In this study of a regionally based sample of ELGANs, survival increased and rates of major developmental impairment detectable at 1 year adjusted age decreased during the 1990's. During this time there was a trend toward increased provision of intensive care to ELGANs. Although evidence exits that increased provision of intensive care can lead to an increased rate of cerebral palsy among survivors , our study provides reassurance about the consequences of intensive care for ELGANs who are born in a tertiary level obstetric center. Our objective was to describe the change in frequency of adverse neurodevelopmental outcome that accompanied the increase in survival of extremely low gestational age newborns during the late 1990's. Because infants who are born outside of our tertiary obstetric referral center do not receive the full benefits of specialized obstetric care, such infants were excluded from our analysis.
There have been few studies of mortality in extremely low gestational age babies and to our knowledge no studies of a geographically based cohort in North America during the 1990's. The survival rate described here is higher than that of two cohorts born in Australia and the British Isles. Doyle et al report a survival rate of 56% (at 5 years of age) in an Australian cohort of infants born at 23–27 weeks gestational age in 1991 to 1992 . In the study by Wood et al, the survival rate among infants born at 23 to 25 weeks in the British Isles in 1995, was 29% . For comparable groups of infants in our study the survival rates, respectively, were 67% and 50%. The higher survival rates reported here might be attributable to several factors, including our exclusion of infants with congenital anomalies, and, more importantly, to the more frequent provision of intensive care to our cohort (91% versus 84% and 86% versus 75% for comparable groups of infants) [24,25]. The survival rate reported here for infants born at 23–26 weeks (63% survival) is similar to that of a comparable group born 1996 to 1997 in The Netherlands (65% survival) .
Comparisons with regard to developmental impairment are more difficult due to differences in the definitions of developmental impairment and the ages at which assessments were obtained. The rates of major developmental impairment described here are similar to those for comparable groups born in Australia and the British Isles. In the Australian study , 20% of ELGANs had a major disability at five years of age (blindness, deafness requiring hearing aids, cerebral palsy affecting the ability to walk, or intelligence quotient score more than 2 SD below the mean), and in the British Isles study , 23% of survivors had a severe disability at 30 months (physical assistance needed to perform daily activities). Our rates for comparable groups were 18% and 23%, respectively. The higher rate of impairment reported by Rijken et al  (36%) is likely due to the classification of infants as impaired if they had cerebral palsy, BSID-MDI more than 2 SD below the mean, or a BSID psychomotor developmental index more than 2 SD below the mean; the last of these criteria was not used in our study. Further, the sample size for that study (36 infants) was small, resulting in a wide 95% confidence interval for the risk estimate (21% to 54%).
We are aware of only two studies of changes in survival and outcome over time of recently born cohorts of ELGANs [10,27]. Our finding that survival increased during the 1990's is consistent with the study by Hoestra et al, who found that survival for infants born at 23 and 24 weeks increased from 56% in 1991–95 to 75% in 1996–2000 . The higher survival rates than those reported here for comparable groups (30% in epoch 1 and 45% in epoch 2) could be due in part to their more aggressive delivery room care (only 2% of infants died in the delivery room) and to not including post discharge deaths. Our findings differ from those reported by Hintz et al. in a multicenter study of 839 infants born before 25 weeks gestation, in which survival rates for those born 1993–96 (40.4% survival) and 1996–99 (43.2% survival) were not significantly different . In that multicenter study, no decrease was found in the rate of neurodevelopmental impairment, and in fact the proportion of infants with MDI more than 2 SD below the mean increased from 40% in 1993–96 to 47% in 1996–99. Important differences between the two studies are the smaller sample size in the current study, which limits the precision of the estimates, and the use of a geographically based sample, which limits bias due to changes in referral patterns. Our finding of improved survival in the late 1990s is consistent with a recent report of by Wilson-Costello et al in which neurodevelopmental outcome at 20 months adjusted age improved in the interval 2000–02, as compared to 1990–99, in a single center cohort of 982 infants with birth weight less than 1000 grams .
In the study by Wilson-Costello, improved outcome appeared to be due in part to more frequent use of antenatal steroids and Cesarean delivery, and a decreased frequency of sepsis, severe cranial ultrasound abnormality, and postnatal steroid use . In the current study the rates of Cesarean delivery and severe cranial ultrasound abnormality did not change. The proportion of our cohort exposed to postnatal steroid (14%) was considerably lower than the range reported by others in the 1990s (40–79%) [3,24,26,27]. Data about sepsis were not available to us. It seems likely that some of the improvement in survival which we observed was due to increased provision of intensive care in the late 1990s. Increased use of antenatal steroids could have contributed both to improved survival as well as improved neurodevelopmental outcome.
The current study is based on the experience in a small geographic area; larger regional studies will be more informative about the outcome of ELGANs. Another limitation is that the estimates of gestational age that were used for this study were not based on prospectively developed criteria. A third limitation of this study is that data about chorioamnionitis and antenatal steroids were not available. These prenatal factors have been associated with, respectively, higher and lower rates of neurological impairment [29,30]. We are able to estimate based on other published studies from our nurseries that the exposure of very low birth weight infants to antenatal steroids increased from 19% prior to the NIH Consensus statement recommending their use  to 75% in 2000.
A potential source of bias in the present study derives from the use of the first edition of the BSID from 1990–93, and the use of the second edition of that test after October 1993. Adjustment of the MDIs  obtained with the first edition of the BSID classified 3 additional infants as neurodevelopmentally impaired in epoch 1. Another possible bias is expectation bias, referring to the possibility that clinicians identify neurological abnormalities more often in infants with abnormalities on cranial ultrasound.
Neurodevelopmental assessments at twelve months adjusted age are strongly predictive of the presence or absence of major developmental impairments persisting at least through eight years of age . However, Hack et al found that of 78 extremely low birth weight infants(i.e. birth weight . Thus it should be noted that 35% of infants with major neurodevelopmental impairments were attributable to MDI > 2 SD below the mean without other impairments. Impairments that are less severe than those studied here (e.g., learning disabilities) can influence a child's adaptive development, school success, and quality of life . In these regards our study is not as informative as would be a study of outcome at school age.