Participants and recruitment

Nulliparous women with singleton pregnancies were recruited from hospitals affiliated with 8 clinical centers: Case Western University; Columbia University; Indiana University; University of Pittsburgh; Northwestern University; University of California at Irvine; University of Pennsylvania; and University of Utah. The Data Coordinating and Analysis Center is RTI International (Research Triangle Park, NC). Appendix A lists the individual hospitals that were involved in recruitment of the cohort. Each site’s local governing Institutional Review Board approved the nuMoM2b protocol and procedures.

Pregnant women who were attending affiliated obstetric clinics and who planned to deliver their baby at 1 of the clinical site hospitals were screened for eligibility in the first trimester of pregnancy. Women were recruited if they had a viable singleton gestation, were between 6 weeks + 0 days gestation and 13 weeks + 6 days’ gestation (based on a documented ultrasound crown-rump length measurement by a certified nuMoM2b sonographer; Table 1 provides the detailed dating criteria that were used), and had no previous pregnancy that lasted ≥20 weeks based on self-report. Study staff confirmed gestational age of any previous deliveries or pregnancy loss through a review of medical records, if available. In addition, recollection of a birthweight for a previous pregnancy of ≥360 g was considered evidence of a pregnancy that had lasted at least 20 weeks.

Exclusion criteria were maternal age <13 years, history of ≥3 spontaneous abortions, likely fatal fetal malformation evident before enrollment, known fetal aneuploidy, assisted reproduction with a donor oocyte, multifetal reduction, participation in an intervention study that was anticipated to influence maternal or fetal outcomes, previous enrollment in nuMoM2b, plan to terminate the pregnancy, or inability to provide consent. Trained study personnel screened for eligibility, explained the study, and administered informed consent. The study was explained in the participant’s native language (including Spanish and Vietnamese). Participants in the nuMoM2b study also were offered participation in Network-wide substudies of sleep disordered breathing, sleep actigraphy, and fetal adrenal gland ultrasound scanning. These nested cohort studies will be reported separately.

Study standardization and design

A common protocol and manual of operations were used for all aspects of the study at all sites. Obstetric care and newborn care were delivered according to local practices and standards at the individual clinical site hospitals. Trained and credentialed study personnel performed all study procedures. Responsibility for general study procedure training rested with the site principal investigator and study coordinator. Formally certified sonographers performed ultrasound studies. A local certification was required for crown-rump length and fetal biometry measurements that included images on at least 3 women that showed proper measurements as defined locally by a Network site expert. Measurements of cervical length and uterine artery Doppler velocimetry were centrally certified by national experts ( Appendix B ).

Study visits, information collected, specimens, and procedures

Study visits were not part of clinical care. Consequently, research data were not made available to the clinical care provider, except for results of sonographic examination for predefined criteria. Study visit timing was chosen to be spread across gestation, to occur before the time of most PTBs, and generally to correlate with times in prenatal care when women normally see caregivers: first trimester, early-mid second trimester, and late second–early third trimester. The timing to correspond to regular clinical care visits was chosen to facilitate ultimate translation into clinical care if predictive markers were to be established. Study visits generally lasted 1-2 hours. Study adherence was maximized by compensation of participants for their time and effort. The exact amount and types of compensation (eg, car seats) were decided by each site. Table 2 provides an overview of interview data collection and clinical measurements; Table 3 provides a list of standardized questionnaire instruments, and Table 4 provides a summary of the biospecimens that were collected by study visit. Visit 1 occurred between 6 + 0 and 13 + 6 weeks’ gestation. Blood pressure, height, weight, and waist, hip, and neck circumference were measured. Detailed interviews were performed to collect demographic characteristics, medical history, physical activity documentation, experience with nausea and vomiting of pregnancy, medication and supplement use, alcohol, tobacco and drug use, nutritional intake, perceived stress, reactions to racism, symptoms of depression, motivation for and intendedness of pregnancy, pregnancy knowledge, and relationship with a partner/father of the baby. In addition, questionnaires regarding sleep, trait anxiety, perceived social support, and difficulties in pregnancy were taken home after visit 1 to be completed at home and returned. Blood and urine samples were obtained. Separate cervical and vaginal fluid samples were obtained by speculum examination for fetal fibronectin and stored for other future assays, such as cytokine and microbiologic analyses. The participant self-collected additional samples of vaginal fluid for fetal fibronectin assays and for storage

Study visit 2 occurred between 16 + 0 and 21 + 6 weeks’ gestation and at least 4 weeks after visit 1 had occurred. Clinical measurements of blood pressure and weight were repeated, and interviews were conducted to gather information on health literacy, perception of experiences with racism, additional demographic characteristics, current physical activity, further details on family history and updates on exposures to alcohol, drugs, and tobacco, and current habits. Updates to record the participants’ experience with emesis and nausea during pregnancy, medications, and relationships with a partner/father of the baby were obtained. A questionnaire on resilience was given to the participant to complete at home. Blood and urine were obtained, as were self-collected vaginal fluid samples for fetal fibronectin assay and storage. Sonographic assessment of fetal biometry, cervical length, and uterine artery Doppler measurements were performed.

Study visit 3 occurred between 22 + 0 and 29 + 6 weeks’ gestation and at least 4 weeks after visit 2. Clinical measures were repeated along with additional interview questionnaires. A survey of sleep patterns and habits, pregnancy experience, and difficulties in pregnancy were completed by the participant. Blood, urine, and self-collected vaginal fluid samples were obtained. Similarly, sonographic assessment of fetal biometry and cervical length were repeated; uterine artery Doppler measurement was repeated, if the study at the second study visit had demonstrated a diastolic notch.

If the participant had a clinical amniocentesis or chorionic villus sample collection, a separate informed consent was administered to collect any excess sample from those procedures to add to the specimen bank. In addition, if a pregnancy loss occurred before 20 + 0 weeks, research personnel collected products of conception if possible.

Criteria for informing the patient’s caregiver of study findings included identification of a major fetal structural malformation, hydrops, fetal death, estimated fetal weight <5th percentile (at second or third ultrasound visit), oligohydramnios (a maximal vertical pocket <2 cm or amniotic fluid index <5 cm), a cervical length <15 mm at <28 weeks’ gestation, fetal bradycardia or tachycardia, or placenta previa or vasa previa found at the third study visit. Given that the study was designed as an observational cohort in which we did not want to increase overtreatment based on information that would have been obtained only at a study visit and wanted to optimize the chance that women with potentially pathologic findings could have their health care modified appropriately, we chose to inform caregivers of information that had been obtained only when it was determined a priori that the knowledge was most clearly necessary to provide adequate clinical care (eg, the discovery of an intrauterine fetal death, estimated fetal weight <5th percentile, and/or cervical length <15 mm). This guided our consensus selection of the trigger parameters as noted earlier.

At the time of delivery, an interview was conducted to update medical history, medication and supplement use, alcohol, tobacco, and drug exposures, perceived reasons for labor admission, and mode of delivery and to collect the intended method of infant feeding. Maternal blood samples were drawn after admission but not later than 1 hour after delivery. We allowed for the collection of maternal blood just after delivery for women who were admitted and delivered before maternal blood could be collected. This decision was made to optimize the chance that maternal blood could be obtained in the peripartum period, even if a woman delivered so quickly that maternal blood could not be obtained before birth. After delivery of the placenta, an umbilical cord blood sample was obtained for DNA, and the placenta, membranes, and umbilical cord were refrigerated until collection of samples. Placenta, membrane, and cord samples were taken for DNA and histologic study. Attempts were made to obtain these samples within 30 minutes of placental delivery; however, collection up to 72 hours after delivery was considered acceptable. When collection and processing within 30 minutes was possible, additional placenta and membrane samples for RNA expression, protein and metabolites, epigenetics, and additional cord blood for plasma were collected if possible ( Table 4 ). These samples were flash frozen in liquid nitrogen with the exception of the samples for RNA expression that were processed with the RNAlater (ThermoFisher Scientific, Inc, Waltham, MA). Appendix C describes details of delivery sample processing. Finally, a neonatal saliva sample for DNA was collected for women with certain APOs if the cord blood sample and the placenta sample for fetal DNA had not been collected.

Samples were all stored locally in –80°C freezers after being processed. Formalin-fixed samples were stored at room temperature. Full processing details are available on request from the Data Coordinating and Analysis Center. Data were entered into a secure web-based data capture system. Multiple data quality checks were performed to ensure high-fidelity information in the database. Specimens for assay of fetal fibronectin were shipped to Hologic, Inc (Sunnyvale, CA). All other specimens were shipped to and stored at a biorepository at Fisher BioServices, Inc (Rockville, MD).

At least 30 days after delivery, a trained certified chart abstractor assessed all participant medical records to record final birth outcomes and any readmissions to the hospital. Certification for chart abstraction was done locally at the Network sites based on study guidelines. The general topics that were abstracted were medical history; prenatal care and medication use; prenatal laboratory and diagnostic studies; maternal and neonatal delivery hospitalization medications, diagnoses, and outcomes; and any other maternal hospitalizations.

Standard survey instruments

Standardized and validated survey instruments were chosen in consultation with experts in the field ( Table 3 ). Instruments were given preference if they had been used and/or validated in previous pregnancy populations or cohorts. Attempts were made to choose 1 instrument to represent a domain of interest theorized to be associated with APOs. Accordingly, we chose surveys to measure perceived stress (the Perceived Stress Scale), the amount of social support and personal resiliency that might buffer external stress (the Multi-Dimensional Scale of Social Support and the Connor-Davidson Resiliency scale), symptoms of depression (the Edinburg Postnatal Depression survey), and anxiety (the trait subscale of the State-Trait Anxiety Inventory). Pregnancy-specific feelings were assessed with the Pregnancy Experiences survey, and perceptions of discrimination were assessed with the Krieger Racism survey. To adequately analyze the association of sleep on APOs, we used 4 instruments that are concerned with different sleep domains: The Berlin Questionnaire to assess for the possibility of sleep apnea, the Women’s Health Initiative Insomnia Rating Scale to ascertain insomnia symptoms, the Epworth Sleepiness Scale to document daytime sleepiness, and the Restless Leg Syndrome Diagnostic Criteria survey. Last, health literacy was assessed with the short version of the REALM instrument.

Outcomes and measures

The primary outcome of the study was defined as pregnancy ending for any cause at <37 + 0 weeks’ gestation. Key study hypotheses involved the outcomes of spontaneous PTB, preeclampsia, and fetal growth restriction.

PTB was defined as delivery of a liveborn or stillborn infant for any cause between 20 + 0 and 36 + 6 weeks’ gestation. Spontaneous PTB was defined as delivery that occurred subsequent to spontaneous onset of preterm labor or premature rupture of the membranes (PROM) or fetal membrane prolapse. Preterm labor was defined as spontaneous uterine contractions (>6 per hour documented by tocodynamometry or maternal history) with the onset before membrane rupture that led to delivery, documented cervical change of at least 1 cm dilation or effacement during the admission, or dilation >2 cm or effacement >80% on admission for contractions. PROM was defined as spontaneous rupture of the membranes before the onset of contractions, regardless of subsequent labor augmentation or cesarean delivery. Rupture of the membranes had to be documented by visible leaking of amniotic fluid from the cervix, the presence of vaginal indigo carmine after intraamniotic instillation, or the presence of any 2 of the following occurrences: pooling of fluid in the vaginal vault, positive nitrazine test results, positive ferning of dried vaginal fluid that was observed microscopically, or a positive commercially available biochemical test for PROM. Fetal membrane prolapse was defined by spontaneous descent of the fetal membranes to or past the external cervical os in the absence of uterine contractions (>6 contractions per hour documented by tocodynamometry or by maternal history), PROM, maternal fever or uterine tenderness, chorioamnionitis (clinical or amniocentesis diagnosis), or abruptio placentae, regardless of the placement of a cervical suture (cerclage). Women who delivered an infant preterm who did not meet any of those criteria were categorized as spontaneous PTB otherwise unspecified.

Indicated PTB was defined as delivery after induction or cesarean delivery between 20 + 0 and 36 + 6 weeks’ gestation. The indication for the delivery was recorded as pregnancy-associated hypertension (including preeclampsia), fetal growth restriction, placental abruption, placenta previa, chorioamnionitis, abnormal fetal test results, fetal conditions that included congenital anomalies, maternal medical conditions, “other” indicated PTB, or “no documented indication.”

Abortion (spontaneous pregnancy loss at <20 weeks’ gestation) was defined as the delivery of a liveborn or fetus who experienced fetal death for any cause before 20 + 0 weeks’ gestation. These were classified as “early” if delivery occurred at <14 + 0 weeks and “late” if it occurred at >14 + 0 weeks’ gestation. Term birth was defined as delivery of a liveborn or stillborn infant for any cause ≥37 + 0 weeks’ gestation.

Other maternal outcomes that were obtained were hypertension (mild, severe), chronic hypertension, baseline and preeclamptic proteinuria (mild, severe), preeclampsia (also eclampsia, HELLP (hemolysis, elevated liver enzymes, and low platelet count) syndrome, and mild and severe preeclampsia), superimposed preeclampsia, gestational hypertension, abruptio placentae, placenta previa, clinical chorioamnionitis, labor induction, labor augmentation, cesarean delivery (scheduled, unplanned, nonurgent, urgent, reason indicated), diabetes mellitus and gestational diabetes categorized by White’s classification, and other outcomes such as deep venous thrombosis, sepsis, acute respiratory distress syndrome, amniotic fluid abnormalities, cerebral vascular accident, severe anemia (hematocrit ≤20% and treated with transfusion), massive transfusion (≥6 units of packed red cells), intensive care unit admission, surgery that was required other than for delivery of the infant, cesarean hysterectomy, and death. Quality control checks (via reabstraction) were performed for primary outcomes by the site principal investigator on a random selection of charts with and without APOs.

Fetus-newborn outcomes included embryonic death, fetal death at <20 + 0 weeks’ gestation, stillbirth or live birth ≥20 + 0 weeks’ gestation, anomalies/malformations, fetal growth restriction, respiratory distress syndrome, ventilator support, continuous positive airway pressure, transient tachypnea of the newborn infant, oxygen therapy, confirmed neonatal sepsis (early onset, late onset), suspected neonatal sepsis (early onset, late onset), necrotizing enterocolitis (stage 1, 2, or 3), intraventricular hemorrhage (grade I, II, III, or IV), chronic lung disease/bronchopulmonary dysplasia, newborn seizures, retinopathy of prematurity, and infant discharge status.

Statistical considerations

NuMoM2b was designed as a prospective cohort study. The target recruitment was 10,000 nulliparous women with singleton pregnancies who would be enrolled in the first trimester. It was estimated that this would yield 1100 PTBs at <37 weeks’ gestation, 330 PTBs at <34 weeks’ gestation, 700 cases of preeclampsia, 300 cases of fetal growth restriction with the use of a strict definition of <3rd percentile, and 70 cases of stillbirth. Power estimates were computed for the detection of a range of odds ratios with a 4:1 ratio of control subjects to cases, which is suitable for nested cohorts and conservative for the full cohort. The power was calculated for Pearson χ ² tests with a 2-sided α = .05, assuming 1100 cases (eg, all PTBs), 770 cases (eg, spontaneous PTBs), 550 cases (eg, white PTBs), 220 cases (eg, PTBs in another racial group), a 100 cases (ie, some other small subset of APOs), and 5% and 20% exposure among the control subjects. With 1100 cases, there would be >80% power to detect a modest effect (odds ratio, >1.25) of a common exposure (>20%); for a rare exposure (5%), the power would be >80% for an odds ratio as low as 1.5. For 100 cases, power would remain >80% for odds ratios of approximately 2.0 and 3.0 with 20% and 5% exposed among the control subjects, respectively.

Standard parametric/nonparametric and asymptotic/exact methods for descriptive and inferential analysis would be used for continuous, binary, ordinal, and polytomous variables to describe characteristics of the cohort and nested subcohorts, control subjects, and cases and to estimate effects and to draw inferences regarding the associations of exposures and predictors with PTB and other APOs, with suitable stratification and modeling adjustments for confounding and assessments of effect modification.

With regard to prediction, we assessed the models based on predictive accuracy (calibration and discrimination). To assess calibration, we examined calibration plots and performed Hosmer-Lemeshow tests. To assess discrimination, we examined risk distributions among cases and control subjects and corresponding true- and false-positive rates for various cutoffs through the use of integrated predictiveness and classification curves. We also used the traditional receiver operating characteristic curve analysis for the area under the curve.