Association of Fetal Growth Restriction With Neurocognitive Function After Repeated Antenatal Betamethasone Treatment vs Placebo

This secondary analysis of a randomized clinical trial evaluates whether fetal growth restriction is associated with the effects of repeated doses of antenatal betamethasone on neurocognitive function in a cohort of children followed up in midchildhood.

The aim of this study is to assess whether there are additional health gains, without adverse effects, in 75 mid-childhood, after exposure to repeat doses of antenatal corticosteroids (ACS). We will study New 76 Zealand children at 6-8 years corrected age who took part in the ACTORDS trial (Australasian 77 Collaborative Randomised Trial of Repeat Doses of Corticosteroids), which randomised mothers at risk 78 of preterm birth at less than 32 weeks, after an initial course of ACS, to either weekly corticosteroids or 79 placebo. ACTORDS found that repeat ACS had clinically important beneficial effects on premature lung 80 disease. However, repeat ACS were also associated with a small negative effect on fetal growth and data 81 from animal studies suggests that increased fetal exposure to ACS may adversely affect brain 82 development and increase the long-term risk of metabolic and cardiovascular disease. Thus the short-term 83 benefits of repeat ACS could be offset by these other harms. There is currently no strong evidence 84 regarding the effects of repeat ACS beyond the age of 2 years. Therefore, we will study a range of health 85 outcomes as well as physiological variables known to be associated with the development of adult disease 86 in order to determine if there is overall benefit from repeat ACS at 6-8 years corrected age. 87

Specific Aims 88
The study consists of three components: 1) general health and neurosensory function (paediatric 89 assessment); 2) cognition and behaviour (psychological assessment); and 3) metabolic, endocrine and 90 cardiovascular function (physiological studies). Parts 1 and 2 will be conducted in the entire Australasian 91 ACTORDS cohort but part 3 will be conducted only in the New Zealand subgroup. More specifically this 92 study will investigate the following outcomes: 93

Study Hypothesis 95
The study hypothesis is that antenatal administration of repeat ACS to women who remain at risk of 96 preterm birth at less than 32 weeks gestation, after an initial course of ACS, has beneficial effects on their 97 children at 6-8 years corrected age with regard to the outcomes in proportion of babies born preterm is actually increasing, a trend that is consistent throughout the 106 developed world (Callaghan 2006;Craig 2002;Joseph 1998;Tucker 2004). The care of these small 107 babies consumes considerable health care resources and places significant stress on families. Prematurity 108 is responsible for 75% of neonatal deaths (Hack 1999) and is the leading cause of infant mortality 109 (Callaghan 2006). 110 ACS substantially reduce the occurrence and severity of respiratory distress syndrome (RDS), which is 111 the principle cause of neonatal mortality and morbidity (Roberts 2006 short and long-term adverse effects in various organ systems, although interpretation of these data is not 120 straightforward. 121 To date, human randomised studies of repeat ACS have not extended beyond 2 years corrected age. Two 122 of these studies found that repeat ACS were associated with small reductions in fetal growth (Crowther 123 2006; Wapner 2006), but the effects were rapidly reversed after birth and no differences in growth were 124 seen at 2 years corrected age. One trial found that repeat ACS may be associated with an increased risk of 125 cerebral palsy, although the result was not statistically significant (Wapner 2007). ACTORDS children 126 exposed to repeat ACS had increased attention difficulties at 2 years corrected age (Crowther 2007). 127 These findings justify caution with regard to the use of repeat ACS and long-term follow-up of 128 randomised cohorts is urgently needed before repeat ACS can be safely recommended for routine clinical 129 use, despite their proven short-term benefit. 130 ACTORDS will be the first and possibly the only randomised trial of repeat ACS to assess children at 131 school age. The Canadian MACS trial, which recently completed recruitment, is also planning long-term 132 follow-up but will assess children only at 5 years corrected age. However, many important outcomes, 133 such as lung function and educational achievement and learning cannot be readily assessed until school 134 age. This study of ACTORDS children will also be the first detailed investigation of the long-term 135 physiologic effects of repeat ACS in humans, which is another important area of concern. 136 2.2 Understanding the Developmental Origins of Adult Disease 137 The developmental origins of health and disease (DOHaD) has become an important paradigm in the 138 study of the pathogenesis of adult disease . A central tenet of the paradigm is 139 that organisms can undergo long-term structural or functional changes in early development in response 140 to nutritional and other environmental cues. These long-term effects may be mediated by epigenetic 141 changes in gene expression (nuclear or mitochondrial), altered stem cell allocation, or resetting of 142 homeostatic mechanisms (Barker 2006;Pike 2008). Early developmental responses can contribute to later 143 disease risk if there is a mismatch between fetal and adult environments, such that early phenotypic 144 adaptations limit an individual's ability to adequately respond to subsequent environmental challenges 145 (Godfrey 2007). 146 Early developmental adaptions appear to occur not only in response to adversity but also within the 147 normal range of growth and development (Barker 2006). For example, the inverse relationship between 148 birth weight and cardiovascular mortality is continuous throughout the normal birth weight range (Seckl 149 2001). Increasingly, early developmental factors are being linked to other chronic illnesses including 150 osteoporosis (Cooper 1997;Godfrey 2001), depression (Thompson 2001), schizophrenia (Gluckman, 151 Cutfield, 2005), and obstructive respiratory disease Shaheen 2004). There is now also 152 evidence to suggest that early developmental adaptations can be inherited by non-genomic mechanisms 153 . 154 An important potential mechanism by which environmental factors may cause long-term effects in utero 155 is fetal overexposure to maternal glucocorticoids. For example, in animals it has been shown that 156 maternal undernutrition impairs the placental barrier to cortisol, thereby exposing the fetus to excess 157 maternal glucocorticoids, despite normal circulating maternal cortisol levels. This in turn leads to lower 158 birth weight, altered postnatal hypothalamic-pituitary-adrenal (HPA) axis function and hypertension 159 (Edwards 1993;Langley-Evans 1997;Seckl 2001). Similar changes in the offspring occur following 160 maternal administration of synthetic corticosteroids, which are not metabolised by placental 11-beta-161 hydroxysteroid dehydrogenase (11 HSD) (Seckl 2001). 162 Given the potential association between corticosteroids and fetal programming it is important that repeat 163 ACS receive careful longitudinal evaluation. This study will provide the first human experimental 164 evidence of the metabolic and cardiovascular effects of repeat antenatal glucocorticoid exposure at school 165 age. These data may offer further insights into the mechanisms underlying the developmental origins of 166 disease. 167

Preterm Birth and the Role of Corticosteroids 169
Preterm babies are at high risk of RDS due to immature lung development. Incomplete alveolar 170 development limits the surface area available for gas exchange, while qualitative and quantitative 171 deficiencies in surfactant cause lung collapse (Roberts 2006). RDS affects 87% of infants less than 28 172 weeks, 64% at 29-30 weeks, 50% at 31-32 weeks and 20% at 33-34 weeks (Boggess 2005). It is the 173 principal cause of early neonatal mortality and morbidity and is an important risk factor for 174 neurodevelopmental impairment. RDS is associated with cerebroventricular haemorrhage and 175 periventricular leucomalacia, both of which contribute to cerebral palsy and other long-term disability. 176 Severe RDS has also been linked to a global decrease in brain volume on volumetric magnetic resonance 177 imaging (MRI) (Thompson 2007). 178 One of the most significant discoveries in perinatal medicine was the recognition that fetal exposure to 179 corticosteroids induces lung maturation, especially surfactant pathways, and thus reduces the incidence 180 and severity of RDS. Liggins  benefit was seen when administered <26 weeks gestation it is possible that this is a type 2 error given the 199 small number of babies in this subgroup and the wide confidence intervals. Because RDS is common <32 200 weeks gestation the absolute benefit of ACS was large, with a number needed to benefit (NNTB) of 5 and 201 9 when ACS were administered between 26-29 weeks and 30-32 weeks respectively. Between 32-33 202 * A complete analysis of the Auckland Steroid Trial data was never published. The paper that is widely quoted (Pediatrics 1972;50(4):515-25) included the results of only 287 women. The remaining data was published in various conference reports. All data was extracted for the most recent Cochrane systematic review (Roberts 2006). ** After the first 717 women were recruited into the Auckland Steroid Trial the dose of betamethasone was increased to a total 48mg but no additional benefit was found. weeks the NNTB was 15. Although ACS reduced the risk of RDS there was no effect on the risk of 203 chronic lung disease. Current evidence is that there is no adverse effect on fetal survival or neonatal 204 outcomes if delivery occurs within 7 days (Roberts 2006 shown that ACS continue to have beneficial respiratory effects at term (Stutchfield 2005). Nevertheless, 210 ACS are not generally used >34 weeks gestation because of the low incidence of RDS. 211

Time Window 212
The benefits of ACS appear to be short lived. The Cochrane systematic review of single course ACS 213 found that there was no benefit if the fetus remained undelivered 7 days after treatment with ACS (table  214 3-2) (Roberts 2006). Peak benefit occurred sometime after 24 hours but before 7 days. While caution is 215 required with post hoc analysis (Gates and Brocklehurst 2007) there is currently no evidence to support 216 benefit from ACS beyond 7 days. Furthermore, delaying delivery by more than 7 days following a course 217 of ACS may be harmful. In the Cochrane review such fetuses were smaller at birth (mean difference in 218 birth weight 147g) (Roberts 2006) and in another meta-analysis ( 95%CI 0.24-1) and cerebral palsy (RR0.6, 95%CI 0.34-1.03) (Roberts 2006). An observational study also 225 found that single course ACS were associated with a higher intelligence quotient (WISC-3 full scale mean 226 difference 6.2, 95%CI 0.8-11.6) at age 14 years in very low birth weight children (Doyle 2000). However, 227 follow-up of a randomised cohort at age 6 years (304 children), found that ACS exposed children had 228 lower Raven Progressive Matrices scores (mean difference 1.2 or 0.3SD, p=0.05), which is a test of 229 general intelligence or abstract thinking (MacArthur 1982). The effect was greatest in boys (mean 230 difference 1.6, p=0.01). This finding could not be replicated in a Dutch follow-up study at 10-12 years 231 (Schmand 1990), although the study was small (90 children; follow-up rate 88%) and not powered to 232 detect clinically important differences in intelligent quotient (IQ). 233 In an observational study ACS increased systolic and diastolic blood pressure at age 14 years (Doyle 234 2000) but a randomised study showed no effect at age 6 years (Dalziel 2004). Childhood growth is 235 unaffected by single course ACS (Roberts 2006). Lung function has been assessed only in an 236 observational study but was also unaffected (Doyle 2000). 237

Long-term Outcomes 238
The use of single course ACS for preterm birth <34 weeks has not been shown to cause any long-term 239 clinical harm, at least into early adulthood. Follow-up of a Dutch randomised cohort at age 20 years (81 240 neonatal survivors; follow-up rate 74%) found that ACS had no effect on growth, blood pressure, 241 cognition, psychopathology, puberty onset, sexual function, education, or socioeconomic status (Dessens 242 2000).  Dalziel 2007). However, ACS exposed subjects showed evidence of insulin resistance on oral 247 glucose challenge, although there was no difference in the incidence of diabetes (Dalziel, Walker, 2005). 248 The clinical significance of this finding for later adulthood remains unknown, but insulin resistance is a 249 risk factor for the later development of diabetes and cardiovascular disease. 250

Maternal Outcomes 251
The Cochrane review found that maternal outcomes are unaffected by single course ACS (Roberts 2006

Trials of Repeat Course Antenatal Corticosteroids 257
Liggins and Howie recognised in the first report of the Auckland Steroid Trial (Liggins 1972) that ACS 258 did not appear to have benefit for RDS beyond 7 days and suggested that there may be benefit in giving 259 repeat doses. In some centres this became standard practice and by the late 1990s the use of repeat doses 260 was quite widespread (Brocklehurst 1999; Quinlivan, Evans, 1998) despite a lack of strong evidence to 261 support their use. Although some animal studies suggested that repeat doses may be more effective, for 262 example, improving lung function in sheep (Ikegami 1997), reports of efficacy in humans were 263 conflicting (Banks 1999;Elimian 2000). A body of evidence also started to emerge that raised concerns 264 about the safety of exposing fetuses to higher doses of glucocorticoids. Animal models demonstrated fetal 265 growth restriction (Ikegami 1997;Quinlivan, Archer, 1998), impaired cerebral development (Dunlop 266 1997; Huang 1999), altered hypothalamic-pituitary-adrenal axis (Ikegami 1997) and emphasematous 267 alveolar development (Tschanz 1995;Willet 2001). Observational data in humans also pointed to 268 impaired fetal growth (Abbasi 2000;Banks 1999;French 1999), increased perinatal mortality (Banks 269 1999), altered brain maturation (Modi 2001), poorer neurodevelopmental outcome (Spinillo 2004), and 270 hyperactivity (French 2004 which used a rescue protocol, was terminated early because of increased severity of RDS in the treatment 290 arm (Peltoniemi 2007). Another trial by Obstetrix Medical Group, USA, which gives the second course of 291 ACS after 14 days, is ongoing (NCT00201643). 292

Neonatal Outcomes 293
The current Cochrane systematic review of repeat ACS (2028 women) includes ACTORDS, the two USA 294 trials and the pilot studies by Aghajafari and McEvoy (Crowther 2011). This review found that the 295 administration of weekly ACS to women at risk of preterm birth <34 weeks gestation reduced all RDS, 296 severe RDS and a composite outcome of serious morbidity, including chronic lung disease, severe 297 cerebroventricular haemorrhage, necrotising enterocolitis, and periventricular leucomalacia (table 3-4). 298 However, there was no effect on individual morbidity outcomes. In keeping with the decrease in RDS, 299 repeat ACS also reduced oxygen and surfactant use, and patent ductus arteriosus (PDA) requiring 300 treatment but not other respiratory support parameters. There were no differences in fetal and neonatal 301 mortality, neonatal or maternal infection, or chronic lung disease. 302 There were no differences in mean birth weight, head circumference, or length. However, two trials 305 showed small reductions in fetal growth in the repeat ACS exposed groups: ACTORDS found lower 306 mean Z-scores for birth weight and head circumference (absolute difference 0.13 and 0.17 respectively); 307 the MFMN trial found an increase in small for gestational age (<10 th percentile) babies, which in post hoc 308 analysis was confined to those infants receiving 4 or more study doses. Both trials reported rapid neonatal 309 catch-up growth so that no differences in size were seen at time of discharge. 310

ACTORDS 312
The Australasian Collaborative Trial of Repeat Doses of Corticosteroids (ACTORDS) assessed whether 313 repeat doses of ACS, given to women who remained at risk of preterm birth at less than 32 weeks 314 gestation, reduced the risk of neonatal respiratory disease without adverse effects ). 315 Women were recruited to the trial if they had ongoing risk of preterm birth 7 or more days after an initial 316 course of ACS and were <32 weeks gestation. Women were excluded if they were in the second stage of 317 labour, had chorioamnionitis needing urgent delivery, had mature lung development, or if further 318 corticosteroid therapy was judged to be essential. Approximately 44% of the eligible women who were 319 asked to take part in the study gave consent. Participants were randomised to weekly doses of 320 betamethasone 11.4mg (celestone chronodose) (n = 489) or saline placebo (n = 493), continuing up to 32 321 weeks gestation. Randomisation was stratified by centre, gestation, and number of fetuses. Of 1146 322 infants alive at randomisation there were 1090 survivors to initial discharge home, of whom 342 were 323 from New Zealand. The mean gestational age of both groups at initial steroids was 26 weeks, and at trial 324 entry and birth it was 28 and 32 weeks respectively. 325 ACTORDS found that repeat doses of ACS had a beneficial effect on respiratory morbidity and PDA 326 ( The hypothalamic-pituitary-adrenal (HPA) axis was studied in two subgroups of infants. In Adelaide cord 332 serum cortisol concentration was measured in 67 infants and salivary cortisol was studied in 51 infants 333 (Ashwood 2006). No significant difference in mean cord serum cortisol concentration was found between 334 those exposed to repeat ACS (34 infants) and those exposed to a single course of ACS in morning plasma cortisol concentration was seen between those exposed to repeat ACS (30 infants) and 340 those exposed to single course ACS (33 infants) on day 2 of life (249 v. 265 mol/L) (Battin 2007). Some 341 of these babies also received a metyrapone challenge, which tests the whole HPA axis. However, no 342 differences in plasma cortisol or adrenocorticotropin (ACTH) concentrations were seen between the 343 groups 3 hours after administration of metyrapone (17 v. 9 infants). 344 Mothers in the repeat ACS group had a higher rate of caesarean section (RR 1.13, CI95% 1.02-1.24) but 346 the overall rate of caesarean section in the trial was high at 62%. There were no differences in other 347 maternal outcomes. 348

Preschool Outcomes 349
To date only two trials have reported outcomes beyond the neonatal period. ACTORDS found that infants 350 exposed to repeat ACS were more likely to be rated by parents as intense ( The MFMN 2 year follow-up study also found a trend to increased risk of cerebral palsy in the repeat 363 ACS group, although the result was not statistically significant (2.9% v 0.5%, RR 5.7 [CI95% 0.7-46.7], 364 p=0.12) and most of the cases of cerebral palsy occurred in those children that received 4 or more study 365 doses (Wapner 2007). This may be a type 1 error given the low absolute event count (6 v. 1 cases of 366 cerebral palsy) and the use of pregnancy as the denominator rather than babies, which is known to 367 overestimate the incidence of outcomes (Gates 2007). 368

Current Evidence for the Beneficial and Adverse Effects of Repeat ACS 369
Corticosteroids can have potent effects in multiple organ systems due to their action as transcription 370 regulators and also through the inhibition of cell growth and DNA replication. Normally the fetus is 371 protected from exposure to maternal cortisol by placental inactivation via 11 HSD. However, synthetic 372 steroids, such as betamethasone, are a poor substrate for 11 HSD and readily cross the placenta. Fetal 373 tissues are particularly sensitive to corticosteroids, whether natural or synthetic, and corticosteroid 374 exposure in early development can potentially cause long-term physiological changes. Therefore, it is 375 essential that studies of repeat ACS include long-term follow-up of health outcomes. 376 The clinical evaluation of ACS is complicated by the fact that prematurity itself is associated with long-377 term physiological effects including elevated blood pressure ( animal studies to support a dose-response relationship, although this cannot be assumed to apply in 385 human pregnancy (Aghajafari, Murphy, Matthews, 2002). 386

Effect on Brain and Neurodevelopment 387
The reductions in premature lung disease associated with the use of repeat ACS could potentially lead to 388 improved neurodevelopmental outcomes, although this benefit may be offset by possible adverse effects 389 of repeat ACS on fetal brain development. In sheep, ACS have multiple effects on brain growth, 390 including reduced brain size (Huang 1999), delayed myelination (Dunlop 1997;Huang 2001) and 391 decreased long-term brain mass . These effects are greatest following repeat ACS (Huang 392 1999; . Guinea pigs and rats exposed to high doses of ACS have altered behaviour into 393 adulthood, including increased anxiety, and impaired learning and memory (Owen 2005;Welberg 2001 ACS had no effect on developmental quotients, developmental testing at this age has limited predictive 402 value for later cognitive ability (Rose 2003). Therefore, assessment at school age is needed to assess the 403 overall effect of repeat ACS on neurodevelopment. 404

Effect on Lung Function 405
Experimental evidence in both animals and humans shows that improvements in respiratory compliance 406 and pulmonary function can be repetitively induced despite prior treatment with corticosteroids (Ikegami 407 1997;McEvoy 2000;Stewart 1998). This suggests that repeat ACS may cause structural maturation of 408 the fetal lung in addition to stimulation of surfactant pathways. However, animal models have shown that 409 this can also lead to emphysematous-like alveolae (Tschanz 1995;Willet 2001). The effect of repeat ACS 410 on later lung function in humans has not been assessed in randomised studies. 411

Effect on Growth & Body Composition 412
A dose-dependent reduction in growth has been well documented in fetal lambs exposed to increasing 413 doses of antenatal betamethasone (Fowden 1996;Ikegami 1997). Permanent changes in body composition 414 have been seen in rats exposed to prenatal dexamethasone, with increased susceptibility to later obesity 415 (Cleasby 2003). Bone cortical thickness is also markedly reduced in adult rats after antenatal exposure to 416 dexamethasone (Swolin-Eide 2002). 417 In humans, repeat ACS have been associated with small reductions in fetal growth but rapid catch-up 418 growth is achieved post-natally (section 3.3.1). At 2-3 years of age no differences in size have been 419 detected in both observational (French 1999) and randomised (Crowther 2007;Wapner 2007) studies. 420 There are no randomised data on the effects of repeat ACS on growth and body composition at school 421 age. 422

Effect on Blood Pressure 423
ACS can cause long-term cardiovascular changes. Sheep exposed to a single course of ACS in early 424 gestation have life-long hypertension (Dodic 1998), as do rats exposed to dexamethasone in late gestation 425 (Levitt 1996). 426 In humans, single course ACS have been associated with late elevation of blood pressure in observational 427 (Doyle 2000) but not randomised studies (Dalziel 2004;Dalziel, Walker, 2005). One observational study 428 showed an association between repeat ACS and elevated neonatal blood pressure (Mildenhall 2006). 429 However, in randomised studies repeat ACS have not been shown to affect blood pressure at 2 years 430 corrected age (Crowther 2007;Wapner 2007). Nevertheless, long-term follow is needed to exclude 431 possible late cardiovascular effects of repeat ACS. 432

Effect on Glucose Homeostasis 433
Insulin resistance describes a state in which target tissues have decreased responsiveness to the actions of 434 insulin on glucose metabolism. It is associated with an increased risk of subsequent glucose intolerance 435 and diabetes (Barker 2005). In rats ACS lead to long-term impairment of glucose tolerance in the 436 offspring (Seckl 2001). Similarly, sheep exposed to repeat doses of betamethasone have altered insulin 437 secretion and increased hepatic glucose-6-phosphatase activity as adults; changes that are consistent with 438 impaired glucose tolerance in later life . 439 In humans, prematurity itself causes insulin resistance and this effect is compounded by other factors such 440 as rapid early weight gain (Regan 2006). A single course of ACS has also been shown to increase this risk 441 in early adulthood (section 3.2.3). It is possible that repeat ACS may have an even greater effect on the 442 insulin-glucose axis. 443

Effect on Cortisol 444
Administration of corticosteroids results in transient down-regulation of the hypothalamic-pituitary-445 adrenal (HPA) axis, due to negative feedback at the level of the hypothalamus and pituitary. The duration 446 of this suppression after postnatal dexamethasone in preterm babies is often several weeks, and may be 447 sufficiently severe to warrant cortisol supplementation in times of stress (Ford 1997;Ng 1997). In 448 experimental animals, the effects of ACS exposure are complex. Rats exposed in utero to dexamethasone 449 have evidence of long-term up-regulation of the HPA axis into adulthood, at least in part due to 450 permanent down-regulation of the hippocampal glucocorticoid receptors leading to resetting of the "set 451 point" of the negative feedback loop. Sheep exposed in utero to repeat doses of betamethasone have up-452 regulation of the HPA axis at one year of age (early adulthood) (Sloboda 2002) but this wanes over time, 453 so that by 3 years (middle age) there is evidence of HPA axis suppression (Sloboda 2007). 454 The effect of ACS on HPA axis function in humans has not been well studied. In two small observational 455 studies of preterm infants, antenatal betamethasone reduced the cortisol response to a physiological 456 stressor at 1 week (Davis 2004;Davis 2006) and at four to six weeks of age (Davis 2006). Long-term 457 follow-up of subjects from the Auckland Steroid Trial did not show differences in fasting plasma cortisol 458 levels at age 30 years (Dalziel, Walker, 2005). 459 Data on the effect of repeat ACS on the HPA axis in humans are sparse. Studies in subgroups of babies 460 from ACTORDS found that repeat ACS were associated with a small temporary reduction in basal 461 salivary cortisol concentration in the first week after birth and decreased stress responses on day 3 of life 462 (section 3.3.2). However another study found no differences in plasma cortisol and ACTH or response to 463 metyrapone challenge at 2-3 days of age (Battin 2007). No other randomised data are available on the 464 effects of repeat ACS on HPA axis function in humans. 465

Summary 466
This study of early school-age outcomes in ACTORDS children is justified for the following reasons: 467 Repeat doses of antenatal corticosteroids (ACS) provide clinical important respiratory and other 468 benefits for babies born preterm. 469 There are no randomised data regarding the effects of repeat ACS beyond 2 years corrected age, 470 but other evidence from animal and human non-randomised studies show that repeat ACS may be 471 harmful in the long-term with regard to neurodevelopment, growth, and cardiovascular and 472 metabolic disease. 473 These possible effects of repeat ACS cannot be fully evaluated before school age and long-term 474 follow is essential to assess overall safety. 475 ACTORDS is currently the largest published randomised trial of repeat ACS and the only trial 476 planning school-age follow-up. 477 The results of this study will directly influence clinical practice and potentially improve outcomes 478 for preterm babies. 479 The additional physiological studies to be performed in the New Zealand subgroup will provide 480 important direct human experimental data regarding the glucocorticoid hypothesis for the 481 developmental origins of adult disease. 482

Design 484
This is a follow-up study from a randomised placebo controlled trial. Children will be assessed without 485 reference to any previous results and study personnel will be blinded to treatment group. All ages will be 486 corrected for gestation at birth as even at 8 years correction for prematurity results in elimination of a 487 small but potentially important bias in cognitive test scores (Rickards 1989). 488

Subjects 489
All surviving children from the ACTORDS trial alive at 6-8 years corrected who reside in New Zealand 490 will be eligible and will be invited to participate in this study. There will be no exclusion criteria. 491 Women were eligible for the ACTORDS trial if they had a singleton, twin or triplet pregnancy at less than 492 32 weeks gestation, 7 or more days after an initial course of ACS, and had no contraindications to the use 493 of further corticosteroids (section 3.3.2). 494 The initial consent for the ACTORDS trial included follow-up to 2-years corrected age. New consent will 495 be sought for this study. It has received ethical approval from the Multi-Regional Ethics Committee of 496 New Zealand (MEC/07/07/101). 497 In the New Zealand subgroup 342 infants survived to discharge from hospital and at 2 years corrected age 498 337 were known to still reside in New Zealand. The ACTORDS study group, based in Adelaide, has had 499 ongoing contact with these families since the 2-year follow-up and has maintained a central database. 500 Currently there are 307 New Zealand children on the database with 153 in Auckland, 94 elsewhere in the 501 North Island and 60 in the South Island. The children are currently aged between 3 to 8 years. Families 502 will be contacted when the child is between 6 to 8 years corrected age and invited to join the study. 503 Assessments will begin in early 2008 and continue until the end of 2010 when the youngest children turn 504 6 years. 505

Primary Outcome 506
The primary outcome will be survival free of sensorineural disability. This will include: cerebral palsy, 507 hearing impairment requiring a hearing aid, blindness or IQ less than 1 SD below the mean. 508 Cerebral palsy (CP) will be defined as a non-progressive loss of motor function with disordered tone or 509 tendon reflexes. The presence of mild upper motor neuron signs in the absence of functional limitation 510 will not be diagnosed as CP. CP will be classified by type and limb distribution (Cans 2007) and graded 511 into three levels of severity based on gross motor function: 512 Mild CP: ambulant with little limitation 513 Moderate CP: ambulant with substantial limitation 514 Severe CP: non-ambulant. 515 Gross motor function will also be classification according to the system of Palisano (1997). 516 Blindness will be defined as visual acuity less than 6/60 in the best eye after best possible correction. 517 The level of neurosensory disability will be further classified as follows: 518 Severe disability: any of severe cerebral palsy, IQ less than -3 SD below the mean, or blindness 519 Moderate disability: any of moderate cerebral palsy, deafness, or IQ from -3 SD to -2 SD below the 520 mean 521 Mild disability: mild cerebral palsy or IQ from -2 SD to -1 SD below the mean 522

Secondary Outcomes 523
Paediatric 524 Mortality 525 Growth: height, weight, head circumference, mid arm circumference (non-dominant), upper: lower 526 segment ratio 527 Sensorineural impairment including blindness, deafness and cerebral palsy as defined above 528 Movement ABC-2 score 529 Blood pressure as measured by an oscillometric device: systolic, diastolic and mean arterial pressure 530 and proportion in the hypertensive range (≥ 95 th percentile or 1.65 SD) 531 Lung function as measured by flow spirometry: FVC, FEV1, FEV1/FVC ratio, PEF, FEF25-75, 532 FEF50, FEF25 as % predicted for age, gender, and height 533 Health related quality of life: parents will complete the Child Health Questionnaire (CHQ) and the 534 Multi-attribute Health Status (MAHS) questionnaire 535 Chronic illness, health services utilisation, and reasons for use 536

537
General cognitive ability as assessed by the Weschler Abbreviated Scale of Intelligence (WASI) 538 Attention & executive function as assessed by subtests from the Test of Everyday Attention for 539 Children (TEACh), the Rey Complex Figure (RCF), the Fruit Stroop task. Parents and teachers will 540 complete the Behavior Rating Inventory of Executive Function (BRIEF). 541 Memory & learning as assessed by the Rey Auditory Verbal Learning Test (RAVLT) 542 Visual perceptual skills as assessed by subtests of the Test of Visual-Perceptual Skills 3 rd edition 543 (TVPS-3) 544 Academic achievement as assessed by the Wide Range of Achievement Test (WRAT4) 545 Behaviour problems: parents and teachers will complete the Strength and Difficulties Questionnaire 546 (SDQ), and the Conner's ADHD/DSMIV Scales (CADS) 547 Oscillatory ambulatory blood pressure (ABP) monitoring, 24 hour recording: 549 24-hour, daytime, and nighttime mean ABPM parameters (systolic and diastolic BP, 550 mean arterial BP, heart rate) and Z-scores (gender and height specific) 551 Blood pressure load and nocturnal dipping 552 Frequently sampled IV glucose tolerance test (FSIGTT) with minimal model analysis (MINMOD): 553 Insulin sensitivity index (S I ) 554 Acute insulin response 555 Glucose effectiveness (Sg) 556 Glucose disappearance coefficient (Kg) 557 Whole body dual energy x-ray absorptiometry (DEXA): 558 Bone area (BA), bone mineral content (BMC), areal bone mineral density (aBMD) 559 BA for height and BMC for BA and height Z-scores (gender specific) 560 Fat mass (FM) and lean mass (LM), FM percentage, FM index (FM/height) 561 Hypothalmic-Pituitary-Adrenal (HPA) axis function: 562 Basal diurnal salivary cortisol and dehydroepiandrosterone (DHEA) 563 Fasting & stress plasma cortisol with FSIGTT 564 Renal function 565 Creatinine 566 Estimated creatinine clearance (CrCL, ml/min/1.73m 2 ) 567

Data management and analyses 568
Statistical analysis 569 Data will be entered into a database and analysed using SAS statistics software with the assistance of the 570 ACTORDS study statistician. Dichotomous outcome data will be contrasted by 2 analysis or log-571 binomial regressions to adjust for confounding variables, and continuous data by t-test or analysis of 572 variance (ANOVA) to adjust for confounding variables. Confounding variables will comprise 573 sociodemographic variables, such as ethnicity, language spoken at home, family structure, mother's 574 marital status, social class, and mother's and father's education, as well as gender. Adjustment will also 575 be necessary to allow for a small design effect caused by non-independence of children from multiple 576 pregnancies. P-values <0.05 will be considered statistically significant. 577

Power and Clinical Significance 578
The best estimate of the primary outcome is 87% survival free of neurosensory disability. This is based 579 on data from the 2-year ACTORDS follow-up. Assuming 90% follow-up and a 2-sided significance level 580 of 5%, the New Zealand subgroup (values for the total Australasian ACTORDS study cohort in brackets) 581 will have 80% power to detect differences in the primary outcome from 87% up to 96% or down to 75% 582 (92% to 80%); differences in cerebral palsy from 4% up to 12% or down to 0% (8% to 1%); and for 583 outcome variables that are continuous and normally distributed (measures of growth, blood pressure, 584 psychological tests), differences as small as 0.34SD (0.18SD). In the case of DEXA, if 200 children are 585 scanned, the study will have 80% power to detect differences of 0.4SD, assuming DEXA parameters are 586 normally distributed. The power of the study to detect larger differences would, of course, be higher. 587 For lung function the total Australasian cohort will be able to detect differences between groups of 3 to 588 4% for values of FVC and FEV 1 (Dalziel, Rea, 2006). 589 For insulin sensitivity, based on data in normal children of this age, the study will have 80% power to 590 detect a 10% difference if 50% of children agree to this test and a 20% difference if only 25% agree. 591 These minimum absolute differences are considered clinically important to detect. For IQ, 0.3 SD 592 represents a 5-point difference in IQ. A 0.3 SD decrease in head circumference in childhood is associated 593 with a 0.7 reduction of IQ points (Gale 2004). For BMI, a 0.3 SD increase in childhood is associated with 594 an 8% increase in the hazard ratio for death from coronary heart disease (Eriksson 1999). For insulin 595 resistance, the risk of developing type 2 diabetes is reported to be 46% over 25 years in a high-risk 596 population (Martin 1992). 597

Structure of assessments 598
Wherever possible, in order to maximise participation rates, all assessments will occur on the same day. 599 The exceptions will occur when DEXA is not available on the assessment site, necessitating a separate 600 appointment, ambulatory blood pressure measurement, which will occur overnight after the other 601 assessments, and salivary cortisol, which will be collected over 3 typical days. 602 Children agreeing to the glucose tolerance test will be asked to fast overnight, and this test will be done 603 first in the early morning. The remaining physical assessments and DEXA will then be completed after 604 the child has had a break and breakfast. A longer lunchtime break will then be followed by the 605 psychological assessments. Children not consenting to blood tests will not be fasted and will start the 606 assessment later in the morning, but follow the same sequence. The entire glucose tolerance test, 607 including setting up, takes approximately 2h, during which the child watches a movie of their choice. 608 DEXA scanning takes <5 minutes. Physical assessment with the paediatrician, including blood pressure 609 and lung function, takes approximately 1 hour. The psychological assessment will take 2.5 to 3 hours, and 610 at least one break will be provided. 611

APPENDIX 1: PAEDIATRIC ASSESSMENT 613
Growth 614 Height and sitting height will be measured by a stadiometer to the nearest 0.1cm (bare feet), weight by 615 electronic scales to the nearest 0.1kg (minimal clothing) and head circumference and mid-arm 616 circumference (non dominant arm at the half-way point between the acromion to olecranon) by tape 617 measure to the nearest 0.1cm. Values will be computed, using corrected age, for the relevant percentile, 618 percent of median, and standard deviation scores from the British Growth Reference Cole 619 1998;. 620

Blood pressure 621
Blood pressure (BP) will be recorded in the right arm with the subject seated in a chair after a 10-minute 622 rest using a Dinamap automated oscillometric device. Three measurements will be made and the average 623 systolic, diastolic and mean arterial pressure will be converted to Z-scores (age, height and gender 624 specific) (Rosner 1993). If the child is hypertensive (>95 th percentile, or approximately >110/75mmHg at 625 6-8 years) a second set of measurements will be taken later in the assessment. American Heart 626 Association guidelines for the correct measurement of blood pressure and selection of cuff size will be 627 followed (Pickering 2005). As a rule of thumb the cuff width should be approximately 75% of arm length 628 (acromion to olecranon distance). 629

Lung function 630
Ventilatory capacity will be measured by spirometry using the EasyOne 2001 flow spirometer (NDD 631 Technologies), which conforms to both American Thoracic Society (ATS) and European Thoracic 632 Society (ERS) diagnostic spirometry standards. The following parameters will be assessed using forced 633 expiration: forced vital capacity (FVC), forced expiratory volume in 1 second (FEV 1 ), peak expiratory 634 flow (PEF), flow rate at 50% and 25% vital capacity (FEF 50 , FEF 25 ), mean flow between 25-75% of vital 635 capacity (FEF 25-75% ). ATS 1994 spirometry guidelines will be observed (American Thoracic Society 636 1995), which defines a satisfactory test as 3 technically acceptable trials including 2 reproducible trials 637 (<0.2L variation in the 2 largest FVC and FEV1 measures). Therefore, a minimum of three forced 638 expiratory manoeuvres will be performed. 639 The largest FVC and FEV1 from all acceptable trials will be recorded, even if they do not come from the 640 same trial, as recommended by the ATS. Other measures will be obtained from the best trial (largest sum 641 of FVC and FEV1) that meets acceptability criteria. Results will be expressed as percent predicted for 642 gender, age and height for healthy, normal birth weight, Caucasian Australian children (Hibbert 1989). 643 The use of an exponential regression form (see appendix 6.4) has ensured that percent predicted is 644 independent of height for each of the lung function parameters. Raw data will not be corrected for 645 ethnicity but later adjustment may be required if ethnic composition differs markedly between study 646 groups. 647 The EasyOne spirometer has been shown to have very stable calibration in clinical use and to not require 648 regular volume calibration (Pe´rez-Padilla 2006; Walters 2006). However, a study investigator will serve 649 as a biological control to monitor spirometer performance on a regular basis. FVC or FEV1 values outside 650 a 95% confidence interval will indicate the need for volumetric calibration (Johns 2003 p21-22). 651

Neurosensory function 652
Visual acuity will be measured using a 3m Snellen chart testing both binocular and monocular vision. 653 Corrective lenses will be worn and the child's eye will be occluded by patch or by the examiner. Visual 654 defects will be deemed 'severe disability' if felt to be subjectively so by the paediatrician but 'legally 655 blind' will be defined as visual acuity <6/60 in the best eye. Further assessment of vision will be advised 656 if fewer than 4 of 6 correct letters are read on the 6/9 line or if there is a 2-line difference between the 657 eyes, as recommended by the American Academy of Pediatrics (2003). 658 Otoscopy will be performed and hearing will be screened using low and high tone whispered numbers. 659 Children will be referred for audiology if they are considered to have language delay or if deafness is 660 suspected. Deafness is defined as hearing impairment requiring hearing aids. 661 Cerebral palsy, defined as a non-progressive loss of motor function with disordered tone or tendon 662 reflexes, will be determined by neurological examination and classified and graded as outlined above. 663 Motor function will also be assessed using the second edition of the Movement Assessment Battery for 664 Children (Movement ABC-2) (Henderson 2007). Children born preterm are at increased risk of motor 665 difficulties, even in the absence of signs of neurological impairment. The Movement ABC is a sensitive 666 test of motor dysfunction in this group (Jongmans 1998). Age adjusted percentiles and Z scores will be 667 determined for the three test components (manual dexterity, aiming and catching, and balance) and the 668 total test using the reference tables in the ABC-2 manual. The ABC-2 has been standardized in a large 669 sample of representative British children. Scores below the 15 th percentile indicate motor impairment and 670 those below the 5 th percentile are associated with severe motor coordination problems. 671

Pubertal Status 672
The onset of puberty will be determined by breast development in girls and testicular volume via an 673 orchidometer in boys (>4ml or more). Pubertal progression will be classified by Tanner stage (Marshall 674 1969; Marshall 1970). 675

General health 676
A paediatrician will formally assess all children by general history and physical examination to determine 677 the presence of any significant chronic illness. Data regarding hospital readmissions will be confirmed, 678 where necessary. The child's caregiver will be asked to complete a questionnaire relating to any 679 respiratory morbidity, history of illness or injury and health service utilization. 680

Health-related quality of life 681
Health-related quality of life will be measured using a paediatric adaptation of a multi-attribute health 682 status (MAHS) classification system (Saigal 1994). The MAHS classification system derives from the 683 oncology literature and describes both the type and severity of functional limitations according to seven 684 attributes: sensation, mobility, emotion, cognition, self-care, pain, and fertility. Each attribute has four or 685 five levels of function. The MAHS clearly assesses outcomes other than neurological function. 686 Children will also be assessed with the Australian Authorised Adaptation of the Child Health 687 Questionnaire (CHQ) (Waters 2000). The CHQ has recently been standardised on over 5000 Australian 688 children aged 5-18 years and provides an assessment of a child's psychosocial health, physical health and 689 well-being. 690

APPENDIX 2: PSYCHOLOGY ASSESSMENT 692
General cognitive ability 693 Children will be assessed using the Wechsler (1999) Abbreviated Scale of Intelligence (WASI). An 694 estimated IQ score reflecting general intellectual ability, will be derived from four subtests: Vocabulary, 695 Similarities, Block Design, and Matrix Reasoning. Each scale/index is age standardised. Intellectual 696 impairment will be classified as follows: 697 Mild intellectual impairment will be an IQ between 70 -84 (from -2 SD to < -1 SD) 698 Moderate intellectual impairment will be an IQ between 55 -69 (-3 SD to < -2 SD) 699 Severe intellectual impairment will be an IQ below 55 (< -3 SD). 700

Attention & Executive Function 701
Preterm children are at increased risk of attention difficulties and executive dysfunction. Both are 702 associated with diffuse white matter injury, which is the predominant form of brain injury in this 703 population. Attention and executive function will be assessed using the following age standardised tests: 704 Subtests from the Test of Everyday Attention for Children (TEACh) (Manly 1999) including Sky 705 Search (selective attention), Score (sustained attention), Sky Search Dual Task (divided attention) and 706 Creature Counting (shifting attention) 707 The Rey Complex Figure (Rey 1993), which assesses spatial organisation and strategic decision-708 making 709 The Fruit Stroop task (Archibald 1999), which assesses impulse control 710 711 In addition, parents and teachers will complete the Behavior Monitor) and scores are age and gender standardised. Internal consistency for the parent form of the 716 BRIEF has been found to be high, ranging from 0.80 to 0.98 (Gioia 2000). Clinical validity has been 717 supported with a variety of diagnostic groups. 718

Memory and Learning 719
The Rey Auditory Verbal Learning Test (RAVLT) (Rey 1964) will be administered to give multiple 720 measures of verbal memory and learning, including proactive inhibition, retention, encoding versus 721 retrieval, and subjective organization. 722

Educational Progress 727
Educational progress will be assessed using the Wide Range Achievement Test (WRAT4) (Wilkinson 728 2005). Three subtests will be administered: word reading, spelling, and arithmetic. Scale scores from 70-729 84 will define mild impairment in these educational domains, while scale scores <70 will define major 730 impairment. 731

Behavioural Problems 732
Children born preterm are at increased risk of attention-deficit hyperactivity disorder (ADHD). This will 733 be assessed using a specific ADHD diagnostic questionnaire, namely, the Conners' (SDQ) will be administered (Goodman 1997). The SDQ is a well-validated questionnaire that assesses 736 overall behaviour problems, emotional symptoms, hyperactivity/inattention, peer relationship problems, 737 and prosocial behaviour. 738 (usually less than 5 minutes) and the radiation dose is low and generally regarded as trivial (Njeh 1999;792 Sanchez 2005). Cortical bone composes 80% of the total skeletal mass, therefore, whole body DEXA 793 reflects predominantly cortical bone mass and dimensions (Leonard 2004). 794 BMC (g) depends on both the size and density of bones. However, DEXA cannot determine true bone 795 density as it measures only cross sectional bone area (BA, cm 2 ) and not bone volume. The increase in 796 BMC that occurs throughout childhood is due mainly to changes in bone size as true volumetric bone 797 density (vBMD, g/cm 3 ), as measured by quantitative computed tomography (QCT), is relatively constant 798 until puberty (Specker 2005). Interpretation of paediatric DEXA is complicated by the fact that large 799 differences can occur in body and bone size within and across different ages. Therefore, BMC values 800 must be adjusted for size but the BMC/BA index, also known as the areal bone mineral density (aBMD, 801 g/cm 2 ), provides only a crude adjustment with larger bones having an artificially inflated aBMD for a 802 given vBMD (Specker 2005). Therefore, total body BMC (TBBMC) should be interpreted in relation to 803 body size as well as BA Molgaard 1997;Prentice 1994). Gender, ethnicity and pubertal 804 status are also important factors affecting BMC, although pubertal status adds little to prediction of 805 TBBMC if adjustment is made for other variables including body size . When compared to 806 QCT, TBBMC-for-height provides a better estimate of bone strength than TBBMC-for-BA and TBBA-807 for-height is the best index of bone dimension (Leonard 2004). 808 Whole body DEXA will also be used to measure body composition including total body fat mass (TBFM,  809 g), lean mass (TBLM, g), fat percentage (FM%), fat distribution (android: gynoid ratio), and fat mass 810 index (FM/ height). DEXA body composition measurements show a close relationship to those obtained 811 by the four-compartment method, although small differences in FM% occur with body size and between 812 manufacturers (Sopher 2004). 813 DEXA results differ between manufacturer and machine. Therefore, a single type of machine will be 814 used, the Lunar Prodigy, which is available at the Liggins (Encore v 8.1) and St George's Radiology in 815 Christchurch (Encore v8.8). Both facilities have the paediatric software module. 816

Cortisol 817
Salivary cortisol measurements are widely used in paediatric research to assess hypopituitarity-pituitary-818 adrenal (HPA) axis function (Hanrahan 2006;Jessop 2007). They allow for determination of basal 819 cortisol levels in a stress-free manner in the home environment but can also be used to assess stress 820 responses as salivary cortisol levels peak within minutes of a plasma cortisol surge (Levine 2007;Vining 821 1983). Saliva and plasma cortisol concentrations are highly correlated (Aardal 1995;Gallagher 2006;Poll 822 2007; Vining 1983; Woodside 1991), especially when salivary cortisol is related to the unbound plasma 823 cortisol fraction (Kirschbaum 1989). Salivary cortisol concentrations are approximately 70% that of 824 unbound plasma cortisol (Rosmalen 2005) and are unaffected by salivary flow rate (Kirschbaum 1989). 825 Basal cortisol concentrations will be determined from saliva collected at home on days representing 826 normal activity, such as a school day. Morning fasting samples will be compared to early evening ones 827 (5pm) to allow assessment of the diurnal pattern. Children will be instructed to collect the morning 828 sample as soon as possible after waking. Because of day-to-day variations in cortisol secretion, samples 829 will be collected over 3 days and averaged. The caregiver will record the time of waking and sample 830 collection. Saliva samples will not be collected within half an hour of eating or brushing teeth. 831 Saliva will be collected by passive drool into Salivette tubes and children will be asked to collect a 832 minimum of 1ml per sample. Cotton swabs will not be used as we have shown that they achieve variable 833 cortisol recovery (unpublished data), a finding that has also been confirmed by other laboratories (Groschl 834 2006; Strazdins 2005). Caregivers will be instructed to freeze the Salivettes immediately after collection. 835 Once all 6 tubes have been collected they will be posted back to the study centre and then stored at -20 C. 836 We have confirmed that cortisol is stable in saliva at room temperature for up to 5 days and that cortisol 837 concentrations are unaffected by 2 freeze-thaw cycles (unpublished data). This has also been shown in 838 other studies (Aardal 1995;Clements 1998;Groschl 2001). 839 Cortisol will be assayed by mass spectrometry using a technique that has been well validated in our 840 laboratory. Mass spectrometry has the advantage of allowing simultaneous measurement of other 841 hormones and we will also assay dehydroepiandrosterone (DHEA) to allow calculation of the cortisol / 842 DHEA ratio. Abnormal cortisol / DHEA ratios have been shown to be have clinical predictive value for 843 psychopathology (Goodyer 2001). 844 In the subgroup of children who agree to undergo a FSIVGTT, we will measure plasma cortisol 845 concentrations at baseline (fasting) and also following the insulin bolus, which may provide further 846 information about HPA axis function in relation to physiological stress. 847

Renal Function 848
Plasma creatinine will be measured in the baseline blood of those children undergoing FSIVGTT via the 849 Liggins Hitachi autoanalyser. Creatinine clearance (ml/min/1.73m 2 ) will be estimated from height using 850 the Schwartz (1976) equation: 851 eGFR = kL/Scr (ml/min/1.73m 2 ) 852 L = height in cm. Scr = creatinine concentration in mg/dl. The value of k is 0.47 for children when using 853 the creatinase method of modern autoanalysers (Schwartz 2007). eGFR shows good correlation with 854 creatinine clearance (r=0.935) (Schwartz 1976). 855 856