What this paper adds
Ten risk factors were identified for CP in term-born infants by all studies examining them.
Few identified risk factors are currently preventable.
Preventive efforts should be extended to include these risk factors
Aim The aim of this study was to conduct a systematic review in order to identify the risk factors for cerebral palsy (CP) in children born at term. The secondary aim was to ascertain if the potential for prevention of these risk factors has been adequately explored.
Method A MEDLINE search up to 31 July 2011 was completed, following the Meta-Analysis of Observational Studies in Epidemiology guidelines. Publications were reviewed to identify those with both a primary aim of identifying risk factors for all children or term-born children with CP and a cohort or case–control study design. Studies were examined for potential chance or systematic bias. The range of point estimates of relative risk is reported.
Results From 21 articles meeting inclusion/exclusion criteria and at low risk of bias, data from 6297 children with CP and 3 804 791 children without CP were extracted. Ten risk factors for term-born infants were statistically significant in each study: placental abnormalities, major and minor birth defects, low birthweight, meconium aspiration, instrumental/emergency Caesarean delivery, birth asphyxia, neonatal seizures, respiratory distress syndrome, hypoglycaemia, and neonatal infections. Strategies for possible prevention currently exist for three of these.
Interpretation Ten consistent risk factors have been identified, some with potential for prevention. Efforts to prevent these risk factors to interrupt the pathway to CP should be extended.
Ten risk factors were identified for CP in term-born infants by all studies examining them.
Few identified risk factors are currently preventable.
Preventive efforts should be extended to include these risk factors
Cerebral palsy (CP) is the most common physical disability of childhood. CP describes a group of disorders of movement and posture that are also often accompanied by associated impairments and secondary musculoskeletal problems.1 Term-born children account for 50 to 65% of children with CP, and they tend to be more severely impaired than children with CP born preterm.2 Moreover, the severity of disability in the term-born group seems to be increasing.3 The incidence of CP among term-born infants ranges between 1 and 1.7 per 1000 live births, suggestive of a rare outcome.2–5 However, with 130 million infants born worldwide each year,6 93% of whom are born at term, such incidence rates suggest that between 120 000 and 217 600 new cases of CP occur each year among term-born children, making this an important yet under-researched group with chronic lifelong disability.
Over the past 50 years real progress has been made in the conceptual understanding of the aetiology of CP. We have confirmed that sentinel events around birth are responsible for a small proportion of CP. More typically, risk factors and multiple events interact in a cascade with additive effects.7,8 However, despite the progress in conceptual understanding, there has been no discernible decrease in overall rates. Therefore, we asked what clinically meaningful messages could be drawn from the literature of the last 50 years to guide us towards prevention in the coming decades.
The research questions for this study were (1) ‘What risk factors for CP have been identified in the literature for infants born at term?’ and (2) ‘Has their preventive potential been adequately explored?’
We followed the recommendations for reporting in the Meta-Analysis of Observational Studies in Epidemiology guidelines.9 The search strategy was developed by two of the authors (SM and SG) and a librarian. We searched MEDLINE and EMBASE for reports published up to 31 July 2011 (with no restriction on earliest search date) using a sensitive methodological filter for aetiological studies. Search terms were (risk factor*) OR (aetiol* OR etiol*) AND cerebral palsy, as well as MeSH (Risk Factor OR Aetiology) AND cerebral palsy with limits of humans.
Inclusion criteria for the study were as follows: (1) CP was the primary outcome (using a known definition appropriate for the time of the original study); (2) the primary aim of the study was to identify risk factors for CP in (a) all births or (b) infants born at term; (3) the study followed a cohort or case–control design; and (4) the publication had been peer reviewed.
Exclusion criteria for the study were as follows: (1) the primary aim was (a) to identify risk factors in those born preterm, (b) to classify aetiology by epoch without identifying specific risk factors, (c) to identify risk factors for a broader definition of adverse outcome that may include CP such as neurodevelopmental impairment, periventricular leukomalacia, learning disability*, epilepsy, etc., from which the risk factors specific to CP cannot be identified, or (d) to identify genetic risk factors, as systematic reviews dedicated entirely to genetic risk factors have been recently published; (2) original observations were not reported (e.g. literature reviews and editorials); (3) there was no control group (e.g. case series, case reports); (4) the study was carried out in developing countries and the risk factor profile pertained to causal pathways currently routinely prevented or of limited applicability in developed countries (e.g. jaundice resulting from rhesus isoimmunization). Figure 1 outlines the final selection process.
The first author (SM) reviewed all 4899 titles/abstracts identified by our search strategy and two other authors (DT and SG) between them reviewed each a second time. The authors selected 122 for full-text review, of which 32 were identified as likely to meet the inclusion criteria. A third independent reviewer (EB) examined these 32 studies against the inclusion and exclusion criteria and recommended that seven further papers be excluded based on exclusion criteria 1a and 4. There was full agreement with this decision. The reference lists of the 25 articles selected were hand searched for further relevant articles, but none was found.
The data extraction tool was designed a priori by the principal reviewer (SM) and included study design, country, birth years, gestational age range considered, numbers with and without CP, numbers and percentages with and without each risk factor, definitions used for each risk factor, and unadjusted and adjusted odds ratios if reported. Extraction of data from each of the 25 articles was carried out independently by two investigators (SM and DT/NB). Discrepancies were discussed and a final decision was made without requiring further consultation (Table I).9-34
|Study details||Country||Study design||Birth years||All CP/term/>2500g||Nr of children with CP||Nr of children without CP||Risk factor||Assessment of risk of biasa|
|Infant selection||Asses. CP and RF||Asses. RF||Assessor blinding||FU/recruitment time||Control for confounding||Nr of observations||Report dose response|
|Dale and Stanley10||Australia||Case–control||1956–1975||All||208||207||P; A; I; N||Adq||Adq||Adq||Inad||Adq||Inad||Adq||NA|
|Blair and Stanley11||Australia||Case–control||1975–1980||All||183||549||P; A; I; N||Adq||Adq||Adq||Adq||Adq||Adq||Adq||Adq|
|Palmer et al.12||Australia||Case–control||1980–1986||All||215||645||A; I||Adq||Adq||Adq||Inad||Adq||Adq||Adq||Adq|
|Dite et al.13||Australia||Case–control||1983–1990||Term||204||816||P; A; I; N||Adq||Adq||Adq||Inad||Adq||Adq||Adq||Adq|
|Walstab et al.14||Australia||Case–control||1991–1992||All||148||291||P; A; I||Adq||Adq||Adq||Adq||Inad||Adq||Adq||NA|
|Walstab et al.15||Australia||Case–control||1983–1992||Term||59||122||N||Adq||Adq||Adq||Adq||Inad||Adq||Inad||NA|
|Fritsch and Haidvogl16||Austria||Case–control||1962–1976||All||178||100||P; A; I; N||Inad||Inad||Inad||Inad||Adq||Inad||Adq||NA|
|Nielsen et al.17||Denmark||Case–control||1982–1990||All||271||217||P; A; I; N||Adq||Adq||Adq||Inad||Adq||Adq||Adq||Adq|
|Stelmach et al.18||Estonia||Case–control||1985–1995||All||153||268||A; I; N||Adq||Adq||Adq||Inad||Adq||Adq||Adq||Adq|
|Petridou et al.19||Greece||Case–control||1984–1988||All||103||254||P; A; I; N||Inad||Inad||Inad||Inad||Adq||Inad||Adq||Adq|
|Curatolo et al.20||Italy||Case–control||Not able to identify||All||64||209||P; A; I; N||Inad||Inad||Adq||Inad||Adq||Adq||Inad||NA|
|Moster et al.21||Norway||Cohort||1967–2001||Term||1938||1 680 503||P; N||Adq||Adq||Adq||Inad||Adq||Adq||Adq||Adq|
|Kulak et al.22||Poland||Cohort||1990–2005||Term||213||280||P; A; I; N||Adq||Adq||Adq||Inad||Adq||Adq||Adq||NA|
|Thorngren-Jerneck and Herbst23||Sweden||Case–control||1984–1998||All||1927||1 574 734||P; A; I||Adq||Adq||Adq||Adq||Adq||Adq||Adq||Adq|
|Gurbuz et al.24||Turkey||Case–control||1990–2000||Term||101||308||P; A; I; N||Adq||Adq||Adq||Adq||Adq||Adq||Adq||Adq|
|Gaffney et al.25||UK||Case–control||1984–1987||Term||141||257||A; I; N||Adq||Adq||Adq||Adq||Adq||Adq||Adq||NA|
|Greenwood et al.26||UK||Case–control||1984–1993||Term||126||290||P; A; I; N||Adq||Adq||Adq||Adq||Adq||Adq||Adq||NA|
|Nelson and Ellenberg27||USA||Cohort||1959–1966||>2500g||170||45 559||Inad||Adq||Adq||Adq||Adq||Inad||Adq||Adq||Adq|
|Nelson and Ellenberg28||USA||Cohort||1959–1966||>2500g||170||45 559||P; A; I; N||Adq||Adq||Adq||Adq||Inad||Adq||Adq||Adq|
|Nelson and Ellenberg29||USA||Cohort||1959–1966||All||189||45 559||P; A; I; N||Adq||Adq||Adq||Adq||Inad||Adq||Adq||Adq|
|Naeye et al.30||USA||Cohort||1959–1966||≥38wks||150||43 287||A; I; N||Adq||Adq||Adq||Adq||Inad||Adq||Adq||Adq|
|Torfs et al.31||USA||Cohort||1959–1966||All||41||19 044||P; A; I; N||Inad||Adq||Adq||Adq||Adq||Adq||Inad||NA|
|Polivka et al.32||USA||Case–control||?||All||112||153||P; A; I; N||Inad||Inad||Adq||Inad||Inad||Adq||Adq||NA|
|Cummins et al.33||USA||Cohort||1983–1985||All||192||155 636||P; A; I; N||Adq||Adq||Adq||Adq||Adq||Adq||Adq||Adq|
|Wu et al.34||USA||Cohort||1991–2003||Term||377||326 006||P; A; I||Adq||Adq||Adq||Inad||Adq||Adq||Adq||Adq|
Two recent systematic reviews have attempted to identify optimal tools to assess the risk of bias in reports of observational research.35,36 Neither review could identify an optimal tool, but both recommended a transparent checklist over a scale, based on weighted and summed scores from individual sources. For this review we adapted a checklist37 which includes criteria for information and selection bias and addresses the domains of participants, measurement of variables, and control of confounding (Appendix SI, supporting information published online). Domains were assessed as adequately or inadequately addressed. If more domains were assessed as inadequate rather than adequate, the study was deemed inadequate for inclusion in the primary findings.
Across studies, risk factors were seldom reported in the same manner. For example, previous live births were reported in individual studies as primipara yes/no; number of previous births 0, 1, 2, 3, 4, >4; parity at least one live birth; parity 1, 2–4, 5+; parity 3+; and mean number of previous deliveries for cases and controls. The manner of reporting dictated which studies could contribute results; for example, proportion of mothers with no previous births could not be obtained for those studies choosing to report mean number of deliveries. In addition to risk factors being reported many ways, the inclusion criterion for cases varied; some studies considered differing subsets of the total CP population, for example spasticity only or exclusion of mild CP cases.
Quantitative meta-analyses were prohibited by such variations in definitions and categorizations of both exposures and outcomes. Instead, we report on the consistency of unadjusted risk associated with each factor and the range of point estimates of measures of relative risk. Adjusted estimates of risk were not considered because, where multivariate analyses were reported, each study controlled for different combinations of covariates, so the results were not comparable. However, most findings were controlled for gestational age, plurality, and sex either by matching in case–control studies or by stratification in cohort studies. Finally, if all studies of term-born infants that investigated a specific risk factor reported an increased risk of CP without any confidence intervals including 1, further literature was searched to identify if methods for preventing the risk factor or its consequence had been reported in systematic reviews.
A total of 25 articles met the inclusion criteria (Fig. 1); none was published in a language other than English. Of these 25, 11 reported on term-born infants and 14 reported on all children with CP. All included studies used a standard definition of CP applicable to the year they were published, and all excluded those with CP as a consequence of a postneonatal event. Nonetheless, the inclusion criteria for CP were not homogeneous across studies. Seven out of 25 studies excluded mild CP, reporting on only those children with moderate to severe CP. Nineteen out of 25 studies reported on all subtypes, whereas five out of 25 were limited to spasticity subtypes, and the one remaining paper reported on only those with CP/intellectual disability, and partial epilepsy. Four out of 25 studies did not include children with a congenital abnormality (birth defect), a factor which is well known to be associated with CP. Whereas 13 out of 25 reported on singleton births only, the others included multiple births. The birth years of cases and control individuals ranged from 1959 to 2005, with the majority (70%) of individuals born between 1980 and 1999.
Articles were appraised for risk of bias (either chance or systematic) with a view to comparing results of studies with a low risk of bias. Four studies were deemed to be at an unacceptably high risk of bias (Table I) and were not included in the primary results.16,19,20,32 Their exclusion reduced the number of children with CP in the primary analyses by 457 (6.7%) and the number of children without CP by 716 (0.02%). None of these were term-specific studies.
The 21 articles considered at a lower risk of bias comprised data from 6297 children with CP and their parents, and 3 804 791 children without CP and their parents.10–15,17,18,21–31,33,34 Populations considered were from Australia (six articles reported on two distinct populations), continental Europe (six articles reported on six distinct populations), the UK (two articles reported on two distinct populations), and the USA (eight articles reported on three distinct populations). Where a study was the subject of more than one publication, care was taken not to count the same participants more than once for each risk factor.
We extracted 808 estimates of risk pertaining to 88 separate risk factors. Risk factors were removed from the primary analysis if (1) fewer than four of the 21 studies reported them, (2) the risk factor was reported only by studies at high risk of bias, or (3) the factor was a consequence of disease and unlikely to be causal. Factors which were removed from primary analysis are reported in Appendix SII (supporting information published online). The primary findings thus comprised seven preconceptional, 13 antenatal, 12 intrapartum, and six neonatal risk factors (Table II).
|Risk factor||All CP||Term or >2500g CP|
|Nr of papers (n=11)||Associationa||Range of point estimates||Nr of papers (n=11)||Associationa||Range of point estimates|
|Preconceptional risk factors|
|Previous live births|
|Previous stillbirths/neonatal death||4||+||1.8–5.4||2||−/+||0.6–1.8|
|Maternal age at delivery|
|Hispanic||1||§||Unable to calculate||1||§||1.0|
|Asian||1||–||Unable to calculate||1||–||0.7|
|Low SES (including education, work, SES and private insurance)||2||§/+||1.0–1.4||3||§/+||1.0–1.6|
|Prior maternal diagnoses|
|Seizures, intellectual disability, thyroid disease||3||+||2.4–9.0||3||+||2.6–10.3|
|Diabetes, asthma, coagulation disorder, surgical history, mental illness, poor obstetric history||3||+||1.2–2.4||3||§/+||1.0–1.9|
|Antenatal risk factors|
|‘Any’ or first trimester||6||−/+||0.9–8||2||−/+||0.5–1.2|
|Second or third trimester||6||++||1.6–3.7||2||+||1.3–1.6|
|Maternal disease in pregnancy|
|Anaemia, rheumatic disease||2||–||0.2–0.7||2||§||0.9–1.1|
|CMV, rubella||1||+||Unable to calculate||0||NA||NA|
|Major and minor birth defects||4||++||2.6–20.7||3||++||2.2–27.4|
|Small for gestational age||5||++||1.9–7.5||4||+||2.1–4.5|
|Large for gestational age||1||+||2.6||1||§/−||0.9–1.6|
|Plurality||Studies were restricted to singletons (n=11), or matched or stratified by plurality (n=6)|
|Sex||Studies were matched, or stratified by sex (n=8)|
|Gestational age||Studies were either ≥37wks (n=8), or matched or stratified by gestational age (n=9)|
|Intrapartum risk factors|
|Length of labour|
|2nd stage >1h||2||++||2.1–2.4||4||+||1.5–2|
|Prolonged and preterm||1||++||3.32||1||+||1.5|
|Induction of labour||1||+||2.5||4||+||1.2–2.1|
|Abnormal fetal presentation|
|Other than vertex||2||++||3.8–4.2||5||+||0.6–5.3|
|Mode of delivery compared with spontaneous|
|Elective Caesarean section||2||§/+||1–2.5||1||§||1|
|Emergency Caesarean section||2||++||1.7–4.3||2||++||1.8–8.1|
|Cord around neck||2||++||1.7–1.9||3||§/+||1–2.3|
|Specifically tight cord||1||+||1.3||0||NA||NA|
|Cord prolapse||1||§||0.4||2||+||Unable to calculate|
|Neonatal risk factors|
|Respiratory distress syndrome||2||++||2.3–18||3||++||2.1–18.1|
|Infections including meningitis, sepsis||4||++||4.1–7.6||4||++||1.8–29|
All six studies considering a prior maternal diagnosis of seizures, intellectual disability or thyroid disease (both hypo- and hyperthyroidism) found these factors to be strongly associated with an increased risk of CP. A range of other maternal disease states were also reported to increase the overall risk of all CP, but their effect was less marked when only term births were considered (Table II).
A maternal obstetric history of stillbirth or neonatal death was more strongly associated with CP in studies investigating all CP than in those investigating term CP. One study of all CP reported an association with an obstetric history of three or more miscarriages. No study limited to term births reported on this variable. No convincing evidence was found for an association between birth order and CP.
Maternal age over 40 years posed a similar risk in all nine studies that reported it. Low maternal age was not consistently reported as a risk. Children of African-American parents living in the USA and children of Australian Aboriginals living in Australia were at increased risk. Low socio-economic status was also reported to be associated with a small risk increase in several continents.
Birth defects were the antenatal risk factor with the highest relative risk for CP. The wide range of central estimates between studies reflects variation in what was included as a birth defect, particularly whether or not minor defects were included. Small for gestational age status and low birthweight also consistently increased risk with an accompanying dose–response effect: the smaller the infant for gestational age and the lower the birthweight, the higher the risk. Only one study found no statistically significant association with measures of reduced intrauterine growth.28 Placental abnormalities (reported as small, calcified, unhealthy, infarcts, and/or complications) increased the risk of CP significantly in studies of term-born infants.
Maternal diseases during pregnancy that presented the highest risks for CP in the child were respiratory and heart diseases, seizures, and incompetent cervix. However, because they were examined individually in only one or two studies, they are reported in Appendix SII with other risk factors that show associations and promise for further studies but to date have rarely been examined. Abnormalities of fluid volume were identified as a statistically significant risk factor in studies of all CP, with conflicting evidence for term-born infants.24,25 Bleeding in the second and third trimesters, hypertension in pregnancy, and pre-eclampsia increased the risk of CP across all gestations. Chorioamnionitis also increased the risk of CP and was the most frequently reported perinatal infection.
Birth asphyxia, despite the variations in reporting, was the strongest and most consistent risk factor for the systematic review. Meconium-stained liquor and particularly meconium aspiration were also strong risk factors for births at all gestational ages. Instrumental deliveries (compared with spontaneous vaginal or elective Caesarean deliveries) were associated with increased risk of CP, as was a breech delivery. Abnormal duration of labour and fetal presentation were reported as statistically significant risk factors in studies considering all CP, but not in studies of term CP.
Overall, the magnitude of association was greatest for factors in the neonatal period. Among the neonatal variables, the presence of seizures was the strongest risk factor across all gestational ages. Respiratory distress syndrome, hypoglycaemia, and infections were similar in their magnitude of risk. There was conflicting evidence whether jaundice increased the risk of CP in term-born infants; however, in the two studies that specified severe jaundice the risk was significant.
Across the four epochs, 10 of the 38 risk factors were consistently associated with a statistically significant (p<0.05) risk of CP in term-born infants. One of these, ‘birth asphyxia’, is currently being treated with hypothermia, reducing the risk of permanent brain injury by 10%.38 There is sufficient evidence to suggest preventive strategies to address some causes of low birthweight, for example reducing alcohol consumption before pregnancy, and meconium aspiration (e.g. curtailing post-term pregnancy). Seven risk factors require further research to identify whether they lend themselves to prevention strategies, and all need further research to maximize their preventive potential (Table III).
|Risk factor||Possibility of prevention||Summary of current high-level evidence|
|Antenatal risk factors|
|Placental dysfunction or abnormalities||?||Insufficient data for the effectiveness of heparin for females considered at risk of placental dysfunction. More research is required39|
|Major and minor birth defects||?||Conflicting evidence regarding the benefit of folic acid and supplementary vitamins40,41|
|Low birthweight||✓||Reducing heavy alcohol consumption during pregnancy reduces low-birthweight rates.42 Insufficient data for magnesium supplementation during pregnancy43 and heparin39|
|Intrapartum risk factors|
|Meconium aspiration||✓||Amnioinfusion in settings with limited perinatal surveillance only;44 curtailment of post-term pregnancy reduces meconium aspiration;45 however, labour induction with prostaglandins is associated with meconium release46 and surfactant47 and steroid therapy48 currently have insufficient evidence|
|Caesarean section, vacuum, breecha||?||Females with continuous support during labour are less likely to have Caesarean or instrumental births.49 External cephalic version (ECV) at term reduces risk of non-cephalic and Caesarean births50 and tocolytics increase success of ECV.51 Insufficient evidence that either moxibustion,52 or postural management53 corrects non-vertex presentation|
|Birth asphyxia||✓✓||Cooling reduces mortality and major neurodevelopmental disability.38 Adjuvant therapies currently being investigated include topiramate, erythropoietin, melatonin, and xenon|
|Neonatal risk factors|
|Neonatal seizures||×||No evidence that cooling prevents seizures.38 Little evidence exists to support the use of anticonvulsants55|
|RDS||?||In preterm infants multiple doses of exogenous surfactant prevents RDS in those at risk and leads to improved clinical outcomes for those with RDS.56 No research specifically for term-born infants|
|Hypoglycaemia||×||Early hypoglycaemia is associated with severe encephalopathy and poor long-term outcomes57|
|Infections including meningitis and sepsis||?||Insufficient evidence for prophylactic vs selective antibiotics for term newborn infants of mothers with risk factors for neonatal infections;58 antibiotics for bacterial meningitis;59 antibiotics for late-onset sepsis;60 IVIG61|
On reviewing the literature we found that lack of standardization and specific reporting prevented optimal use of available data (e.g. meta-analyses even of uncontroversial continuous variables such as maternal age, parity, or birthweight could not be conducted). Acknowledging these limitations, 38 risk factors for CP were identified across four epochs and the consistency of findings was determined. Study findings are reported for two groups: those whose denominator was all CP and those whose denominator was only term CP. In the case of most risk factors, the strength of association with term CP and all CP was different. However, risk factors such as older maternal age at delivery, prior maternal disease, pre-eclampsia, and birth defects were not associated with different risk, and thus may be on causal paths to CP irrespective of gestational age at birth. In fact, these factors may work together, as high maternal age is also associated with a higher incidence of birth defects, pre-eclampsia, vascular disease, and pre-existing maternal diseases such as thyroid disorders.62,63 Interestingly, the incidence of intrapartum complications and perinatal death at term is also higher in older mothers.54
The primary findings consist of 10 risk factors in term-born infants that were consistently reported as statistically significant predictors of CP. These were placental abnormalities, birth defects, low birthweight, meconium aspiration, instrumental/emergency Caesarean delivery, birth asphyxia, neonatal seizures, respiratory distress syndrome, hypoglycaemia, and neonatal infection. The literature was reviewed to identify if any of these factors are currently preventable or if strategies exist to reduce the likelihood of their occurrence. The only risk factor for term-born infants for which a post-event treatment is currently available is birth asphyxia. It is estimated that one in every six to nine cases of CP due to birth asphyxia can be prevented if the infant receives hypothermia within 6 hours of the causal event.38 Within a causal pathway paradigm this is an example of intervening on a proximal cause.64
Preventing more distal factors on the path may interrupt the pathway earlier, before an injury occurs. The presence of a non-cerebral birth defect is one distal factor that may lie on a causal path. Presently, there is conflicting evidence concerning the prevention of birth defects (other than neural tube defects) with folate and supplemental vitamins.40,41 Another example of a distal risk factor is placental abnormality during pregnancy. Interventions such as heparin show promise, but long-term data are sparse and insufficient.39 The placenta is very rarely sent for pathology; however, in a recent total population case–control study of CP, in those born at or after 35 weeks’ gestation, macroscopic examination identified 8.8% of those with spastic quadriplegia as having placental infarcts compared with 2% of matched comparison children.65 The placenta is an area of great interest for future CP research.
This review draws attention to adverse modes of delivery as a potential risk for CP. Most instrumental deliveries are a result of prolonged second stage labour, and emergency Caesarean sections are undertaken mainly because of a failure to progress. Thus, there is an argument for early introduction of strategies to prevent prolonged labour such as active management, mobilization, and doulas (continuous support).49 Elective Caesareans are another means of avoiding prolonged labour, but it is clear that the large increase in elective Caesarean sections in developing countries over the past 30 years has not equated to a reduction in CP.66 Research efforts should be focused on the means of accurately identifying females most at risk who may benefit from an elective Caesarean section (e.g. intrauterine growth restriction and multiple pregnancy) who at this point may not be offered one, acknowledging that Caesarean sections themselves are not risk-free.
When developing trials aimed at identifying strategies to reduce the birth prevalence of CP in term-born infants, it is imperative that strategies are investigated for both potential distal (including placental abnormalities, low birthweight, reasons for hypoglycaemia) and proximal (meconium aspiration, birth asphyxia, neonatal seizures, respiratory distress syndrome, and neonatal infections) causes on the pathway to CP.
The major limitation of this study is that both risk factors and population samples were variably defined. Currently available published papers on term CP were not of sufficient uniformity to allow a meta-analysis to be performed. Comparisons between studies would be greatly facilitated if there were standardized methods of reporting key factors such as gestational age, birthweight, maternal age, previous maternal disease, and socio-economic status. Birth asphyxia was the strongest risk factor identified in this systematic review. However, the term ‘birth asphyxia’ is non-specific, and infants considered to be asphyxiated at birth may actually be found to have other diagnoses such as sepsis or non-asphyxial encephalopathy.
Combination risk factors (e.g. small for gestational age and true knot in umbilical cord) were not reported in this review.11 To date, very few papers have reported combinations of risk factors and their interactions. It is acknowledged that some combined risk factors may strengthen and explain associations, and within a causal pathway framework we encourage researchers to investigate factors that may act synergistically.
Another limitation of the present review is that several studies did not report effect sizes for associations that failed to achieve statistical significance. This selective reporting of statistically significant outcomes means that publication bias is inevitable. Pragmatically, it is difficult for researchers to report all their findings; however, failure to do so threatens the validity of any systematic review. Similarly, a number of studies used combination or overarching outcome variables such as ‘maternal disease during pregnancy’ in response to the small numbers of patients with any one specific risk factor in any study. CP is a heterogeneous disability, so it can be anticipated that it may be the result of a large number of causal pathways which still need to be separately identified. But for this to occur, and with the goal of prevention in mind, it is imperative that we start to report specific risk factors rather than general risk areas. We acknowledge that there is great difficulty reporting the many specific and often non-significant findings; however, if all research data sets prior to grouping and analysis were available centrally, preferably as individual patient data, then infrequently occurring risk factors that cannot be examined in any one study may be successfully investigated in combined data sets. We therefore propose that, to successfully identify causal pathways, a central clearing house of observational and experimental data aimed at prevention should be considered.
Himmelmann et al.67 recently published a systematic review of risk factors for CP in children born at term. These authors restricted their review to publications at in or after 2000 with the aim of identifying research breaking new ground. Our review did not have a lower publication year limit, ensuring that some of the earlier and most comprehensive studies on risk factors for CP were not excluded. Himmelmann et al. searched comprehensively for individual risk factors and included studies looking at the outcomes of a risk factor, with CP being one of many outcomes. In contrast, our review reported only on studies whose primary aim was to identify risk factors for CP. We were able to identify if risk factors were reported consistently, and the strength of the estimated risk. Our review complements theirs, with the limitations of each being at least in part addressed by the other, except that meta-analyses were not possible in either review. Taken together, the two reviews give a thorough overview of the state of the literature regarding risk factors for CP in those born at term and highlight the limitations inherent in utilizing the data as they are currently available.
Following this review we make the following recommendations: (1) 10 risk factors for term-born infants have been confirmed in all studies investigating them and focused preventive efforts should be directed towards them; (2) standardized definitions and categorizations of major risk factors should be agreed upon internationally then utilized in further research; (3) a clearing house should be established so that infrequently occurring specific risk factors can be reported (e.g. as anonymised individual patient data allowing meta-analysis of such risk factors).
Preventive strategies for CP in term-born infants are urgently required because infants born at term contribute up to 65% of CP cases, and their impairments tend to be more severe than those of children born preterm. The last 50 years of CP research have culminated in a much greater understanding of its diverse aetiology. The next 50 years could culminate in prevention and cure if we are prepared to pool our efforts.
North American usage: mental retardation.
The Cerebral Palsy Research Foundation supported SM and SG; Macquarie Group Foundation and Cerebral Palsy Research Foundation supported NB; NHMRC program grant no. 353514 supported EB.