OBJECTIVES. It is unclear whether declines in neonatal and infant mortality have led to changes in the occurrence of cerebral palsy. We conducted a study to examine and investigate recent temporal changes in the prevalence of cerebral palsy in a population-based cohort of very preterm infants who were 24 to 30 weeks of gestational age.
METHODS. A population-based cohort of very preterm infants who were born between January 1, 1993, and December 31, 2002, was evaluated by the Perinatal Follow-up Program of Nova Scotia. Follow-up extended to age 2 years to ascertain the presence or absence of cerebral palsy and for overall survival. Infant survival and cerebral palsy rates were compared by year and also in two 5-year periods, 1993–1997 and 1998–2002. Logistic regression analyses were used to identify factors that potentially were responsible for temporal changes in cerebral palsy rates.
RESULTS. A total of 672 liveborn very preterm infants were born to mothers who resided in Nova Scotia between 1993 and 2002. Infant mortality among very preterm infants decreased from 256 per 1000 live births in 1993 to 114 per 1000 live births in 2002, whereas the cerebral palsy rates increased from 44.4 per 1000 live births in 1993 to 100.0 per 1000 live births in 2002. Low gestational age, postnatal dexamethasone use, patent ductus arteriosus, severe hyaline membrane disease, resuscitation in the delivery room, and intraventricular hemorrhage were associated with higher rates of cerebral palsy, whereas antenatal corticosteroid use was associated with a lower rate.
CONCLUSION. Cerebral palsy has increased substantially among very preterm infants in association with and possibly as a consequence of large declines in infant mortality.
Controversy exists regarding the relationship between the declining rate of neonatal mortality and the increasing cerebral palsy (CP) rate.1 Population-based studies from Scotland, Norway, and the United States2–4 have failed to show an increase in the prevalence of CP, although more recent studies from England, Sweden, Italy, and Finland5–8 have shown such an increase (more so in the very low birth weight or very preterm infant). Of the population-based studies that have evaluated very preterm infants, Tin et al9 showed that although the mortality rate decreased, the severe disability rate remained unchanged. However, specific details regarding CP were not included in this publication. The Victorian Infant Collaborative Study Group10 also showed a decrease in mortality from 1985–1987 to 1991–1992 but did not show an increased CP rate.
Temporal increases in CP prevalence may be anticipated given the large declines that have been observed in neonatal and infant mortality rates. With more infants, especially very preterm infants, surviving past infancy, it is not surprising to observe increases in population rates of CP and other disabilities. Conversely, factors that are known to be inversely associated with CP risk, such as antenatal corticosteroids,11,12 have increased in frequency in recent years. Other factors that are associated with higher rates of CP, such as multiple births and preterm births, have increased, however. We, therefore, conducted a population-based study to evaluate temporal changes in the prevalence of CP and the rate of infant death in a birth cohort of 24- to 30-week infants who were born between 1993 and 2002.
Nova Scotia is a province on the east coast of Canada and has a population of 942691 (2001 census); there were 98755 births (including stillbirths) between January 1, 1993, and December 31, 2002. All liveborn very preterm infants who were 24 to 30 weeks’ gestational age, including all delivery room deaths, and born to mothers who were resident of Nova Scotia were enrolled in the Perinatal Follow-up Program of Nova Scotia. The majority (85%) of very preterm infants were born at the 1 pediatric tertiary care hospital, the Izaak Walton Killam (IWK) Health Centre, 8% were transferred to the IWK for some of their neonatal care, and 7% received their neonatal care in other hospitals. Two Nova Scotia mothers were visiting outside Nova Scotia and delivered in tertiary care hospitals in Alberta and Florida. Early neonatal deaths in hospitals other than the IWK were tracked through the provincial registry of deaths. Each infant was entered in the Perinatal Follow-up Program database, which included details about maternal illnesses and procedures; newborn illnesses and procedures; demographic information; and long-term outcome data about neurodevelopmental performance, sensory impairments, and neuropsychological test results.
The assignment of gestational age for very preterm infants was based on 1 of the following 4 criteria (in order of priority): conception date, maternal last menstrual period date, early ultrasound assessment, or a clinical assessment (either a Ballard assessment13 performed by the housestaff or clinical assessments performed by the attending neonatologist). The only conflicts were between the ultrasound assessments and gestational age on the basis of the maternal last menstrual period. In such cases, ultrasound estimates were used when the difference between menstrual-based gestational age and ultrasound was >10 days.12
To identify CP, all surviving infants were enrolled in the Perinatal Follow-up Program and were evaluated by a neonatologist or developmental pediatrician who performed a general physical and neurodevelopmental examination at 0, 4, 8, 12, 18, and 24 months’ corrected gestational age (corrected for gestational age at birth using the mother’s expected date of delivery). Children who received a diagnosis of or were suspected of having CP were examined by a pediatric neurologist to confirm or exclude the diagnosis. CP was defined as a disorder of control of movement or posture secondary to a nonprogressive brain lesion.14
For the purposes of the present study, a liveborn infant was defined as an infant who was ≥24 weeks and displaying any signs of life at birth, whereas total births included all births at ≥24 weeks, either liveborn or stillborn. A stillbirth was defined as a fetal death before birth in a fetus who was 24 to 30 weeks (the gestational age was the gestation at birth). Infant death was defined as the death of a very preterm infant (24–30 weeks’ gestational age) before 1 year’s corrected gestational age. Total mortality included stillbirths and infant deaths. Gestational age–specific CP and/or mortality rates were calculated using traditional and fetuses at risk approaches.15
Logistic regression was used to determine crude and adjusted temporal trends in CP. Risk factors for CP were entered into an initial model along with calendar period (1998–2002 vs 1993–1997), and a final model for period effects was estimated using a stepwise backward elimination process. Cystic periventricular leukomalacia was not included in the model because it occurs commonly in CP and might realistically be considered a part of the disorder.
Between January 1, 1993, and December 31, 2002, there were 672 liveborn infants who were 24 to 30 weeks’ gestational age and born to women who were resident in Nova Scotia. The rate of very preterm infants in Nova Scotia was 6.8 per 1000 live births. Gestational age was assigned using the date of conception in 11 (2%), menstrual dates in 504 (75%), ultrasonographic gestational age in 142 (21%), and a clinical estimate of gestation in 15 (2%). There were 111 infant deaths (97 <28 days of age and 14 between 28 days and 1 year) and 561 survivors. Of the 561 survivors, 66 had CP and 21 (3.1% of the initial cohort of 672) were lost to follow-up. There was no change in the stillbirth rate between 1993 and 2002 (P = .59; Table 1) but a significant decline in infant mortality (P = .003; Table 1). Table 1 also shows that the prevalence of CP per 1000 live births increased between 1993 and 2002 (P = .002). This increase in prevalence also was significant when the denominator was restricted to survivors only (59.7 per 1000 survivors in 1993 increasing to 112.9 per 1000 survivors in 2002; P = .007). Neither CP and infant death (P = .74) nor CP and fetal-infant death (P = .46) showed a change between 1993 and 2002.
Table 2 shows the results of logistic regression analysis. The final logistic regression model included gestational age <28 weeks versus >28 to 30 weeks, postnatal dexamethasone use, patent ductus arteriosus, severe hyaline membrane disease, resuscitation in the delivery room, grades 3 and 4 intraventricular hemorrhage, and antenatal corticosteroid use (Table 3). Both crude and adjusted models showed that CP increased between 1993–1997 and 1998–2002. Factors that were associated with CP could not explain the temporal increase in CP rates (Table 3).
Our study shows that the prevalence of CP increased among a population-based cohort of very preterm infants between 1993 and 2002. This increase in CP was matched by a corresponding decline in infant mortality rates. Particularly noteworthy was the absence of any temporal change in the frequency of combined outcome of CP or infant death. Finally, a logistic regression model that was based on risk factors for CP could not explain the temporal increase in CP rates.
For accurately addressing the issue of the relationship between increasing CP rates and declining mortality rates among high-risk infants, high-quality data from the appropriate high-risk group are essential. Three issues need to be addressed: (1) defining the appropriate high-risk group, (2) eliminating bias from hospital-based data, and (3) reducing bias from large losses to follow-up.
With regard to defining the appropriate high-risk group, 2 overlapping groups, namely very low birth weight infants (<1500 g) and very preterm infants (generally <31 weeks’ gestational age) have been the focus of study. The advantage of the former is that birth weight is easy to obtain and highly reproducible, whereas gestational age often is difficult to obtain and may not always be accurate. However, it should be noted that very low birth weight infants have an estimated fetal growth rate of 75% of the expected rate and 26% are small for dates.16 Therefore, defining cohorts of high-risk infants on the basis of birth weight alone is confusing because low birth weight can be the consequence of low gestational age, poor fetal growth or both. It therefore is preferable to define the birth cohort using gestational age criteria.
The second issue is that most studies of high-risk infants tend to use hospital-specific data because of the ease of following such a cohort compared with following a complete geographically based cohort. Nevertheless, strong recommendations to report outcome for complete regionally based populations have been made to eliminate referral bias.17–20
Third, many centers have relatively low follow-up rates (<85%) because of the difficulty of following a complete cohort of children for long periods of time. This bias leads to underestimations of the true rate of children with severe disabilities. Tin et al21 showed a higher rate of disability among children whose parents failed to keep multiple follow-up appointments compared with those who routinely returned for follow-up appointments. Wolke et al22 noted that the general developmental quotient at 5 months was lower in those who failed to show for a subsequent assessment at 20 months compared with those who showed at 20 months. Hille et al23 noted that preterm children with severe disabilities or those who required special education failed to respond to requests for long-term follow-up. These data indicate that it is imperative to place resources in follow-up programs to track as close to 100% of the high-risk group as possible. Certainly, follow-up rates of 85% or less are thought to be inadequate for assessing CP and disability rates.21
The Perinatal Follow-up Program of Nova Scotia meets all 3 conditions to answer this important question because it uses a gestational age cutoff in a population-based cohort with a very low loss to follow-up rate (3% loss). Therefore, the data in our study address the issue of increasing CP rates associated with reduced mortality.
Some comment is required regarding the variables that were noted to be associated with either a higher or a lower prevalence of CP in the regression model. Antenatal corticosteroids predicted a lower CP rate, which is not surprising. Meta-analysis shows that antenatal steroids are associated with a lower incidence of intraventricular hemorrhage and a trend toward a lower long-term disability rate.24 Severe hyaline membrane disease25 and severe intraventricular hemorrhage26 also are known risk factors for CP. Postnatal dexamethasone has been shown to be an important risk factor and possibly even a cause of CP.27 Finally, the latter era of birth (1998–2002 vs 1993–1997) predicted higher rates of CP (P = .004) even after adjustment for known risk factors. If era of birth had not been significantly associated with CP in the adjusted model, then the determinants that were entered in the regression model would have been said to explain the temporal increase in CP. However, because era of birth continued to be significant, even after adjustment for all risk factors, the declining infant mortality rate has to be considered a probable explanation for the increasing CP rate.
Whether the rising rate of CP in very preterm infants means a higher overall prevalence of CP in the entire population is unclear. It is possible that obstetric interventions that result in deliveries at earlier gestational ages select infants who are at a greater underlying risk for CP. Therefore, rather than being born at a later gestational age and developing CP, it may be that these infants now are being delivered at earlier gestational ages and thus the CP among very preterm infants is increasing. Alternatively, it is possible that some of the secular demographic trends in pregnancy such as older maternal age or greater use of reproductive technology may have an influence on the overall prevalence of CP and what is seen in preterm infants is part of a larger population trend. Population-based CP registers are better suited to address this possibility.
The current data are derived from assessments that were performed up to 24 months’ corrected gestational age, which could underestimate the true prevalence of CP. This is not a serious problem because the diagnostic protocol used was the same throughout the duration of this study; therefore, the underestimation would be similar in both eras. Furthermore, diagnosis at this age is very effective at identifying children with CP as noted by the Victorian Infant Collaborative Study Group, a population-based follow-up of extremely low birth weight infants that noted changes in the diagnosis of CP between 2 and 5 years of age occurring in only 3 of 209 survivors.28 In the current study, of the 66 children with CP, 50 had the diagnosis suspected from abnormal neurologic findings on a previous assessment. An additional 95 infants had suspicious neurologic findings that resolved in all but 2 infants, who remained hypotonic and delayed in the gross motor domain when they were evaluated for their final assessment in the Perinatal Follow-up Program.
We have shown that recent declines in infant mortality among very preterm infants coincide with increases in prevalence of CP among such infants. This is consistent with other studies that show that technologic advances in recent years have resulted in reduction in death at the expense of higher rates of neurodevelopmental disability.5,29
Dr Joseph is supported by a Peter Lougheed New Investigator award from the Canadian Institutes of Health Research.
We acknowledge the assistance of the Reproductive Care Program of Nova Scotia and in particular John Fahey in providing data regarding total births and stillbirths in Nova Scotia. The assistance of the staff in the Perinatal Follow-up Program was essential in evaluating children and collecting data that were used in this article. Most important, we extend our appreciation to all of the children and their families who contributed time and patience so that this article could be written.
- Accepted July 7, 2006.
- Address correspondence to Michael J. Vincer, MD, IWK Health Centre 5850/5980 University Ave, PO Box 9700, Halifax, Nova Scotia, Canada B3K 6R8. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
Dr Vincer conceived the study, analyzed and interpreted the data, and was the principal author; Dr Joseph provided suggestions for statistical analysis; Drs Vincer and Wood acquired long-term outcome data of infants who were enrolled in the Perinatal Follow-up Program; all authors contributed to the design and analysis of the study and approved the final version of the manuscript; and Dr Vincer is the guarantor.
- ↵Kudrjavcev T, Schoenberg BS, Kurland LT, Groover RV. Cerebral palsy: trends in incidence and changes in concurrent neonatal mortality: Rochester, MN, 1950–1976. Neurol.1983;33 :1433– 1438
- ↵Colver AF, Gibson M, Hey EN, Jarvis SN, Mackie PC, Richmond S. Increasing rates of cerebral palsy across the severity spectrum in north-east England 1964–1993. The North of England Collaborative Cerebral Palsy Survey. Arch Dis Child Fetal Neonatal Ed.2000;83 :F7– F12
- ↵Tin W, Wariyar U, Hey E. Changing prognosis for babies less than 28 weeks’ gestation in the north of England between 1983 and 1994. Br Med J.1997;314 :107– 111
- ↵Arnold CC, Kramer MS, Hobbs CA, McLean FH, Usher RH. Very low birth weight: a problematic cohort for epidemiologic studies of very small or immature neonates. Am J Epidemiol.1991;134 :604– 613
- ↵Escobar GJ, Littenberg B, Petitti DB. Outcome among surviving very low birthweight infants: a meta-analysis. Arch Dis Child.1991;66 :204– 211
- Heinonen KM. Assessing the effectiveness of care for very low birth weight infants: do we really need population-based data? Pediatrics.1989;84 :912– 913
- Mutch LMM, Johnson MA, Morley R. Follow up studies: design, organisation, and analysis. Arch Dis Child.1989;64 :1394– 1402
- ↵Tin W, Fritz S, Wariyar U, Hey E. Outcome of very preterm birth: children reviewed with ease at 2 years differ from those followed up with difficulty. Arch Dis Child Fetal Neonatal Ed.1998;79 :F83– F87
- ↵Hille ET, Elbertse L, Gravenhorst JB, Brand R, Verloove-Vanhorick SP. Nonresponse bias in a follow-up study of 19-year-old adolescents born as preterm infants. Pediatrics2005;116(5) . Available at: www.pediatrics.org/cgi/content/full/116/5/e662
- ↵Crowley P. Prophylactic corticosteroids for preterm birth. Cochrane Database Syst Rev. 2000;CD000065
- ↵Murphy DJ, Hope PL, Johnson A. Neonatal risk factors for cerebral palsy in very preterm babies: case-control study. Br Med J.1997;314 :404– 408
- ↵Ancel PY, Livinec F, Larroque B, et al. Cerebral palsy among very preterm children in relation to gestational age and neonatal ultrasound abnormalities: the EPIPAGE cohort study. Pediatrics.2006;117 :828– 835
- ↵Ross G, Lipper EG, Auld PAM. Consistency and change in the development of premature infants weighing less than 1,501 grams at birth. Pediatrics.1985;76 :885– 891
- ↵Wilson-Costello D, Friedman H, Minich N, Fanaroff AA, Hack M. Improved survival rates with increased neurodevelopmental disability for extremely low birth weight infants in the 1990s. Pediatrics.2005;115 :997– 1003
- Copyright © 2006 by the American Academy of Pediatrics