Study design

We performed a prospective longitudinal cohort study of women with a previous preterm birth at <35 weeks of gestation from the Magee-Womens Hospital Maternal-Fetal-Medicine practice between October 2006 and November 2008. The study was approved by the University of Pittsburgh Institutional Review Board. Eligible women were enrolled between 15 weeks and 23 weeks 6 days of gestation. Exclusion criteria were cerclage, twin gestation, placenta previa, autoimmune disease, human immunodeficiency virus, vaginal bleeding, current diagnosis of sexually transmitted infections ( Neisseria gonorrhoeae, Chlamdydia trachomatis, Trichomonas vaginalis ), or lethal fetal anomaly. These conditions were selected for exclusion because these women may have altered immune status or biologically distinct pathways of cervical shortening that would confound the association that we proposed to explore. Seven women were excluded a priori from this analysis because of a history of a loop electrocautery excision procedure. Women with a history of loop electrocautery excision procedure may have a unique pathway to cervical shortening. We further speculated that there may be alterations in cervical inflammation related to cervical remodeling or from the cervical disease caused by human papillomavirus that could confound the relationship that we were trying to investigate.

Gestational age was based on menstrual dating in conjunction with a first or early second trimester ultrasound scan. Women had cervical swabs for N gonorrhoeae and C trachomatis at their initial prenatal visit or, subsequently, if they were symptomatic. Bacterial vaginosis screening was performed routinely in the early second trimester and was diagnosed by vaginal pH ≥4.7 and a score of 7-10 from a Gram-stained vaginal smear, which was interpreted by the Nugent method. C trachomatis and N gonorrhoeae were identified with nucleic acid amplification tests.

At each visit, we performed a pelvic examination for collection of cervical fluid and a transvaginal ultrasound scan for cervical length at intervals of 2-3 weeks until a gestational age of 24 weeks. At the time of the speculum examination, Dacron swabs were placed in the cervix and left for 10 seconds to achieve saturation. The swab was placed in a plastic container with 350 μL of phosphate-buffered saline solution (final dilution 1:5). The specimens were stored at –80°C until processed. Transvaginal ultrasound scans were performed by the study investigators who used the methods of Iams. Three cervical length measurements were performed at rest and were recorded at each visit. The images were reviewed for quality before they were included in an averaged measurement for each encounter.

Laboratory analysis/measurement of cytokines

Cervical fluid specimens were thawed at room temperature. Swab and diluent were placed in a centrifuge filter unit (Spin-X; Costar, Cambridge, MA) and centrifuged at 12,000 rpm for 20 minutes at 4°C.

Commercially available multiplexed fluorescent bead-based immunoassays (Millipore Corp, Billerica, MA) were used to determine cytokine (IL1-β, -4, -6, -10, and -13) concentrations in accordance with manufacturer’s instructions. Briefly, 25 μL aliquots of a known standard concentration or sample were added in duplicate to individual wells of a 96-well multiscreen plate. Multiplexed fluorescent capture beads were added to each well, and plates were incubated on a plate shaker in the dark at 4°C overnight. Plate wells were washed with assay buffer, and the beads were resuspended in 25 μL of detection antibody. The plate was incubated in the dark at room temperature for 60 minutes. A streptavidin-phycoerythrin solution was then added for 30 minutes. Plates were washed again, and the beads were resuspended in 150 μL of sheath fluid. Beads were analyzed on Luminex 100 (Luminex, Austin, TX) multiplex instrumentation. Specimens were run in duplicate, and the average mean fluorescent intensities were extrapolated from 7-point standard curves with the use of 4-parameter logistic algorithms (Applied Cytometry Systems, Sacramento, CA). Each assay was run with an intra- and interassay variation on <10%. For all analytes, the lower limit of sensitivity was 6.9 pg/mL. Values below the lower limit of detection were recorded as the lower limit of the assay for the purpose of the analysis.

Statistical analysis

To standardize each cytokine for each woman, cytokine concentrations were evaluated as multiples of the median unitless values. We used the sum of the multiples of the median for the 2 proinflammatory cytokines to generate a proinflammatory score and did the same for the 3 antiinflammatory multiples of the median to generate the antiinflammatory score. The pro-/antiinflammatory ratio was the independent variable; the dependent variable was cervical length. Because there were repeated measures among the women, which violates the statistical assumption of independence, the linear regression model was fitted by the method of generalized estimating equations. We used an exchangeable working correlation structure because the visit history did not correspond to the same gestational age among different women. The correlation structure used a maximum of 6 possible visits per patient. Robust estimates of variance were used for correlated data. Variables were included for adjustment in the model if they were significant in univariate analysis or had strong biologic plausibility. Potential confounders that were considered for adjustment included smoking, race, gestational age, body mass index (BMI), bacterial vaginosis, N gonorrhoeae, C trachomatis, T vaginalis , and 17-hydroxyprogesterone caproate therapy. In this population of women, BMI was the only significant covariate. Outliers were evaluated by residual plots. Two leverage points were influential and were removed from the analysis. The criteria for statistical significance was fixed as a type I error rate of 5%. The data analysis was performed with Stata 10 statistical software (Stata Corp, College Station, TX).