Table of Contents
The study, which was initiated by the National Institute of Child Health and Human Development Network of Maternal–Fetal Medicine Units, was conducted at 11 network sites between 1994 and 1996 and was approved by the human-subjects review board at each institution. All women provided written informed consent.
To enroll a population with an increased likelihood of preterm delivery, we recruited women with singleton pregnancies who had a history of spontaneous preterm delivery between 20 and 36 weeks of gestation or bleeding in the second trimester of the current pregnancy — both risk factors for preterm delivery. We also enrolled a limited number of women without any risk factors to allow comparison of the frequency of contractions in women with and without risk factors for preterm delivery. The sample size was based on a univariate comparison of the mean hourly contraction rate at any given time between women delivering their infants before term and those delivering at term. We calculated that, for an assumed preterm-delivery rate between 10 percent and 15 percent, a sample of 300 would give the study approximately 80 percent power to detect a difference of 0.5 standard deviation (e.g., a difference of one contraction per hour if the standard deviation was no more than two contractions per hour9). Because it contained patients at high risk, this sample of 300 was likely to include women delivering before 35 weeks, when neonatal morbidity is increased. The goal was to recruit 50 to 70 women at low risk and 230 to 250 women at high risk.
The women were screened for eligibility before 22 weeks of gestation. All women underwent an ultrasound examination before enrollment to confirm or establish the duration of gestation. Women who had received or were scheduled to receive an ambulatory monitor or tocolytic medication or to undergo cerclage were not eligible, nor were women whose pregnancies were complicated by placenta previa or a major fetal anomaly detected by ultrasonography. Women who did not have telephones were not enrolled, because the transmission of data collected by the monitoring system required a telephone.
Once enrolled, women were evaluated every two to three weeks at outpatient visits so investigators could reinforce compliance with study protocols, perform examinations with a speculum to collect swabs of cervicovaginal fluid for the measurement of fibronectin, and perform digital and transvaginal ultrasound assessments of the cervix. Fetal fibronectin is an extracellular-matrix protein normally found at the interface between the fetal membranes and the maternal decidua; its presence in cervicovaginal secretions has been related to an increased risk of preterm delivery.11,13 The risk of preterm delivery has also been related to cervical length as measured by ultrasonography12 and to the Bishop score for digital examination of the cervix.14 The Bishop score is a composite measure of cervical length, dilatation, position, consistency, and the degree of descent (station) of the presenting part of the fetus. The results indicate the degree of readiness for labor; values from 0 to 4 indicate not ready for labor, and values from 9 to 13 indicate ready for labor. These procedures were performed by study personnel according to previously established protocols.11-14
The visits were scheduled to collect data at 22 to 24 weeks (visit 1), 25 to 26 weeks (visit 2), 27 to 28 weeks (visit 3), 29 to 30 weeks (visit 4), 31 to 32 weeks (visit 5), and 33 weeks or later (visit 6). Swabs of cervicovaginal fluid for detection of fibronectin were collected at all six visits. Digital and ultrasound assessments of the cervix were performed only at visits 1, 3, and 5.
Monitoring of Uterine Contractions
Research nurses at each center were trained in the use of a home contraction monitor (Healthdyne System 37, Matria, Marietta, Ga.). Monitors and assistance with data collection were provided by the manufacturer under a contract chosen by competitive bidding in accordance with guidelines of the Department of Health and Human Services.
A research nurse visited each woman at home to instruct her in the use of the monitor and transmission of data. Women were asked to record uterine activity for a minimum of one hour at least twice daily in two sessions at least two hours apart, one between 4 a.m. and 3:59 p.m. and the other between 4 p.m. and 3:59 a.m., on 2 or more days per week from enrollment to 28 weeks. After 28 weeks, two additional monitoring sessions per week were required. Data were transmitted immediately after collection to a dedicated computer system at the data-coordinating center at the George Washington University Biostatistics Center. Compliance with the protocol, including the time, date, and quality of the recordings for each woman, was monitored weekly. Data collected from medical records, interviews, and ultrasound examinations were transmitted weekly to the center.
The enrolled women were followed to delivery. The duration of pregnancy and the reasons for preterm delivery, if it occurred, were recorded. The health care providers, investigators, and patients were blinded to the results of testing. None of the tests included in the analysis were performed outside the study.
Analyses of Recordings
The recordings of uterine activity were analyzed according to a standard protocol by four research nurses. Any session lasting at least 30 minutes that produced a satisfactory tracing was reviewed. A research nurse first screened each recording to identify those that showed no contractions. All recordings with one or more contractions were read jointly by two nurses who were unaware of the outcome of the pregnancy. A contraction was defined as a deflection that occurred from a clear base line and that had a rounded peak and lasted from 40 to 120 seconds. The guiding principle in formulating this definition was that it include all deflections that any experienced obstetrician or obstetrical nurse would easily agree to identify as a contraction. Inverted, “double-peak,” and “camel-back” contractions were included. “Possible” contractions — that is, those that were subtle, had a variable base line, had a flat peak, or were accompanied by an artifact that obscured the above criteria — were not considered contractions. Regular audits, in which a sample of recordings with contractions was reanalyzed, were conducted throughout the study to ensure consistent interpretation. Although discordant interpretation occurred in 14 to 28 percent of recordings, the discrepancy was never greater than one contraction per hour and was evenly distributed between increased and decreased numbers of contractions. Ten percent of recordings with no contractions were also reviewed again as a quality-control measure; the rate of discordance among observers was 1.4 percent for these recordings.
Analyses of uterine activity were performed to identify relations with the duration of gestation, the time of day (4 a.m. to 3:59 p.m. [designated as daytime] vs. 4 p.m. to 3:59 a.m. [designated as nighttime]), risk status (high vs. low), and the timing of delivery (term delivery vs. spontaneous preterm delivery at less than 35 weeks of gestation, after preterm labor or preterm rupture of the membranes). Hourly contraction rates were calculated for each woman and for each week of gestation and were analyzed by a repeated-measures, random-effects model. Because each woman could contribute data only as long as her pregnancy continued, more data were collected for women who did not deliver prematurely. Contraction-frequency data from women who delivered their infants after 35 weeks were therefore compared with data collected between 24 weeks 0 days and 28 weeks 6 days for women who delivered between 29 weeks 0 days and 32 weeks 6 days, and with data collected between 24 weeks 0 days and 32 weeks 6 days for women who delivered between 33 weeks 0 days and 34 weeks 6 days.
The maximal daytime and nighttime contraction rates per hour for each interval of weeks (22 to 24, 25 to 26, 27 to 28, 29 to 30, 31 to 32, and 33 or more weeks) were evaluated as predictors of preterm delivery by logistic regression.
To compare uterine activity with the results of other tests as predictors of preterm delivery, we determined the sensitivity, specificity, and predictive values of each test at each gestational interval for spontaneous preterm delivery before 35 weeks of gestation. The results were dichotomized for the Bishop score (according to published criteria),14 cervical length (≤25 vs. >25 mm),12 and cervicovaginal fibronectin (<50 vs. ≥50 ng per milliliter).13 Contraction frequency was dichotomized as less than four or four or more contractions per hour, as previously described.3
We performed multivariate logistic-regression analyses that included maximal uterine-contraction frequency, presence or absence of fetal fibronectin in cervicovaginal secretions, cervical length, and Bishop score as dichotomous variables. Evaluation of collinearity, including the standard errors and the correlation of the parameter estimates, indicated that correlations between the results of the screening tests did not cause a problem with the logistic-regression analyses.
Receiver-operating-characteristic curves were constructed for each test for each two-week gestational interval. The curves were compared by calculating the area under the curve with use of a nonparametric approach based on the theory developed for generalized U statistics.15