Leukorrhea and Bacterial Vaginosis as In-Ofce Predictors of Cervical Infection in High-Risk Women

Michele M. Hakakha, MD, Jane Davis, MD, Lisa M. Korst, MD, PhD, and Neil S. Silverman, MD
OBJECTIVE: To evaluate 1) whether microscopic detection of leukorrhea or bacterial vaginosis identies patients at high risk for cervical infection with Chlamydia trachomatis or Neisseria gonorrhoeae, and 2) if pregnancy alters the predictive value of these ndings. METHODS: Wet-mount screening examination of vaginal discharge was performed on all new patients seen at two resident-staffed clinics serving primarily indigent women. Leukorrhea was dened as >10 white blood cells per high-power eld on microscopic examination; Amsel criteria were used to determine the presence of bacterial vaginosis, with a positive clue cell test result dened as >20% of epithelial cells. The diagnoses of C trachomatis and N gonorrhoeae infection were established by deoxyribonucleic acid amplication tests. RESULTS: The study population consisted of 194 women, 118 (61%) of whom were pregnant. Overall, 11% of women had positive cultures for chlamydia or gonorrhea. Although both leukorrhea and clue cells were independently associated with positive cervical cultures, multivariate analysis found that clue cells did not contribute to the predictive value of leukorrhea alone among both pregnant (relative risk [RR] 15.7) and nonpregnant (RR 58.7) women. Negative predictive values for the screening test were comparably high (98 100%), independent of pregnancy status. CONCLUSION: Leukorrhea, in the presence or absence of bacterial vaginosis, was strongly associated with cervical infections with C trachomatis or N gonorrhoeae among both pregnant and nonpregnant patients. In settings where patient follow-up is uncertain, on-site screening tests identify women for whom empiric antibiotic therapy for sexually transmitted diseases may be appropriate. (Obstet Gynecol 2002;100:808 12. 2002 by The American College of Obstetricians and Gynecologists.)

Infections with Chlamydia trachomatis and Neisseria gonorrhoeae remain major causes of morbidity among women, especially those of childbearing age. Approximately 12 million new cases of sexually transmitted diseases
From the Department of Obstetrics and Gynecology, Burns and Allen Research Institute, Cedars-Sinai Medical Center, Los Angeles, California.

(STDs) occur annually in the United States. In 1997, there were over 200 new cases of chlamydia infection per 100,000 people, a rate exceeding that for all other reportable infectious diseases in the United States.1 In addition, 87% of the top ten communicable diseases reported to the Centers for Disease Control and Prevention (CDC) in 1995 were sexually transmitted diseases.1 This underrecognized epidemic has enormous health and economic consequences for women of all ages. Women of childbearing age are at particular risk for STD infection and subsequent complications. Many infections are asymptomatic and under-diagnosed, and their resulting adverse health effects, including pelvic inammatory disease (PID), tubo-ovarian abscess, infertility, chronic pelvic pain, and ectopic pregnancy can occur years after the initial infection. Among American women who suffer from infertility, for example, at least 15% can be attributed to complications of STDs, with up to 50% of these women unaware of their past exposure.2 In addition to these reproductive health issues, infection with chlamydia or gonorrhea during pregnancy has been associated with a variety of adverse outcomes, including premature rupture of membranes, chorioamnionitis, and postpartum infection.3 In the past, STD treatment goals have been particularly difcult to attain in settings serving more transient, lower-socioeconomic-group patient populations, who may also have a higher intrinsic risk of infection.4 Poor compliance and inconsistent follow-up after the initial clinic visit limits adequate assessment of treatment efcacy. Therefore, a rapid, sensitive, on-site screening test for cervical STDs would be a valuable adjunct to complete obstetric/gynecologic care for these women, allowing a focused point-of-service approach to management of this issue. Working with a high-risk, transient population of primarily adolescent/young-adult inner-city women, we sought to assess the utility of easily recognizable microscopic components of vaginal secretions as a screening tool for the detection of cervical STDs. In addition, because some authors and clinicians have described a

808

VOL. 100, NO. 4, OCTOBER 2002 2002 by The American College of Obstetricians and Gynecologists. Published by Elsevier Science Inc.

0029-7844/02/$22.00 PII S0029-7844(02)02147-6

relative leukorrhea of pregnancy5,6 we examined wet mounts from both pregnant and nonpregnant women in this setting to determine the following: 1) whether microscopic detection of leukorrhea or bacterial vaginosis identies a group of patients at high-risk for cervical infection with C trachomatis or N gonorrhoeae, and 2) if pregnancy alters the predictive value of those ndings compared to those for nonpregnant patients.

MATERIALS AND METHODS The study population consisted of consecutive new obstetric and gynecologic patients seen in two residentstaffed, inner-city clinics in Los Angeles between August 2000 and April 2001. All patients invited to participate consented to inclusion in this study. Two clinic sites were utilized: 1) the Los Angeles Free Obstetric Clinic in Hollywood, and 2) the Cedars-Sinai Medical Center outpatient Obstetrics and Gynecology Clinic in West Los Angeles. The study protocol was approved by the Medical Centers institutional review board for human subjects and did not require written informed consent. All new patients undergoing routine pelvic examinations and baseline cervical cultures also had wet-mount examination of vaginal discharge performed by the obstetrics and gynecology resident staff. Deoxyribonucleic acid (DNA) amplication tests were used to diagnose cervical infection with C trachomatis and N gonorrhoeae (LCx STD Swab System, Abbott Laboratories, Abbott Park, IL). A cotton swab was used to obtain vaginal secretions or discharge with placement into both normal saline and potassium hydroxide solutions for microscopic evaluation. For saline evaluation, the swab was placed into a prepared test tube containing 0.2 mL of normal saline solution. One drop of the suspension was applied to a glass slide and examined at 400 via light microscopy. Potassium hydroxide evaluation was performed by mixing one drop of potassium hydroxide on a glass slide with secretions from a separate vaginal swab and examining microscopically. Leukorrhea was dened as 10 white blood cells per high-power eld on microscopy.7 Amsel criteria, including three of the following four ndings, were used to determine the presence of bacterial vaginosis on saline wet mount: a pH of 4.5, a positive whiff test, 20% clue cells, and a homogeneous discharge.8 Clinical ndings regarding discharge type, presence or absence of mucopurulent cervicitis, pelvic pain, and cervical friability were recorded by residents on a standardized data collection form. Information regarding discharge pH, whiff test results, and wet-mount ndings was recorded on a standardized data collection form. Results of cervical cultures for C trachomatis and

N gonorrhoeae were later abstracted from the patients charts. Categorical variables were examined using 2 analysis with Yates correction, and continuous variables were analyzed with nonparametric testing as appropriate. Means are expressed plus or minus the standard deviation, and statistical signicance is dened at the P .05 level. Odds ratios (OR) and relative risks (RR) are expressed as appropriate. After performing univariate analyses, multivariate logistic regression analysis was used to construct an equation to adjust for potential confounding and effect modication. The data sets were created and analyses performed using Epi-Info 6 (CDC, Atlanta, GA). Sample size calculation was based on an assumption that leukorrhea would be more difcult to diagnose among pregnant women and that the sample would be comparably divided by pregnancy status. Therefore, more conservative estimates were used for the calculation assumptions, and the derived sample size doubled to account for both pregnant and nonpregnant women. We assumed that 10% of women screened would have a positive screening test (ofce microscopy test for leukorrhea) and that the incidence of disease (a positive cervical culture) would be 5% in the screen-negative population and 50% in the screen-positive population. For a .05 two-sided signicance level and a power level of 80%, 76 patients were determined to be necessary to detect that difference in group proportions, which led to a total desired study population of 152 women. RESULTS Of the 194 women evaluated for the study, 118 (60.8%) were pregnant. An overview of patient characteristics identied for the entire study population and stratied by pregnancy status is shown in Table 1. The overall rate of cervical infection, the primary outcome measure for this study, was 11% (10/194); only one woman was infected with gonorrhea alone, and four women (2%) were co-infected with both chlamydia and gonorrhea. Table 1 demonstrates that, except for patient age, there were no important differences in the patient characteristics studied between pregnant and nonpregnant women. The results of the univariate analysis are seen in Table 2, with a positive culture result for either C trachomatis or N gonorrhoeae dened as the outcome variable of interest. For the purposes of this analysis, vaginal pH was redened as a dichotomous variable using 4.5 as a cutoff and alkaline pH dened as a vaginal pH 4.5. Aside from age and pregnancy, all other patient characteristics identied were seen to be associated with an increased risk of a positive culture result. Patients with leukorrhea, in

VOL. 100, NO. 4, OCTOBER 2002

Hakakha et al

Leukorrhea Predicting STDs

809

Table 1. Characteristics of the Study Population Characteristic Age (median, range) (y) Abnormal discharge pH (mean SD) Friable cervix Whiff test () Clue cells () Leukorrhea Both cultures () Either culture () All patients (N 194) 25 (1664) 51 (26) 4.6 0.3 6 (3.1) 20 (10.3) 22 (11.3) 27 (13.9) 4 (2.1) 21 (10.8) Pregnant (n 118) 25 (1642) 33 (28) 4.6 0.3 3 (2.5) 10 (8.5) 10 (8.5) 15 (12.7) 2 (1.7) 9 (7.6) Nonpregnant (n 76) 26.5 (1764) 18 (24) 4.6 0.3 3 (3.9) 10 (13.2) 12 (15.8) 12 (15.8) 2 (2.6) 12 (15.8) P .002* .62 .48 .84 .42 .12 .54 .83 .07

SD standard deviation; positive. Values are numbers (percentages) unless otherwise stated. * Wilcoxon rank-sum test. Student t tests.

particular, were shown in this univariate analysis to be 124 times more likely to have a positive cervical culture than those without that characteristic; the presence of clue cells conferred a six-fold increase in risk of cervical infection. Among the 22 patients with a positive test for clue cells, 19 (86.4%) also met Amsel criteria for the diagnosis of bacterial vaginosis. The three patients with clue cells who did not meet Amsel criteria were positive for leukorrhea. With the exception of one case, bacterial vaginosis as dened either by clue cells alone or by Amsel criteria was noted only in the presence of leukorrhea (Table 3). Amsel criteria would not have been more helpful than clue cells alone in dening the overriding contribution of leukorrhea for the prediction of a positive cervical culture. Therefore, the presence of clue cells in the absence of leukorrhea could not be considered to be an independent risk factor for cervical infection. A multivariate logistic regression analysis using maternal age (continuous), pregnancy status (yes/no), and the presence of leukorrhea (yes/no) was constructed. This analysis demonstrated no substantial association of maternal age with the likelihood of having a positive cervi-

cal culture result. However, nonpregnant status was found to be independently associated with having a positive culture. For this reason, the analysis was simplied to test the predictive value of leukorrhea for a positive culture, with the patient population stratied by pregnancy status. This revised model showed that among nonpregnant women, those who had leukorrhea were 58.7 times more likely to have a positive culture result than women without leukorrhea (11/12 [91.7%] versus 1/64 [1.6%], P .001; RR 58.7 [95% condence interval (CI) 8.3, 413.2]). Among pregnant women, leukorrhea was still associated with an increased risk of a positive cervical culture, although the relative risk could not be calculated (9/15 [60%] versus 0/103 [0%]; P .001). Because in our study population the majority of the positive cervical cultures were for chlamydia alone, the data were also examined using a positive chlamydia culture as the outcome variable. Leukorrhea remained a strong predictor of chlamydial infection in this model among both nonpregnant women (RR 58.7; 95% CI 8.3, 413.2) and pregnant women (RR 15.7; 95% CI 7.7, 32.2). Comparing pregnant and nonpregnant women, both

Table 2. Association of Clinical Factors in the Study Population and the Risk of a Positive Cervical Culture Positive culture (%) Clinical factor Pregnancy Abnormal discharge Alkaline pH Pelvic pain Friable cervix Whiff test () Clue cells () Leukorrhea Patients with factor 9/118 (7.6) 15/51 (29.4) 13/53 (24.5) 11/16 (68.8) 4/6 (66.7) 7/20 (35.0) 9/22 (40.9) 20/27 (74.1) Patients without factor 12/76 (15.8) 6/143 (4.2) 8/141 (5.7) 10/178 (5.6) 17/188 (9.0) 14/174 (8.0) 12/172 (7.0) 1/167 (0.6) RR for positive culture* (95% CI) 0.5 (0.2, 1.1) 7.0 (2.9, 17.1) 4.3 (1.9, 9.8) 12.2 (6.2, 24.3) 7.4 (3.6, 15.2) 4.4 (2.0, 9.5) 5.9 (2.8, 12.3) 123.7 (17.3, 884.2)

RR relative risk; CI condence interval; other abbreviation as in Table 1. Values are numbers (percentages) unless otherwise stated. * Comparisons for all clinical factors except pregnancy signicant at P .001 level.

810

Hakakha et al

Leukorrhea Predicting STDs

OBSTETRICS & GYNECOLOGY

Table 3. Comparison of Leukorrhea and Clue Cells as Risk Factors for a Positive Cervical Culture Microscopy result Leukorrhea () only (n 19) Clue cells () only (n 14) Both tests () (n 8) Both tests () (n 153)
negative; other abbreviation as in Table 1. Values are numbers (percentages).

Either culture () 12 (63) 1 (7) 8 (100) 0 (0)

the sensitivity (100% versus 92%, respectively) and specicity (95% versus 98%) of leukorrhea for predicting cervical infection were comparable. The positive predictive value for leukorrhea, however, was higher among nonpregnant patients (92% versus 60%), whereas the negative predictive value of leukorrhea for both groups was similarly high (100% for pregnant and 98% for nonpregnant patients). DISCUSSION We have demonstrated that the presence of leukorrhea, in the presence or absence of bacterial vaginosis on microscopic examination in an outpatient setting, is strongly associated with concomitant cervical infection with C trachomatis or N gonorrhoeae. Multivariate analysis revealed that the presence of leukorrhea in particular was a very strong predictor of cervical infection, particularly with chlamydia, in both nonpregnant (RR 58.7) and pregnant (RR 15.7) patients, after considering age, pregnancy, and the microscopic presence of clue cells. Bacterial vaginosis, dened for this outpatientsetting study as either the presence of clue cells alone or as the presence of all of Amsel diagnostic criteria, was shown in our univariate analyses to carry a six-fold increased risk of a positive cervical culture but had no independent contribution to risk above that found for leukorrhea. Current screening tests used to detect cervical infection with STDs include microbiologic and cell cultures, direct immunouorescent antibody assays, enzyme immunoassays (EIAs), and DNA probes. More recently, nucleic acid amplication tests have become increasingly employed, because they have been shown to have sensitivity and specicity rates of 99% and 90%, respectively.9,10 The barrier encountered with most of these diagnostic tests, however, is that they require at least 24 48 hours to yield results. Although the EIA for chlamydia requires only 30 minutes to complete, most studies indicate that the sensitivity of the EIA is only 60% when used in populations with an infection rate of at least 6%.11 In addition to their poor sensitivity, EIA tests are not suitable for processing large numbers of patient samples

because they are labor intensive and cannot be automated. In an attempt to avoid these commonly encountered problems associated with current screening methods, a number of investigators have explored the predictive value of a variety of vaginal markers of cervical infection. Moscicki et al12 found that 10 white blood cells per high-power eld on a Gram stain of cervical secretions was 98% accurate in identifying the absence of C trachomatis when compared with culture results. In another study, Bohmer et al13 studied a population of pregnant women and found that the absence of leukorrhea on wet mount could accurately predict the absence of cervical chlamydia and gonorrhea infection. Others have postulated that bacterial vaginosis may be a marker for cervical STD infection, with the thought that the increased prevalence of anaerobic bacteria in the vagina may predispose certain women to develop upper genital tract infections with C trachomatis and N gonorrhoeae.14,15 Our study evolved as an extension of some of these earlier studies, hoping to conrm the strong predictive values of negative test results. In addition, and even more importantly for women at highest risk for STD infection, we attempted to evaluate the positive predictive value of rapid, on-site screening tests for cervical STD infection. Our results clearly demonstrate a powerful association between microscopic leukorrhea in particular and infection with either C trachomatis or N gonorrhoeae in a cohort of high-risk, inner-city women. The predictive value of these readily available, in-ofce tests were slightly affected by pregnancy status, with a lower positive predictive value for leukorrhea in pregnant compared to nonpregnant women (60% versus 92%). The high negative predictive value of this screening test was unaffected by pregnancy status, however (100% versus 98%). In addition, the high relative risk of cervical infection demonstrated in the presence of leukorrhea independent of pregnancy status supports the utility of ofce microscopy as a screening test for a high-risk population with unpredictable compliance with follow-up visits. In such populations, like ours, a diagnose-and-treat protocol employing wet mount examinations could attempt to identify up to 75% of women with cervical STDs at the time of initial visit. In addition, in health care settings with limited resources, our results suggest that a negative ofce test for both leukorrhea and clue cells could predict the absence of STD infection with 9299% accuracy, suggesting the option of not utilizing more expensive diagnostic tests if cost and payment are significant issues. Further cost benet studies may help to better dene the value of such targeted testing protocols. The majority of the cervical infections identied in this study population were due to chlamydia rather than

VOL. 100, NO. 4, OCTOBER 2002

Hakakha et al

Leukorrhea Predicting STDs

811

gonorrhea, suggesting that empiric treatment based on screening tests might be best targeted against that single agent. A larger study might have permitted accrual of a larger number of women with gonococcal infection as well. Still, the sample size of our study population surpassed our initial power calculation assumptions, with 99% power to detect the differences we had proposed based on stratication by pregnancy and the presence or absence of leukorrhea. Moreover, the fact that gonococcal infection was still identied in this population, even at low levels, would seem to warrant single-dose treatment targeted against both agents in this high-risk cohort if leukorrhea is identied on screening. Testing of outcomes stratied by individual infectious agents might be warranted in evaluations of the screening tests in more compliant patient populations; in our less-compliant patient population, however, we chose to use a test result positive for either organism as the primary outcome variable. In summary, we have demonstrated that in-ofce wet mounts may serve as sensitive, rapid screening tests that could help eliminate the problems associated with noncompliance and poor follow-up in high-risk, transient populations. Therefore, in settings where follow-up may be uncertain, these inexpensive on-site screening tests, particularly for the presence of leukorrhea, appear to identify women for whom empiric antibiotic therapy may be appropriate.
REFERENCES 1. Division of STD Prevention. Sexually transmitted disease surveillance, 1998. Atlanta, GA: Centers for Disease Control and Prevention, 1999. 2. Eng TR, Butler WT. The hidden epidemic: Confronting sexually transmitted diseases. Washington, DC: National Academy Press, 1997. 3. Romero R, Mazor M. Infection and preterm labor. Clin Obstet Gynecol 1988;31:553 84. 4. Katz BP, Zwicki BW, Caine VA, Jones RB. Compliance with antibiotic therapy for Chlamydia trachomatis and Neisseria gonorrhoeae. Sex Transm Dis 1992;19:351 4. 5. McGregor JA, French JI. Bacterial vaginosis in pregnancy. Obstet Gynecol Surv 2000:55:119.

6. Cunningham FG, Gant NF, Leveno KJ, Gilstrap LC, Hauth JC, Wenstrom KD, eds. Williams obstetrics. New York: McGraw Hill, 2001:221 47. 7. Eltabbakh GH, Eltabbakh GD, Broekhuizen FF, Griner BT. Value of wet mount and cervical cultures at the time of cervical cytology in asymptomatic women. Obstet Gynecol 1995;85:499 503. 8. Amsel R, Totten PA, Speigel CA, Chen KC, Eschenbach D, Holmes KK. Non-specic vaginitis: Diagnostic techniques and microbial and epidemiologic associations. Am J Med 1983;74:14 22. 9. Rome ES. Sexually transmitted diseases: Testing and treating. Adolesc Med 1999;10:231 41. 10. Bass CA, Jungkind DL, Silverman NS, Bondi JM. Clinical evaluation of a new polymerase chain reaction assay for detection of Chlamydia trachomatis in endocervical specimens. J Clin Microbiol 1993;31:2648 53. 11. Centers for Disease Control and Prevention. Chlamydia screening: A new HEDIS measure important to your members [fact sheet]. Atlanta, GA: Centers for Disease Control and Prevention, 2000. 12. Moscicki B, Shafer M-A, Millstein SG, Irwin CE Jr, Schachter J. The use and limitations of endocervical Gram stains and mucopurulent cervicitis as predictors for Chlamydia trachomatis in female adolescents. Am J Obstet Gynecol 1987;157:6571. 13. Bohmer JT, Schemmer G, Harrison F, Kreft W, Elliot M. Cervical wet mount as a negative predictor of gonococciand Chlamydia trachomatis-induced cervicitis in a gravid population. Am J Obstet Gynecol 1999;181:2837. 14. Soper DE, Brockwell NJ, Dalton HP, Johnson D. Observations concerning the microbial etiology of acute salpingitis. Am J Obstet Gynecol 1994;170:1008 14. 15. Eschenbach DA, Hillier S, Critchlow C, Stevens C, DeRousen T, Holmes KK. Diagnosis and clinical manifestations of bacterial vaginosis. Am J Obstet Gynecol 1988; 158:819 28. Address reprint requests to: Neil S. Silverman, MD, CedarsSinai Medical Center, Department of Obstetrics and Gynecology, 8635 W. 3rd Street, Suite 160 West, Los Angeles, CA 90048; E-mail: silvermann@cshs.org. Received December 12, 2001. Received in revised form April 16, 2002. Accepted May 13, 2002.

812

Hakakha et al

Leukorrhea Predicting STDs

OBSTETRICS & GYNECOLOGY