This study included participants from the Pregnancy Outcomes and Community Health (POUCH) Study (30
), which enrolled women in their 15th
week of pregnancy from 52 clinics in five Michigan communities from 1998 to 2004. Eligibility criteria included maternal age >14 years, English-speaking, a singleton pregnancy with no known congenital or chromosomal abnormalities, no pre-pregnancy diabetes, and screening at 15th
weeks’ gestation for maternal serum alpha-fetoprotein (MSAFP), a biomarker of particular interest in the POUCH Study. Eligible women were randomly sampled into the study and those with unexplained high MSAFP (> 2 multiples of the median) were over-sampled (7% of cohort). Of the 3,038 women enrolled, 19 were lost to follow-up, leaving 3,019 (99.4 %) in the cohort. Study protocols were approved by institutional review boards at Michigan State University and eight hospitals in the five communities. All participants provided written consent prior to enrollment. Comparisons of POUCH study women with aggregate birth certificate data specific to the five study communities showed that the POUCH sample was very similar to community mothers on most factors measured, including age, parity, education levels, the proportions of women with Medicaid insurance, PTD, previous stillbirth, previous preterm infant, and previous low birth weight infant. The one exception, the percentage of African Americans over 30 years of age, was 14% in the POUCH study and 21% in the birth certificate data (30
A sub-cohort was selected to maximize resources. The sub-cohort included all women who delivered preterm, all women with unexplained high MSAFP levels, and a random sample of women who delivered at term with normal MSAFP, with an over-sampling of African Americans in this latter category. In the POUCH study sub-cohort (N=1,371), 692 (50.5%) women reported their race/ethnicity as non-Hispanic White, 579 (42.2%) as African American, and 100 (7.3%) as another racial/ethnic group. Because of the small number of participants and the broad racial/ethnic mix, the “other” group was excluded from the following analyses. In addition, 235 (18%) non-Hispanic White and African-American women declined vaginal fluid sampling and 5 had samples with insufficient volume, leaving a sub-cohort sample of 1,031 women for the current analyses (787 term, 244 preterm).
At enrollment, study participants met with a study nurse and provided information about demographics, current pregnancy, reproductive history, health behaviors such as smoking, and psychosocial factors. Vaginal fluid samples were collected via a fetal fibronectin specimen collection kit (Adeza International, Sunnyvale, CA, USA). After placement of a vaginal speculum, vaginal fluid was collected from the posterior fornix with a sterile Dacron swab. The swab was placed in 1 ml extraction buffer and refrigerated (4°C) for at least 24 hours. After the refrigeration period, buffer was expressed from the swab, the specimen was filtered using a 10.25mm × 4” serum filter (Fisherbrand Serum Filter System, Fisher Scientific, Pittsburgh, PA, USA), divided into 0.5 ml aliquots, and stored at −80°C.
Vaginal fluid specimens were later thawed and 50 μl aliquots were removed, refrozen, and shipped to the University of Alabama at Birmingham for assay. An ELISA kit measured total HNP level which included HNP 1, 2 and 3 (HyCult Biotechnology, The Netherlands) and was reported out from the assay as one value. This approach has been used in other studies, including one that assessed defensin levels and presence of sexually transmitted pathogens in urethral lavages (31
). Thirteen women had a vaginal HNP 1-3 level below the lower limit of detection and these values were imputed as half the detection level.
At the time of vaginal fluid sampling, study nurses also obtained vaginal smears. Smears were Gram-stained and evaluated for bacterial vaginosis by a microbiologist using the Nugent scoring system (5
). Scores ranged from 0 to 10, with bacterial vaginosis status categorized as negative (0-3), intermediate (4-6), and positive (7-10).
Information on three major vaginal infections, Chlamydia, Gonorrhea, and Trichomonas, was ascertained in two ways. At enrollment, just prior to vaginal fluid sampling, women were asked about diagnoses of these vaginal infections for two different periods, the year before the pregnancy and during pregnancy. Prenatal and labor and delivery records were abstracted and women were classified as being culture positive or negative for any of the three infections. Culture results from the pregnancy period were available in 99% of women. In addition, the maternal interview asked women to report antibiotic use from the beginning of the pregnancy up through the time of enrollment.
Preterm delivery was defined as birth before 37 completed weeks’ gestation. Gestational age was determined by the last menstrual period (LMP) or by ultrasound data when the LMP-derived gestational age differed from the ultrasound estimate by at least two weeks. Based on information abstracted from labor and delivery medical records, PTD was divided into two groups: 1) Spontaneous PTD included women with preterm labor defined as regular contractions that led to cervical change (≥2 cm dilatation), or spontaneous premature rupture of membranes; and 2) medically indicated PTD included women who had labor induced or who were delivered by C-section before either preterm labor or premature rupture of membranes.
The overall analytic goals were first to assess relations between vaginal HNP 1-3 levels and maternal characteristics including bacterial vaginosis, and second to examine associations between vaginal HNP 1-3 levels and risk of PTD. All analyses incorporated sampling weights to account for the sampling schemes used to construct the cohort and sub-cohort, with preterms and African-American women weighted to their original proportion in the cohort, and unexplained MSAFP weighted to the original proportion in the population. The proportional sampling weights remove any bias introduced by oversampling at risk groups into the sub-cohort. For the comparison of mean vaginal HNP 1-3 levels between African Americans and non-Hispanic Whites, HNP 1-3 levels were transformed to the natural log scale and analyzed using SAS Surveymeans and Surveyreg procedures (32
HNP 1-3 levels were dichotomized using the median value (high ≥ median, low < median) from the distribution of HNP 1-3 levels in women who delivered at term and had normal MSAFP levels. The median was selected because effects were similar for women in the upper third and fourth quartile of HNP 1-3 levels. Race-specific analyses were performed to evaluate the bacterial vaginosis Nugent score and other maternal characteristics (age, education level, smoking, Medicaid Insurance status, week of pregnancy at study enrollment, and parity) in relation to high and low HNP 1-3 levels (SAS surveyfreq). The Rao-Scott chi-square test, used in complex survey designs (33
), was applied to test for statistically significant associations. Maternal characteristics associated with HNP 1-3 levels at p<0.10 were considered potential confounders to be included in final models.
Polytomous logistic regression models (34
) (SAS surveylogistic) tested associations between covariates (high/low HNP 1-3 levels, bacterial vaginosis, and race) and a three-level outcome variable: Term (referent group), spontaneous PTD, and medically indicated PTD. Results showed a statistically significant three-way interaction between race, HNP 1-3 levels, and bacterial vaginosis status. This led to the development of race-specific models to calculate unadjusted and adjusted odds ratios for the associations between HNP 1-3 levels and PTD subtypes. Models with a four-level outcome; i.e., same as above but spontaneous PTD subdivided into < 32 weeks and 32-36 weeks, explored the specificity of the relation between HNP 1-3 levels and timing of spontaneous PTD. Beta estimates from race-specific models were used to calculate probabilities of each PTD subtype for six groups defined by bacterial vaginosis (normal, intermediate, high) and HNP 1-3 (high, low) status and results were displayed in graphs.