Methods overview: the available data
4.1 The first EPICure study – 1995
The EPICure studies were two prospectively collected observational studies of ex-tremely premature births. The first was carried out between March 1st and December 31st 1995, inclusive, with data collected on all births that were believed at the time of birth to have occurred between 20 and 25 completed weeks of gestation. There were no exclusion criteria, and the 276 hospitals where data were collected represented 100% of the services providing maternity care across the whole geographic area of Great Britain and Ireland. Follow-up of the cohort has occurred at 2.5, [56] 6,[57] 11 [58] and 16 years of age [180], with a further follow-up at 19 years currently under way.[181]
Data collection for the cohort was carried out in two phases. So as to avoid dupli-cation of effort, data relating to all births under 26 weeks gestational age were collected in conjunction with the the Confidential Enquiry into Stillbirths and Deaths in Infancy (CESDI) into log books containing only nine variables (identification number, hospital of birth, date of birth, gestation, birth weight, gender, number of fetuses, signs of life at delivery, and whether admitted to into a neonatal intensive care unit) located on the labour wards. For babies that were admitted onto a neonatal intensive care unit, a comprehensive “case record form” was then commenced. This comprised a 24 page form with questions relating to:
• entry criteria;
• maternal data;
• demographic data;
• status at discharge or death (divided into pre- or post-estimated date of delivery (EDD));
• drugs at death or discharge;
4.1 The first EPICure study – 1995
• nutrition and growth;
• ROP;
• head ultrasound scan data;
• other problems; and,
• contact information for follow-up.
A complete list of data items included is listed by Costeloe et al in a supplementary appendix to their paper,[54] and is replicated in appendix A.
All data from the study were collected by health care staff involved in the provision of care, with the labour ward log books being returned to the main research office in London on a monthly basis.1 Where gaps or data ambiguities were identified, one of the principal investigators visited local hospitals to review individual sets of case notes; only if the case review forms could not be completed following this were the data considered “missing”. Data were then single-entered into a computerised database by two study nurses, with double-entry being used for a random 10% sample in order to verify accuracy. This was reported as finding an error rate of 0.1% – 17 errors among a total of 15,280 data points entered.[54]
Available data
Of the 4,004 fetuses delivered during the study period, 843 live born babies were ad-mitted into a neonatal intensive care unit. Gestational age for these babies was then reassessed by the principle investigators using the criteria shown in table 4.1. Eight hundred and eleven babies were confirmed as having a gestational age that met the inclusion criteria, with subsequent attrition as shown previously in table 2.2.[54]
For this thesis, data were provided on the 668 babies who were born to an English mother in an English hospital.2 This data set had been used extensively prior to this point, and contained over 600 original and derived variables as well as information relating to subsequent follow-up of survivors at 2.5, 6 and 11 years of age. Cleaned variables were noted at a meeting between myself and Ms. Enid Hennessy in April
1Personal communication with Professors Kate Costeloe and Neil Marlow.
2Data were provided to me by Ms. Enid Hennessy, trial statistician, based at the Wolfson Institute of Preventive Medicine, Queen Mary, University of London.
Table 4.1: Hierarchy of criteria used to determine gestational age in the EPICure studies.
Adapted from Costeloe et al (2000 and 2012) [54, 55]. LMP: last menstrual period; USS:
ultrasound scan (performed at < 20 weeks in 1995, or the earliest scan performed at any time in 2006).; EDD: estimated date of delivery.
Situation in which EDD to be determined
EPICure EPICure 2
(1995) (2006)
If LMP and USS both present but differ by
>14 days
USS USS
If LMP and USS both present LMP USS
If LMP only present LMP LMP
If USS only present USS USS
If neither USS or LMP present clinical judgement clinical judgement
2011, and a copy of the original case record form annotated with variable names was provided. Data were formatted as a Stata [182] “.dta” file.
4.2 EPICure 2 – 2006
EPICure 2 was conducted during the entire calendar year of 2006, and was limited to births that occurred between 22+0 and 26+6 weeks gestation to mothers who were usually resident in England and delivered at one of the 182 maternity hospitals there, with local midwifery units individually contacted early the following year to check for any missed deliveries.[55, 68] As for the first EPICure study, data collection was per-formed by clinical staff on the labour wards in conjunction with the confidential enquiry investigating perinatal mortality – renamed in 2003 to the CEMACH[116] – with site visits once per month from the principle investigators to ensure data completeness.[55]
Each fetus delivered required a new notification pack to be opened; these contained a
“perinatal notification” form (“PN:E2” form), and bar coded sticky labels to aid unique identification. The birth was recorded in a labour ward log book using one of the sticky labels, specimens were collected from the placenta or umbilical cord where possible, and the labour ward staff completed as much of the detail in the PN:E2 form as possible.
For babies admitted to neonatal intensive care units, the form remained with the baby throughout the duration of the hospital stay for completion by subsequent care givers, else for still births or babies that had died in the delivery room the relevant parts of the form were completed and returned to the study office at the Homerton Hospital in London.[55, 68, 183] Consequently, many more detailed data were available for those