Incidence, causes and phenotypes of acute seizures in Kenyan children post the malaria-decline period

This study was conducted in Kilifi Health and Demographic Surveillance System which is situated on the Kenyan coast about 50 kilometres from Mombasa and has a population of about 270,000. The population is comprised of a Mijikenda speaking community, who are mainly subsistence farmers and a few fishermen. The transmission intensity of malaria has reduced this area, resulting in reduction in morbidity and mortality due to malaria [5], in particular seizures-attributable to malaria (SAM) [2].

A total sample of 3074 children aged 13 years and below with a history of seizure(s) during that period of the presenting illness were admitted to Kilifi County Hospital (KCH) from the year 2009–2013. Children with severe malaria were treated with first-line intravascular artesunate [8], with broad-spectrum antibiotics. Children admitted with seizures lasting 5 minutes were first treated with intravenous diazepam, and for those whose seizures did not stop within 10 minutes a second dose of diazepam or intramuscular paraldehyde was given [9]. Seizures lasting 15–20 minutes were managed with phenobarbital, and those continuing further with phenytoin. Seizures refractory to phenytoin were managed in consultation with senior clinicians, who would often recommend another dose of phenobarbital or midazolam since artificial ventilation was not available.

Clinical files for these children were reviewed to determine which admissions had acute seizures. Data for 1063 records was obtained from reviewing the patient’s case notes, for the remaining 2011 records data was obtained from the hospital inpatient database. Case notes were independently reviewed by Serem KG (who examined all case notes) and Kariuki SM (who randomly reviewed half of the case notes); with any differences in phenotypes resolved through consensus. The data extracted includes among others: the age in months of the patient, diagnosis, seizure type, medications received and blood workups. Data were extracted into Excel spread sheet, before being transferred into the analysis software.

Phenotypes of seizures are defined in Table 1, and were based on ILAE recommendations [10‐12], that have been validated in previous studies in this area [13, 14]. Neonates were defined as those aged 28 days or younger [7]. Malnutrition was defined as a weight for age z-scores (WAZ) < −2 or mid-upper arm circumference (MUAC) <11 centimetres or a clinical diagnosis on discharge. Meningitis was defined as cerebrospinal white cell count >50/μL [15]. Anaemia defined as a haemoglobin concentration <50 g/L, thrombocytopenia as a platelet count <150,000/μL, hypoglycemia as glucose concentration of <3 Mmol/L, hyponatremia as sodium concentration of <125 Mmol/L, and hyperkalemia as potassium concentration of >4.5 Mmol/L.

The data were analysed using the STATA 13 statistical software (Stata Corporation, Texas, USA). The association between complex seizures and social demographic characteristics and other clinical variables was measured using logistic regression, and variables with a p-value <0.25 were used to build a multivariate model in a stepwise regression technique. Hosmer-Lemeshow test was used to test the goodness of fit in logistic regression model. For comparison of continuous variables, the Student t-test was used and if data did not have a Gaussian distribution, Mann–Whitney U-test was used. For comparison of categorical variables, the Pearson’s Chi-square test or Fisher’s exact (where observations were infrequent) were used. Proximate causes of acute seizures mentioned in the literature were reported here as percentages of occurrence among admissions.

For different phenotypes of acute seizures, we computed overall incidence, then stratified incidence by year, age group and falciparum malaria status. Incidence was computed by dividing number of children with acute seizures by person-years of observation for all admissions and by live births for neonatal admissions. The trend of incidence of acute seizures was determined using the Poisson regression technique. The trend for all admissions was repeated for neonatal seizures, non-malarial seizures, malarial seizures, and malarial seizures with a parasite density ≥2500 μ/L, a sensitive and specific cut-off for defining seizures attributable to malaria that was modelled with logistic regression techniques in previous studies [2, 16].

This study was approved by the Kenyan Ethical Review Committee (SSC No 2599) and parents of children participating in the study gave a written informed consent.

These data are stored securely within the KEMRI-Wellcome Research Programme servers as was agreed with participants during consenting and can be availed upon formal request.

Over the 5 year period, 2742 children with acute seizures were admitted to KCH, of whom 1567 (57 %) were males (Table 2) and 182 (7 %) had neonatal seizures. Nearly half (46 %) of children were admitted with complex seizures, and different phenotypes of complex seizures showed significant overlap with each other (Fig. 1). Malaria parasitemia was detected in 38 % of all admissions and in 36 % of complex seizures (36 % of prolonged seizures, 39 % of convulsive status epilepticus, 31 % of repetitive seizures and 22 % of focal seizures). Malnutrition was present in 23 % of all admissions, and was more frequent in children with complex seizures (27 % vs. 21 %; p < 0.001), but less frequent in those with malaria parasitemia (16 % vs. 27 %; p < 0.001). Complex seizures were independently associated with mean corpuscular volume in a multivariable logistic regression model (odds ratio = 0.92 (95 % CI, 0.85-0.99); p = 0.033).

The most common causes of the 2742 admissions with acute seizures were malaria (n = 899, 33 %)), respiratory tract infection (n = 532, 19 %)), fever of unknown origin (n = 467, 17 %)) and hypoglycemia (n = 274, 10 %). Other less common causes of acute seizures were as shown in Table 3. Malaria was an important co-morbidity in those with malnutrition (23 %).

A number of several possible causes were significantly more frequent in complex seizures compared to simple seizures: malaria (p < 0.001), respiratory tract infections (p < 0.001), fever of unknown origin (p < 0.001), encephalopathy of undetermined cause (p < 0.001), neonatal sepsis (p < 0.001), hypoglycemia (p = 0.006) and hyponatremia (p = 0.003). On the other hand, gastroenteritis was more common in simple than complex seizures (p = 0.025). Hypoglycemia was significantly more common in those with neonatal seizures compared to those without (54/133 (41 %) vs. 153/1881 (8 %); p < 0.001) and as was sepsis (95/182 (52 %) vs. 16/2565 (1 %) p < 0.001) and hypoxic-ischemic encephalopathy (39/182 (21 %) vs. 0/2565 (0 %); p < 0.001).

The overall incidence of admissions with acute seizures was 312 per 100,000/year (95 % CI, 295–329), being 170 per 100,000/year (95 % CI, 157–182) in males. The overall incidence of acute seizures significantly increased with years of admission over the study period (incidence rate ratio (IRR) = 1.13; p < 0.001) (Fig. 2). Similarly, the overall incidence over the study period appeared to decrease with age (IRR = 0.651; p < 0.001) (Fig. 3). The incidence of all acute seizures appeared to decrease with distance from the hospital and was 257 per 100,000/year (95 % CI, 242–272) for those living within 5 kilometres from KCH, and 55 per 100,000/year (95 % CI, 48–62) for those living ≥5 kilometres from the hospital. The incidence of acute seizures among the neonates was 443 per 100,000 live births (95 % CI, 383–512), and remained stable over the five year period (IRR = 0.99 (95 % CI, 0.90-1.10); p = 0.905).

The incidence of all seizures attributable to malaria (SAM) was 75 per 100,000/year (95 % CI, 67–83), and that for non-SAM was 237 per 100,000/year (95 % CI, 223–252). The incidence of SAM increased over the period (IRR = 1.22 (95 % CI, 1.13-1.31); p < 0.001) while that for non-SAM remained unchanged (IRR = 1.03 (95 % CI, 0.96-1.06); p = 0.569).

The overall case fatality proportion over the 5-year period was 6 % and remained stable throughout the study (p = 0.166). Death was more frequent in complex seizures than simple seizures (8 % vs.4 %; p < 0.001) and in neonatal seizures than seizures in those aged >28 days (10 % vs. 5 %; p = 0.005). In particular, 12 % of the deaths were in prolonged seizures and convulsive status epilepticus, 6 % were in repetitive seizures and 8 % in focal seizures. Mortality following acute seizures was reported in 20 % with malaria, 11 % with respiratory tract infections, 23 % with malnutrition and 12 % of encephalopathy with undetermined causes.

The incidence of acute seizures in this study (312 per 100,000/year) is lower than that reported in a previous study (425 per 100,000/year) [1], which used data on admissions during a time when malaria was a common cause of morbidity and mortality in this area. Despite the recent decline in malaria, our present data reveals that the incidence has been increasing between 2009 and 2013, an important epidemiological finding which is not apparent if the present incidence was compared with earlier years when incidence of malaria was high.

This increase in incidence was particularly observed for SAM, being highest for the complex seizure phenotypes; supporting increasing burden for acute seizures post the malaria decline era. There are three explanations for the increasing incidence of seizures following the earlier decline. First and most plausible, a reduction in malaria may have resulted in less exposure, thereby reducing the acquisition of naturally acquired immunity to malaria, resulting in more severe illness, including seizures, that requires hospitalisation [5]. Secondly, this could be related to proportional increase in neonatal causes of seizures such as hypoglycemia, sepsis and hypoxic-ischemic encephalopathy compared to an earlier study [1]. Finally, the increasing incidence of acute seizures could be related to possible improvements in health care services not measured in this study, which would encourage admissions.

The incidence is grossly underestimated since only 28 % of it was from areas >5 kilometres. This suggests that poor infrastructure and transport costs may have resulted in some (if not most) seizures not being treated in hospital. In a recent study, only 56 % of persons with convulsive status epilepticus occurring with epilepsy were treated in KCH [17]. Other cultural reasons such as preference of traditional healers over biomedical practitioners may have contributed [18]. Despite the conservative incidence our estimates for prolonged and/or status epilepticus is higher than that reported in developed countries e.g. in London [19]. This may be explained by the high prevalence of infectious causes in this area e.g. malaria parasitemia, which was found in 36 % of all prolonged seizures; over 90 % of complex seizures in parasitemic children are caused by malaria [2]. Delays in management of acute seizures in peripheral clinics that may result in admission of seizures already refractory to treatment [14], adding to the high burden of prolonged seizures in this study.

Malaria remains the most important cause of admission with seizures, with malaria parasitemia observed in over a third of admissions. This proportion however is slightly less than that reported in previous studies in this area (>50 %) [1, 14], suggesting a possibility of emergence of other important causes of acute seizures particularly those common in the neonatal period. Falciparum malaria may cause seizures by: (i) sequestration-induced brain damage that manifests as seizures [20]; (ii) down-regulation of GABA receptors thereby increasing susceptibility to seizures [21]; (iii) metabolic complications such as hyponatremia or hypoglycaemia and; (iv) induction of inflammatory molecules that lower seizure threshold [22]. Susceptibility to seizures may be determined by genetic polymorphisms [13].

Hypoglycemia, unknown encephalopathy and neonatal sepsis, were more common in the neonates in greater proportions than in a previous study [1], yet some of these can be easily prevented and managed. It is however, unclear if a proportion of encephalopathy of unknown origin reflects neurotropic viruses that were not investigated in this study [23].

Malaria was more common in complex seizures compared to simple seizures, suggesting a causal role for these complex phenotypes, as demonstrated in previous studies [2]. Additionally, other causes of seizures were significantly more frequent in complex seizures, perhaps because it represents a severe phenotype of seizures, as supported by its association with increased mortality in this study and neurocognitive impairments in previous studies [24, 25]. However, gastroenteritis appeared to be more frequent in simple seizures compared to complex seizures as would be expected.

Nearly half (46 %) of the admissions were complex seizures, a significant reduction from 70 % reported in previous studies in the same population [4, 13]. Since this phenotype is particularly attributable to malaria, the low frequency may be related to lower incidence of malaria during the study period compared to the early 2000s [5]. Complex seizures were associated with mean corpuscular volume, which may be a. marker of: (i) iron deficiency; (ii) red blood cell redistribution following sequestration into the deep capillary bends of the brain [26]; or (iii) presence of alpha thalassaemia [27].

The mortality from acute seizures was greater than in previous studies (6 % vs. 3 %) [1], probably suggesting a change in aetiology related to increase in neonatal seizures, in which mortality was highest (10 %). Additionally, case fatality was common (12 %) of those with encephalopathy of unknown origin which could be important in the neonatal period. Future studies should screen for viral aetiology of acute seizures, which may be associated with mortality but were not assessed in this study [28].

In terms of causes of acute seizures, mortality was highest in malaria (20 %) and malnutrition (20 %). Control of malaria and prevention of malnutrition would reduce just under half of mortalities associated with acute seizures.

These data were prospectively collected over a long time period with clinical and laboratory information documented in standardised proformas, thus findings are robust and reliable. Careful classification of acute seizures allowed objective relation of these phenotypes of seizures with causes, incidence and outcomes. These data may not represent the burden and situation of acute seizures in the community, since some seizures in this area are not treated at hospital. It was not possible to perform electroencephalography, owing to logistical constraints.