Prevalence of cerebral palsy/1000 (top panel) and daily number of births (bottom panel) according to gestational age at birth from 37 weeks to 44 weeks among the 1 682 441 members of the study cohort. The smoothed curve (top panel) was fitted using the locally weighted scatterplot smoothing (LOESS) method.
Information on cerebral palsy and digestive diseases (International Classification of Diseases, Ninth Revision [ICD-9] codes: 520-579, ICD-10 codes: K00-K93) is from the study cohort of 1 682 441, whereas information on congenital anomalies is based on 1 720 443 (original study cohort without exclusion of congenital anomalies). Error bars indicate 95% confidence intervals. Reference category for all plots is 40 weeks.
Relative risk of cerebral palsy (adjusted for year of birth, infant sex, maternal age, single motherhood, mother and father's level of education, and immigrant status of the parents) related to gestational age in a subcohort of 139 976 individuals born from December 1998 through 2001. LMP indicates last menstrual period; errors bars indicate 95% confidence intervals. Reference category for both plots is 40 weeks.
Moster D, Wilcox AJ, Vollset SE, Markestad T, Lie RT. Cerebral Palsy Among Term and Postterm Births. JAMA. 2010;304(9):976-982. doi:10.1001/jama.2010.1271
Author Affiliations: Department of Public Health and Primary Health Care (Drs Moster, Vollset, and Lie) and Department of Clinical Medicine, Section for Pediatrics (Dr Markestad), University of Bergen, Bergen, Norway; Department of Pediatrics, Haukeland University Hospital, Bergen, Norway (Drs Moster and Markestad); Epidemiology Branch, National Institute of Environmental Health Sciences, National Institutes of Health, Durham, North Carolina (Dr Wilcox); and Medical Birth Registry of Norway, Norwegian Institute of Public Health, Bergen (Drs Vollset and Lie).
Context Although preterm delivery is a well-established risk factor for cerebral palsy (CP), preterm deliveries contribute only a minority of affected infants. There is little information on the relation of CP risk to gestational age in the term range, where most CP occurs.
Objective To determine whether timing of birth in the term and postterm period is associated with risk of CP.
Design, Setting, and Participants Population-based follow-up study using the Medical Birth Registry of Norway to identify 1 682 441 singleton children born in the years 1967-2001 with a gestational age of 37 through 44 weeks and no congenital anomalies. The cohort was followed up through 2005 by linkage to other national registries.
Main Outcome Measures Absolute and relative risk of CP for children surviving to at least 4 years of age.
Results Of the cohort of term and postterm children, 1938 were registered with CP in the National Insurance Scheme. Infants born at 40 weeks had the lowest risk of CP, with a prevalence of 0.99/1000 (95% confidence interval [CI], 0.90-1.08). Risk for CP was higher with earlier or later delivery, with a prevalence at 37 weeks of 1.91/1000 (95% CI, 1.58-2.25) and a relative risk (RR) of 1.9 (95% CI, 1.6-2.4), a prevalence at 38 weeks of 1.25/1000 (95% CI, 1.07-1.42) and an RR of 1.3 (95% CI, 1.1-1.6), a prevalence at 42 weeks of 1.36/1000 (95% CI, 1.19-1.53) and an RR of 1.4 (95% CI, 1.2-1.6), and a prevalence after 42 weeks of 1.44 (95% CI, 1.15-1.72) and an RR of 1.4 (95% CI, 1.1-1.8). These associations were even stronger in a subset with gestational age based on ultrasound measurements: at 37 weeks the prevalence was 1.17/1000 (95% CI, 0.30-2.04) and the relative risk was 3.7 (95% CI, 1.5-9.1). At 42 weeks the prevalence was 0.85/1000 (95% CI, 0.33-1.38) and the relative risk was 2.4 (95% CI, 1.1-5.3). Adjustment for infant sex, maternal age, and various socioeconomic measures had little effect.
Conclusion Compared with delivery at 40 weeks' gestation, delivery at 37 or 38 weeks or at 42 weeks or later was associated with an increased risk of CP.
Cerebral palsy (CP) is the most common cause of physical disability in childhood, with limitations that persist throughout life.1- 3Quiz Ref IDCerebral palsy is characterized by nonprogressive disorders of movement and posture, presumed to result from insult to the brain during fetal or early infant life.4 These motor problems are often accompanied by disturbances of cognition and other neurologic difficulties.1,3,4 Cerebral palsy can be a severe disability and a substantial burden for the affected individual's family and society.
The underlying causes of CP remain largely unknown.5 Cerebral palsy is associated with complicated labor and delivery, but most cases have little apparent association with delivery care.6,7 One of the strongest predictors of CP is preterm birth, with the risk of CP increasing steadily with earlier delivery.8 Although risk is lower among term births, about three-fourths of all infants with CP are born after 36 weeks.8 Within this range of term births, there are few data on the possible association of CP with gestational age. We explored the relation of CP risk with gestational age among term and postterm births using national health and insurance registries in Norway.
Each Norwegian citizen has a unique identification number that allows linkage of an individual's data among the national registries. The Medical Birth Registry of Norway has collected information since 1967 on all births with a gestational age of at least 16 weeks.9 Information on gestational age is based on the last menstrual period (LMP). Using the birth registry, we identified all singleton live births from 1967 through 2001 with a gestational age of 37 to 44 weeks. Follow-up data were available through 2005. To remove likely errors of gestational age based on LMP, we excluded infants with birth weights more than 3 standard deviations from the mean for a given gestational-age week, stratified by infant sex.10 Birth defects are associated with CP and also with gestational age at birth.11 To help remove possible confounding by birth defects, we also excluded infants with any registered congenital anomalies. Birth defects were coded using International Classification of Diseases (ICD) versions 8 and 10 in the Medical Birth Registry and versions 9 and 10 in the insurance registry.12
Cerebral palsy cannot be diagnosed at birth. Our data on CP came from the Norwegian compulsory health insurance system.13 Individuals in Norway are entitled to benefits for a disability that involves significant expenses, requires special attention or nursing, or reduces working capacity by at least 50%.14 These benefits are provided without regard to income or wealth and are considered to provide a fairly complete registry of disabled individuals in Norway.
Benefits are based on physician diagnosis, with diagnoses registered according to ICD-9 or ICD-10. We obtained information on CP (ICD-9 codes: 342-344; ICD-10 codes: G80-G83) using the medical diagnoses linked to the disability benefits defined previously. A validation study15 has shown that registered CP diagnoses correspond well with medical records, with some underregistration of the mildest cases.
There is controversy regarding the youngest age at which CP can be reliably diagnosed.16,17 Most cases are established by 4 years, and so we required our cohort to survive and be followed up to at least the age of 4 years. Information on deaths was obtained through the Cause of Death Registry. Information on parents' education and immigration status came from the Norwegian National Education Database and Statistics Norway.18
The study was approved by the Norwegian Data Inspectorate, the Norwegian Labor and Welfare Organization, the Office of the National Registrar, and the Norwegian Directorate of Health. The approval included a waiver of individual study participant consent.
We first estimated the prevalence of CP according to gestational day of birth for all births from 37 to 44 weeks. In further analyses, we categorized gestational age into 7 groups: 37 weeks (ie, 259 to 265 days), 38 weeks, 39 weeks, 40 weeks, 41 weeks, 42 weeks, and 43 through 44 weeks. Ratios of prevalences were used to estimate relative risks (RRs), with 40 weeks as the reference. We estimated the RR of CP for each gestational group using log-binomial regression and adjusted for year of birth, sex, maternal age, single motherhood, mother's level of education, father's level of education, and immigrant status of the parents. We plotted the association of gestational age with congenital anomalies and diseases of the digestive system that are unlikely to be caused by time of delivery to explore the possibility of reverse causation or selection.
Date of LMP is an error-prone measure of gestational age.19 Since at least 1994, an estimated 98% of pregnant women in Norway have received a routine ultrasound examination by the mid-second trimester.20 These data have been reported in the Medical Birth Registry starting in December 1998. We explored the quality of LMP data by repeating the overall analysis using ultrasound measures of gestational age for the subset of infants born since December 1998.
Year of birth, mother's level of education, and father's level of education were entered as continuous variables; the remaining were categorical. Maternal age was divided into 3 categories (<18 years, 18-39 years, and ≥40 years). Father's and mother's education was defined on a 9-level scale, with 0 for no education and 8 (the highest) for a doctoral degree. A child was considered to have immigrant parents if both parents were born outside of Norway. Since changes in CP prevalence and gestational–age distribution over the 35 years of the study might have affected our results, we repeated the analyses in 3 smaller time windows (1967-1977, 1978-1989, and 1990-2001).
SPSS version 17.0 was used for statistical analyses (SPSS Inc, Chicago, Illinois). All tests were 2-sided with a 5% significance level.
There were 2 024 215 live births in Norway from 1967 through 2001. We excluded those missing data on gestational age (6.0%), those born preterm (5.5%), those with gestational age more than 44 weeks (1.0%), those whose birth weights were incompatible with their gestational age (0.7%), twins or other multiple births (1.3%), those with malformations recorded either in the birth registry or the insurance registry (2.0%), and those who died younger than 4 years of age (0.4%). This left an analysis cohort of 1 682 441 singleton births with a gestational age of at least 37 weeks.
Figure 1 shows the distribution of these births by gestational day at delivery, with a summary smoothed curve. There were 1938 individuals subsequently registered with CP (1.15/1000 births; 95% confidence interval [CI], 1.10-1.20). Delivery at 40 weeks was associated with lowest CP risk (prevalence, 0.99/1000; 95% CI, 0.90-1.08). Compared with delivery at 40 weeks, prevalence of CP at 37 weeks was 1.91/1000 (95% CI, 1.58-2.25) and RR was 1.9 (95% CI, 1.6-2.4); the prevalence at 38 weeks was 1.25/1000 (95% CI, 1.07-1.42) and RR was 1.3 (95% CI, 1.1-1.6); the prevalence of CP at 42 weeks was 1.36/1000 (95% CI, 1.19-1.53) and RR was 1.4 (95% CI, 1.2-1.6); and after 42 weeks, the prevalence was 1.44 (95% CI, 1.15-1.72) with RR of 1.4 (95% CI, 1.1-1.8) (Figure 2).
Table 1 and Table 2 show characteristics of the parents, infants, and deliveries for children with and without subsequent CP. Children with CP were slightly more likely to have single mothers (P = .03) and parents with less education (P < .001). Complications of labor and delivery (breech presentation, cesarean delivery) were up to twice as likely for infants who were later diagnosed with CP (P < .001). Boys were overrepresented among CP children (P < .001). Children with CP had lower mean birth weight (3437 g vs 3585 g; P < .001) and smaller head circumference (35.1 cm vs 35.3 cm; P < .001). Children with CP were 82 times more likely to have had an Apgar score below 4 (P < .001) and were 8 times more likely to have been transferred to a pediatric unit after delivery (P < .001). Most of these associations were present at each gestational age of delivery.
Adjustment for year of birth, infant sex, maternal age, presence or absence of a partner, educational level of the mother and father, and immigrant status of the parents produced identical RRs by gestational age (Figure 2). Congenital malformations had a weak U-shaped association with gestational age, while diseases of the digestive system did not (Figure 2; supplementary data in eTable).
Ultrasound-based measurements of gestational age were available for the subset of 139 976 children born in December 1998 and later, including 85 with CP. Among this subgroup, the pattern of CP risk with gestational age was stronger with ultrasound-based gestational age than with LMP (Figure 3). Compared with infants born at 40 weeks (prevalence, 0.36/1000; 95% CI, 0.18-0.54), infants born at 37 weeks had a CP prevalence of 1.17/1000 (95% CI, 0.30-2.04) and RR of 3.7 (95% CI, 1.5-9.1), while infants at 42 weeks had a CP prevalence of 0.85/1000 (95% CI, 0.33-1.38) and RR of 2.4 (95% CI, 1.1-5.3), based on ultrasound dating of gestational age.
The prevalence of CP among births from 37 weeks onward decreased from 1.4/1000 (95% CI, 1.3-1.5) in 1967-1971 to 0.7/1000 (95% CI, 0.6-0.8) in 1997-2001 (eFigure 1). The association of CP with gestational age, however, was consistent over time (eFigure 2).
Much attention has been given to the high risk of CP with preterm delivery.21- 25 The risk of CP among term infants is much lower, but nonetheless most CP occurs among term deliveries. Furthermore, the risk at term and beyond is not uniform. The risk of CP is lowest at 40 weeks, with the highest risks at 37 weeks and at 42 weeks or later. A recent report supports our finding of increased CP risk among postterm births.16 Other neurologic conditions have also been found to vary by gestational age at term. A U-shaped pattern for low IQ was recently reported among term and postterm births.26
Our findings do not appear to be due to errors in LMP. The U-shaped risk was even stronger among the subgroup with ultrasound measures of gestational age (which are still imperfect, but with less error than LMP).19 Such a strengthening of effect would be expected if the association with gestational age were true but blunted by measurement error related to LMP.
We had information on several parental and birth characteristics associated with a subsequent diagnosis of CP. These risk factors are consistent with results from other studies.15,27- 31 The distribution of these risk factors across 37- to 44-week births was generally similar for the CP and non-CP group. Adjustment for parental characteristics had no influence on the results. We did not adjust for circumstances of labor, delivery, or neonatal period because these may be part of the causal pathway or an early expression of CP.
Cerebral palsy risk has a robust U-shaped association with gestational age among infants reaching term. The biological mechanisms that underlie this association are not as clear. One possible interpretation is that delivery too early or too late, even within the limited range of term and postterm births, increases the risk of CP.
However, an equally plausible interpretation is that fetuses predisposed to CP have a disturbance in the timing of their delivery, which causes them to be more often delivered early or late. This apparently happens with other fetal conditions: there is a U-shaped pattern in the risk of congenital anomalies with gestational age after 37 weeks. Quiz Ref IDSince congenital anomalies are not caused by the timing of delivery, the most plausible explanation is reverse causation; malformed infants experience disruptions in their time of delivery, with increased chance of delivery either earlier or later than 40 weeks.
Although the forces that regulate timing of a normal delivery are poorly understood,32,33 it appears that the types of malformations most likely to disrupt the timing of delivery often involve cerebral function. For example, anencephalic fetuses have a tendency to be born postterm,34,35 children with Trisomy 18 to be born preterm or postterm,36 and children with Down syndrome to be born early.37 It is possible that the cerebral damage later expressed as CP similarly disrupts time of delivery.
The question of causal interpretation has been a long-standing puzzle for CP in another important respect. Labor and delivery complications are commonly found in infants subsequently diagnosed with CP. This could indicate that CP is caused by damage to the infant in the course of delivery, but it could also mean that the prenatal factors leading to CP cause problems of delivery.16,38 The lower birth weight and head circumference at birth among CP cases suggest that these children differ from non-CP infants even before birth. As with most previous studies, our study has no possibility of discerning prenatal from perinatal or postnatal causes of CP.
A strength of this study is its population-based cohort design. This provides statistical power for gestational-age–specific estimates of risk during the term and postterm periods, when risk of CP is low. The study design also avoids problems of selective participation, detection bias, recall bias, and loss to follow-up.
We were able to adjust for possible risk factors including maternal age, single motherhood, mother's and father's level of education, immigrant status of the parents, and infant sex. These factors made little difference to the results.
One notable weakness of the study is the lack of any information on subtypes of CP. We are unable to determine, for example, whether the risks with gestational age are more pronounced for certain subtypes. Furthermore, by using a disability register to identify CP, we may have missed information on some of the mildest CP cases, as suggested by an earlier validation study.15 The CP cases in the present study may in general be more severe than if they had been identified by hospital or clinic records.
The exclusion of children who died before 4 years of age could potentially bias the results by excluding the most severe cases of CP. However, of the 7096 children who died before 4 years of age, only 3 had a diagnosis of CP. Repeated analyses after inclusion of all children who died before the age of 4 years had no effect on the results.
Quiz Ref IDAnother possible difficulty is the accumulation of births over such a long time interval (35 years). Changes in obstetric practice, CP diagnosis, and the recording of gestational age or other study variables may have affected the results. However, when stratifying by time period, there was no evidence that the association of gestational age with CP differed over time.
Neonatal mortality declined by 75% during the study period. Some children who survived with CP in more recent years might have died if they had been born in earlier years, which could lead to an increase in severe cases of CP over time. What we observe, however, is the opposite—the prevalence of CP has declined slightly over time. Factors that have reduced neonatal mortality may also have reduced the risk of CP.
Quiz Ref IDClinicians typically regard term births (37-41 weeks12) as low risk, with the possibility of increased risk with postterm delivery.39 This standard definition of term does not correspond well with the period of lowest risk for CP in this study or with the weeks when most infants are born. Weeks 37 and 38 seem more to resemble weeks 42 and 43, both in CP risk and in the general likelihood of delivery, leaving 39 to 41 weeks as the optimum time for delivery.
If the time of delivery affects CP risk, then intervention at 40 weeks might reduce CP risk, while elective delivery at 37 or 38 weeks might increase it. If infants prone to CP are disrupted in their delivery times, the prevalence of CP would be unchanged regardless of time of delivery. A definitive answer would require a randomized clinical trial of deliveries at various gestational ages—an impractical option, given the very low prevalence of CP. A clue might come from the U-shaped risk of congenital anomalies with gestational age, which is definitely not causal. Quiz Ref IDIf the same occurs with CP, then a change in time of delivery would have no influence on a child's underlying risk of CP. Until the biological mechanisms for these patterns of risk in term and postterm births are better understood, it would be hasty to assume that interventions on gestational age at delivery could reduce the occurrence of CP.
Corresponding Author: Dag Moster, MD, PhD, Department of Public Health and Primary Health Care, University of Bergen, PO Box 7804, N-5020 Bergen, Norway (Dag.Moster@smis.uib.no).
Author Contributions: Dr Moster had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Moster, Wilcox, Vollset, Markestad, Lie.
Acquisition of data: Moster.
Analysis and interpretation of data: Moster, Wilcox, Vollset, Markestad, Lie.
Drafting of the manuscript: Moster.
Critical revision of the manuscript for important intellectual content: Moster, Wilcox, Vollset, Markestad, Lie.
Statistical analysis: Moster, Lie.
Administrative, technical, or material support: Moster.
Study supervision: Moster, Wilcox, Vollset, Markestad, Lie.
Financial Disclosures: None reported.
Funding/Support: This research was supported in part by the Intramural Research Program of the National Institutes of Health, National Institute of Environmental Health Sciences; the Unger-Vetlesen Charitable Fund; The US-Norway Fulbright Foundation; the Norwegian Society of Pediatricians; the University of Bergen; and the Research Council of Norway.
Role of the Sponsors: The funding agencies had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.