Objective. Advances in neonatology have contributed to improved survival for extremely low birth weight (ELBW) infants. Neurodevelopmental outcome is usually reported for a single large group of infants rather than according to smaller birth weight groups because of small numbers. Our purpose was to review the neurodevelopmental outcome of a large group of ELBW infants and examine differential outcome according to birth weight.
Study Design. A total of 446 infants born between 1979 and 1991, with a birth weight of 500 to 999 g, were followed to mean age 55 months ± 33 standard deviation. Univariate analyses of medical risk factors of birth weight, gestational age, year of birth, growth retardation, gender, inborn/outborn status, days on oxygen, intracranial hemorrhage, and social risk in relation to outcome were conducted on the group as a whole. Neurologic/developmental outcome was also analyzed by 100-g weight groups.
Results. A total of 61% of all infants were completely normal, with no neurologic, neurosensory, or cognitive deficits. There was no association between outcome and birth weight. There was a strong association between intracranial hemorrhage (ICH) grade III or IV and/or cystic periventricular leukomalacia (PVL) and abnormal outcome (Somers' D = .17) and ICH III/IV and/or cystic PVL and cognitive outcome (Kendall's tau = .15). Mild to moderate cognitive delays were associated with chronic lung disease (oxygen >60 days) (Kruskal-Wallis χ2 = 17.53) or high social risk (Kruskal-Wallis χ2 = 22.17).
Conclusion. In this study of ELBW infants, low birth weight was not associated with abnormal outcome. The risk factors of ICH III-IV/cystic PVL, chronic lung disease, and high social risk were associated with abnormal outcome.
During the last decade, tertiary centers have reported decreasing mortality for extremely low birth weight (ELBW) infants.1 Declining mortality rates have been most dramatic for the smallest infants, specifically those of <750 g birth weight. Advances in general perinatology and neonatology as well as specific new therapies such as exogenous surfactants and antenatal steroids have all contributed to this improved survival.2,3
At our institution, the number of very low birth weight survivors that have been entered into our Intensive Care Nursery (ICN) Follow-Up Program increased after 1985, especially in the lowest birth weight groups (Table 1). Although increasing in numbers, these infants still represent only a small portion of all the survivors of <1500 g birth weight. They frequently have prolonged, complicated and expensive hospitalizations with uncertain outlooks at the time of discharge.4,5 It has been suggested that with these increases in survival, it would be likely that 50% of children with cerebral palsy will have been of extremely low birth weight.6 Because of small sample sizes, most survival and outcome studies continue to group all infants under a certain birth weight, perhaps overlooking differences in a growing population of infants in the lowest surviving weight groups. It is not at all clear that the 500-g and 999-g birth weight infants are comparable in their medical and developmental risks.
In 1981, Bennett-Britton et al reported the outcome of ELBW infants separated into 50- to 100-g weight groups.7 They found that outcome deteriorated significantly with declining birth weights. With virtually 100% of survivors who weigh <700 g having severe neurologic abnormalities, they concluded that intensive care was not justified for these extremely small infants. In the mid-1980s Kitchen et al separated infants by birth weights and found improved quality of outcome in the more recent survivors.8 A current study of infants at mean adjusted age 19 months by La Pine found no differences in outcome by 100-g weight groups (400 to 800 g).9 Yet in long-term follow-up, Hack has raised important concerns about the long-term outcome of infants weighing <750 g at birth.10 We have examined our own large group of ELBW infants to examine the question of differential outcome according to birth weight. Medical and social risk factors that have been identified by previous research to correlate with poor neurodevelopmental outcome were determined for each of the infants in this study.11
In the period between January 1, 1979, and December 31, 1991, 518 infants with birth weights <1000 g were discharged from the ICN at Mount Zion Hospital and Medical Center and the University of California, San Francisco. As the sole referral ICN for contract hospitals, all outborn infants with a <1000 g birth weight were referred to our institutions. Fourteen infants died after discharge. A total of 58 infants were lost to follow-up before 1 year of age, leaving 446 (88%) infants who were followed to or beyond 1 year. One infant was removed from the analyses because of a late genetic diagnosis. Ethnic origins of infants were typical for our geographical area: 47% white, 22% African-American, 17% Hispanic, 7% Asian/Pacific, and 5% other.
Infants were lost for a variety of reasons, including moving to foreign countries, moving to other states, and transience related to social disorganization.
Infant medical charts were reviewed for presence of risk factors for outcome.
Risk factors included birth weight, gestational age, year of birth, appropriateness of weight for age (AGA), gender, inborn/outborn status, days of oxygen, and intracranial hemorrhage (ICH). Gestational age determination was made using a combination of maternal dates, early ultrasonography (≤24 weeks gestation), and neonatal examination. Small for gestation infants were those whose birth weights were >2 standard deviations (SD) below the mean on the California Growth Chart. A rating of the socioeconomic factors and social risk factors was obtained at follow-up visits. For purposes of statistical analyses, some variables were collapsed. The following variables were grouped for different analyses as follows:
Birth weight was treated as a numeric variable for analysis of differences between lost and followed infants. For outcome analyses, infants were separated into five weight groups, by 100-g increments and treated as a categorical variable.
Year of Birth
Year of birth was treated as an numeric variable for some analyses. Other analyses were conducted by splitting the years of birth into three categories: presurfactant, (1979 to 1985), transitional (1986 to 1988), and surfactant era (1989 to 1991). During the transitional period, infants at both hospitals were enrolled in a double-blind study testing the efficacy of an artificial surfactant (Exosurf, Burroughs Wellcome). Therefore, some infants received surfactant during this era because of the study protocol.
Days in Oxygen
Because some infants went home on oxygen and remained oxygen-dependent for long periods of time, a few infant outliers could have an untoward effect on analyses. Therefore, days on oxygen were treated as a categorical variable. For a four-factor analysis, days on oxygen was grouped as follows: 0 to 30 days = no chronic lung disease (CLD); 31 to 59 days = mild CLD; 60 to 89 days = moderate CLD, and >90 days = severe CLD. For a two-factor analysis, infants were grouped as follows: 0 to 59 days = no CLD; ≥60 days = CLD. Thirty-two infants were discharged from the ICN in cannula O2. The need for O2 at the time of discharge has been identified previously as a risk factor for poor outcome.12 Need for O2 after discharge increased with decreasing birth weight. A total of 20% of infants in the 500-g weight group required oxygen after discharge, compared with 4% in the 900-g weight group. The mean age of infants at the time of discharge who were sent home in oxygen was 4.5 weeks adjusted age (range, term to 10 weeks adjusted age). For this high-risk subset, total months in oxygen including the inpatient period ranged from 4 to 36 months. No infants were discharged on mechanical ventilation.
The severity of ICH was determined using the four-level grading system described by Papile et al.13 For infants born before 1981, evaluation of ICH was followed using cranial computed tomography scans. Infants born after 1981 were evaluated with serial bedside ultrasonography. Both computed tomography and ultrasonography were initiated within the first week of life as part of the nursery protocol to identify infants at risk for ICH. The studies were repeated routinely on a weekly or every-other-week basis for the first 4 weeks of life. Beyond 4 weeks, if abnormal findings were present, the studies were repeated periodically until abnormal findings either were stable or disappeared. Classification of hemorrhage was determined by a neuroradiologist. For statistical analyses, grades I and II ICH were considered as one group and called uncomplicatedhemorrhages. Grades III and IV and cystic periventricular leukomalacia (PVL, defined as persistent periventricular echodensities, progressing onto cyst formation) were grouped as complicatedhemorrhages. To look more closely at the effect of large ICHs or cystic leukomalacia, a binary analysis was also conducted, grouping No ICH and grades I and II ICH together versus grades III and IV and cystic PVL. None of the 15 500-g weight group infant survivors had complicated ICH; infants in this weight category with severe ICH/PVL died before discharge.
Social Risk Factor
At follow-up visits, families were asked to provide information relating to educational status of parents, occupation, and type of health insurance (private, none, or government-assisted [ie, Medicaid program]). Because many families were welfare-dependent, traditional occupational categories were underrepresented in our group. A general rating of employed/unemployed was a better descriptor for differentiating socioeconomic differences in this group of infants. Because not all mothers or infants were tested routinely for illegal drugs, a notation was made for positive toxicology screens where performed. For many infants, this was a reason for placement in foster care at discharge from the hospital. Infants in foster care >1 year were rated as being at social risk. Infants who had a smooth transition from foster care for 1 year to adoption were rated at low social risk, if there was also no maternal drug history or fetal exposure to drugs and if adoptive parents met criteria for low social risk. Maternal education <12 years, parent(s) unemployed, and/or dependence on government assistance for insurance were additional criteria for the infant being at high social risk. Followed infants were distributed approximately evenly for the social risk factor, with 55% being at relatively low social risk and 45% being at high social risk.
Characteristics of the infants are outlined in Table2.
Infants were assessed according to a protocol for very low birth weight infants in the Follow-Up Clinic. They were seen frequently in the first year of life for neurodevelopmental examinations. Infants were assessed by a neonatologist or pediatric nurse practitioner with training in developmental and behavioral pediatrics. Formal developmental testing was carried out by a psychologist at adjusted age 12 months, 18 months, and 2½ years, and chronologic age 4½ years and 7 to 8 years. Most infants received both a neurodevelopmental examination and formal testing at visits at adjusted age 12 months or older. However, some received only one of the two evaluations. Neurologic, neurosensory, and cognitive outcomes were considered as follows.
Cerebral palsy or significant abnormalities in muscle strength, tone, reflexes, or movement patterns causing functional impairment were classified as abnormal outcome. These included such diagnoses as hemiplegia, diplegia, quadriplegia, and ataxia. Infants with abnormalities in tone, strength, or reflexes who did not present with a clear diagnosis after 1 year were given a suspect classification. These mild neuromotor abnormalities included findings such as tremors and clumsiness.
Bilateral blindness or hearing loss requiring hearing aids or other communication aids were considered moderate to severe abnormal neurosensory outcome. Infants with vision in only one eye or normal hearing in only one ear were classified as having a mildly abnormal neurosensory outcome.
The Bayley Scales of Infant Development was administered at 12 and 18 months of age; the Stanford–Binet Intelligence Scale was administered at 2½ to 4 years; the McCarthy Scales of Children's Abilities was administered at age 4 to 6 years, and the Wechsler Intelligence Scale for Children, Revised, was administered at age 7 and older. Age was adjusted for prematurity until three years. Mean age at last assessment was 55 months ± 33 SD. The percent of infants seen at different age ranges was as follows: <18 months, 8%; 18 to 24 months, 7%; 25 to 36 months, 11%; 37 to 48 months, 11%; 49 to 60 months, 24%; 61 to 72 months, 8%; >72 months, 31%.
The age at last assessment by era (presurfactant, transitional, and surfactant) is shown in Fig 1. A similar number of infants in each era had their last assessment at <30 months, when cognitive testing scores are considered less reliable. Infants in the most recent era (surfactant) were well represented at the older age of assessment groups.
Children were assessed with the test instruments if they had functional ability to complete the assessment. Children with severe motor limitations and severely limited communication abilities were not administered the cognitive assessment but were placed in an a priori abnormal category because of their special educational needs. Reports of functional skills were obtained from their special education programs, if possible.
Scores on the Bayley; Stanford–Binet; McCarthy General Cognitive Index; and Wechsler Intelligence Scale for Children, Revised, Full-Scale Intelligence Quotient that were between 1 and 2 SDs below the mean were considered borderline range; scores >2 SDs below the mean were considered abnormal. It should be noted that developmental test scores do not reflect the school and life problems that some ELBW infants may encounter. Infants obtaining normal intelligence quotient scores at school-age may have learning disabilities. Infants obtaining test scores in a borderline range at school-age can show great variability in function. Those at the higher end of the range (80 to 84) may be able to function in a normal classroom in a public school with some additional “resource room” help sporadically or more chronically. Those at the lower end of the range (69 to 79) are likely to require substantial special educational assistance to make progress. Family circumstances will tend to exacerbate or ameliorate to some extent the quality of life for those functioning in a borderline range of intelligence.
For statistical analysis, we used the score from the last test administered. Some ELBW children show a pattern of gradual improvement over time, whereas others begin to show subtle deficits not detectable on infant tests. For this reason, the last test in a series generally will be most predictive of school-age outcome. Test scores can change over time; however, very low scoring infants tend to stay very low scoring on later assessments.14,15
The use of the developmental test scores on a categorical basis (number of SDs from the mean) rather than as numeric scores allowed these scores to be treated similarly for the logistic regression analysis. Neurologic/neurosensory outcome was available on 442 of 445 infants, and cognitive data were available on 441 of 445 infants.
Differences in characteristics of lost and followed infants were analyzed using pooled variance t tests, after analyzing the data sets with the Kolmogorov–Smirnov Two-Sample Test to test for equality of distributions on ordinal variables of birth weight, gestational age, and days spent on oxygen. Likelihood ratio–χ2 analyses were used to look at differences between lost and followed infants for categorical variables of small for gestational age (SGA), sex, inborn, complicated ICH/PVL, and social risk factor. Pearson χ2 and Kruskal-Wallis tests on mean ranks and Kendall's tau and Somers' D measures of association were used to test the association between outcome and risk factors, depending on whether the ordered categorical variables were dichotomous or polytomous.
The lost infants did not differ significantly from those who were followed for birth weight, gestational age, percent small for gestational age, percent male, and percent inborn. No differences were found between lost and followed groups for oxygen requirements or percent of infants with complicated ICH. There was a significant difference between lost and followed infants for the social risk factor. More infants in the lost group were at high social risk than infants in the followed group (likelihood ratio–χ2= 12.345; P = .0004).
In the total group of followed infants, 61% had a completely normal outcome, with no neurologic, neurosensory, or cognitive deficits. Of the children who were not completely normal, 12% had a combination of neurologic/neurosensory and cognitive deficits, 3% had a neurologic/neurosensory deficit alone, and 23% had a cognitive deficit alone.
Because some children with a physical handicap such as diplegia may have normal intelligence, we looked separately at neurologic/neurosensory outcome and cognitive outcome. Considering neurologic/neurosensory outcome alone without regard to cognitive deficits, 85% of infants had a normal neurologic/neurosensory outcome. Considering cognitive outcome alone, without regard to neurologic/neurosensory outcome, 64% of infants had a normal cognitive outcome (Table 3). Five percent of infants had a mild neurologic or mild neurosensory deficit, and 22% had a borderline cognitive deficit. Moderately to severely abnormal neurologic or neurosensory outcome was found in 11% of infants, and abnormal cognitive deficit was found in 14% of infants.
Of the 47 infants who had a moderate to severe neurologic/neurosensory handicap, 41 had cerebral palsy, 5 were blind, and 1 was deaf. The deaf infant was born in the presurfactant era. One blind infant was born in the presurfactant era, and 4 were born in the transitional era. Eighteen infants with cerebral palsy were born in the presurfactant era, 9 in the transitional era, and 14 in the surfactant era. The rate of cerebral palsy in all three eras was similar (10%, 8%, and 10%).
Outcome by 100-g weight groups is seen in Table4. Assuming all lost infants were normal or all lost infants were not normal, a percent range of normal outcome is shown.
We examined the independent effect of risk factors on outcome.
There was no association between neurologic/neurosensory outcome and birth weight. We grouped birth weight by 100-g weight groups, with birth weight as an explanatory variable, and neurologic/neurosensory outcome as a dependent variable (Somers' D = −.002;P = .95, not significant [NS]). There was no association between cognitive outcome and birth weight groups (birth weight explanatory, cognitive outcome dependent, Somers'D = .054; P = .13, NS). If all cognitive scores >1 SD below the mean were considered to be abnormal, there was still no association with birth weight (Somers'D = −.0045, NS).
Year of Birth
Categorical analyses were used to look at the association of birth year with birth weight (100-g weight groups) by neurologic/neurosensory outcome. Using a three-category analysis (presurfactant era, transitional era, and surfactant era), a positive association of era with birth weight for 371 infants who had normal neurologic outcome was seen (Kendall's tau = .18; P = .0001). There were too few infants in each category to conduct a similar analysis for infants with mild (N = 24) or moderate (N = 47) neurologic deficits.
Linear logistic regression analyses of outcome using birth year category did not show a relation to neurologic/neurosensory outcome.
Birth year category approached significance for cognitive outcome (χ2 = 3.49; odds ratio; 1.4; P = .06). This analysis included all outcome groups: normal, mild deficit, and moderate deficit. This may indicate a trend for improved cognitive outcome for infants of all birth weights over time.
There was no association of place of birth with neurologic outcome (Pearson χ2 = .5188; P = .47, NS) or cognitive outcome (Pearson χ2 = .6203; P= .43, NS). A total of 90% of inborn infants had normal or suspect neurologic/neurosensory outcome, compared with 88% of outborn infants, and 87% of inborn infants had normal or borderline cognitive scores, compared with 85% of outborn infants.
There was no statistically significant association of AGA/SGA status with neurologic outcome (Pearson χ2 = 3.1087;P = .08, NS); however, the direction of the trend was for AGA infants to have a poorer neurologic outcome. There was no association of AGA/SGA status with cognitive outcome (Pearson χ2 = .0390; P = .84, NS).
There was no association between sex of infant with neurologic outcome (Pearson χ2 = 2.5668; P = .11, NS). There was an association of sex with cognitive outcome, favoring females (Pearson χ2 = 3.8089; P= .05).
Gestational age was found to have predictive power for neurologic outcome in a logistic regression analysis (r = −.1233; P = .01; β = −.2686; standard error/β = .1049). This negative association may be related to the fact that infants of the smallest gestational ages with severe intracranial pathology generally did not survive to discharge, whereas infants of more mature gestational ages did. Gestational age was not related to cognitive outcome (Pearson χ2 r = .0000; P = .25, NS).
For 437 children with complete data (scans and outcome), there was a significant association of ICH/PVL with abnormal neurologic/neurosensory outcome. (Somers' D = .17;P = .0000).
Twenty-two of 59 infants (37%) with complicated ICH and/or cystic PVL had moderate to severely abnormal neurologic/neurosensory examinations. Five children with complicated ICH and/or cystic PVL (9%) had mildly abnormal examinations. Slightly more than half (54%) of the children with complicated ICH and/or cystic PVL had normal neurologic and neurosensory findings. Figure 2 shows the distribution of outcome related to ICH and/or cystic PVL. With increasing grade of bleed, the percent of children with normal outcome declines and abnormal outcome increases. This was also true for the association between cognitive outcome and ICH (Kendall's tau b = .15;P = .0008).
CLD showed a trend toward significant association with neurologic/neurosensory outcome when days of oxygen were analyzed as a four-category variable (Somers' D = .051;P = .05). When the data were reanalyzed using CLD as a two-category variable (O2 fewer than or more than 60 days), no association was seen (Kruskal-Wallis χ2 = 1.17, NS).
CLD did show significant association with poor cognitive outcome, however (Somers' D = .13; P = .0009). With increasing days of oxygen requirement, fewer children had a normal outcome. Categorizing 02 as a four-factor variable, 70% of children with ≤30 days of oxygen had a normal outcome, compared with 51% normal outcome for children with >90 days of oxygen. This association was strengthened when oxygen was dichotomized at greater than or less than 60 days of oxygen (Kruskal-Wallis χ2 = 17.53; P = .0000) (Table5).
We obtained social risk data for most but not all infants for whom we had neurodevelopmental data (N = 429) and cognitive data (N = 430). As might be expected, there was no association between neurologic/neurosensory outcome and social risk factors (Kruskal-Wallis χ2 = 1.20, NS). There was a strong relationship between social risk and cognitive outcome for these very low birth weight infants (Kruskal-Wallis χ2 = 22.17;P = .0000). As noted above, we did not have equal representation of all social classes in our study sample. This is partly attributable to other factors associated with premature birth: teen-age pregnancy, substance abuse, no prenatal care. Three quarters of our infants who came from families of relatively low social risk were in a normal range for cognitive skills. Only 53% of those at higher social risk were performing in a normal range.
We are unaware of methods for conducting power analysis for Somers' D, which is an asymmetric (ie, one predictor and one criterion) measure of association for ordinal categorical variables. However, an analogous parametric coefficient would be the Pearson r for simple regression. Our sample was large enough to detect a Pearson r as weak as .12 with power = .80. In our data, a Somers' D of .05 (as for the prediction of neurologic outcome by CLD), which was significant at P= .05, corresponded to a Pearson r of .07, showing that the asymmetric Somers' D was more powerful than the Pearsonr would be. Most of our tests for the prediction of outcome from ICH and CLD were significant, making power analysis irrelevant. For those analyses of other predictors that were not significant, such as birth weight predicting neurologic and cognitive outcome and social risk predicting neurologic outcome, it seems likely that the effect sizes are quite small in the population, given the power of the other tests.
Because ICH and CLD are known risk factors for developmental disabilities, we examined the outcome of children who had neither risk factor. There were 110 infants with no CLD (O2 < 30 days) and no ICH or PVL. At discharge, these infants might be considered at lowest risk for subsequent deficits. These infants did not all turn out to be deficit-free however. Two infants had mild neurologic or neurosensory deficits alone, and two had moderate to severe neurologic/neurosensory deficits alone. Four other infants had mild or moderate neurologic/neurosensory deficits accompanied by mild or moderate cognitive deficits.
Of the 102 infants with no neurologic or neurosensory deficits, which might contribute to lowered cognitive scores, 80 had normal cognitive scores. Two children had marked cognitive delays. Twenty children had cognitive scores in a borderline range.
Although survival in the ELBW infant has improved dramatically because of recent advances in perinatal care, the critical question is that of the quality of such survival. Despite surfactant treatment, infants born at 24 to 28 weeks' gestation are at risk for developing bronchopulmonary dysplasia and CLD leading to possible developmental delays. The very immature brain is at risk for serious ICH leading to cerebral palsy and other neurologic abnormalities. Advanced retinopathy may also be more prevalent in infants with extreme prematurity.
Our study evaluated a large cohort of ELBW born in a 12-year period. Our smallest infants were within the range of outcome similar to that for larger infants. However, it should be noted that no infants in the 500- to 599-g weight group who survived to discharge had complicated ICH or PVL, although six infants in the 600- to 699-g weight group did. Infants of lower birth weight did not have poorer neurologic/neurosensory outcomes than infants of greater birth weight in the group of <1000-g infants that we studied. For cognitive outcome, there was a trend toward improved outcome over the 12-year period we studied for infants of all weight groups. Our smaller infants in general showed a lower percentage of significant cognitive deficits, but this was not statistically significant. Our findings suggest that although birth weight alone is not related to poor outcome, several other known risk factors are related to poor outcome: ICH, CLD, and social risk.
There was a strong association between complicated grades of ICH (III and IV) and/or cystic PVL and abnormal neurologic outcome. Infants with CLD (oxygen requirements >60 days) had an increased incidence of moderate cognitive delays compared with infants who required oxygen <60 days. Whereas as many as 20% of the 500-g infants required supplemental oxygen at discharge, the incidence of bronchopulmonary dysplasia in ELBW infants becomes a significant risk factor for poor outcome. There was also a significant association between higher social risk and moderate cognitive delays.
Because birth weight alone in this study does not identify poor outcome, it is clear that as the numbers of ELBW infants increase, the task is to decrease the other known associated risk factors for poor outcome: ICH/PVL, CLD, and social risk. We acknowledge that the low incidence of severe ICH/PVL in our population may have contributed to the optimistic neurologic outcome, particularly in the lower weight groups.
As neonatology masters the advances available in medical care for ELBW infants, it is important to know which variables are reliably related to long-term outcome, and birth weight alone may not be a reliable indicator.
We thank Dr Roderic Phibbs for his review and input.
- Received November 18, 1996.
- Accepted March 6, 1997.
Reprint requests to (R.E.P.) Department of Pediatrics, Box 0748, UCSF, San Francisco, CA 94143.
- ELBW =
- extremely low birth weight •
- ICN =
- intensive care nursery •
- AGA =
- appropriateness of weight for age •
- ICH =
- intracranial hemorrhage •
- SD =
- standard deviation •
- CLD =
- chronic lung disease •
- PVL =
- periventricular leukomalacia •
- SGA =
- small for gestational age •
- NS =
- not significant
- Horbar JD,
- Wright EC,
- Onstad L
- Schwartz RM,
- Luby AM,
- Scanlon JW,
- Kellogg RJ
- Pomerance JJ,
- Pomerance LJ,
- Gottleib JA
- McCormick MC,
- Bernbaum JC,
- Eisenberg JM,
- Kustra SL,
- Finnegan E
- La Pine TR, Jackson C, Bennett FC
- Campbell LR,
- McAlister W,
- Volpe JJ
- Copyright © 1997 American Academy of Pediatrics