Copyright 1998 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1998
To summarize the literature on mortality rates and prevalences of major neurodevelopmental disabilities and to examine trends of these outcomes over time in extremely premature neonates.
MEDLINE was used to search the English literature for studies published since 1970 reporting on both mortality and disability in infants born at or before 26 weeks' gestation (extremely immature [EI] cohort), with a birth weight of 800 g or less (extremely small [ES] cohort), or subgroups of these.
Studies were included in the analysis if all of the following were reported: mortality; direct examination of 75% or more of the survivors; and the proportion of patients with at least 1 of the following disabilities: cerebral palsy, mental retardation, blindness, and deafness. Studies reporting cohorts included as a subset of cohorts in another study were excluded. Forty-two studies providing mortality and disability data for 20 cohorts of 4116 EI infants and 38 cohorts of 4345 ES infants born after 1972 met the inclusion criteria.
Data were abstracted from all studies that met these criteria by two of us (J.M.L. and D.E.W.), independently; the data were then cross-checked to ensure accuracy.
Survival averaged 41% for EI infants and 30% for ES infants, and it increased significantly with time. In contrast to mortality, the prevalences of major neurodevelopmental disabilities among survivors have not changed over time. The most common major disability was mental retardation, found in 14% of EI and ES survivors. Cerebral palsy was found in 12% of EI survivors and 8% of ES survivors, blindness was found in 8% of EI and ES survivors, and deafness was found in 3% of EI and ES survivors. Overall, 22% of EI survivors and 24% of ES survivors were classified as having at least 1 major disability. Each 100 EI or ES livebirths yielded 7 children with major disabilities; this prevalence was correlated with survival across cohorts.
The prevalence of disabilities had not changed among EI or ES survivors with increasing survival. However, increasing survival of these infants has resulted in a steadily increasing prevalence of children with disabilities.
THE DEVELOPMENT of the technology of newborn intensive care (NIC) and its application in premature infants has been one of the most successful of all technological innovations in medicine in reducing mortality.1 However, its effects on the prevalence of later disability are more controversial. In recent years, NIC increasingly has been used in extremely immature infants whose chances of survival were negligible before the development of NIC. These infants are therefore a relatively new type of survivor whose prognosis cannot be known from historical experience. Although rare, these survivors can use a considerable portion of NIC resources. In 1993 in the United States, 11950 infants were born weighing 501 to 800 g, and 13531 infants were born at 23 to 26 weeks' gestational age (John L. Kiely, PhD, National Center for Health Statistics, written communication, November 13, 1996).
There is considerable disagreement on the advisability of instituting NIC in such infants and much variation internationally and within North America in actual practice.2,3 Reluctance to intervene stems in part from concerns about the risks of the long-term disabilities, principally neurodevelopmental, to which premature infants in general are especially susceptible. These disabilities include severe conditions, such as cerebral palsy (CP), mental retardation, deafness, and blindness, and milder but more common problems, such as learning disabilities and attention-deficit hyperactivity disorder.4- 6 Because widespread application of NIC management at or below the thresholds of 800 g and 26 weeks is of recent origin, available data on outcomes are generally restricted to the first few years of life. We, therefore, set out to perform a review of the mortality and the prevalence of severe neurodevelopmental disabilities in survivors because these outcomes usually can be determined in the first 2 years of life. The purpose of this analysis was to quantitatively summarize mortality and severe disability in extremely immature (EI) and extremely small (ES) infants, to describe trends in these outcomes over time and with varying survival, and to estimate the public health burden that might be expected to result from providing NIC to these infants.
MEDLINE was used to search the English literature from January 1970 through August 1997 using the MESH headings "follow-up" or "outcome" or "cerebral palsy" and "infant, premature" or "infant, low birth weight." All studies reporting disabilities in infants or children who weighed 800 g or less at birth, whose gestational age was 26 weeks or less, or whose birth weight or gestational age ranges were below these thresholds were reviewed. The bibliographies of these studies also were reviewed to identify other reports of outcome in these birth weight and gestational age ranges. Studies were included in the analysis if all of the following were reported: mortality in the cohort (at least to hospital discharge); direct examination and formal testing of 75% or more of the survivors; and the proportion of patients examined with at least 1 of the following disabilities: CP, mental retardation, blindness, and deafness. Data were abstracted from all studies that met these criteria by two of us (J.M.L. and D.E.W.), independently; these data were then cross-checked to ensure accuracy.
A feature of the literature is that series of studies from the same institution or group of institutions commonly publish results in successive birth cohorts over time. However, overlapping reports, in which subgroups of infants are included in more than 1 report and cannot be separated out, are frequent. It is not unusual to find, for example, a study reporting outcomes for births from 1973 to 1978 and a second study reporting on births from 1977 to 1980. In addition, when institutions are grouped for reporting purposes, the infants described sometimes have previously been included in other publications from the same group. Thus, some infants are reported on more than once in this quantitative review. This was avoided to the extent possible by systematically cross-referencing all reports.
For ease of presentation, we refer to cohorts that are defined by gestational age as EI cohorts and to those that are defined by birth weight as ES cohorts. Outcomes of interest are survival and the prevalence of CP, mental retardation, blindness, deafness, and at least 1 of these disabilities. Data are presented as means and 95% confidence intervals (CI). Outcomes are presented as percentage of ES and EI livebirths and, except survival, as percentage of ES and EI survivors.
Comparisons of survival and prevalence of disabilities between EI and ES infants were made using χ2 tests with Yates correction. Marginal effects logistic regression analysis using the Huber-White robust variance estimator and weighted for cohort sample size was used to examine the relationships between prevalence of disabilities and mid-year of birth (ie, the mid-year of the interval in which patients included in a cohort were born) and survival. In these analyses, the upper gestational age limit used to define EI cohorts was included as a factor because survival was significantly lower and the prevalences of several disabilities among survivors were significantly higher in cohorts defined by an upper gestational age limit of 25 weeks than in those defined by an upper limit of 26 weeks. No such difference was found between ES cohorts defined by an upper birth weight limit of 749 or 750 g and 799 or 800 g. In addition, EI and ES cohorts were clustered by institution or geographic region to reduce heterogeneity.
Study criteria were fulfilled by 42 studies7- 48 describing 58 cohorts (Table 1 and Table 2); infants were classified by gestational age in 16 reports and by birth weight in 25 reports; 1 report11 described cohorts categorized each way.
In most EI cohorts, gestational age was determined by last menstrual period or ultrasonography before 20 weeks' gestation. Among 20 EI cohorts (Table 1), 16 were defined by an upper limit of 26 weeks and 4 were defined by an upper limit of 25 weeks. Seven EI cohorts were defined by a lower limit of 24 weeks and 11 were defined by a lower limit of 23 weeks; no lower limit was specified for 2 EI cohorts. Extremely small cohorts (Table 2) were more heterogeneous. Among 38 ES cohorts, 21 were defined by an upper limit of 799 or 800 g and 17 were defined by an upper limit of 749 or 750 g. Among the 21 cohorts whose upper limit was 799 or 800 g, 7 specified no lower weight limit, 2 were defined by a lower limit of 401 g, 9 were defined by a lower limit of 500 g, and 3 were defined by a lower limit of 501 g. Of the 17 cohorts defined by an upper limit of 749 or 750 g, 5 specified no lower limit, 8 were defined by a lower limit of 500 g, 2 were defined by a lower limit of 501 g, and 2 were defined by a lower limit of 600 g.
Survival differed between EI and ES cohorts (Table 1 and Table 2). In EI cohorts, overall survival was 40.9% (95% CI, 39.4%-42.4%); in ES cohorts, overall survival was 29.7% (95% CI, 28.4%-31.1%) (P<.001). Survival was significantly lower in EI cohorts defined by an upper gestational age limit of 25 weeks (28.1%; 95% CI, 25.2%-31.1%) than in EI cohorts defined by an upper limit of 26 weeks (44.6%; 95% CI, 42.9%-46.5%) (P<.001). Interpretation of these differences is confounded by differences in the birth years of the cohorts (because reports defining cohorts by gestational age have been more common recently) and by the population from which the cohort was drawn, ie, livebirths in a geographic region, livebirths in reporting hospital(s), admissions to reporting NIC unit(s), etc. Survival varied from 3.4% of livebirths in a geographic area (in a cohort of infants weighing 500-749 g born in 1978-1979)30 to 80.8% of livebirths in reporting hospitals (in a cohort of infants weighing 600-750 g treated with surfactant born in 1988-1989).47 In the 1 nationwide report11 of all liveborn infants in 1983 of 26 weeks' gestation or less and all liveborn infants weighing 499 to 749 g, survival did not differ significantly between the 2 groups (21.7% vs 27.7%) (P = .47).
Overall, the cohorts born more recently experienced improved survival. Survival was reported from the same institution(s) for the same birth weight ranges for 2 or more successive time intervals in 10 cases15,26,29,30,32,34,36- 38,40,44 and it improved over time in 8 of these. Figure 1 illustrates cohort survival in relation to mid-year of cohort birth for ES infants. Weighted logistic regression analysis showed a significant increase in survival over time in EI and ES infants. Survival increased 2.2% per year from 1980 to 1990 among EI infants and 2.1% per year from 1976 to 1990 among ES infants (P<.001 for both).
The survival rate for extremely small cohorts in relation to mid-year of birth. Only cohorts drawn from livebirth populations are included. The size of each white circle is in proportion to the number of livebirths in the cohort.
As with mortality, substantial variation was seen in the prevalence of CP at follow-up (Table 3). Criteria for inclusion as a patient with CP were variable. Most commonly (9 cohorts), spastic diplegia, hemiplegia, or quadriplegia was specified. In 7 cohorts, patients had to have "moderate to severe" CP for inclusion. In 2 cohorts, the criteria for patients with CP was nonambulation by age 2 years. In 1 cohort, only patients with "incapacitating" CP were included; in another, only patients with "severe" CP were included. In 3 cohorts, CP was defined as spasticity. In 5 cohorts, patients with "any" CP were included. The lowest risk of CP among survivors in any cohort with more than 10 survivors examined was 0.0%30,47; the highest was 26.4%.17 Criteria in these 3 studies were similar for age at assessment and for the years in which infants were born. Some variation across studies must be caused by differing criteria for the diagnosis of CP. In 1 of the cohorts with the lowest rate of CP, cases had to be of at least moderate severity for inclusion, but in the cohort with the highest rate of CP, any CP was counted. Nonetheless, it is unlikely that the differences among cohorts can be completely accounted for by classification issues. Although the criteria for age of assessment varied among studies, there was no significant correlation between age at assessment and CP prevalence.
The overall prevalence of CP among ES survivors was 7.5% (95% CI, 5.5%-10.0%), not significantly different from the 8.0% prevalence of CP among survivors in a recent meta-analysis of studies of infants weighing less than 1500 g.49 Thus, it seems that ES infants are not subject to a much higher risk of CP than are very low birth weight infants. Cerebral palsy prevalence in EI survivors, 12.2% (95% CI, 10.2%-14.6%), was higher than in ES survivors (P = .005). Cerebral palsy prevalence was significantly higher in the 1 EI cohort defined by an upper gestational limit of 25 weeks (26.4%; 95% CI, 19.4%-34.6%) than in the 8 EI cohorts defined by an upper limit of 26 weeks (9.7%; 95% CI, 7.7%-12.1%) (P<.001). However, the 1 EI cohort defined by an upper limit of 25 weeks defined CP as any; most of the other cohorts had more stringent criteria for this diagnosis. Each 100 EI livebirths contributed 2.7 (95% CI, 1.7-4.0) cases of CP, whereas each 100 ES livebirths yielded 1.7 (95% CI, 1.0-2.8) cases of CP.
Cognitive functioning was most commonly assessed using the Bayley Scales of Infant Development, which are suitable only for infants younger than 30 months of age. Cohorts with older children most commonly used the Stanford-Binet Scale. Four cohorts used the Wechsler Intelligence Scale for Children, 1 cohort used the Griffiths Scale, and 1 used the McCarthy General Cognitive Index.
Mental retardation was defined most commonly as a performance quotient less than 70 or a quotient more than 2 SDs below the mean for the cognitive test used. Unlike the case for CP, mental retardation did not differ between EI and ES cohorts (Table 4). The prevalence of mental retardation averaged 13.6% (95% CI, 11.5%-16.1%) among EI survivors and 14.3% (95% CI, 11.7%-17.4%) among ES survivors. The range in cohorts with more than 10 survivors examined was 0.0%47 to 23.3%.22 Mental retardation was more common than CP in ES survivors, indicating that mental retardation without accompanying diagnosable motor deficit was not rare in these infants. Each 100 livebirths of EI infants produced 3.9 (95% CI, 2.7-5.4) mentally retarded children, and each 100 livebirths of ES infants produced 4.5 (95% CI, 3.3-6.2) such children.
No definition of vision impairment was provided in most studies; affected infants were simply described as blind. In 5 cohorts, corrected visual acuity of less than 20/200 in the better eye was specified. The term "visual impairment" was used for 1 cohort, with the description "severe" added in another 2. Two studies22,44 describing 3 cohorts included monocular blindness in the definition. The criterion for deafness was the requirement of a hearing aid in 10 cohorts. In another 8 cohorts, the terms "deaf" or "hearing impaired" were used. "Severe" deafness or hearing impairment was specified in 4 cohorts. One study44 describing 2 cohorts included unilateral deafness in the definition.
Blindness proved to be a more common problem than deafness in EI (7.5% [95% CI, 5.9%-9.5] vs 2.6% [95% CI, 1.7%-3.8]) and ES (7.8% [95% CI, 6.1-10.0] vs 2.9% [95% CI, 1.8-4.6]) survivors (Table 5). The prevalence of blindness (but not deafness) was higher in the 1 EI cohort defined by an upper gestational age limit of 25 weeks (14.7%; 95% CI, 9.5%-22.0%) than in the 8 EI cohorts defined by an upper limit of 26 weeks (6.2%; 95% CI, 4.6%-8.3%) (P = .002). Each 100 EI livebirths yielded 1.3 (95% CI, 0.7-2.4) cases of blindness and 0.8 (95% CI, 0.3-1.6) cases of deafness in childhood. Each 100 ES livebirths yielded 2.1 (95% CI, 1.4-3.4) cases of blindness and 0.7 (95% CI, 0.3-1.6) cases of deafness in childhood.
Because disabilities tend to cluster in the same patient, the total of all disabilities exceeds the total number of children with at least 1 disability. The proportion of survivors with at least 1 disability was available for all but 1 of the cohorts. Mental retardation, CP (of varying severity), blindness, and deafness were included as disabilities in almost all of these 57 cohorts; 1 cohort was not routinely screened for deafness.33 In some cohorts, however, hydrocephalus and seizures also were included as disabilities. The individual prevalence of these 2 conditions could not be estimated, however, because they were not reported separately in most cohorts.
Overall, 22.1% (95% CI, 20.0%-24.5%) of EI survivors (Table 6) and 24.2% (95% CI, 21.9%-26.6%) of ES survivors (Table 7) were classified as disabled. In studies with more than 10 survivors examined, the prevalence of disability ranged from 0.0%11,30 to 52.4%.16 The prevalence of disability was higher in the EI cohorts defined by an upper gestational age limit of 25 weeks (30.6%; 95% CI, 25.1%-36.6%) than in the EI cohorts defined by an upper limit of 26 weeks (22.5%; 95% CI, 17.7%-22.5%) (P<.001). Each 100 livebirths of EI infants yielded 7.1 (95% CI, 6.1-8.2) disabled children and each 100 livebirths of ES infants yielded 6.7 (95% CI, 5.7-7.8) disabled children.
Among EI survivors and livebirths, the prevalences of CP and mental retardation increased significantly over time. The prevalences of blindness and deafness did not change significantly over time among either EI survivors or livebirths.
Among ES survivors, the prevalences of these specific disabilities did not change significantly over time. Among ES livebirths, the prevalence of CP significantly increased and mental retardation tended to increase over time; the prevalences of blindness and deafness did not change.
Figure 2, top, illustrates the prevalence of at least 1 major disability among ES survivors in relation to mid-year of birth. The prevalence of disability among survivors did not change by weighted logistic regression analysis. Results were similar for EI survivors. By contrast, the prevalence of disability among ES livebirths increased significantly over time (P<.004), as illustrated in Figure 2, bottom. Results were similar for EI livebirths. An additional 6 and 4 disabled children were estimated to have been added to the population per year for every 1000 EI and ES livebirths, respectively.
Top, The prevalence of disability among extremely small survivors in relation to mid-year of birth. The size of each circle is in proportion to the number of survivors in the cohort. Bottom, The percentage of extremely small livebirths surviving with at least 1 disability in relation to mid-year of birth. Only cohorts drawn from livebirth populations are included. The size of each circle is in proportion to the number of livebirths in the cohort.
Among EI survivors, the prevalences of CP, mental retardation, blindness, and deafness did not differ significantly with higher survival. However, among EI livebirths, the prevalences of CP and mental retardation were significantly higher with higher survival. The prevalence of blindness did not differ significantly; the prevalence of deafness tended to decrease.
Among ES survivors, the prevalences of CP, blindness, and deafness did not differ significantly with higher survival; the prevalence of mental retardation was significantly lower with higher survival. Among ES livebirths, the prevalences of CP and blindness were significantly higher and the prevalence of deafness tended to be lower with higher survival; the prevalence of mental retardation did not differ significantly.
Figure 3, top, illustrates the prevalence of at least 1 major disability among ES survivors in relation to survival. The prevalence of disability among ES survivors did not differ significantly with higher survival by weighted logistic regression analysis. By contrast, the prevalence of disability among ES livebirths was significantly higher with higher survival (P = .005), as illustrated in Figure 3, bottom. Results were similar for EI livebirths. An additional 2 disabled children were estimated to have been added to the population per 1000 EI or ES livebirths for each 1% increment in survival.
Top, The prevalence of disability among extremely small survivors in relation to survival. The size of each circle is in proportion to the number of survivors in the cohort. Bottom, The percentage of extremely small livebirths surviving with at least 1 disability in relation to survival. Only cohorts drawn from livebirth populations are included. The size of each circle is in proportion to the number of livebirths in the cohort.
Information about neurodevelopmental outcomes in ES and EI infants is particularly important because of the increased ability and willingness to provide them NIC and of the striking improvement in survival with which this has been associated during the past 20 years.50- 52 The largest and most mature of this group of infants (those with birth weights of 700-800 g or gestational ages of 25-26 weeks) are now surviving at a rate of approximately 80% in some medical centers.53 Ideally, then, counseling of parents about prognosis should be based on outcomes among survivors born no more than 5 years ago. In reality, however, we must rely on older data because the availability of reliable outcome data lags considerably behind current practice.
In addition to this shortcoming, interpretation of this analysis is complicated by many other factors. First, the studies are heterogeneous. Cohorts vary in important characteristics: the population from which the cohort was drawn (livebirths in a geographic region, livebirths at a single hospital, and NIC unit admissions), the gestational age and birth weight truncation points, the age at follow-up, the criteria for diagnosis of specific neurodevelopmental outcomes, and the neurodevelopmental outcomes reported. Second, overlapping cohorts are sometimes reported in multiple studies. Thus, great care is required to minimize the number of patients counted more than once, and double-counting of patients cannot entirely be avoided. Third, information about important characteristics that may affect neurodevelopmental outcome is frequently not reported. Most studies do not report the distribution of sex, inborn or outborn status, appropriateness of birth weight for gestational age, or socioeconomic status. Although individual studies report the effect of some of the above characteristics on outcome, a systematic analysis of the effect of these potentially confounding factors is not possible. Fourth, publication bias cannot be excluded. It is possible that institutions with unusually poor outcomes would not report them and that reports of unusually good outcomes would not be accepted for publication because of skepticism of reviewers. However, reports of unusually poor and unusually good outcomes were identified. Fifth, most of the studies available are before the use of surfactant. However, studies available have not yet found a relationship between surfactant use and neurodevelopmental outcome15,16,44,46,54- 57; and antenatal corticosteroid use has not been reported to improve neurodevelopmental outcome.58
Irrespective of these shortcomings, such outcome information as is available, with its limitations well understood, is essential for patient care. Parents require these data to make informed management decisions. Professionals must be aware of these data also, but there are indications that many are not.59,60 Finally, outcomes must be known so that the resources that these increasing numbers of survivors will require can be anticipated and their neurodevelopmental outcomes optimized.
We limited our review to major disabilities because most reports make their assessments at 18 to 36 months of age, when only major disabilities can be assessed with any degree of reliability. Outcomes in school-aged survivors are available in few studies and then only for infants born more than 10 years ago. The most frequent neurodevelopmental disability was mental retardation, occurring in 14.0% of EI and ES survivors. The prevalence of CP was 12.2% in EI survivors and 7.5% in ES survivors (P = .005). The prevalence of blindness was 8.0% and of deafness was 3.0% in EI and ES survivors. Overall, 22.1% of EI survivors and 24.2% of ES survivors were disabled, the latter defined as any combination of mental retardation, CP, blindness, and deafness (and in some reports hydrocephalus or seizure disorder).
Among EI survivors, the prevalences of CP and mental retardation increased significantly from 1980 to 1990; however, this increase does not seem to be related to an increase in survival because there was no significant relationship between the prevalences of CP or mental retardation among EI survivors and survival. The prevalences of blindness and deafness among EI survivors did not change over time and did not vary with survival. Neither did the prevalence of any disability. Among ES survivors, the prevalences of most disabilities among survivors did not change over time and did not vary with survival; however, the prevalence of mental retardation among ES survivors was lower with higher survival.
Among EI and ES livebirths, the prevalence of disabilities increased with improvement in survival over time, although this was not true in all cases for all specific disabilities. At the same time, the prevalence of survivors without major neurodevelopmental disabilities increased in parallel.
In summary, provision of NIC to EI and ES infants has resulted in steadily increasing survival of children at high risk for neurodevelopmental disabilities. It is critical that outcomes of EI and ES infants be continually assessed over time so that prognostic information can be as relevant as possible to current NIC. Standardized reporting of risk factors and major outcomes, at least at age 2 years, as has been recommended by others,61- 64 would greatly facilitate this process. We submit that NIC of such infants requires a parallel commitment to providing the resources necessary for long-term assessment of their outcomes and the resources required to optimize their quality of life.
Accepted for publication December 1, 1997.
Supported by grant RO1 HS 08396 from the Agency for Health Care Policy Research, Washington, DC.
We thank the Sparrow Health System Medical Library, Lansing, Mich, for assistance in obtaining the references reviewed for the manuscript.
Editor's Note: The good news is that the vast resources expended in perinatology have resulted in decreases in mortality among micro-preemies; the bad news is that the same resources are not forthcoming to address adequately the problems of these survivors.—Catherine D. DeAngelis, MD
Reprints: John M. Lorenz, MD, Sparrow Hospital, PO Box 30480, Lansing, MI 48909-7980 (e-mail: firstname.lastname@example.org).
Lorenz JM, Wooliever DE, Jetton JR, Paneth N. A Quantitative Review of Mortality and Developmental Disability in Extremely Premature Newborns. Arch Pediatr Adolesc Med. 1998;152(5):425-435. doi:10.1001/archpedi.152.5.425