Address correspondence to Christin M. Eltze, Neurosciences Unit, UCL-Institute of Child Health, 4/5 Long Yard, London WC1N 3LU, U.K. E-mail: firstname.lastname@example.org
Purpose: Most published data on infants presenting with epilepsy originate from hospital/specialist clinic settings and may therefore not be representative of the general population. We carried out a population-based study to estimate the incidence of epilepsy onset in infants, to characterize the range of phenotypes and associated structural brain abnormalities, and to determine whether specific epilepsy diagnoses could be established at onset.
Methods: Children between 1 and 24 months of age with new-onset epilepsy were ascertained over 13 months from the residents in 15 boroughs of North London. Classification based on clinical information, electroencephalography (EEG), and neuroimaging data was undertaken independently by two pediatric neurologists. Neuroimages were reviewed by two neuroradiologists blinded to clinical details.
Key Findings: A total of 57 children were enrolled giving an ascertainment-adjusted incidence of 70.1 (95% CI [56.3, 88.5])/100,000 children ≤2 years of age/year (ascertainment 76%). The incidence was highest among Asian children. An electroclinical syndrome was identified in 24 (42%) cases of which 21 were epileptic encephalopathies. Magnetic resonance (MR) images of 51 cases (89% of the total cohort) were reviewed. These demonstrated positive findings in 37 (72%) of 51 cases, of which 26 (51%) of 51 were etiologically relevant, and included developmental malformations in 11 (21%) of 51.
Significance: In a population setting infantile onset epilepsy presents mostly with complex phenotypes commonly associated with structural brain abnormalities. Routine MR imaging at presentation is therefore justified. However, identification of specific electroclinical syndromes remains difficult at onset.
To date there has been a paucity of prospectively acquired data in the literature reporting incidence, breadth of epilepsy presentations, and associated structural brain abnormalities identified on magnetic resonance imaging (MRI) from a population-based setting. Such information, however, is important for clinicians to recognize early children who are most at risk for adverse outcomes and decide on investigations as well as therapeutic strategies that will be most effective. For example, recent international guidelines for neuroimaging of children with newly diagnosed epilepsy recommend MRI for all patients under the age of 2 years, although the validity of this recommendation from a population perspective remains unclear (Gaillard et al., 2009).
We carried out a prospective population-based study, the “North London Epilepsy in Infancy Study,” in order to estimate the incidence of infantile-onset epilepsy, and to characterize the range of phenotypes and associated structural brain abnormalities, and to determine whether specific epilepsy syndrome diagnoses could be established at onset.
The study design involved a regional population survey with interception of incident cases for enrollment into a cohort observation study. Children with onset of epilepsy in the first 2 years of life were ascertained by surveying an infant population resident in a defined geographic area of North London to determine the incidence of the condition, and then recruited into an observational patient-specific cohort study for detailed evaluation. Eligible for inclusion in the study were all infants between 1 and 24 months of age with newly diagnosed epilepsy, defined as recurrent unprovoked seizures. Excluded were children with provoked seizures (i.e., by fever, infections, trauma, and electrolyte disturbances, transient metabolic or endocrine disorders) and those with seizures confined to the neonatal period only. Between September 1, 2005 and October 1, 2006, eligible cases resident in 15 adjacent boroughs (area: 494.29 km2) located in North London (part of London, United Kingdom, North of the river Thames) at the time of their diagnosis were notified to the research team by pediatricians using two main notification systems, an active—(monthly “postal questionnaire”) and a passive surveillance system (direct by “telephone, e-mail or verbally”), as detailed below. In the following we refer to the surveyed geographical area as “North London.” Postal codes unique for each borough were used to verify that eligible cases were resident in North London. To maximize our ascertainment, we raised awareness of the study prior to the start of enrollment. Consultant pediatricians based in hospital and community clinics in North London were informed in writing about the study, inclusion criteria for eligible cases, and notification methods that were going to be used. In addition we promoted the “Epilepsy in Infancy Study” at regional educational meetings attended by pediatricians and at departmental meetings in hospitals in North London. In addition, we established a network of collaborating consultant pediatricians based in each hospital in the surveyed area as contact for the research team.
Notification methods previously successfully applied in a different epidemiologic study carried out in North London were adapted for the current study (Chin et al., 2006). Cases were notified using two systems. First, a “postal questionnaire” was sent monthly from the central study center to all consultant pediatricians based in hospitals and community clinics in North London asking them whether they had seen any infants with new onset epilepsy. Second, patients could be directly notified by “telephone/email/verbally.” A telephone line (24 h automated service) was available to medical and nursing staff at North London hospitals and community clinics to inform the study center when they saw eligible infants in emergency departments, hospital wards, or outpatients clinics. The telephone number was displayed on posters in collaborating hospitals and distributed in “business card” format to medical staff. In addition they were also able to notify eligible cases to the first author directly verbally or by e-mail. The research team did not receive any self-referrals from parents/guardians of eligible cases directly.
When the parents/guardians, who were initially informed by medical or nursing staff about the study, gave consent to share their contact details with the research team, these were also provided with the notification of the case (by “postal questionnaire” and/or “telephone/e-mail/verbal”) to the central study center. If parents/guardians did not consent, anonymized notification and clinical information was send to the study center.
Clinical information to verify that cases met inclusion criteria, ethnic origin of the child (categorized into “White,”“Asian” [included Indian, Pakistani, Bangladeshi, other Asian], “Black” [included Caribbean, African, and other black], “Other non-white” [included Chinese and other non-white ethnic groups], and “Mixed” White [with one of the other non-white groups]), and investigation results were obtained using a standardized history proforma or anonymized clinical correspondence from the notifying pediatricians.
The parents/guardians were approached by the research team for written consent for entry of their child into the cohort observation study. This involved a detailed interview with carers, clinical examination, and neurodevelopmental assessment using standardized tests carried out by members of the central research team.
Investigation results of infants enrolled into the observational cohort including electroencephalography (EEG) recordings and neuroimages were obtained for review. Neuroimages were independently reviewed by two neuroradiologists blinded to clinical details, and differences were resolved by consensus. Findings were categorized as “etiologically relevant” (structural abnormalities that are commonly reported in children with epilepsy), “not etiologically relevant or uncertain” (commonly observed in children presenting with a broader range of neurologic and neurodevelopmental problems) and “normal – no positive findings.”
Clinical, EEG, and neuroimaging data were reviewed by two pediatric neurologists, who exclusively manage patients with epilepsy in their clinical practice (RCS, JHC). They independently classified seizures and epilepsy syndromes using the international classification of seizures and epilepsy syndromes proposed by the task force of the International League Against Epilepsy (ILAE) in 2001 (Engel, 2001). Cases were first categorized under one of the epilepsy syndrome groups (Engel, 2001) and then allocated a specific epilepsy syndrome diagnosis if possible. Discrepancies between raters were subsequently resolved by discussion of the individual cases.
The study was approved by the Great Ormond Street Hospital for Children NHS Trust and UCL- Institute of Child Health Research Ethics Committee and also registered with the local research and development departments of the collaborating acute hospital/primary care organizations.
The crude incidence of new onset epilepsy in children to the age of 2 years was calculated using 2006 mid-population estimates provided by the Office for National Statistics for the surveyed area as the denominator (98,090 children age 2 years and under). The 95% confidence intervals were calculated by applying the exact method for single proportions. A two-source capture–recapture model was applied to report an ascertainment-adjusted incidence (Source 1 “postal questionnaire”: cases notified by the monthly postal questionnaire, Source 2 “telephone/e-mail/verbal”: cases notified by telephone, verbally or by e-mail to the first author; Brenner, 1995). Poisson regression was used to identify factors associated with increased incidence. Poisson regression models were fitted with log-population at risk as an offset term to obtain rates: main effect terms for sex, age group (“<12 months,”“12–24 months”) and ethnicity, and their interaction terms were included in the model.
Interrater agreement was assessed by calculating kappa (κ), a reliability measure that is corrected for agreement occurring by chance (“chance corrected proportional agreement”). A value of κ much below 0.5 generally indicates poor agreement (Altman, 1991).
All statistical analyses were performed using SPSS version 15 (SPSS Inc, Chicago, IL, U.S.A.), R version 2.13 (R Foundation for Statistical Computing, Vienna, Austria) and Stata Version 9 (Stata Corp., College Station, TX, U.S.A.).
Incidence and risk factors
Fifty-seven cases were enrolled giving a crude incidence estimate of 53.6 (95% CI 41.4–69.5)/100,000 children ≤2 years/year. Applying the two source capture recapture model the ascertainment adjusted incidence was 70.1 (95% CI 56.3– 88.5)/100,000 children ≤2 years/year, with 76% completeness of ascertainment (see Fig. 1: details of enrollment and Table 1: details of capture–recapture estimates). The incidence of newly diagnosed epilepsy younger than 1 year of age was 82.1 (95% CI 61.4–109.8)/100,000 children ≤2 years/year, which was higher compared to that of children aged 12–24 months (23.4, 95% CI 13.3 – 40.8)/100,000 children ≤2 years/year). The incidence figures for male and female patients were similar in both age bands (Fig. 2). In order to examine the ascertainment in the different age groups (“<12 months,”“12–24 months”) further capture recapture models were calculated stratified by age (Table 2). The ascertainment in the age group 12–24 months was lower at 50%. Poisson regression models established age and ethnicity as significant predictors of incidence. Using ascertainment adjusted case numbers for both age groups, the risk ratio for epilepsy between the “<12 months” and “12–24 months” groups was 2.33 (95% CI 1.44 – 3.76; p < 0.005).
Table 1. Capture–recapture estimates
Source 2: Telephone/e-mail/verbal
Source 1: Postal questionnaire
aCalculated: d= n − 19 − 22 − 16.
bn=35 × 41/19; 95% CI n ± 1.96 √(35 × 41 × 16 × 22/193) = 58–92, under-ascertainment was 24%.
Table 2. Two-source capture–recapture analysis (stratified by age groups)
Cases ascertained (N)
Source 1: Postal questionnaire (N)
Source 2: Telephone/e-mail/verbal (N)
Cases notified by sources 1 and 2 (N)
Ascertainment adjusted case number (N)
Completeness of ascertainment (%)
Sum of adjusted case numbers from age bands
Data on ethnicity were available for 56 infants: 23 (41%) were White, 21 (37.5%) Asian, 10 (18%) Black, and 2 (3.5%) were Mixed White and Asian. The risk of presenting with epilepsy before the age of 2 years was significantly higher for patients belonging to one of the non-White groups compared to the White group (risk ratio [RR] 2.37, 95% CI 1.4–4.0, p=0.002). Subgroup analysis using White ethnic group as a reference (compared with each group: “Asian,”“Black,” and “Mixed Asian-White”) demonstrated that this increased risk among non-White infants was due to the increased risk in Asian infants (RR 2.84, 95% CI 1.57–5.13, p < 0.001; “Black” group: RR 1.85, 95% CI 0.88–3.9, p = 0.1; “Mixed Asian-White” group RR 1.71, 95% CI 0.4–7.24, p = 0.47). Additional capture recapture analyses demonstrated that ascertainment in the Asian group versus all other ethnic groups was similar at 81% and 72%, respectively (Table S1).
Neuroimaging and etiologies
Fifty infants were enrolled in the observational cohort and were clinically assessed after a median time interval of 7 weeks (interquartile range 4–11 weeks) following the diagnosis of epilepsy. In addition, anonymized clinical information including reports of investigations was obtained for a further seven cases. The reported cohort therefore comprised 57 children (31 boys) with a mean age of epilepsy onset at 6.9 months (range 0.1–22 months; see Table 3 for further clinical details).
Table 3. Clinical features of North London infancy epilepsy cohort (n = 57)
Abnormal development prior to epilepsy onset (when onset after neonatal period, N = 48)
Severity of seizures at baseline
Monthly (<12 weeks seizure-free)
Seizure-free for >12 weeks
Number of antiepileptic drugs trialed
Neuroimaging was performed in 55 children (96% of the total cohort) of which 54 underwent magnetic resonance MRI and one child computed tomography (CT). Eight infants had both CT and MRI scans. Neuroimages of 52 children were obtained for review (MR images, available for 51, 89% of the 57 children in the total cohort). MR images demonstrated an etiologically relevant abnormality in 26 (51%) of the 51 cases. Acquired lesions were identified in 14 of the 26 cases (53%, secondary to central nervous system [CNS] infections, hypoxic ischemic injury, metabolic disorders, or signal abnormalities thought to be related to seizure activity) and developmental brain malformations in 11 (42%, see Table 4). MRI findings in a further 11 children (21%) were of uncertain etiologic relevance. These included delayed myelination, lack of white matter bulk, thin corpus callosum, or arachnoid cysts.
Table 4. Etiologies identified in the North London infancy epilepsy cohort
N (%) n = 57
Details (number of cases)
aInfants with recurrent unprovoked seizures owing to CNS insults, such as infection or perinatally acquired injuries including hypoxic ischemic brain injuries, resulting in a static encephalopathy.
bLow respiratory chain enzymes in two patients: one low complex IV in muscle tissue and skin fibroblasts, one with low complex I, II, III, and IV on muscle biopsy.
Overall, an underlying etiology for the epilepsy was identified in 29 children (51% of the cohort, n = 57), with developmental brain malformations being the largest etiologic category (21%), followed by acquired brain insults (16%, see Table 4).
Distribution of electroclinical syndromes and epilepsy types
Following consensus discussion, all but three cases could be categorized in one of the broad epilepsy syndrome groups suggested in the 2001 ILAE proposal (Table 5). A specific electroclinical syndrome regardless of etiology was identified in 24 (42%) of the 57 cases enrolled and in a further 14 (24%) cases, the epilepsy type could be attributed to a structural brain abnormality (“neocortical epilepsies defined by location and etiology”). For 19 (33%) of the 57 children, however, no specific electroclinical syndrome or epilepsy type defined by seizure type and etiologic category could be allocated (Table 5). The value of κ regarding inter-rater agreement was 0.48 (95% CI 0.30–0.66) for the categorization in broad syndrome groups and 0.5 (95% CI 0.34 –0.66) for classification of cases into electroclinical syndromes and other types of epilepsy.
dUnable to classify (16); no agreement between raters (3).
West syndrome/infantile spasms (seven structural/metabolic, one presumed genetica, eight unknown cause)
Ohtahara syndrome (one structural/metabolic, one unknown cause)
Dravet syndrome (SMEI)
Dravet syndrome (SMEI; two genetic: SCN1A mutation confirmed)
Benign infantile seizures (nonfamilial)
Benign infantile seizures (nonfamilial)
Benign myoclonic epilepsy in infancy
Myoclonic epilepsy in infancy
Neocortical epilepsies Other types defined by location and etiology
Nonsyndromic epilepsies: Structural/metabolic
Probable mesial temporal lobe epilepsy with hippocampal sclerosis
Probably symptomatic focal
Unable to determine epilepsy syndrome group
Unable to determine specific epilepsy syndrome
Unknown cause and no identifiable electroclinical syndrome
Since the classification of cases was carried out, revised etiologic categories have been published by the ILAE commission on classification and terminology (Berg et al., 2010). The first author applied these new categories based on review of the etiology data available for each case. All of the specific electroclinical syndrome diagnoses made remained; changes predominantly involved etiologic grouping (see Table 5).
To our knowledge this is the first prospective population-based study focusing on epilepsies with onset in the first 2 years of life. We have confirmed that epilepsy in this age group is common (adjusted point incidence of 70.1, 95% CI 56.3– 88.5/100,000 children ≤2 years/year), presents in the majority with complex epilepsy phenotypes, and is commonly associated with structural brain abnormalities. Although the majority of infants with new-onset epilepsy can be classified using current ILAE classification schemes, in more than one fourth of cases further syndromic and etiologic delineation is required.
There are no other recent studies in the literature that applied methods similar to those used in the “North London Epilepsy in Infancy Study.” Considering the methodologic differences and limited comparability, our incidence estimate of infantile epilepsy is similar to estimates that can be derived from older retrospective studies of wider childhood cohorts with case ascertainment from one source; 81 (95% CI 67–93)/100,000 children <2 years/year from a Canadian study (cases identified from regional EEG department; Camfield et al., 1996), and 61 (95% CI 39–95)/100,000 children <2 years/year from the United Kingdom 1958 National Child Development Study (screening questionnaires to families enrolled in birth cohort; Kurtz et al., 1998).
We adjusted for under-ascertainment using a two-source capture–recapture approach. Positive dependency between the two sources was likely (meaning that cases identified by one source are more likely to be enrolled in the other source), and therefore the degree of underestimation is less compared to using a traditional case–registration approach (Brenner, 1995).
In the United Kingdom it is common practice and also recommended in national guidelines that children presenting with seizures or suspected new-onset epilepsy to primary care physicians and emergency departments are referred to a pediatrician (National Institute for Clinical Excellence, 2012). We therefore aimed to maximize ascertainment by establishing a network of collaborating pediatricians based in hospitals and community clinics that provided services to North London. The participation in the postal survey involving consultant pediatricians was high (mean monthly responder rate 77%). A limitation of our study is that we cannot exclude, similar to other regional surveys, that some eligible cases residing close to the boundaries may have been seen by pediatricians outside the surveyed area. The relative high ascertainment of >70% would, however, support that the majority of infants were captured.
Two source capture–recapture analyses stratified by age group demonstrate lower ascertainment in the 12–24 month age band, suggesting variability in case reporting through the applied notification systems. Taking this into consideration by using ascertainment adjusted case numbers for the age groups in the Poisson regression model, our study showed that the risk of epilepsy in the first year of life was more than double the risk in the second year of life. A similar observation has been reported in other studies with a steep decrease in the incidence of newly diagnosed epilepsy in the second year compared to the first year of life (Hauser et al., 1993; Camfield et al., 1996; Olafsson et al., 2005). The propensity of the immature brain for seizures has been well established in animal models (Haut et al., 2004). The reduced risk of seizures in the second year of life may reflect brain maturation processes that result in an increased seizure threshold. This is further supported by the fact that a greater number of age-specific electroclinical syndromes with onset in the first year have been delineated.
Ethnic group was an additional relevant factor for the risk of epilepsy in our study. Compared to the “White” group, Asian infants (including children from Indian, Pakistani, and Bangladeshi origin) had three times the risk of presenting with epilepsy, whereas the risk was not significantly higher in the other non-white ethnic groups (“Black” or “Mixed Asian and White”). The ascertainment in the Asian compared to all other ethnic groups was similar. Data suggesting an impact of ethnicity on the incidence of unprovoked seizures in children younger than the age of 5 have been reported from other multiethnic urban populations (Annegers et al., 1999). Such variability in seizure susceptibility between different ethnic groups is most likely mediated by genetic factors, although environmental factors may act independently or interactively to contribute to this phenomenon.
The proportion of identified etiologies was higher in our cohort (51%) compared to previously reported population-based figures; for example, in a retrospective Finnish infancy epilepsy cohort with onset before 2 years of age (31%, total cohort, n=72; Rantala & Ingalsuo, 1999) and childhood epilepsy cohorts including infants and older children (18–36%; Arts et al., 1999; Berg et al., 1999; Larsson & Eeg-Olofsson, 2006). However, lesional and genetic etiologies may have been under-recognized in these older studies given the recent advances in neuroimaging and molecular genetic investigations. The proportion of identified etiologies in our cohort would now probably be higher as gene alterations associated with early onset epileptic encephalopathies including STXBP126, CDKL527, SLC25A2228, and PLCB129 (testing not available at the time of study), are increasingly being recognized.
The high diagnostic yield of etiologically relevant MRI lesions (51%) observed in our study has also been reported by others in this age group. In a large community-based childhood cohort (1 month to 16 years) the yield of MRI was approximately 25% in the <2 year onset group, with 16% of etiologically relevant lesions in the overall group (Berg et al., 2009). Hsieh et al. (2010) found 57% abnormal MRI scans in a hospital-based cohort of children younger than 2 years presenting with one or more afebrile seizures from a multiethnic inner city population (MRI was performed in only 57% of the total cohort, n=317).
Developmental brain lesions/malformations were the most frequently identified etiologies in our study (21%). Nearly two thirds of children with developmental brain malformations had extensive unilateral or bihemispheric cortical abnormalities and presented with seizures in the first 6 months of life. Such findings support both international and national guidelines recommending MRI in all children with new-onset epilepsy before the age of 2 years (Gaillard et al., 2009).
Most of the infants presented with epilepsies frequently associated with poor seizure outcomes and developmental impairment (“focal symptomatic epilepsies”: 16 [28%] and “epileptic encephalopathies”: 22 [39%]), whereas cases with idiopathic epilepsy syndromes presumed to have a more benign evolution (9 cases, 16%, see Table 4) were relatively infrequent. This suggests that in a population-based setting, self-limiting epilepsies are rare in this very young age group. We minimized referral bias and underascertainment of such milder epilepsy types by designing our study as a collaborative project involving hospital and community pediatricians.
Two pediatric neurologists agreed following discussion on one of the broad epilepsy syndrome groups suggested in the 2001 ILAE proposal in the majority of children in this cohort (Table 5). Further delineation was possible by allocation to one of the defined electroclinical syndromes (42%) or by description as epilepsy associated with a structural brain abnormality (24%) (Engel, 2001). In one third of cases the epilepsy type could not be further determined beyond seizure type and associated clinical findings (Engel, 2001). The interrater agreement between the two pediatric neurologists, when applying the 2001 ILAE classification according to their own judgement, was only moderate with regard to epilepsy syndrome diagnosis, further highlighting diagnostic difficulties in infancy, and confusion in application of the terminology. As expected, discrepancies between the raters occurred especially with classification of cases with unknown etiology as “undetermined,”“idiopathic focal,” or “probable symptomatic focal.” When defining syndrome groups by the etiological categories “idiopathic” or “probable symptomatic,” the 2001 classification proposal (Engel, 2001) presumes outcome is known, especially when “idiopathic” is understood to imply a “benign” evolution with responsiveness to antiepileptic medication. Little information can be inferred from such syndrome group diagnoses compared to the more narrowly defined electroclinical syndromes and these diagnoses may also be misleading. In the latest revision of the classification, the ILAE commission discourages use of these categories and recommends a descriptive approach using seizure type, age of onset, and etiology if a specific electroclinical syndrome cannot be identified (Berg et al., 2010). New etiologic categories have been defined to take into consideration progress made in molecular genetics and neuroimaging: “Genetic,”“structural/metabolic,”“unknown cause” (Berg et al., 2010). Because the specific electroclinical syndromes have not changed in the new proposed organization, the etiologic categories were relatively easy to apply (see also Table 5).
Our study provides prospective population-based information that will alert clinicians who are managing infants with epilepsy in a nonspecialist setting of the high risk of complex epilepsy presentation in this age group. These data will help in the appropriate allocation of resources for this vulnerable group, including availability of adequate neuroimaging and need for specialist intervention. Because those with focal structural lesions may benefit from epilepsy surgery (Hemb et al., 2010), early referral for evaluation should be considered.
We would like to thank the members of the collaborating group of consultant pediatricians and pediatric neurologists in North London (Dr. Jacqueline Taylor, Dr. Peggy O’Driscoll, Dr. Nicholas Cavanagh, Dr. Ruby Schwartz, Dr. Adnan Manzur, Dr. MAS Ahmed, Dr. Denise Gurtin, Dr. S Mathew, Dr. Emma Devereux, Dr. Caroline Oren, Dr. S Prasad, Dr. Andrew Lloyd-Evans, Dr. Diane Smyth, Dr. Adelaida Martinez, Prof. M Gardiner, Dr. Simon Whitmarsh, Dr. Corina O’Neill, Dr. Sanjay Bhate, Dr. Paola Nicolaides, Dr. Vijay Ganesan, Dr. Lucinda Carr, Dr. Sarah Aylett, Dr. Carlos De Souza, Dr. Cheryl Hemingway, Dr. Robert Robison, Dr. Sophia Varadkar, the late Dr. Sarah Benton and the late Prof Robert Surtees), as well as their junior doctors and nursing and administrative staff for their participation in this project. We also thank Dr. Stewart Boyd and Dr. Steven White for their help with the review of EEG data, as well as Dr. Michelle De Haan and Anna Merrit for obtaining data on developmental status of the children enrolled in the cohort study.
The authors declare no conflicts of interest. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
Epilepsy Research UK, Foyle Foundation, Bailey Thomas Charitable Trust, and UCB Pharma. RFMC held a National Institute for Health Research (NIHR) Academic Clinical Lectureship. RCS is supported by Great Ormond Street Hospital Children’s Charity, the NIHR, the Norwegian Medical Council, and BUPA Foundation, and has received travel grants from SPL Ltd, Jansen-Cilag, UCB, Sanofi, and Glaxo-Wellcome. This work was undertaken at GOSH/UCL Institute of Child Health, which received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centre funding scheme. The Centre for Paediatric Epidemiology and Biostatistics, UCL-Institute of Child Health, also benefits from funding support from the Medical Research Council in its capacity as the MRC Centre of Epidemiology for Child Health.
CM Eltze, JH Cross, RC Scott RC, and RFM Chin participated in the design and conduct of the study and writing of the manuscript. WK Chong and T Cox reviewed neuroimages. A Whitney assessed patients and collected data for the study. CM Eltze, RC Scott, RFM Chin, and M Cortina-Borja carried out the statistical analysis.