally required to prevent sample loss due to acid sticking, and specialized solvent mixtures are required for HPLC. Headspace sampling obviates the need for sample preparation or derivatization and the use of internal standards. Other headspace sampling systems exist, such as headspace GC-MS, but these are vastly more complex and time-consuming solutions to what can be a relatively simple analysis. In conclusion, this study describes an alternative method for the diagnosis of BV on the basis of analysis of vaginal acetic acid levels that, by using the established threshold of 900 ppm, provides a clear and unequivocal answer as to the presence of BV. In this study the vaginal acetic acid levels for BV-positive patients undergoing treatment with metronidazole dropped as the clinical symptoms resolved. The technology is under development for the production of a device that is suitable for use by nonexpert personnel in a point-of-care setting in which extended calibration routines will not be required and with which the results can be generated within the clinical consultation time frame. Test system and day TABLE 3. Agreements between the results obtained by OMA and those obtained by use of the Amsel criteria and the Nugent classified results a No. of samples: No. of samples with the following result with OMA: Positive Negative True positive False negative True negative False positive % Agreement Amsel criteria Nugent system a Samples with the intermediate class of results according to the Nugent scoring system were classified as negative.

6 VOL. 42, 2004 TRACKING VAGINAL ACETIC ACID 5175 ACKNOWLEDGMENTS We are grateful to H. Johnson, J. Giles, and P. Turrell for excellent technical assistance. REFERENCES 1. Al-Mushrif, A. Eley, and B. M. Jones Inhibition of chemotaxis by organic acids from anaerobes may prevent a purulent response in bacterial vaginosis. J. Med. Microbiol. 49: Amsel, R., P. A. Totten, C. A. Spiegel, K. C. Chen, D. Eschenbach, and K. K. Holmes Nonspecific vaginitis. Diagnostic criteria and microbial and epidemiologic associations. Am. J. Med. 74: Centers for Disease Control and Prevention Sexually transmitted diseases treatment guidelines. Morb. Mortal. Wkly. Rep. 51(RR-6): Chandiok, S., B. A. Crawley, B. A. Oppenheim, P. R. Chadwick, S. Higgins, and K. C. Persaud Screening for bacterial vaginosis: a novel application of artificial nose technology. J. Clin. Pathol. 50: Chaudry, A. N., T. M. Hawkins, and P. J. Travers A method for selecting an optimum sensor array. Sensors Actuators B 69: Chen, K. C., R. Amsel, D. A. Eschenbach, and K. K. Holmes Biochemical diagnosis of vaginitis: determination of diamines in vaginal fluid. J. Infect. Dis. 145: Eschenbach, D. A., S. L. Hillier, C. Critchlow, C. Stevens, T. DeRouen, and K. K. Holmes Diagnosis and clinical manifestations of bacterial vaginosis. Am. J. Obstet. Gynecol. 158: Eschenbach, D. A Bacterial vaginosis and anaerobes in obstetricgynecologic infection. Clin. Infect. Dis. 16(Suppl. 4):S282 S Ferris, D. G., M. S. Litaker, L. Woodward, D. Mathis, and J. Hendrich Treatment of bacterial vaginosis: a comparison of oral metronidazole, metronidazole vaginal gel and clindamycin vaginal cream. J. Fam. Pract. 41: Gardner, J. W., and P. N. Bartlett A brief history of electronic noses. Sensors Actuators B 18 19: Greaves, W. L., J. Chungafung, B. Morris, A. Haile, and J. L. Townsend Clindamycin versus metronidazole in the treatment of bacterial vaginosis. Obstet. Gynecol. 72: Hay, P., A. Tummon, M. Ogunfile, A. Adebiyi, and A. Adefowora Evaluation of a novel diagnostic test for bacterial vaginosis: the electronic nose. Int. J. STD AIDS 14: Hay, P Recurrent bacterial vaginosis. Curr. Infect. Dis. Rep. 2: Hay, P Diagnosing bacterial vaginosis. CME bulletin. Sex. Transm. Infect. HIV 3: Hay, P. E., R. F. Lamont, D. Taylor-Robinson, D. J. Morgan, C. Ison, and J. Pearson Abnormal bacterial colonisation of the genital tract and subsequent preterm delivery and late miscarriage. BMJ 308: Hillier, S., and K. K. Holmes Bacterial vaginosis, p In K. K. Holmes, P. A. Mardh, P. F. Sparks, and P. J. Wiener (ed.), Sexually transmitted diseases, 2nd ed. McGraw-Hill Book Co., New York, N.Y. 17. Hillier, S. L., R. P. Nugent, D. A. Eschenbach, M. A. Krohn, R. S. Gibbs, D. H. Martin, M. F. Cotch, R. Edelman, J. G. Pastorek II, A. V. Rao, et al Association between bacterial vaginosis and preterm delivery of a low-birth-weight infant. N. Engl. J. Med. 333: Ison, C. A., and P. E. Hay Validation of a simplified grading of Gram stained vaginal smears for use in genitourinary medicine clinics. Sex. Transm. Infect. 78: Jackson, J. E Principal component analysis and factor analysis. Part 1. Principal components. J. Qual. Technol. 13: Kanatzidis, M. G Conductive polymers. Chem. Engineer. News 3: Korn, A. P., G. Bolan, N. Padian, M. Ohm-Smith, J. Schachter, and D. V. Landers Plasma cell endometritis in women with symptomatic bacterial vaginosis. Obstet. Gynecol. 85: Larsson, P. G., J. J. Platz-Christensen, U. Forsum, and C. Pahlson Clue cells in predicting infections after abnormal hysterectomy. Obstet. Gynecol. 77: Livengood, C. H., III, J. L. Thomason, and G. B. Hill Bacterial vaginosis: diagnostic and pathogenetic findings during topical clindamycin therapy. Am. J. Obstet. Gynecol. 163: Nagase, H., K. Wakabayashi, and T. Imanaka Effect of doping anions in polypyrrole gas sensors. Sensors Actuators B 13 14: Nugent, R. P., M. A. Krohn, and S. L. Hillier Reliability of diagnosing bacterial vaginosis is improved by a standardized method of Gram stain interpretation. J. Clin. Microbiol. 29: Persaud, K. C., and P. Pelosi Semiconducting organic polymers for gas sensors. European patent EP Schwebke, J. R., S. L. Hillier, J. D. Sobel, J. A. McGregor, and R. L. Sweet Validity of the vaginal Gram stain for the diagnosis of bacterial vaginosis. Obstet. Gynecol. 88(4 Pt 1): Sewankambo, N., R. H. Gray, M. J. Wawer, L. Paxton, D. McNaim, F. Wabwire-Mangen, D. Serwadda, C. Li, N. Kiwanuka, S. L. Hillier, L. Rabe, C. A. Gaydos, T. C. Quinn, and J. Konde-Lule HIV-1 infection associated with abnormal vaginal flora morphology and bacterial vaginosis. Lancet 350: Spiegel, C. A., R. Amsel, D. Eschenbach, F. Schoenknecht, and K. K. Holmes Anaerobic bacteria in nonspecfic vaginitis. N. Engl. J. Med. 303: Spiegel, C. A., R. Amsel, and K. K. Holmes Diagnosis of bacterial vaginosis by direct Gram stain of vaginal fluid. J. Clin. Microbiol. 18: Spiegel, C. A Bacterial vaginosis. Clin. Microbiol. Rev. 4: Stanek, R., R. E. Gain, D. D. Glover, and B. Larsen High performance ion exclusion chromatographic characterization of the vaginal organic acids in women with bacterial vaginosis. Biomed. Chromatogr. 6: Taylor-Robinson, D., D. J. Morgan, M. Sheehan, I. J. Rosenstein, and R. F. Lamont Relation between Gram-stain and clinical criteria for diagnosing bacterial vaginosis with special reference to Gram grade II evaluation. Int. J. STD AIDS 14: Thomason, J. L., S. M. Gelbart, R. J. Anderson, A. K. Walt, P. J. Osypowski, and F. F. Broekhuizen Statistical evaluation of diagnostic criteria for bacterial vaginosis. Am. J. Obstet. Gynecol. 162: Thomason, J. L., S. M. Gelbart, J. A. James, J. M. Edwards, and P. R. Hamilton Is analysis of vaginal secretions for volatile organic acids to bacterial vaginosis of any diagnostic value? Am. J. Obstet. Gynecol. 159: Wold, S., K. Esbensen, and P. Geladi Principal component analysis. Chem. Int. Lab. Syst. 2:35 37.