The role of bacterial vaginosis in infection after major gynecologic surgery.

PURPOSE: Previous studies have reported an association between bacterial vaginosis (BV) and postoperative fever and infection. This prospective study investigated whether the intermediate or definite stages of BV are risk factors for postoperative infection after major gynecologic surgery. METHODS: Vaginal cultures were obtained preoperatively from 175 women undergoing gynecologic surgery. The diagnostic criteria for BV were based on Nugent's standardized method of Gram stain interpretation. Postoperative fever was defined as at least one temperature equal to 101.0 degrees F or greater, or two or more temperatures more than 6 hours apart equal to 100.4 degrees F or greater. RESULTS: Thirty-six percent of the positive-BV group developed a postoperative fever, compared with 20% of the Lactobacillus-predominant group and 12% of the intermediate-BV group (P = 0.017). The differences between the positive-BV group and the Lactobacillus-predominant group, and between the positive-BV group and the intermediate-BV group, with respect to postoperative fever, were statistically significant (P = 0.045 and P = 0.007, respectively). The difference between the intermediate-BV group and the Lactobacillus-predominant group was not statistically significant (P = 0.28). CONCLUSIONS: Although the association between BV and postoperative febrile morbidity could be a spurious result of confounding with other variables, it may be prudent for the surgeon to identify patients with BV and treat them preoperatively.

iostoperative fever and infection continue to be significant complications of major gynecologic surgery. The reported incidence of postoperative infection in women without antibiotic prophylaxis ranges from 9 to 50%. 1,2 Factors related to infection include age, obesity, indication for surgery, type of surgery, duration of procedure, and amount of perioperative bleeding. 3,4 Soper et al. s,6 and Larsson et al. 7 have reported an association between bacterial vaginosis and vaginal cuff cellulitis after abdominal hysterectomy. Person et al. s found that infection after gynecologic surgery was associated with bacterial vaginosis (BV).
One of the major challenges in the evaluation of BV is the widely varying and subjective clinical criteria used for diagnosis. The literature describes BV in a variety of ways. The most widely accepted clinical method of evaluation requires any three of the following four criteria to establish a diagnosis: a vaginal pH greater than 4.5, the presence of clue cells in the vaginal fluid, a milky homogenous discharge, and the release of an amine (fishy) odor after the addition of 10% hydrogen peroxide to the vaginal fluid. 9 The subjective interpretation of these signs can be very different, particularly with respect to the recognition of clue cells in the wet mount preparation of the vaginal fluid. This often leads to discrepant diagnoses. 1 A more objective and reproducible measure than the direct wet-mount exam is the Gramstained vaginal smear, which has 62 to 100% sensitivity and a positive-predictive value of 76 to 100%. 1

MATERIALS AND METHODS
A total of 175 women were consecutively chosen and followed prospectively in this study. Vaginal cultures were obtained preoperatively for nonpregnant patients undergoing major gynecologic surgery, which included benign gynecologic, urogyne-cologic, and gynecologic oncology cases. All women who presented for such surgery from July 1997 through October 1997 were candidates for this study. Data were obtained for the following variables: age, gravidity, parity, diagnosis, surgical procedure, preoperative and postoperative antibiotic use, preoperative and postoperative complete blood count, length of surgery, estimated blood loss, complications, and postoperative fever. Each patient was identified only by a code to ensure confidentiality. Specimens were obtained in a uniform fashion from the lower third of the vagina using a Culturette II transport medium (Becton-Dickinson, MD). Specimens were taken prior to any preoperative povidone iodine preparation and were transported to the laboratory at ambient temperature for analysis within 1 hour of collection. The vaginal smear was air-dried and Gram stained using safranin as the counterstain.
The diagnostic criteria for BV were based on Nugent's standardized method of Gram-stain interpretation. Each Gram-stained smear was evaluated under oil immersion (1,000 magnification) for the following three bacterial morphotypes" large grampositive rods (Lactobacilli), small gram-negative or small gram-variable rods (bacteroides or Gardnerdla vaginalis), and curved gram-negative to gramvariable rods (Mobiluncus). Each morphotype was quantitated from 1+ to 4+ with respect to the number of morphotypes per oil immersion field (0 no morphotypes; 1+ less than morphotype; 2+ 1 to 4 morphotypes; 3+ 5 to 30 morphotypes; 4+ 30 or more morphotypes). 6 The weighted quantitative sum of the morphotypes was then used to develop a 0to 10-point scoring system for the diagnosis of BV. 6 The criterion for BV was a score of 7 or higher; a score of 4 to 6 was considered intermediate (mixed flora), and a score of 0 to 3 was considered normal vaginal flora (Lactobadllus predominant). 6 The surgical procedures were classified into four types with respect to invasiveness: 1) peritoneum not entered vaginally or abdominally (e.g., Burch procedure); 2) peritoneum entered vaginally but not abdominally (e.g., vaginal hysterectomy); 3) peritoneum entered abdominally but not vaginally (e.g., exploratory laparotomy); and 4) peritoneum entered both vaginally and abdominally (e.g., total abdominal hysterectomy). Postoperative fever was defined as at least one temperature equal to 101.0 F or greater or two or more temperatures more than 6 hours apart equal to 100.4F or greater. Preoperative antibiotics were given based on risk factors, according to physician discretion. Most patients (86%) received preoperative antibiotics. When given, antibiotics were administered within an hour of incision time. The specific antibiotic administered was at the discretion of the patient's surgeon.
SPSS for Windows (version 7.5)was used for data management and statistical analysis. The chisquare test of association was used to compare groups with respect to nominal variables, and the nonparametric Mann-Whitney test was done to compare groups with respect to non-nominal variables that were statistically non-normal. A 0.05 sig-nificance level was used for all statistical tests.
Means are presented as mean standard deviation.

RESULTS
A total of 199 women had major gynecologic surgery during the 4 months of study. Twenty-four women were excluded from the analysis because of missing data, leaving a total sample of 175 women. Characteristics of the study sample are summarized in Table 1. Most women (89%) received preoperative or perioperative antibiotics, with cefazolin alone (41%) and cefoxitin alone (33%) the most commonly given antibiotics.
Forty-six percent of the patients had Lactobacillus-predominant vaginal microflora, 27% had an intermediate BV vaginal microflora, and 27% had a BV vaginal microflora. There were no statistically significant differences between the three BV groups with respect to age, gravidity, parity, race,  (Table 1). Table 2 describes relationships between preoperative and operative characteristics and postoperative fever. Although the percentage of patients with postoperative fever was higher for procedures in which the vagina was entered than for other procedures (26% versus 18%), this difference was not statistically significant. Postoperative fever was also more common in patients who received preoperative or perioperative antibiotics (24% versus 10%), but this difference was not statistically significant. The higher rate of postoperative fever associated with pre/perioperative antibiotic use may reflect a tendency to use antibiotics in patients who are at higher risk of postoperative infection. No statistically significant difference was found between cefazolin alone, cefoxitin alone, and other antibiotics with respect to the postoperative fever rate.

INFECTIOUS DISEASES IN OBSTETRICS AND GYNECOLOGY 171
There was a statistically significant difference between the three BV groups with respect to postoperative fever (P 0.017): 36% of the positive-BV group developed postoperative fever, compared with 20% of the Lactobacillus-predominant group and 12% of the intermediate-BV group ( Table 2).
Further significance testing found that the differences between the positive-BV group and the Lactobacillus-predominant group, and between the positive-BV group and the intermediate-BV group, with respect to postoperative fever, were statistically significant (P 0.045 and P 0.007, respectively). The difference between the intermediate-BV group and the Lactobacillus-predominant group was not statistically significant (P 0.28).
The relationship between BV and postoperative fever was further evaluated by separately analyzing patients with procedures in which the vagina was entered versus other patients, as well as patients who were given preoperative antibiotics versus other patients (Table 2). For procedures in which the vagina was entered, the postoperative fever rate was higher in the positive-BV group (39%) than in the Lactobacillus-predominant group (28%) and the intermediate-BV group (15%). For procedures in which the vagina was not entered, the postoperative fever rate remained higher in the positive-BV group (33%) than in the Lactobadlluspredominant group (14%) and the intermediate-BV group (8%). Although these differences were not statistically significant, this may be due to the loss of statistical power when the sample was subdivided for these analyses.
Only 19 patients did not receive preoperative or perioperative antibiotics, an insufficient sample size for comparison of the BV groups within this subgroup. For the 156 patients who received pre/ perioperative antibiotics, the postoperative fever rate was higher in the positive-BV group (38%) than in the Lactobacillus-predominant group (21%) and the intermediate-BV group (14%), a statistically significant difference (P 0.024). Further significance testing found that the difference between the positive-BV group and the intermediate-BV group with respect to postoperative fever was sta-tistically significant (P 0.009). The differences between the Lactobacillus-predominant group and the positive-BV group and the intermediate-BV group were not statistically significant (P 0.056 and P 0.30, respectively).

DISCUSSION
Bacterial vaginosis comprises many different species, each with a different impact on the vaginal flora. Lactobacillus exerts a protective effect in the vagina by producing hydrogen peroxide and bacteriocins and by lowering the pH, thereby inhibiting the colonization of the vagina by BV-associated organisms. 17 The serious morbidity of pelvic infection associated with gynecologic surgery makes identification of risk factors a high priority in reducing the patient's risk for adverse outcome. This study found a statistically significant, positive association between the presence of BV at the time of surgery and the incidence of postoperative febrile morbidity. Although we cannot rule out the possibility that this finding may be a spurious result of confounding with other variables, it may be prudent for the surgeon to identify patients with BV and treat them preoperatively. An effort to return the vaginal ecosystem to one that is dominated by Lactobacillus at the time of surgery may decrease the incidence of postoperative febrile and infectious morbidity. The questions that remain to be studied are: (1) Does BV really expose the patient to significant risk of developing postoperative pelvic infection? (2) Is the patient with BV undergoing gynecologic surgery and receiving standard cephalosporin prophylaxis likely to fail antibiotic prophylaxis? and (3) Would the patient with BV undergoing gynecologic surgery benefit from receiving metronidazole rather than a first-generation cephalosporin for antibiotic prophylaxis?