Materials and Methods

A retrospective cohort study was conducted. The cohort included all singleton deliveries in the Coombe Women and Infants University Hospital between January 1, 2000, and December 31, 2007. This is a tertiary care maternity hospital that is located in a deprived inner-city area. The hospital serves a catchment area with a population of >500,000. All women have a booking interview with a midwife at the first antenatal visit. Comprehensive records are available routinely for all deliveries of >24 weeks’ gestation or 500 g weight. This study was approved by the hospital’s research ethics committee.

Maternal sociodemographic, medical and obstetric characteristics, and perinatal outcomes were recorded from electronic antenatal and delivery records. Information on the following maternal characteristics was extracted: age at delivery, socioeconomic group, nationality, marital status, parity, planned/unplanned pregnancy, booking gestation, receipt of publicly funded antenatal care, smoking during pregnancy, alcohol use before pregnancy, and serologic status. Hepatitis B–positive serologic status was defined as a current or past positive result for hepatitis B surface antigen, core antibody, E antigen, or E antibody; hepatitis C was defined as current or past positive result for hepatitis C antibody or viral RNA; and human immunodeficiency virus was defined as a positive result for human immunodeficiency virus antigen or antibody.

Opiate-dependent women are referred routinely to a Drug Liaison Midwife who coordinates the care, liaising between the addiction and obstetric services. Multiple sources of ascertainment were used to identify women who were using methadone at the time of delivery, which included antenatal records, controlled drug registers, and prescription records. Methadone dose at delivery was recorded from controlled drug registers or prescription records. Controlled drug registers must be used to record all doses of methadone that are dispensed in the hospital. Women were categorized as methadone-exposed based on use at delivery.

Maternal and perinatal outcome measures included mode of delivery, gestational age at delivery, weight for gestational age, infant condition at birth (Apgar scores at 1 and 5 minutes and whether resuscitation was required), admission to neonatal special/intensive care, congenital anomalies, and perinatal death. Every woman had an ultrasound scan at the booking antenatal visit. Gestational age was estimated from the calculation based on first day of the last menstrual period; however, the booking ultrasound scan estimate was preferred if the dates were uncertain or there was a discrepancy of >7 days. Preterm birth was defined as the birth of a live baby at <37 weeks’ gestation. Very preterm birth was defined as the birth of a live baby <32 weeks’ gestation. Preterm births were further categorized according to whether preterm labor was spontaneous. Birthweight percentiles were calculated with the GROW-Centile calculator that yields percentiles that are customized for maternal height, weight, ethnicity, parity, infant sex, and delivery gestation. Population mean maternal height and weight were used for percentile calculations where data were missing for any individual pregnancy. Birthweights below the customized 10th percentile were defined as small-for-gestational age. Further neonatal outcome information was available for neonates who were admitted to the neonatal unit, including the main reason for admission and congenital anomaly diagnoses. The perinatal death register was used for the ascertainment of stillbirths and neonatal deaths, along with electronic delivery suite records. Stillbirth was defined as delivery of a baby that showed no signs of life ≥24 weeks’ gestation. Neonatal death was defined as the death of a baby within the first 7 days of life. Congenital anomalies were identified from records of a physical examination of all babies after delivery and from neonatal unit discharge records. Congenital anomalies were categorized as major, minor, or chromosomal based on the EUROCAT classification system (EUROCAT Central Registry, Newtownabbey, Northern Ireland).

During the study period, neonates were monitored initially for NAS on the postnatal ward with the use of a 17-item scoring system that had been adapted from Finnegan and Ehrlich by removing the moro reflex, sweating, and mottling items. Neonates with suspected NAS were admitted to the neonatal unit for assessment and treatment, if necessary. NAS was normally diagnosed if the neonate had ≥2 successive Finnegan scores of ≥8. Diagnosis of NAS was ascertained from multiple fields in the electronic neonatal discharge register, which included main diagnosis, central nervous system diagnosis, unresolved problems at discharge, other diagnoses, and discharge medications. Neonatal urine toxicologic screening for opiates (excluding methadone), benzodiazepines, cannabis, cocaine, amphetamines, and alcohol was performed when clinically indicated in the first days of life.

Basic descriptive statistics were used to describe maternal characteristics and perinatal outcomes for the exposed and unexposed groups. Differences in means were assessed with independent samples Student’s t test. Univariable and multivariable logistic regression analyses were used to determine crude and adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for the association between methadone exposure status and maternal characteristics, and between methadone exposure status and a range of perinatal outcomes. Only maternal characteristics that differed significantly between the exposed and unexposed groups were included in the final multivariable models. The analysis was repeated excluding women with diabetes mellitus during the pregnancy or a history of epilepsy. Separate univariable and multivariable logistic regression analyses were conducted to examine the determinants of NAS in methadone-exposed neonates. Variables that were found to have an impact on the incidence of NAS in the univariable analyses were included in the multivariable models. Likelihood ratio tests were used to test for evidence of interactions. Subgroup analyses were carried out examining the determinants of NAS excluding infants with urine toxicologic results that were positive for opiates, benzodiazepines, cannabis, or cocaine exposure. A logistic regression analysis with methadone dose group as an ordered categoric variable was used to test for a dose-response relationship between methadone and the occurrence of NAS.

STATA IC 10 (Stata Corp, College Station, TX) was used for all statistical analysis. The STATA command xtmelogit was used to carry out mixed effects logistic regression to generate odds ratios that were adjusted for the lack of independence in perinatal outcomes for women with >1 delivery in the study period. These regression models adjusted for the fact that some women may have had >1 infant in the cohort during the study period.