Cervicovaginal micro flora and pregnancy outcome: Results of a double-blind, placebo-controlled trial of erythromycin treatment James A. McGregor, MDCM: Janice I. French, CNM, MS: Ruth Richter, CNM, MSN: Margo Vuchetich, BSN: Vaundeen Bachus, BSN: Kyung Seo, MD: Sharon Hillier, PhD,b Franklyn N. Judson, MD," John McFee, MD," Judith Schoonmaker, PhD: and James K. Todd, MDd Denver, Colorado, and Seattle, Washington Available information suggests that some instances of preterm birth or premature rupture of membranes are associated with clinically unrecognized infection and inflammation of the lower uterine segment, decidua, and fetal membranes. Various cervicovaginal microorganisms have been recovered from these sites. Many of these microorganisms produce factors that may lead to weakening of the fetal membranes, release of prostaglandins, or both. This study evaluated the presence of various lower genital tract microflora and bacterial conditions in 229 women enrolled in a double-blind, placebo-controlled trial of short-course erythromycin treatment at 26 to 30 weeks' gestation to prevent preterm birth. Demographic, obstetric, and microbiologic parameters were prospectively evaluated. Premature rupture of membranes occurred less frequently (p < 0.01) among women who received erythromycin (6%) versus placebo (16%). Preterm premature rupture of membranes also occurred less frequently, although not significantly (p = 0.3) in patients who received erythromycin (2%) versus placebo (5%). Erythromycin treatment significantly decreased the occurrence of premature rupture of membranes among women who were initially positive for Chlamydia trachomatis infection. Logistic regression analysis demonstrated that C. trachomatis (p = 0.05; odds ratio, 9), vaginal wash phospholipase C (p = 0.08; odds ratio, 6) and prior preterm birth (p = 0.007; odds ratio 17) were associated with increased risk of preterm birth. Bacterial vaginosis, Mycoplasma hominis, Ureaplasma urealyticum were not significantly associated with increased risk of preterm birth or preterm rupture of membranes. These findings support a role for selected lower genital tract microflora in preterm birth and premature rupture. Large controlled treatment trials of specific infections or conditions associated with preterm birth and premature rupture of membranes are required to confirm the value of antimicrobial treatments in prevention of microbial-associated preterm birth. (AM J OBSTET GVNECOL 1990;163:1580-91.)

Key words: Prematurity, premature rupture of membranes, infection, antibiotic, treatment, prevention

Gestational age and birth weight are the most important biologic determinants of an individual's chances for survival and healthy growth and development.' Approximately 300,000 infants are born in the United States each year with the biologic disadvantages of prematurity. Of these approximately 15% to 20% die in the newborn period.' From the Department of Obstetrics and Gynecology, University of Colorado Health Sciences Center,' the Department of Obstetrics and Gynecology, University of Washington School of Medicine,' Denver Public Health and Hospitals and the Department of Medicine and Preventive Medicine, University of Colorado Health Sciences Center,' and the Department of Pediatrics, The Children's Hospital." Received for publication March 23, 1990; revised May 25, 1990; accepted July 27, 1990. Reprint requests: James A. McGregor, MDCM, 4200 E. 9th Ave., Box B198, Denver, CO 80262.

611124110

1580

To be maximally effective in prevention of preterm birth, intervention strategies should be cause specific and based on understanding of pathophysiology. Much information indicates that infectious processes and resultant inflammation in and around the utero placental unit are involved in the pathogenesis of a proportion of women who suffer preterm labor, premature rupture of membranes, and premature birth. 2 Mechanisms by which cervicovaginal microflora and host inflammatory cells may act to mediate preterm birth require further elucidation. Cervicovaginal microorganisms may gain access to the lower uterine segment through the cervix and directly mediate infection involving the decidua, the chorioamnion, and possibly the placenta. Such infections may be initially unrecognized, and if uncontained, can progress to clinically apparent chorioamnionitis with intraamniotic fluid in-

Volume 163 Number 5, Part 1

fection and possible fetal sepsis,:! Cervicovaginal microflora and host inflammatory cells may also mediate or promote preterm labor or pre term rupture of membranes through the release of various phospholipases and proteases. 1 . 5 Attempts to improve pregnancy outcome with antimicrobial treatment to eliminate specific lower urogenital tract microorganisms remain inconclusive but encouraging. Elder et al. 6 noted increased birth weight among newborns of women without bacteriuria who were treated ante natally with a 6-week course of tetracycline. 6 In an attempt to eliminate genital infections caused by Ureaplasma urealyticum, Mycoplasma Hominis, or both organisms among pregnant women, Kass et al. 7 and McCormack et al. 8 demonstrated increased birth weight among newborns of pregnant women who received erythromycin for up to 6 weeks during the third trimester. Neither study specifically discussed the occurrence of preterm premature rupture of membranes or the presence of other various genitourinary tract microflora. Recently Nugent et al. were unable to demonstrate benefits from antenatal erythromycin treatment in women colonized with U. urealyticum but not M. hominis, group B streptococcus, or Chlamydia trachomatis (Nugent RP, et aI., Society for Epidemiologic Research Abstracts, 1989;929-30). In a study focused on women with group B streptococcal urinary tract colonization, Thomsen et al. 9 reported a reduction in the occurrence of premature rupture of membranes and preterm birth among women treated with penicillin. Among women hospitalized with pre term labor who received parenteral tocolytic therapy, several studies have shown mixed results of erythromycin or ampicillin treatment or a combination of the two drugs in prolonging pregnancy. 10. II We undertook this study to prospectively evaluate associations of cervicovaginal microflora and selected lower genital tract microbe-associated factors with pregnancy outcomes. We used a short course of orally administered erythromycin to "probe" the relationship of these microbial factors or conditions with obstetric and neonatal outcomes including premature rupture of membranes, preterm premature rupture of membranes «37 weeks' gestation), gestational length, birth weight, and newborn infection in a group of inner-city pregnant women.

Methods From October 1985 to August 1988, 235 women between 26 and 30 weeks' gestation who attended publicly supported antenatal clinics in the Denver metropolitan area were enrolled in this prospective double-blind, placebo-controlled study. Prior evaluations of unselected women at these same inner-city sites showed rates of preterm birth between 16% to 20%, with the major-

Cervicovaginal microflora and pregnancy outcome

1581

ity of these births occurring after 26 weeks' gestation (John McFee, MD, personal communication). In the studies by Kass et aJ.7 and McCormack et al.,8 only erythromycin given in the third (vs second) trimester of pregnancy was associated with increased birth weight. Therefore we enrolled and treated patients between 26 and 30 weeks' gestation. Written informed consent was obtained as approved by the institutional human subjects committees. Women were excluded if they had a recognized cause for prematurity (multiple gestation, placenta previa, or cervical incompetence); if they had severe cardiac, renal, respiratory or immunologic diseases; or if they were allergic to erythromycin. Subsequently six women were excluded from analysis: four women were lost to followup after enrollment, one was treated for premature labor on the day of enrollment, and one woman experienced intrauterine fetal death at 30 weeks' gestation. The total evaluated sample included 229 women and their newborns. Clinical evaluation. Women were enrolled by one of three experienced clinicians. Participants were interviewed to obtain demographic, social, behavioral, sexual, medical, and obstetric information. Specific information regarding urogenital symptoms was recorded. A vaginal speculum examination was performed and characteristics of vaginal discharge (consistency, amount, color, odor) were recorded on standard forms. Samples were obtained as follows. One milliliter of saline solution was inoculated with a cotton swab containing vaginal fluid for microscopic examination for Trichomonas vaginalis, yeast pseudohyphae, and release of amine odor with addition of 10% potassium hydroxide "whiff test." This sample was then held at 4° C and evaluated in the laboratory for nonspecific protease activity using Azocoll substrate. 5 Vaginal fluid pH was assessed directly with Nitrazine paper (Squibb, Princeton, N.J.) and recorded as >4.5 or :$4.5. An additional sample of vaginal fluid was allowed to air dry on a slide for Gram stain evaluation of bacterial vagi nos is and Mobiluncus spp. morphotypes. 12 Separate endocervical swabs were immediately inoculated onto a chocolate agar and a modified ThayerMartin split plate for the recovery of Neisseria gonorrhoeae organisms, and into sucrose-phosphate glutamate transport media for recovery of C. trachomatis. Cervicovaginal swabs were inoculated onto 5% sheep blood agar (tryptic soy base, Remel, Denver) for recovery of group A streptococcus, group B streptococcus, Staphylococcus aureus, Gardnerella vaginalis, and yeast organisms. These microorganisms were identified with standard techniques. Additional vaginal swabs were inoculated into a standard mycoplasma medium for recovery of M. hominis and U. urealyticum. Cultures for C. trachomatis, M. hominis, and

1582

McGregor et al.

U. urealyticum specimens were frozen at - 70° C and processed in batches. After 48 hours' incubation at 35° C, inoculated cycloheximide-treated McCoy cell monolayers were treated with fluorescence-conjugated anti-chlamydia monoclonal antibody (Syva Microtrak, Syva Co., Palo Alto) and examined microscopically for the presence of stained inclusions. Negative cultures were blind passaged once and reprocessed. M. hominis and U. urealyticum organisms were cultured and identified with the Mycotrim GU Triphasic Culture System (Hana Biologics, Inc., Alameda, Calif.) After microbial samples were obtained and vaginal pH was determined, the posterior vaginal fornix was irrigated with 4 ml of sterile Dulbecco's buffered saline solution and after standardized mixing was aspirated into a sterile syringe. The vaginal wash specimens were frozen immediately at - 20° C, then frozen at -70° C within 4 hours of collection. Wash samples were evaluated in batches for nonspecific protease,' phospholipase C,13 and total protein (Biorad, Richmond, Calif.). Proline aminopeptidase was quantified in these samples by a modification of the method of Thomasen et al. 14 Vaginal samples were incubated for 4 hours with proline nitroanilide (0.2 gm/ 100 ml Tris buffer) at 37° C, 6% carbon dioxide. Samples were read at 450 nm and compared with standard leucine aminopeptidase for quantitation. Treatment and follow-up. After initial examination the women were given a sealed bottle containing 21 tablets and were instructed to take one tablet three times a day for 1 week, and to record when they took the tablet on a medication diary. Treatment was prepared and randomized by means of a computer-generated random number list by The Upjohn Company (Kalamazoo). Women received identical-appearing treatment bottles and tablets that were either erythromycin base (E-Mycin, Upjohn) 333 mg, or placebo. Patients continued routine prenatal care after enrollment in the study. Two to four weeks after enrollment the women were seen by one of the investigators during a routine prenatal visit; diaries and bottles were collected and unused pills were counted. Women were questioned regarding side effects, sexual activity, and signs and symptoms of vaginitis before repeat vaginal sampling. Only positive N. gonorrhoeae cultures were reported to health care providers. Except for one patient with N. gonorrhoeae infection, sexual partners of patients were not referred for evaluation and treatment. Maternal and newborn medical records were reviewed to obtain additional information on antenatal, intrapartum, and postpartum maternal course and the newborn course through 3 months of age. If records were unavailable the data were obtained by telephone calls and written correspondence to private physicians and patients.

November 1990 Am J Obstet Gynecol

Definitions. Gestational age was determined before data analysis and based on best clinical criteria, which included last menstrual period, fetal growth, and ultrasonographic measurements, and newborn gestational age assessment. 15 Preterm labor was defined as regular uterine contractions (two or more in 10 minutes) treated with hospitalization and at least one subcutaneous injection or intravenous infusion of tocolytic agent followed by oral tocolytic therapy, or labor that resulted in a preterm birth. Patients with preterm labor that followed pre term premature rupture of membranes were not counted as preterm labor patients. Premature rupture of membranes was defined as rupture of membranes at least 1 hour before the onset of uterine contractions. Preterm premature rupture of membranes was defined as rupture of membranes before labor (as above) before 37 weeks' gestation. Preterm birth was considered as birth before completion of 37 weeks' gestation. Infants who weighed 10 cigarettes I day) Alcohol (any) Street drug use (any)

Erythromycin

Placebo

(N = 119)

(N = 110)

23.0 ± 4.3 13-37

23.2 ± 4.2 16-34

45 42 30 2 66

43 42 21 4 60

(37.8%) (35.3%) (25.2%) (1.6%) (55.5%)

(39.1%) (38.2%) (19.1%) (3.6%) (54.5%)

2.9 ± 1.8 1-11 29 (24.4%) 15 (12.6%) 61 (53%) 24 (20.9%)

2.8 ± 1.6 1-8 30 (27.3%) 12 (10.9%) 57 (54.3%) 20 (19.2%)

29 (24.4%) 25 (21.0%) 23 (19.3%)

36 (32.7%) 30 (27.3%) 14 (12.7%)

*One or more prior births 4.5 Proline aminopeptidase Phospholipase C Nonspecific protease

Erythromycin (N = 119)

I

1 (1%)t 13 (12%) 74 (76%) 24 (24%) 32 (27%) 15 (13%) 27 (23%) 4 (4%) 8 (7%)

P osttreatment* Placebo (N = 110)

13 82 17 25 14 32 5 4

0 (12%) (84%) (18%) (23%) (13%) (29%) (5%) (4%)

Erythromycin (N = 75)

0 7 (10%) 42 (64%) 8 (12%) 19 (26%) 9 (12%) 20 (27%) 1 (1%) 2 (3%)

I

Placebo (N = 81)

9 58 12 21 12 20 3 3

0 (12%) (84%) (17%) (27%) (16%) (26%) (4%) (4%)

6 (5%)

1 (1%)

6 (8%)

2 (3%)

1 (1%) 42 (35%) 23 (21 %) 36 (32%) 7 (6%) 26 (24%)

1 (I %) 44 (40%) 37 (37%Jt 37 (36%) 9 (9%) 32 (30%)

1 (1%) 22 (29%) 20 (28%) 20 (32%) 3 (5%) 17 (25%)

1 (1 %) 18 (24%) 22 (29%) 20(31%) 7 (10%) 21 (28%)

*There were no differences in the overall frequency of recovered agents or related factors among patients sampled before and after study treatments. tNumber and percent positive. *p = 0.01; relative risk, 0.6; 95% confidence interval, 0.4 to 0.9.

up for intrauterine growth retardation. Pregnancyinduced hypertension was diagnosed among four (3%) erythromycin patients and four (4%) placebo patients antenatally. Each was delivered at term. Two patients experienced intrauterine fetal death, one of undetermined cause at 30 weeks in a patient in the placebo group and one fetus with Down's syndrome at 42 weeks in a women in the erythromycin treatment group. The patient with the 30-week-gestation fetal death was not considered in analysis for preterm outcome variables; all other patients with concurrent or intercurrent antenatal medical or obstetric problems were included in the analysis. Sixty-six percent of the erythromycin group and 69% of the placebo group received nonprotocol antimicrobial therapy during antenatal care. No patient received antibiotics within 2 weeks of study treatment. The preponderance of treatments occurred before study entry (first trimester: 21%; second trimester: 40%; third trimester: 37%). Vaginitis was the most common reason for antimicrobial treatment. Approximately 20% of both groups received treatment during pregnancy for a single urinary tract infection. Compliance. The mean number of study pills taken differed between groups (erythromycin, 15.9 ± 7.8; placebo, 17.8 ± 5.9; P = 0.05). Only 73% of women randomized to receive erythromycin and 84% of women who received placebo completed 4 or more days of study treatment (p = 0.04). Side effects, including nausea, vomiting, and diarrhea, were reported sig-

nificantly (p < 0.05) more often among women who received erythromycin (nausea: erythromycin = 27%; placebo = 13%; P = 0.01; vomiting: erythromycin = 13%; placebo = 4%; P = 0.02; diarrhea: erythromycin = 5%; placebo = 0; p = 0.03). Pregnancy outcome. Spontaneous vaginal delivery occurred among 80% of patients in the erythromycin group and 71 % of placebo patients. Cesarean section delivery was performed for 8% and 12% of patients who received erythromycin and placebo, respectively. Assisted vaginal delivery occurred in the remaining patients. There were no overall differences in the occurrence of preterm labor, preterm birth, or low-birth-weight infants among treatment groups when the data were analyzed by univariate analysis. The mean gestational age at delivery was 39.3 ± 1.9 weeks among patients in the erythromycin group and 39.2 ± 1.6 weeks for those who received placebo. Mean birth weight was 3132 ± 588 gm among erythromycin patients and 3099 ± 548 gm among placebo patients (p = 0.7). Further analysis showed no difference in gestational age or birth weight at delivery even for those patients who had completed at least 4 days of study treatment [gestational age at delivery: erythromycin (N = 87),39.3 ± 1.7 weeks; placebo (N = 93), 39.4 ± 1.6 weeks; birth weight: erythromycin, 3191 gm; placebo, 3161 gm] or those who completed the full 1 week of study treatment [gestational age at delivery: erythromycin (N = 57), 39.2 ± 1.7 weeks; placebo (N = 60), 39.4 ± 1.6

Volume 163 Number 5, Part I

weeks; birth weight: erythromycin, 3195 gm, placebo, 31S0 gm]. Two patients in the erythromycin group (2%) and three (3%) patients in the placebo group were treated for clinically diagnosed chorioamnionitis. Postpartum endometritis occurred among six (6%) patients in the erythromycin group and eight (S%) in the placebo group. There were no significant differences between study groups for any neonatal outcome evaluated. Premature rupture of membranes. Ru pture of membranes before labor occurred significantly more often among women who received placebo (16%) compared with those who received erythromycin (6%), (P = 0.01; relative risk, 0.4; 95% confidence interval, 0.2 to O.S). Preterm premature rupture of membranes occurred among seven women and was also decreased among women in the erythromycin group, although this failed to reach statistical significance (erythromycin, 2%; placebo, 5%; p = 0.3; relative risk, 0.4; 95% confidence interval, 0.1 to I.S). Trends were noted between selected medical factors and premature rupture of membranes; more women who experienced premature rupture of membranes gave a history of prior sexually transmitted disease (erythromycin, 15%; placebo, 7%; p = 0.06, relative risk, 2; 95% confidence interval, 1.0 to 4.9) and women who experience preterm premature rupture of membranes more often had a history of a single prior abortion (erythromycin, 7%; placebo, 2%; p = O.OS; relative risk, 4; 95% confidence interval, 0.9 to 14.S). None of the other demographic, behavioral or other medical problems analyzed were significantly associated with increased risk of premature rupture of membranes or preterm premature rupture of membranes. Overall, premature rupture of membranes (any gestational age) occurred significantly more often among women from whom C. trachoma tis was recovered at study enrollment (p = 0.03; relative risk, 2.5; confidence interval, 1.1 to 5.7) and also among those with vaginal pH >4.5 (p = 0.03; relative risk, 2.5; 95% confidence interval, 1.1 to 5.5). Women with U. urealyticum infection experienced premature rupture of membranes more frequently; however, this did not reach statistical significance (p = O.OS, relative risk, 5.0, 95% confidence interval, 0.9 to 27.S). To test for effects of erythromycin treatment on pregnancy outcomes, data were stratified according to the presence of susceptible microorganisms or abnormal vaginal conditions. None of the 13 women treated with erythromycin from whom C. trachomatis was initially recovered experienced premature rupture of membranes. Conversely, six of 12 women with positive C. trachomatis cultures who received (50%) placebo experienced premature rupture of membranes (p = 0.03, relative risk, 0.4; 95% confidence interval, 0.2

Cervicovaginal microflora and pregnancy outcome

1585

to O.S). The Breslow-Day test for homogeneity of results of odds ratios confirmed this significant difference (p = 0.03). Among women from whom U. urealyticum was recovered, the occurrence of premature rupture of membranes was significantly less in those who received erythromycin (erythromycin, 7%; placebo, IS%, p = 0.03; relative risk, 0.4; 95% confidence interval, 0.1 to 0.9); among women who were negative for U. urealyticum, only a single patient in the placebo group experienced premature rupture of membranes (Breslow-Day test, NS). Among the women with vaginal pH >4.5, women in the erythromycin group experienced premature rupture of membranes less often, although not significantly (erythromycin, 4%; placebo, 24%; p = 0.07; relative risk, 0.2; 95% confidence interval, 0.03 to 0.96). Pre term premature rupture of membranes occurred more often among women whose initial cultures were positive for S. aureus (erythromycin, 2112 [17%]; placebo, 5/209 [2%]); (p = 0.04; relative risk, 7; 95% confidence interval, I.S to 2S.4). Women with positive findings at enrollment for C. trachomatis, T. vaginalis, group B streptococcus, yeast species, phospholipase C, positive whiff test, and pH >4.5 also experienced preterm premature rupture of membranes more frequently; however, these associations failed to reach statistical significance. Preterm birth. Preterm birth occurred among eight (7%) women in the erythromycin group and nine (S%) women who received placebo. History of a prior preterm birth was the only historical, demographic, or behavioral variable significantly associated with preterm birth in the current pregnancy (p = 0.002; relative risk, 4; 95% confidence interval, 1.7 to 9.9). Recovery of C. trachomatis was significantly associated with increased risk of preterm birth (p = 0.03; relative risk, 3; 95% confidence interval, 1.2 to S.I). A positive whiff test was also significantly associated with increased risk of preterm birth (p = 0.04; relative risk, 3; 95% confidence interval, 1.1 to 7.6) as was the presence of both pH >4.5 and a positive whiff test (p = 0.03; relative risk, 4; 95% confidence interval, 1.3 to 10.6). Vaginal pH >4.5 alone was not significantly associated with increased risk of preterm birth (p = 0.2; relative risk, 2; 95% confidence interval, 0.7 to 5.1). Overall, bacterial vaginosis (determined by Gram stain) and positive proline aminopeptidase were not associated with preterm birth. Preterm birth occurred more frequently, but not significantly, among the small number of women with T. vaginalis, S. aureus, and increased phospholipase C found at initial examination. Preterm birth occurred before completion of 34 weeks' gestation in five patients. C. trachomatis (p = 0.01; relative risk, 11; 95% confidence interval, 2.S to 46), M. hominis (p = 0.008; relative risk, 26; 95% con-

1586 McGregor et al.

November 1990 Am J Obstet Gynecol

Table III. Cervicovaginal findings and selected risk factors for five patients with delivery :5:34 weeks Prior preterm birth Prior abortion

C. trachomatis

Patient no. 1

Patient no. 55

Patient no. 109

+

+ + +

+

NO

+ + +

+ + +

NO NO

+ +

U U

T. vaginalis

Bacterial vaginosis Proline aminopeptidase pH >4.5 Whiff test M obiluncus species U. urealyticum M. hominis

+ +

Patient no. 119

Patient no. 141

+ + + + + + + + +

+ + + + + +

ND, Not done; U, unevaluable specimen.

fidence interval, 1.4 to 494), proline aminopeptidase (P = 0.04; relative risk, 8; confidence interval 95%, 1.2 to 48), pH >4.5 (P = 0.07; relative risk, 8; 95% confidence interval, 1.1 to 50), and positive whiff test (P = 0.02; relative risk, 11; 95% confidence interval, 2.5 to 44) were significantly associated with increased risk of birth :5:34 weeks' gestation. Of note, four of these five women had findings of multiple microbiologic factors thought to be associated with increased risk of prematurity (Table III). Preterm birth in the absence of premature rupture of membranes occurred among nine women and significantly more often among women with both vaginal pH >4.5 and positive whiff test (p = 0.03; relative risk, 6; 95% confidence interval, 1.6 to 19.5). Similarly, preterm birth without prior rupture of membranes occurred more frequently, but not significantly, among women who were positive for C. trachomatis (p = 0.07), T. vaginaZis, M. hominis, proline aminopeptidase (p = 0.09), whiff test, and phospholipase C. Multivariate analysis. Logistic regression analysis was performed to control for the effects of multiple independent variables on the occurrence of preterm birth, low birth weight, or premature rupture of membranes (Table IV). As a preliminary step, a logistic regression analysis was performed to identify significant independent variables from among maternal demographic, behavioral, and past medical history variables that were predictive of preterm birth, low birth weight, or premature rupture of membranes, Among these independent variables (maternal age, ethnic group, marital status, gravidity, prior abortion, prior preterm birth, use of alcohol, street drugs, smoking, and history of pelvic inflammatory disease), only prior preterm birth was significantly associated with preterm birth and low birth weight (p = 0.001). Second, on the basis of results of previously published studies or theoretical associations, another model was constructed that included bacteriologic findings

(c. trachomatis, T. vaginaZis, U. ureaZyticum, M. hominis, aureus, bacterial vaginosis), vaginal enzymes (high

s.

vaginal nonspecific protease, phospholipase C), vaginal pH >4.5, treatment group, use of other antenatal antibiotics, and prior preterm birth. In this second model (Table IV), C. trachomatis (p = 0.05; OR, 9), phospholipase C (p = 0.08; OR, 6) and prior preterm birth (p < 0.01; OR, 17) were associated with preterm birth. After adjusting for the effects of the other independent variables in the model, erythromycin treatment was associated with decreased risk of preterm birth (p = 0.06; OR, 0.1). A final model was created that included only the independent variables that were statistically significant or bordered on statistical significance (C. trachomatis, phospholipase C, prior preterm birth, and study treatment). In this final model, C. trachomatis (p = 0.02; OR, 4) and prior preterm birth (p = 0.009; OR, 5) were the only variables associated with increased risk of pre term birth. In similar analyses, low birth weight «2500 gm) was positively associated with prior preterm birth (p = 0.003, OR, 8) and negatively associated with erythromycin treatment (p = 0.02, OR, 0.2); and erythromycin treatment was associated with decreased risk of premature rupture of membranes (p = 0.01; OR, 0.1) (Table IV). Because of the small number of women who experienced preterm premature rupture of membranes, preterm birth :5:34 weeks' gestation, and preterm birth in the absence of prior premature rupture of membranes, these outcomes were not evaluated with multivariate techniques. In this study, cervical infection with C. trachomatis accounted for 20% of the preterm birth (population attributable risk). Such calculations are probably oversimplifications because an individual mother's host susceptibility and the presence of different microorganisms are not considered.

Cervicovaginal microflora and pregnancy outcome

Volume 163 Number 5, Part 1

1587

Table IV. Results of multivariate analysis Dependent variable

Pre term birth

Pre term birth

Pre term birth

Independent variable

Demographic, historical, medical factors* Prior pre term birth Cervicovaginal microorganisms* Erythromycin treatment Prior pre term birth C. trachomatis

Birth weight