Materials and Methods

With an institutional review board–approved protocol, singleton pregnancies with risk factors for spontaneous preterm birth that attended the Lehigh Valley Hospital Perinatal Testing Center between April 1998 and December 2006 were screened with serial transvaginal ultrasound scans beginning at 16 weeks’ gestation. Risk factors for preterm birth included a history of spontaneous preterm birth, a second-trimester pregnancy loss, previous cervical surgery (conization or loop excision), or documented uterine anomaly. Also, low risk asymptomatic singleton pregnancies between 16 and 24 weeks’ gestation were screened for evidence of cervical shortening with transabdominal ultrasound scans as part of routine anatomic survey. If the cervix appeared short (≤25 mm) transabdominally, a transvaginal ultrasound scan was performed. Transvaginal cervical length measurement was obtained by the standardized technique described by Rust et al.

Patients with ultrasonographic evidence of a short cervix (defined as a transvaginal distal cervical length ≤25 mm [shortest cervical length] with or without transfundal pressure) were offered enrollment into randomized controlled trials of cerclage vs either no cerclage or progesterone therapy for treatment of short cervix. All patients who were enrolled in the study underwent an ultrasound-guided transabdominal amniocentesis to exclude intraamniotic infection (low glucose level, elevated white blood cell count, and aerobic/anaerobic culture) before study enrollment. Five milliliters of unspun AF was aliquoted into 15 mL polypropylene tubes and stored at –70°C for future cytokine analysis. As part of the randomized trial protocol, all patients received 48-72 hours of indomethacin and clindamycin after amniocentesis and before random assignment.

The cohort in this study includes patients who received no intervention for short cervix and had available outcome data. This cohort therefore included patients who were assigned randomly to the “no intervention” arm of the randomized trial by Rust et al and patients who were not assigned randomly (who did not to meet randomization criteria or who, after enrollment, declined random assignment). Patients were not excluded from this cohort if the amniocentesis revealed evidence of infection (low glucose level or positive culture). We included these patients because they represented the full spectrum of patients with short cervix and because the presence or absence of infection was not known before amniocentesis. Patients were excluded for the following conditions: any known fetal chromosomal or structural anomaly, multiple gestation, ruptured membranes, vaginal bleeding, or the need for an obstetrically indicated delivery. Data on maternal demographics, gestational age at study entry, risk factors for spontaneous preterm birth, cervical length measurements, and perinatal outcome were recorded from the Lehigh Valley Hospital cerclage database.

A comprehensive AF cytokine analysis of all patients with a short cervix was performed with the Bio-Plex Array system (Bio-Rad Laboratories Inc) simultaneously to assay and quantify 25 different cytokine concentrations. The cytokines that were evaluated were interleukin (IL)-1β, -1ra, -2, -4, -5, -6, -7, -8, -9, -10, -12, -13, -15, and -17, eotaxin, granulocyte colony-stimulating factor, interferon gamma, inducible protein-10, monocyte chemotactic protein–1, macrophage inflammatory protein–1a and –1b, platelet-derived growth factor–bb, tumor necrosis factor alpha, vascular endothelial growth factor, and regulated on activation, normal T cell expressed and secreted.

To accomplish our first objective (creation of an AF inflammatory score), we performed a univariate analysis to examine the distribution of cytokine levels, to define the upper quartile, and to determine the proportion of samples that had a cytokine measurement below the level of detection for the assay. Patient data were then stratified according to gestational age at delivery (<34 vs ≥34 weeks). We selected this gestational age because birth at <34 weeks is associated with the potential for significant neonatal morbidity. The median cytokine values were compared for these 2 groups with the use of the nonparametric Wilcoxon’s rank-sum test. For a mediator to be included in the AF inflammatory score, the probability value needed to be < .002 (Bonferroni correction for simultaneously comparing 25 mediators). In addition, the mediator needed to be detected in at least 25% of patients.

Data from the aforementioned analysis were used to determine which mediators reached statistical significance and therefore should be included in the AF inflammatory score. Patients were assigned 1 point for each significant mediator if their level was in the upper quartile. The AF cytokine score was determined for each patient and its distribution, and its relationship to other clinical characteristics were examined.

To accomplish the second objective (to determine the predictive ability of the AF inflammatory score), a receiver-operator characteristic curve was used to determine the AF inflammatory score cutoff that would be predictive of spontaneous preterm birth at <34 weeks’ gestation. The sensitivity, specificity, positive predictive value, and negative predictive value for the prediction of spontaneous delivery at <34 weeks’ gestation was then calculated. Patient data were then stratified as a high or low inflammatory score based on that cutoff. A Kaplan-Meier pregnancy survival curve was constructed to examine the relationship between the AF inflammatory score and the interval to delivery. Baseline clinical characteristics were compared with the use of either the Wilcoxon’s rank-sum test or Fisher’s exact test, where appropriate. Spearman correlation was used to determine the relationship between the inflammatory score and cervical length. A probability value of < .05 was required for statistical significance. Calculations were performed with SAS software (version 9.2; SAS Institute Inc, Cary, NC).

Our third objective was to test the discriminatory ability of the AF inflammatory score in a separate cohort of normal pregnant patients who were assumed to have a normal second-trimester cervical length. Chow et al have published cytokine data on 100 asymptomatic women who underwent genetic amniocentesis with the use of the same Bio-Plex platform (Bio-Rad Laboratories Inc). In a collaborative agreement, we applied the AF inflammatory score to their patient-level data to determine the distribution of scores in this normal population. This cohort was used because it was the only large publication that we were aware of to measure AF cytokine levels on the same Bio-Plex platform (Bio-Rad Laboratories Inc) from asymptomatic patients who had undergone genetic amniocentesis. Although cervical length was not measured specifically in this group, the incidence of mid-trimester short cervix was approximately 10%. In addition, all of their patients delivered at term, with the exception of 1 pregnancy that delivered at 35 weeks’ gestation. This outcome further supports the assumption that most of their cervical lengths were normal at the time of amniocentesis. If the proposed scoring system is predictive of outcome, we would expect most of these patients to be classified as having a low inflammatory score.