Gynecology

Correlation between cervical cytologic results and Gram stain as diagnostic tests for bacterial vaginosis
John D. Davis, M D , a Erin E. Connor, MD," Penny Clark, PhD, a Edward J. Wilkinson, MD, b and Patrick Duft, MD"
Gainesville, Florida
OBJECTIVE: Our purpose was to determine the reliability of the Papanicolaou smear in making the diagnosis of bacterial vaginosis with the vaginal Gram stain used as the diagnostic standard. STUDY DE81GN; We conducted a prospective, blinded, cross-sectional study of 210 consecutive patients referred to the Colposcopy Clinic for evaluation of abnormal cervical cytologic results. Each patient had a standard Papanicolaou smear and Gram stain of vaginal discharge. The sensitivity, specificity, positive predictive value, and negative predictive value of the Papanicolaou smear were determined with the Gram stain used as the standard for diagnosis of bacterial vaginosis. RESULTS: Of the 210 patients, 80 (38.1%) had Gram stains that were positive for bacterial vaginosis and 118 (56,2%) had negative Gram stains. Twelve (5.7%) had intermediate Gram stains that were also considered negative. Of the 80 patients with positive Gram stains, 44 had cervical smears consistent with bacterial vaginosis and 36 had negative smears. Of the 1.30 patients with negative Gram stains, two had a positive cervical smear. Therefore, compared to the Gram stain, cervical cytologic test results had a sensitivity of 55% and a specificity of 98%. The respective positive predictive and negative predictive values were 96% and 78%. CONCLU81ON" Compared to Gram stain of vaginal secretions, the cervical Papanicolaou smear has fair sensitivity (55%) and excellent positive predictive value (96%) in diagnosing bacterial vaginosis. (Am J Obstet Gynecol 1997;177:532-5.)

Key words: Bacterial vaginosis, Papanicolaou smear, Gram stain

Bacterial vaginosis is an infection of the lower genital tract characterized by a shift from normal, p r e d o m i n a n t l y aerobic vaginal flora to p r e d o m i n a n t l y anaerobic flora. Bacterial vaginosis is associated with a n u m b e r of obstetric and gynecologic complications, including p r e t e r m labor, p r e t e r m p r e m a t u r e r u p t u r e of membranes, chorioamnionitis, p u e r p e r a l e n d o m e tritis, pelvic inflammatory disease, urinaI T tract infection, postoperative cuff cellulitis, and cervical dysplas i a ) ' 2 A p p r o x i m a t e l y 50% of patients with bacterial vaginosis do n o t have any symptoms3; the chief complaint of patients who do have symptoms is a malodorous vaginal discharge. The diagnosis is usually m a d e
From the Department of Obstetrics and Gynecology ~ and the Department of Pathology,~ University of Florida College of Medicine. Received for publication Janua~y 2, 1997," revised Februmy 27, 1997; accepted April 28, 1997. Reprint requests:John D. Davis, MD, Director, Division of Gynecology, Department of Obstetricsand Gynecology, UniversityofFlorida Collegeof Medicine, PO Box 100294, Room M-302E, Gainesville, FL 326100294. Copyrigt~t 1997 by Mosby-Year Book, Inc. 0002-9378/97 $5.00 + 0 6/1/82913
532

according to the criteria set forth by Amsel et al. 3 Patients taust have at least three of the following four signs: (1) thin, h o m o g e n e o u s vaginal discharge, (2) vaginal p H >4.5, (3) clue cells on wet smear, and (4) presence of an amine o d o r with the addition of potassium hydroxide. Other methods have been used to establish the diagnosis of bacterial vaginosis. The use of vaginal cultures for Gardnerella vaginalis is limited by the test's poor specificity. The vaginal Gram stain has been evaluated extensively as a diagnostic test for bacterial vaginosis. 4-6 The clinical criteria described by Amsel et al. 3 correlate well with the Gram stain results interpreted according to the criteria of Spiegel et al. 4 A high level of reproducibility and reliability in identifying bacterial vaginosis has resulted when the criteria of Spiegel et al. 4 were used in conjuncdon with the standardized scoring system proposed by Nugent et al. 5 Recently, Schwebke et al. 6 confirmed the value ofvaginal Gram stain with the use of the criteria of Nugent et al. 5 for the diagnosis of bacterial vaginosis. The Papanicolaou smear also has been evaluated as a

Volume 177, Number3 AmJ Obstet Gynecol

Davis et dl,

533

diagnostic test for bacterial vaginosis. At least one study has shown that the Papanicolaou smear is an excellent diagnostic test for bacterial vaginosis, whereas other reports have indicated that the test is less specific than Gram stain or wet smear diagnosis. 7-m The objective of this smdy was to determine the reliability of the Papanicolaou smear in making the diagnosis of bacterial vaginosis, with the Gram stain used as the standard.

Material and methods

This study was conducted at the University of Florida between October 1994 and April 1995. The study design was a prospective, blinded, cross-sectionalt study of 210 conseeutive patients referred to the Colposcopy Clinic for evaluation of abnormal cervical cytologic results. There were no exclusion criteria. Both n o n p r e g n a n t and pregnant women were included in the smdy. The study was conducted u n d e r a protocol approved by the institutional review board. Patients were evaluated by obstetrics and gynecology residents, family praetice residents, medical students, and nurse practitioners working n n d e r the direct supervision of an attending physician. A Papanieolaou smear of the cervix and a sample of vaginal diseharge for Gram stain were obtained before the colposcopic examination. Papanicolaou smears were obtained in the routine m a n n e r ; endocervical sampling was performed with either an endocervical brush or a moistened cotton-tip applicator in n o n p r e g n a n t and pregn a n t patients, respectively. The sample of vaginal discharge was obtained from the posterior fornix and vaginal sidewalls with a cotton-tip applicator. Patients then u n d e r w e n t colposcopic examination, with biopsies performed if indicated. Determinations of vaginal pH, the amine test, and saline solution preparations were n o t routinely used unless the patient had overt symptoms of a vaginal infection. The Papanicolaou smears were evaluated by our cytopathology laboratory in the routine m a n n e r according to the Bethesda System recommendations,s A physician (EJ.W.) who is board cerfified in both obstetrics and gynecology and pathology serves as the supervisor of the laboratory. The slides were not identified as part of a smdy so that the routine evaluation methods of the cytopathology laboratory could be assessed. The certified cytotechnologists and cytopathologists were blinded to the results of the Gram stains. If there was clear evidence of a filmy background of small coccobacilli, individual squamous cells with a layer of eoccobacilli along the margins of the cell membranes, and conspicuous absence of lactobacilli, the smear was reported as positive for bacterial vaginosis. All three findings had to be present for the diagnosis of bacterial vaginosis to be made. The vaginal swab specimens were smeared directly on

clean glass slides immediately after collection. The slides were air-dried and processed in the Obstetrics and Gynecology Researeh Microbiology Laboratory. The specimens were stained in batches according to the standard procedure for Gram staining and microscopieally reviewed by one examiner (P.C.). Briefly, each slide was examined u n d e r an oil immersion objective (1000 magnification) for the following morphotypes: large gram-positive rods (LactobaciIIusmorphotypes), small gram-variable and gram-negative rods (GardnereUa and Bacteroides morphotypes), and cured gram-variable rods (Mobiluncusmorphotypes). Each morphotype was quantitated as 0 to 4+ and weighted to yield a score of 0 to 10, according to the criteria of Nugent et dl. A vaginal smear with a store of > 7 was defined as bacterial vaginosis, a score of 4 to 6 was classified as intermediate bacterial vaginosis, and a score of 0 to 3 was considered normal. For the purposes of this smdy patients with intermediate scores were considered to be negative. In studies of obstetric patients, women u4tb intermediate bacterial vaginosis scores have had an increased incidence of adverse pregnancy outcome, Howerer, the significance of an intermediate score in gynecologic patients has not been fully delineated; therefore for the purposes of this smdy we chose to consider these patients to be "negative." The person who interpreted the Gram stains was blinded to the resuhs of the Papanicolaou smears. The resuhs of the Papanicolaou smears and Gram stains were compared and analyzed. The sensitivity, specificity, positive predictive value, and negative predictive value of the Papanicolaou smear were determined with the Gram stain used as the diagnostic standard.

Results
Of the 210 patients, 80 (38.1%) had Gram stain results that were positive for bacterial vaginosis. 118 (56.2%~ had negative Gram stares, and 12 (5.7%1 had intermediate Gram stains. Of the 80 patients with positive Gram stains. 44 had cervical smears that were consistent with bacterial vaginosis and 36 had negative smears t senmtivity 55%). Of the 130 patients with negative and intermediate Gram stains, two had positive cervical cytologic findings (specificity 98%). Of the 45 pafients ~~th a positive Papanicolaou smear, 44 had positive Gram stains. Of the 165 patients with negative cervical smears. 129 had negative or intermediate Gram stains. The respective positive predictive and negauve predictive values were 96% and 78%.

Comment
Evaluation of a wet smear according to the clinical criteria of Amsel et dl. 3 is the most c o m m o n m e t h o d used today to diagnose bacterial vaginosis. The possibility of using the Papanicolaou smear as a screening

534

Davis et al.

September 1997 Am J Obstet Gynecol

method for bacterial vaginosis is attractive because the Papanicolaou smear is routinely performed and no addional specimens would need to be obtained or additional cost incurred. The results of our study show that the Papanicolaou smear has only a 55% sensitivity for the diagnosis of bacterial vaginosis. Therefore it is not an adequate screening test for bacterial vaginosis. However, with a positive predicve value of 96%, Papanicolaou smear findings consistent with bacterial vaginosis certainly are reliable enough that, in a patient population with a prevalence of bacterial vaginosis similar to that in ours, a decision about treatment could be made without the need for another diagnostic test. Others have studied the ability of the Papanicolaou smear to detect bacterial vaginosis. Platz-Christensen et al. 7 reported that the Papanicolaou smear showed a respective sensitivity and specificity of 90% and 97% in detecting bacterial vaginosis when compared to clinical criteria. In o u t study the specificity of the Papanicolaou smear in detecng bacterial vaginosis was equivalent to that found in their study; however, we could not reproduce their sensitivity. There a r e a n u m b e r of possible explanations for the decreased sensitivity in our study. First, our cytotechnologists may not have had the same expertise in diagnosing bacterial vaginosis by Papanicolaou smear as the single person reading the Papanicolaou smears in their study. We did not evaluate interreader variability in our study, so we do not know if one or more of our cytotechnologists had difficulty diagnosing bacterial vaginosis on Papanicolaou smear. However, the cytotechnologists were trained and supervised by one of the authors (EJ.W.) who is board certified in both obstetrics and gynecology and pathology, and laboratory protocol dictates that every Papanicolaou smear be examined for evidence of genital tract infection, including bacterial vaginosis. The cytologic findings listed in the preceding text are straighfforward and easy to recognize after an appropriate period of training. In addition, we wanted to evaluate the Papanicolaou smear as a detection method under routine cytologic processing conditions. A second possible explanation for the low sensitivity in our study c o m p a r e d with the study by Platz-Christensen et al. 7 is that they used the presence of clue cells alone on Papanicolaou smear to diagnose bacterial vaginosis whereas we used the more rigorous diagnostic criteria of the Bethesda System. 8 We believe the Bethesda requirements for diagnosing bacterial vaginosis on Papanicolaou smear (i.e., clear evidence of a filmy background of small coccobacilli, individual squamous cells with a layer of coccobacilli along the margins of the cell membranes, and conspicuous absence of lactobacilli) are preferable to the finding of clue cells alone because, even by cJincal criteria, clue cells alone are not sufficient to diagnose bacterial vaginosis. A third possible explanation for our lower sensivity may

be that we obtained our specimens for Papanicolaou smears from the cervix, not the vagina. The authors of the article mentioned do not specify whether they obtained their specimens from the cervix or the vagina. If they did indeed sample the vagina, this would be a very reasonable explanation for the discrepancy in our sensitivities, because bacterial vaginosis is a vaginal infection. We likely could have improved the sensitivity in our study by obtaining a specimen of vaginal secretions; however, again, we wanted to evaluate the Papanicolaou smear as a detection method under routine cytologic processing conditions. A final difference between our study and the one by Platz-Christensen et al. 7 warrants consideration. They used the clinical criteria of Amsel et aLs to diagnose bacterial vaginosis, whereas we elected to use the Gram stain as our diagnostic standard. In their smdy only two examiners were responsible for clinically diagnosing bacterial vaginosis. In our study we had multiple examiners with varying levels of expertise in the clinical diagnosis of bacterial vaginosis. We thought that using the Gram stain as the diagnostic standard would be the best way to eliminate the interpractitioner variability in the clinical diagnosis, because our microbiologist is weil versed in this method and becanse other studies have confirmed the reliability of the Gram stain in diagnosing bacterial vaginosis.4-6 In c o m p a r i n g different m e t h o d s of diagnosing bacterial vaginosis Hillier 9 states that the "diagnosis of bacterial vaginosis by examination of Papanicolaou smears is less specific than the Gram stain of vaginal fluid because standardized criteria for the evaluation of Pap smears have not been routinely applied." The Bethesda guidelines indicate that Papanicolaou smear changes consistent with bacterial vaginosis are n e i t h e r specific n o r sufficient for the diagnosis. 8 According to the July 1996 Technical Bulletin of The American College of Obstetricians and Gynecologists, "Vaginitis, ''1 bacterial vaginosis cannot be accurately diagnosed by Papanicolaou smear. In o u t study we f o u n d that, when using the standardized diagnostic criteria of the Bethesda System, the positive predictive value of the Papanicolaou smear was 96% in diagnosing bacterial vaginosis. Therefore, when the Papanicolaou smear is positive, an additional confirmatory test is unnecessary. However, because the sensitivity of the Papanicolaou smear is only 55%, this test is an inadequate screen for bacterial vaginosis.

REFERENCES

1. ThomasonJL, Gelbart SM, Scaglione NJ. Bacterial vaginosis: current review with indications for asymptomatic therapy. Am J Obstet Gynecol 1991;165:1210-7. 2. Clark P, Kurtzer T, Duft P. Role of bacterial vaginosis in peripartum infections. Infect Dis Obstet Gynecol 1994;2: 179-83. 3. Amsel R, Totten PA, Spiegel CA, Chen KC, Eschenbach D, Holmes KK. Nonspecific vaginitis: diagnostic criteria and

Volume 177, Number 3 Am J Obstet Gynecol

Davis et al,

535

microbial and epidemiologic associations..~n J Med 1983; 74:14-22. 4. Spiegel CA, Amsel R, Holmes KI~ Diagnosis of bacteriat vaginosis by direct Gram stain of vaginal fluid. J Clin Microbiol 1983;18:170-7. 5. Nugent ~~, 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. 6. 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;88:573-6.

7. Platz-Christensen J-J, Larsson P-G, Sundstrom E, Bondeson L. Detection of bacterial vaginosis in Papanicolaou smears. Am J Obstet Gynecol 1989;160:132-3. 8. Current Issues. The 199t Bethesda System for reporting cervical/vaginal cytologica diagnosis. Diagnostic Cytopathol 1993;9:23-43. 9. Hillier SL. Diagnostic microbiology of bacteriaI vaginosis. kan J Obstet Gynecol 1993;169:455-9. 10. American College of Obstetricians and Gynecologists Committee on Technical Bulletins. Vaginitis. Washington: The College; 1996 Jul. Technical Bulletin No.: 226.