Objective. To analyze whether the increasing survival of very low birth weight infants during the 1980s and 1990s has increased the risk of cerebral palsy among survivors.
Methods. The study cohort consisted of 2076 consecutively born infants, with birth weights of 500 to 1500 g and no major anomaly, born July 1, 1982, through June 30, 1994, to residents of a 17-county region in North Carolina. These infants had a mean birth weight of 1096 g (standard deviation, 251 g) and a mean gestational age of 29 weeks (standard deviation, 3 weeks). One thousand five hundred sixty-eight infants (76%) survived to 1 year adjusted age, at which point 1282 infants (82%) were examined at our medical center. The diagnosis of cerebral palsy was made only if the examining pediatrician and a pediatric physical therapist agreed on the diagnosis. To analyze trends across time, the Cochran-Armitage χ2 test and logistic regression were applied to data for infants categorized into six 2-year epochs according to year of birth.
Results. Mortality did not change significantly through 1990, and then began to decrease in 1990 to 1994. During the study period, mortality decreased from 36.8% between 1982 and 1984, to 13.8% between 1992 and 1994. The prevalence of cerebral palsy among survivors was constant from 1982 to 1988 (11.3%), decreased slightly from 1988 to 1990 (9.2%), and was lowest in 1990 to 1994 (5.2%). These secular trends in mortality and cerebral palsy risk remained significant when adjusted for gestational age, gender, and race. When adjusted for surfactant use, the trend in mortality was no longer significant, whereas the trend in cerebral palsy risk persisted.
Conclusions. The increasing survival of very low birth weight infants in the 1980s and 1990s has not resulted in an increased prevalence of cerebral palsy among survivors.
Very low birth weight infants have a risk of cerebral palsy that is between 40 and 100 times greater than that of normal birth weight infants.1 Because of improving rates of survival, this group now accounts for more than 25% of all new cases of cerebral palsy.2 3 In Western Australia,3England,4 and Sweden5 during the 1970s and 1980s the decrease in neonatal mortality rate was accompanied by an increase in the risk of cerebral palsy among survivors. In contrast, Meberg and Broch6 observed a decline in the risk of cerebral palsy in Norway, comparing births that occurred between 1970 to 1974 and 1985 to 1989. Bhushan et al7 have pointed out that even if the risk of cerebral palsy among survivors remains constant throughout time, a decreasing neonatal mortality rate will result in an increasing cerebral palsy prevalence (as a function of live births).
In the current study, secular trends in survival and cerebral palsy were analyzed among a geographically-defined cohort of very low birth weight infants. The study period included 8 years before, and 4 years after, surfactant treatment was made a part of routine care for infants requiring mechanical ventilation with birth weights less than 1350 g. Studies from other locations indicate that routine use of surfactant replacement therapy has resulted in a dramatic decrease in mortality for very low birth weight infants.8-10 Thus we investigated associations between this intervention and outcome (mortality and cerebral palsy prevalence) in the cohort that we studied.
The very low birth weight infants constituted a birth cohort of 2076 infants with the following attributes: 1) birth weights between 500 to 1500 g and no major congenital anomaly; 2) date of birth July 1, 1982, to June 30, 1994; 3) born to a mother residing in 1 of 17 counties in northwest North Carolina; and 4) admission to one of two neonatal intensive care units that are the only sources of neonatal intensive care in the 17-county region in which the infants' mother resided. More than 85% of the study infants were born at a level III obstetric referral center. The obstetric care is provided by perinatologists, obstetric residents supervised by perinatologists, and, occasionally by private obstetricians. Neonatal care is supervised by neonatologists in one of two level III neonatal intensive care units. Before 1991 all very low birth weight infants who required assisted ventilation were stabilized at Forsyth Memorial Hospital (the delivering hospital) and then transferred to the Brenner Children's Hospital (the university hospital), approximately 3 miles away. After 1991, only infants with a birth weight <700 g were transferred. Many of the study infants participated in randomized placebo-controlled trials of surfactant (Exosurf, Burroughs Wellcome, Research Triangle Park, NC) between July 1989 and August 1990. After August 1990, surfactant was given prophylactically to infants with birth weights less than 1350 g who required assisted ventilation within the first several hours of birth. Rescue surfactant was given to other infants with respiratory distress syndrome if an arterial blood gas measurement indicated that their arterial to alveolar PO 2 ratio was less than 0.28.
Neonatal Data Including Surfactant Receipt
Randomization assignments for infants who participated in randomized trials of surfactant were obtained from Burroughs Wellcome, and were used to classify these infants with respect to receipt of surfactant. All other neonatal data were obtained from a computerized database maintained by a research assistant who reviewed medical records to ascertain infants' birth weight, gestational age, gender, race, and hospital of birth. Gestational age usually was estimated by the admitting physician, based on the date of the mother's last menstrual period. The attending neonatologist completed a data collection form regarding treatments and diagnoses made in the neonatal intensive care unit. The two factors considered in this analysis were surfactant receipt (for infants who did not participate in placebo-controlled trials) and the results of neonatal cranial ultrasound examinations. These examinations were obtained using either a Diasonics (Santa Clara, CA) or an Acuson (Mountain View, CA) scanner with a 5 megahertz transducer. Most infants underwent cranial ultrasonography at least twice, first at 4 to 7 days and then again at 14 days of age. Infants whose ultrasound was abnormal at 14 days usually had additional ultrasound examinations. For this study, we defined a major cranial ultrasound abnormality as persistent ventricular enlargement, posthemorrhagic hydrocephalus (progressive ventricular enlargement requiring surgical intervention), or intraparenchymal echodensity or echolucency (intracerebral echodensity with or without subsequent development of cystic changes). We used the term intraparenchymal echodensity to refer to any abnormal echodensity in the cerebrum, recognizing that this finding may represent hemorrhagic as well as ischemic lesions.11 The diagnosis of ventricular enlargement was based on the opinion of the radiologists, not on a measurement of the ventricular size.
Ascertainment of Cerebral Palsy
Of 2076 infants, 508 (24%) were known to have died in the first year of life. Of the remaining infants, 199 were discharged alive from the neonatal intensive care but were lost to follow-up after discharge. Thus data about survival through 1 year of age (adjusted for prematurity) was available for 1877 infants. Of the 1369 infants who were known to have survived to 1 year adjusted age (adjusted for prematurity), 1282 were evaluated at our medical center and 87 were evaluated at other developmental clinics. Of the infants evaluated at our medical center, 105 (8.2%) were diagnosed as having cerebral palsy, as compared with 8 (9.2%) of the 87 infants evaluated at other sites.
Details of the multidisciplinary evaluation at our medical center have been described elsewhere.12 A pediatrician, who was aware of important aspects of the neonatal history (eg, results of cranial ultrasound examinations) performed a neurologic examination on each infant, with attention to abnormal active and passive posture of extremities (eg, fisting, toe-standing), deep tendon reflexes, and passive muscle tone.
All infants whose neonatal cranial ultrasound indicated an abnormality other than an uncomplicated subependymal hemorrhage were examined by a developmental pediatrician. Infants whose neonatal cranial ultrasound was normal or indicated only a subependymal hemorrhage were examined by a developmental pediatrician, a neonatologist with special interest in developmental outcome of high-risk infants, or a second-year pediatric resident (supervised by the developmental pediatrician). As a general rule, nonambulating infants, infants suspected by the pediatric resident of having a neurologic abnormality, and infants with delayed motor development (based on the Bayley Scales Psychomotor Developmental Index <68 [Bayley N. Bayley Scales of Infant Development.New York, NY: Psychological Corporation; 1969 and Bayley N.Bayley Scales of Infant Development. 2nd ed. San Antonio, TX: Psychological Corporation; 1993]), were seen by the developmental pediatrician or the neonatologist. Cerebral palsy was diagnosed only if both a pediatrician and a pediatric physical therapist agreed on the presence of definitely abnormal position and posture, and delayed motor development because of impaired neuromotor function. Three physical therapists were involved in confirmation of diagnoses of cerebral palsy. All three are full-time pediatric physical therapists who staff a pediatric rehabilitation hospital and an orthopedic clinic specifically for children with cerebral palsy.
With a few exceptions, infants with cerebral palsy had a spastic form of this disorder. Abnormalities common to these infants were hypertonicity, accentuated deep tendon reflexes, lack of expected protective responses, and delayed motor development. Based on subjective judgment by the examining pediatrician, three-quarters of cases were classified as having moderate or severe neurologic abnormality and one-quarter as having mild abnormality. More than three-quarters of cases of cerebral palsy had a Bayley Scales Psychomotor Developmental Index <68.
To analyze trends across time in mortality rate and the prevalence of cerebral palsy we categorized infants into 2-year epochs according to date of birth. A χ2 test was applied to the 3 × 6 contingency table containing the observed frequencies for the three-level outcome (dead, alive with cerebral palsy, and alive without cerebral palsy) at each of the six epochs. This contingency table includes cerebral palsy prevalence as a proportion of live births, as contrasted to cerebral palsy prevalence among surviving infants. The Cochran-Armitage test13 for trend throughout time was applied to the two 2 × 6 tables for the binary outcomes mortality and cerebral palsy prevalence among survivors, respectively. Logistic regression models were used to further analyze trends in mortality and cerebral palsy prevalence, adjusting for covariates such as birth weight, gestational age, gender, and race of the infant, and whether the infant was born at the level III obstetric referral center versus another site of birth.
Associations between surfactant replacement therapy and outcomes (mortality and cerebral palsy) were analyzed by using logistic regression models to estimate the independent associations with surfactant receipt, adjusting for birth weight, gender, ethnicity, and hospital of birth. For these analyses we used data from a subset of the study cohort, ie, 716 infants born July 1, 1988, to June 30, 1994, with birth weights 500 to 1349 g, who required assisted ventilation. Clinical protocol in our nurseries specifies that all infants with these characteristics be treated with surfactant as soon as possible after delivery, and this has been routine practice since August 1990. We assume, therefore, that infants born between July 1, 1988, and August 1, 1990, who had the aforementioned characteristics, are the most appropriate historical controls for infants treated with prophylactic surfactant in the first hours of life. Inclusion of infants with birth weights >1350, most of whom do not have respiratory distress syndrome and therefore do not require surfactant therapy, or infants who did not require assisted ventilation and have a low risk of mortality and cerebral palsy, would distort the analysis of the effects of surfactant.
Attributes of the Study Cohort
Attributes of the study cohort are summarized in Table1. Most characteristics were similar throughout the time period, with one exception, that being an increase throughout time in the proportion of survivors who had birth weights below 1000 g (29.1%, 36.3%, and 40.5% for infants born between 1982 to 1986, 1986 to 1990, and 1990 to 1994, respectively). The 199 infants who were discharged alive from neonatal intensive care but did not return for developmental evaluation (lost to follow-up), as compared with the 1282 infants whom we examined at 1 year adjusted age, tended to be healthier; those lost to follow-up were less likely to have had birth weights <1000 g (12% vs 36%), to have required assisted ventilation (46% vs 72%), and to have had a major cranial ultrasound abnormality (3% vs 9%), and their average gestational age was greater (30.3 weeks vs 28.8 weeks). The 87 infants who were evaluated at 1 year of age at other medical centers had a mean gestational age of 30.0 weeks, and the percentages with birth weights <1000 g, assisted ventilation, and major cranial ultrasound abnormalities were, respectively, 15%, 56%, and 6%.
Trends in Mortality
The distribution of outcomes, shown in Table2, changed during the six consecutive 2-year epochs from 1982 to 1994 (Figure)1. The mortality rate is fairly constant through 1990, and then begins to decrease in subsequent epochs, to a low of 13.8% in the most recent epoch. The unadjusted trend test is significant (P < .001), implying a decrease in mortality during the time period. After adjusting for birth weight, gender, ethnicity, site of birth (level III obstetric center versus other), and assisted ventilation, the odds ratio (95% confidence limits) for mortality in the most recent two epochs, relative to 1988 to 1990 were as follows: 1990 to 1992: 0.63 (0.43, 0.93) and 1992 to 1994: 0.32 (0.21, 0.48), suggesting a 37% and 68% reduction in the odds of mortality, respectively. When we included in the analysis the 199 infants who were discharged alive from neonatal intensive care, and assumed that these infants survived to 1 year of age, the change in survival throughout time remained significant.
Trends in Cerebral Palsy Risk Among Infants Surviving to 1 Year Adjusted Age
Excluding 3 cases of cerebral palsy that were attributable to postneonatal events, 113 infants were diagnosed as having cerebral palsy (8.3% of those examined at 1 year of age). The prevalence of cerebral palsy, as a proportion of live births and as a proportion of survivors, is shown for 2-year epochs in Table 2. The cerebral palsy prevalence among survivors was constant between 1982 to 1988, decreased slightly between 1988 to 1990, and was lowest between 1990 to 1994. Adjusting for birth weight, gender, ethnicity, site of birth (level III obstetric center versus other), and assisted ventilation, the risk of cerebral palsy between 1990 to 1992 was lower than in 1988 to 1990 [odds ratio, 0.39 (0.17, 0.86)]. Cerebral palsy prevalence between 1992 to 1994 was lower than in 1988 to 1990, but this difference was not significant [odds ratio, 0.73 (0.38, 1.40)], and the prevalence in epochs before 1988 to 1990 did not differ significantly from that between 1988 to 1990. If the cerebral palsy prevalence among survivors had remained constant throughout time, then the decreasing mortality rate since 1990 would have been accompanied by an increase in the cerebral palsy prevalence as a proportion of live births. However, as shown in Table 2, the decrease in the cerebral palsy prevalence among survivors was accompanied by decreases in both the cerebral palsy prevalence as a proportion of live births and the mortality rate.
Associations Between Surfactant Replacement and Outcome (Mortality and Cerebral Palsy)
Associations between surfactant therapy and outcome (mortality and cerebral palsy) were analyzed for a subset of the study cohort who were born July 1, 1988 to June 30, 1994, had birth weights 500 to 1349 g, and who required assisted ventilation. Table3 shows the results of logistic regression models in which the outcome of interest was mortality. When surfactant receipt was not included in the regression models infants born in the 2-year epochs after the introduction of surfactant for routine clinical use were at a lower risk of dying. However, when adjustment was made for surfactant receipt, date of birth was no longer a statistically significant predictor of mortality, although the interval 1990 to 1992 was nearly associated with an increased mortality risk (P = .09). This suggests that the lower mortality in 1990 to 1994, as compared with 1988 to 1990, can be explained by surfactant therapy. In contrast, as shown in Table4, the decline in cerebral palsy risk in the interval 1990 to 1992 does not seem to be attributable to surfactant therapy because the association persisted when adjustment was made for surfactant.
We explored other possible explanations for the observed downward trend in cerebral palsy, including the decreasing frequency with which infants born at our tertiary obstetric center were transported to our children's hospital nearby and the decreasing frequency of major cranial ultrasound abnormalities. Throughout the study period, the more severely ill Infants were transported to our children's hospital; not surprisingly, transport was associated with a higher risk of cerebral palsy. On the other hand, this association did not persist when adjusting for birth weight, gender, race, need for assisted ventilation, and date of birth. As would be expected, major cranial ultrasound abnormality was strongly associated with cerebral palsy; however, in multivariate logistic regression, statistically significant associations were found between cerebral palsy and both major cranial ultrasound abnormality (odds ratio: 19.5; 95% confidence limits: 8.6, 44.3), as well as date of birth (odds ratio for 1990 to 1992: 0.19; 95% confidence limits: 0.07, 0.56; odds ratio for 1992 to 1994: 0.52; 95% confidence limits: 0.24, 1.14).
The major finding from this study is that the decline in mortality among very low birth weight infants during the 1990s has not resulted in an increase in the prevalence of cerebral palsy. Our analysis suggests that the decrease in mortality risk after July 1, 1990 can be explained by the introduction of surfactant replacement. On the other hand, the decrease in the prevalence of cerebral palsy among survivors that has occurred in our region in the 1990s does not seem to be attributable to this new therapy.
We are aware of no other published study of changes in the prevalence of cerebral palsy in a population-based cohort during the 1990s. Several, but not all, previous studies describe an increase in the prevalence of cerebral palsy as mortality decreased among very low birth weight infants born in the 1970s and 1980s. The declining mortality risk during those decades was most likely because of improvements in obstetric care, such as antenatal steroids, and various interventions, referred to collectively as neonatal intensive care. If replicated in other geographic regions, our findings would indicate that surfactant replacement, in contrast to the collection of factors responsible for declining mortality before 1990, has not resulted in an increasing risk of cerebral palsy among surviving infants.
Our finding with respect to the effect of surfactant on mortality is consistent with the results of randomized controlled trials14-16 as well as observational studies from other locations.17 18 In agreement with follow-up studies from randomized controlled trials of surfactant, our results suggest that surfactant does not affect the risk of cerebral palsy among surviving very low birth weight infants.19 20
Several possible limitations of our study should be noted. First, we cannot exclude the possibility that selection bias explains our finding of a decreased risk of cerebral palsy during the 1990s. Although the two hospitals in which study infants were identified were the only sources of level II and level III neonatal intensive care in the region from which our sample was drawn, some very low birth weight infants may have died before transport to our referral center. On the other hand, it seems unlikely that the number of infants dying before transport has increased throughout this time; our results probably cannot be explained by this source of bias. It is also possible that selection bias resulted from incomplete follow-up at 1 year of age. This bias could explain our results only if infants with cerebral palsy were less likely to return for evaluation in the years 1990 to 1994, but not in the earlier years. A third limitation is that an association between surfactant and the risk of cerebral palsy could have been masked if its effect depended on the infant's age at initiation of therapy, the type of surfactant (natural or synthetic), or the number of doses.
Finally, we must consider the possibility that the criteria for the diagnosis of cerebral palsy changed during the study period. In an effort to diagnose cerebral palsy reliably, we have restricted this diagnosis to those infants with definite clinical abnormalities and motor impairment, about which two examiners agree. Data from the United States Collaborative Perinatal Project, suggest that limiting the diagnosis of cerebral palsy to more severely affected infants increases the likelihood that the diagnosis will persist through early childhood. In that study, 26% of mild cases diagnosed at 1 year resolved during early childhood, but 56% of moderate and severe cases persisted.21 Some reassurance of the validity of our prevalence may be taken from comparison with a recent prospective population-based study of cerebral palsy—the New Jersey Brain Hemorrhage Study. In that study Pinto-Martin et al22 found that 12.2% of infants with birth weights <1500 g, born between 1984 to 1987, had disabling cerebral palsy at 2 years of age. In our study 10.6% of infants born between 1982 to 1988 had cerebral palsy at 1 year of age.
Our findings have important implications regarding the etiology and prevention of cerebral palsy in very low birth weight infants. Stanley2 has pointed out that the increase in cerebral palsy that occurred during the 1970s and 1980s could be attributable to increased survival of infants with brain damage occurring before birth or an increase in the incidence of damage because of complications of prematurity. The increased neonatal survival and decreased risk of cerebral palsy, after introduction of surfactant, implies that a decrease in brain damage attributable to complications of prematurity has occurred. Otherwise one must postulate that infants with prenatally acquired brain damage have a survival probability equal to that of an infant without such damage and that the effect of surfactant is differential with respect to this damage, with a greater effect in infants without damage.
Further study is needed to assess whether our findings are generalizable and if so, what factors account for the decrease in cerebral palsy. Our analysis suggests that this decrease is due, at least in part, to factors other than the decreasing frequency of major cranial ultrasound abnormalities observed in the present study and in a previous study by Philip et al.23 In addition to surfactant, medications that have been used more frequently in the 1990s, as compared with the previous decade, include indomethacin and dexamethasone. However, a large randomized trial of indomethacin24 and observational studies of dexamethasone25 26 have identified no improvement in neurologic outcome in infants treated with these medications, as compared with controls. A growing body of information suggests that intrauterine infection increases the risk of cerebral palsy in premature infants,27-29 and that inflammatory cytokines may be involved in the pathogenesis.30-32 Other studies suggest that prenatal treatment of mothers with magnesium sulfate is associated with a decreased risk of cerebral palsy.33-35It is plausible, then, that a downward trend in cerebral palsy could be attributed, at least in part, to changes in prenatal factors. More study of these factors could lead to the development of strategies to reduce the risk of cerebral palsy in very low birth weight infants.
Support for this research was provided by the Brenner Children's Hospital, Winston-Salem, North Carolina and the North Carolina Department of Environment, Health, and Natural Resources.
- Received February 3, 1997.
- Accepted July 11, 1997.
Reprint requests to (T.M.O.) Department of Pediatrics, Bowman Gray School of Medicine of Wake Forest University, Medical Center Blvd, Winston-Salem, NC 27157.
- Pharoah POD,
- Cooke T,
- Cooke RWI,
- Rosenbloom L
- Bhushan V,
- Paneth N,
- Kiely JL
- O'Shea TM,
- Dillard RG,
- Klinepeter KL,
- Goldstein DJ
- Liechty EA,
- Donovan E,
- Purohit D,
- et al.
- Horbar JD,
- Wright EC,
- Onstad L,
- and the Members of the National Institute of Child Health and Human Development Neonatal Research Network
- Corbet A,
- Long W,
- Schumacher R,
- Gerdes J,
- Cotton R,
- and the American Exosurf Neonatal Study Group
- Nelson KB,
- Ellenberg JH
- Pinto-Martin JA,
- Riolo S,
- Cnaan A,
- Holzman C,
- Susser MW,
- Paneth N
- Philip AGS,
- Allan WC,
- Tito AM,
- Wheeler LR
- Ment LR,
- Vohr B,
- Oh W,
- et al.
- Cooke RWI
- Nelson KB,
- Grether JK
- Hauth JC,
- Goldenberg RL,
- Nelson KG,
- BuBard MB,
- Peralta MA,
- Gaudier FL
- Copyright © 1998 American Academy of Pediatrics