The study was conducted at the outpatients’ services of the child development centre in Dhaka Shishu (Children’s) Hospital. For most, it would be the first contact with services for seizures and related problems as none are usually available at primary and secondary level. Most families were self-referred. A child-health physician with basic training in child neurology, a developmental therapist trained in physiotherapy, occupational and speech therapy within a developmental framework, and a child psychologist were the core team members of the service. They received 1 month’s training in childhood epilepsy, associated psychomotor comorbidities, and their management. Weekly meetings with the principal investigator were used to monitor clinical information, diagnosis, drug treatment, and for problem-solving.
Children aged 2 months to 15 years, presenting with two or more unprovoked seizures with or without motor and/or cognitive comorbidities, were included. Families were classified by monthly income in taka (1 US dollar≈70 taka), as low (<10 000 taka per mo), middle (10 000–20 000 taka per mo), and high (more than 20 000 taka per mo);2 residence was classified as urban or rural. Birth history included place and mode of delivery; ‘perinatal asphyxia’ was defined in this setting as a failure to start regular respiration within a minute of birth because that was the available description3 and thus not necessarily including a neonatal encephalopathy.4
General and neurodevelopmental examination findings were collected on a pre-coded, structured form. Details of seizures were recorded as ‘age at onset’, which was categorized as ‘early’ when unprovoked seizures started at or before 12 months of age, or ‘late’; ‘seizure type(s)’ were categorized as ‘single’ or ‘multiple’ type seizures when there was a history of more than one type of seizure such as generalized tonic–clonic and myoclonic or head-drops or absences; ‘rate of seizures’ was categorized as ‘high’ when there were one or more attacks per week.
Epilepsies were classified for descriptive purposes using the International League Against Epilepsy syndromic classification current at the time of the study.5–9 Based on examination findings, EEG, and available neuroimaging features, seizures were categorized as ‘generalized’ or ‘partial/focal’, and epilepsies as ‘idiopathic’, ‘symptomatic or cryptogenic’, and ‘unclassifiable’. A history of unprovoked, repeated seizures among the siblings, parents, grandparents, and paternal or maternal cousins was coded as a positive family history.
Comorbidities were recorded as ‘motor disability’ when a motor functional deficit caused significant limitation of daily living activities, and cognitive impairment when the IQ was <70.
Psychometric assessment was performed close to enrolment using standardized age-appropriate assessments: the Bayley Scales of Infant Development,10 Wechsler Intelligence Scales for Children, revised,11 and the Independent Behaviour Assessment Scale,12 which had been adapted for use in Bangladesh.
A routine digital EEG (a minimum of 30min with photic stimulation and hyperventilation where possible but without sleep) was recorded and reported independently by two paediatric neurophysiologists, one of whom was blind to the child’s clinical data. The interrater reliability was very good on an unweighted kappa measure (κ=0.93; 95% confidence interval [CI] 0.90–0.95;13Table I). EEG findings were categorized as ‘normal’ and ‘abnormal’ for age and state of arousal. Abnormal brain activities were subcategorized for further descriptive analysis into (1) epileptiform discharges (discharges with spike, sharp, and slow wave complexes) with normal background activities between discharges, (2) non-epileptic background abnormality (diffuse or localized irregular slow waves, excessive beta waves, unorganized non-reactive background activities), and (3) both (epileptiform discharges with background abnormality). Neuroimaging, such as ultrasonography, computed tomography or magnetic resonance imaging of the brain, was performed when clinically indicated and financially feasible but was not used for predicting seizure control.
Table I. Kappa analysis of EEGs to determine the interrater reliability (383 EEGs were reported by one blinded neurophysiologist and one unblinded)
|Blinded neurophysiologist||Unblinded neurophysiologist||Total||CI|
|Total|| || ||383|| |
Treatment with antiepileptic drugs was given following standard treatment procedures for specific seizures and epilepsy syndromes.14,15 Locally available antiepileptic drugs, phenobarbital, carbamazepine, sodium valproate, phenytoin, nitrazepam, and clonazepam were used. In addition, short courses of oral prednisolone were used for infantile spasms.14 Parents with children having prolonged generalized tonic–clonic seizures were trained to use rectal diazepam.16–19 Follow-up was for a minimum of 12 months, at 1- to 3-month intervals depending on the seizure rate and travel distances.
Compliance was encouraged in various ways: for example, travel costs were provided to very poor families, and telephone or postal communications and home visits were used when these failed. Monitoring of medication was by regular follow-up, tablet counts, and, very occasionally, for example when there was no response in seizure rate, by blood levels. Family members were taught the importance of regular medication and follow-up in epilepsy. Diaries of seizures were maintained by families.
All analyses used Stata Intercooled (StataCorp, version 12). Of the 390 children originally enrolled, 71 were lost to follow-up. The clinical features in the enrolled and lost to follow-up groups were compared to investigate whether there were differences between the groups that could introduce important bias. Chi-squared analyses were used for categorical data, and a t-test was used for continuous data. Significant predictors of outcome were investigated using logistic regression. The OR (which approximates a relative risk20) and 95% CIs were calculated using bootstrapping with 100 replications.21 We analysed this 10 times to identify weak and strong predictors of outcome. Significant predictors were divided into strong (significant [p<0.05] in each of the 10 analyses) and weak (significant in fewer than all 10 analyses).22 As the CIs differ on each analysis, only ORs are presented. Potential predictors of seizure outcome included in the model were (1) ‘early onset’, (2) ‘multiple seizures types’, (3) ‘high rate of seizures’, (4) severe epilepsy syndrome with regression, (5) ‘motor disability’, (6) ‘IQ<70’, (7) ‘positive family history of epilepsy’, and (8) ‘abnormal EEG’. All potential predictors were included in each analysis.
The study protocol was approved by the ethical review committee of Dhaka Shishu Hospital, Bangladesh Institute of Child Health, and the research ethics committee of Great Ormond Street Hospital for Children and University College London Institute of Child Health. Patients were recruited from April to October 2001. Parents’ written consent was obtained at the initial consultation.