Cerebral Malaria and Sequelar Epilepsy: First Matched Case-Control Study in Gabon


Address correspondence and reprint requests to Pr. Pierre-Marie Preux at Institute of Neuroepidemiology and Tropical Neurology EA 3174, Faculty of Medicine, 2 rue Dr. Marcland, 87025 Limoges, France. E-mail: preux@unilim.fr


Summary: Purpose: Cerebral malaria (CM) is suspected to be a potential cause of epilepsy in tropical areas. The purpose of this article was to evaluate the relationship between CM and epilepsy in Gabon.

Methods: A matched case–control study was carried out on a sample of subjects aged six months to 25 years and hospitalized between 1990 and 2004 in three hospitals in Libreville, Gabon. Cases were defined as patients suffering from epilepsy and confirmed by a neurologist. Controls were defined as patients without epilepsy. The exposure of interest was CM according to WHO criteria.

Results: In total, 296 cases and 296 controls were included. Of these, 36 (26 cases and 10 controls) had a CM history. The adjusted odds ratio (aOR) to develop epilepsy after CM was 3.9 [95% CI: 1.7–8.9], p < 0.001. Additional risk factors were identified: family history of epilepsy: aOR = 6.0 [95% CI: 2.6–14.1], p < 0.0001, and febrile convulsions: aOR = 9.2 [95% CI 4.0–21.1], p < 0.0001.

Conclusions: This first case–control study on that issue suggests that epilepsy-related CM is an underrecognized problem. It emphasizes the need for further studies to better evaluate the role of convulsions during CM.

Cerebral malaria (CM), a major complication of Plasmodium falciparum (PF) malaria is a febrile encephalopathy responsible for the majority of deaths related to this disease (Snow et al., 1999). Convulsions frequently occur during the acute stage of encephalopathy (Akpede et al., 1993; Waruiru et al., 1996). These are either focal or generalized and sometimes cause a tonic–clonic status epilepticus (Crawley et al., 1996). CM should be suspected when a patient presents an unrousable coma (WHO, 2000). In addition to being associated with a high mortality rate, CM leads to cognitive (Holding et al., 1999; Boivin, 2002), neuropsychiatric (Varney et al., 1997) and neurological (Newton and Krishna, 1998) sequela in survivors.

Epilepsy is one of the most often cited neurological sequela in the literature (de Bittencourt et al., 1996; Newton and Krishna, 1998). However, few studies have been done to quantify the possible relationship between these two pathologies (CM and epilepsy). In a retrospective cohort study in children aged between six and nine, which took place in Kenya, Carter et al. (2004) observed that the prevalence of epilepsy was high in children with a previous history of CM. In a longitudinal study done in Mali, where a cohort of children aged six months to 14 years were followed after a severe or a simple episode of malaria, the incidence rate of epilepsy after CM was higher than subjects who had suffered from symptomatic non-CM (Ngoungou et al., 2006).

The prevalence of epilepsy is particularly high in tropical African regions (Preux and Druet-Cabanac, 2005). Several parasitoses of the central nervous system (CNS) promoted by geoclimatic conditions have been described as etiological factors for epilepsy (Preux and Druet-Cabanac, 2005). In Gabon, malaria is endemic and its transmission is stable. CM represents 24% of the severe clinical forms of the disease in children (Dzeing-Ella et al., 2005). Epilepsy is the third largest cause of hospitalization in neurology wards (Le Bigot, 1994).

The objective of this study was to evaluate the impact of CM and associated convulsions on the course of epilepsy in Libreville, Gabon.


Study site

This study was conducted between December 2003 and June 2004, in the Departments of Neurology, Pediatrics and Neurosurgery of three hospitals in Libreville, Gabon [Libreville Hospital Centre (CHL), Jeanne Ebori Foundation (FJE), and Paediatric hospital Owendo (HPO)]. The CHL is the country's largest hospital. The FJE and HPO are two hospitals run by the “Caisse Nationale de Securité Sociale (CNSS)” essentially treating civil servants and patients ensured by the social security system. Libreville is the administrative and political capital of Gabon, with about 600,000 inhabitants. The climate is equatorial, hot and humid, as it is throughout the country, encouraging the endemic and stable transmission of malaria (Koko et al., 1997).

Population and type of study

This study was a matched case–control study, including patients between the ages of six months and 25 years admitted to one of the three departments involved in this study between January 1990 and May 2004.


The cases were patients hospitalized for the diagnosis of epilepsy, defined according to the epidemiological definition by the International League Against Epilepsy (Commission on Epidemiology and Prognosis, International League Against Epilepsy, 1993) as patients having at least two unprovoked epileptic seizures (i.e., not related with particular situations or medical conditions; febrile convulsions; seizures occurring only during a metabolic or a toxic condition; caused by alcohol; antimalaria drugs or eclampsia) and occurring over a period of more than 24 hours. Seizures during an episode of CM were not considered as epilepsy. A diagnosis of epilepsy, confirmed by a neurologist, had to be written in the patient record.


The controls were patients free of epilepsy and neurological disease, who were admitted to the same hospitals for reasons other than epilepsy. As all cases were hospitalized, we choose not to include controls from the community and not hospitalized. Each control was matched to a case according to age (± five years) and the year of hospitalization (± one year). Like for the cases, they were then recruited according to their hospital file.

Exposure factors

The exposure factors were obtained from patient records and interviews.

Neurological malaria

  • - An antecedent of CM was defined as a profound or unrousable coma (if done, Blantyre coma score scale of two or less), with PF parasitemia without any other explanation for the coma (CSF studies were done searching for meningitis). The presence of a convulsion or recurrent convulsions was noted during the course of a malaria episode with PF parasitemia, with or without coma.
  • - Therefore, in this study, each subject was classified under one of the following categories: CM with coma alone, CM with coma and convulsions, convulsions alone, nothing (neither coma nor convulsions). CM should then occur before the onset of epilepsy.

Other exposure factors

  • - An antecedent of head trauma with loss of consciousness.
  • - A family history of epilepsy defined as a first- or second-degree family member suffering from epilepsy.
  • - An antecedent of birth complications defined as a lengthy and difficult birth, a caesarean section or the use of a mechanical intervention (i.e., forceps).
  • - An antecedent of febrile convulsions, defined by the existence of one or more convulsions during fever appearing before the age of six years, without any evidence of CNS infection.
  • - An antecedent of another infection of the CNS, defined by neurological deficits accompanied by a fever.
  • - An antecedent of another serious disease (confirmed by clinical and/or biological examinations).
  • - Sickle cell disease (SCD), confirmed by clinical and biological examinations (hemoglobin electrophoresis).

Data collection

This case–control study was carried out in two phases. The data were initially obtained by reviewing the records, and then, through personal face to face interviews or phone calls, using a standardized questionnaire for investigation of epilepsy in tropical countries (Preux, 2000). The records were systematically reviewed by two investigators, one a physician. Records with insufficient information on past/current medical history (diagnosis, clinical evaluation, parasitological exams) were excluded. Incomplete records were constituted in fact of records with only the diagnosis written on the record without any details. Specific criteria for inclusion based upon adequacy of documentation were applied to both cases and controls. In this case, the following eligible file on the list was examined, until a suitable file could be included. The medical records were reviewed for data verification and to gather additional data namely electroencephalogram (EEG), other examination reports (CT scan or arteriography), treatment compliance and clinical and biological data on exposure factors. Interviews of patients and/or parents were carried out by a physician or a neurologist and the principal investigator, with the objective of gathering information related to the medical history and antecedents (patient history). In addition, any records existing in other departments (e.g., pediatrics) were reviewed to get information about exposure factors. If records concerning the same patient were found in two different departments (e.g., Neurology at CHL and Neurosurgery at FJE), only one file was included. The same procedure was applied for controls. A selection of the controls was done from registers using hospital identification numbers. Notes relating to the first eligible control on the list that matched with a case on age and year of hospitalization were brought up from the medical records store for inspection.

Ethical issues

An informed oral consent was obtained from the subjects or parents of the children included. The authorization for the study was obtained from the Ministry of Health of Gabon.

Statistical analysis

Data was collected by means of standardized questionnaires. The software packages Epi-info 6 [Centers for Disease Control and Prevention (CDC); World Health Organization (WHO)], and Statview 5.0 (SAS Institute, Cary, NC, U.S.A.) were used for descriptive, and comparative uni- and multivariate statistical analysis. Standard statistical tests were used for univariate analysis: Pearson's Chi-square test or Fisher's exact test for the frequency comparisons and the Mann–Whitney test for the comparison of distributions. The analysis of the association between CM or other risk factors and the occurrence of epilepsy was made by the calculation of the matched odds ratios and their 95% confidence intervals. The multivariate analysis was performed using a logistic regression model including the variables for which the significant threshold was less than 0.25 in bivariate analysis. A step-by-step descending analysis was done (Hosmer and Lemeshow, 1989). The level of significance was fixed to 5% for the whole analysis.

The minimum number of subjects needed for the study, taking into consideration one control per case, with a 5% alpha error risk and a power of 90% to detect a minimum OR = 3.0, in a region with CM estimated prevalence of 5%, was 518 subjects in total (259 cases and 259 controls). The odds ratio (OR) was fixed to three because of the previous studies from Carter et al. (2004) who found an OR around four. No study reported up to now the prevalence of CM in Gabonese general population. CM occurred in about 10% of cases with malaria in children from Kenya (Marsh et al., 1995). The peak of prevalence of Plasmodium falciparum infection was about 42% in Libreville in children (Richard-Lenoble et al., 1986). So we fixed the prevalence of CM in controls to 5% to calculate the minimal number of subjects.


In total, 592 subjects (296 cases with epilepsy and 296 controls without epilepsy) were included. There were no differences between the cases and the controls according to the matching criteria (age and year of hospitalization). The maximum difference between the age of a case and his matched control was one year and among the cases less than one year old, the mean difference was only one month. Controls were hospitalized for: Burkitt lymphoma (2), diabetes (2), orthopedic surgery (4), kidney failure (5), dermatology (8), urology (8), otorhinolaryngology (9), urinary infection (16), asthma (13), cardiomyopathy (17), digestive parasitoses (27), viral infections (19), bacterial infections (38), gastroenteritis (59), pulmonary infections (69).


In the 296 cases with epilepsy, 173 (58.4%) were male and 123 (41.6%) were female therefore the sex ratio (M/F) was 1.4. The average age was 11.4 (±6.3) years. The majority of patients (95.6%) lived in Libreville or its suburbs. The epilepsy was generalized in 213 (72%) cases and partial in 83 (28%). Among the cases of partial epilepsy, 65.1% had partial seizures with secondary generalization, 24.1% had simple partial seizures, and 10.8% had complex partial seizures. Etiologies that could explain the occurrence of epilepsy were: a previous history of meningitis, 10 (3.4%), head trauma, 13 (4.4%), prematurity, 2 (0.7%) including one with a hydrocephalus and five (1.7%) with cerebral toxoplasmosis (HIV positive). EEG was performed in 273 (92.2%) patients and was noted abnormal in records of 208 patients, although the abnormalities were mainly diffuse slow-waves. Very few paroxysmal abnormalities (spike-waves) were observed. Arteriography was abnormal in 11 (3.7%) patients. A scan was performed in 12 (4.1%) patients. All were abnormal, with seven showing a head trauma; five were suggestive of cerebral toxoplasmosis. An antiepileptic treatment, mainly monotherapy, was prescribed at the end of the hospitalization, in 171 (57.8%) patients consisting of phenobarbitone, in 108 (36.5%) of valproic acid and in 17 (5.7%) of carbamazepine. CSF study was carried out systematically in the children suspected to have meningitis and if positive, they were not included in the study.


A previous history of neurological malaria (coma and/or convulsions) was found in 67 patients (45 cases with epilepsy and 12 controls), i.e., 11.3% of the total subjects in the study. Recurrent convulsions (average: 3.5 ± 2.0) during the course of malaria episodes were found in 55 patients (44 cases and 11 controls) and an unrousable coma in 36 patients. 14 of these patients (10 cases and 4 controls) were evaluated according to the BCS and the depth of coma was classified as “unrousable coma” in the records of 22 others.

Relationship between cerebral malaria, other risk factors and epilepsy

In total, CM (coma with or without convulsions) was found in the history of 36 subjects of which 26 were cases with epilepsy and 10 were controls. The matched OR was 3.4 (95% CI: 1.6–7.4), p = 0.002. Only one patient suffering from coma without convulsions was observed in each group. Table 1 gives the comparative results for each exposure factor between the cases and the controls. The distribution of the type of epilepsy in the 26 cases with a history of CM was as follows: 18 (69.2%) patients had generalized epilepsy and 8 (30.8%) had partial epilepsy. Among the latter, two had simple partial seizures, one with complex partial seizures and five had partial seizures with secondary generalization. The mean interval between CM and diagnosis of epilepsy in the 26 cases was 13 ± 10 months.

Table 1. Comparisons between cases and controls (nonadjusted analysis), Libreville, Gabon, 2004
 Cases (n = 296)Controls (n = 296)OR95% CIp
  1. FC, febrile convulsions; CI, confidence interval; CM, cerebral malaria.

Sex (male)173 (58.4%)  169 (57.1%)  1.10.7–1.5  0.40 
Family history of epilepsy 37 (12.5%)   8 (2.7%)5.82.4–17.0<0.0001
History of FC (without CM) 63 (21.3%)  10 (3.4%)8.64.2–17.4<0.0001
Severe measles14 (4.7%) 5 (1.7%)3.21.1–11.60.02 
Meningitis10 (3.4%) 6 (2.0%)1.80.5–6.8 0.21 
Sickle cell disease 5 (1.7%)19 (6.4%)0.30.1–0.7  0.003
Neurological malaria
 CM (coma alone) 1 (0.3%) 1 (0.3%)1.20.1–19.70.90 
 CM (coma with convulsions)25 (8.4%) 9 (3.0%)3.41.5–7.5  0.003
 Convulsions19 (6.4%) 2 (0.7%)11.6 2.6–50.7 0.001

Analysis by logistic regression showed that the febrile convulsions (excluding CM), family history of epilepsy and CM (coma and convulsions) remained independent risk factors and sickle cell disease (SCD) remained a protective factor (Table 2). The fraction of children with CM who subsequently developed epilepsy was 61.5% (95% CI: 47.2–75.9) and in the general population 5.0% (95% CI: 3.4–7.4) of all epilepsy in children was attributable to CM.

Table 2. Results of multivariate logistic regression, final model, Libreville, Gabon, 2004
 aOR95% CIp-value
  1. FC, febrile convulsions; aOR, adjusted odds ratio; CI, confidence interval; CM, cerebral malaria.

History of FC (without CM)
 (Reference: negative)9.24.0–21.1<0.0001
Family history of epilepsy
 (Reference: negative)6.02.6–14.1<0.0001
CM (coma with convulsions)
 (Reference: negative)3.91.7–8.9  0.001
Sickle cell disease
 (Reference: negative)0.20.1–0.6  0.006


The results of this first matched case–control study searching for a link between CM and epilepsy suggest that the occurrence of convulsions associated with coma in malaria is a significant risk factor for epilepsy along with history of febrile convulsions and a family history of epilepsy.

Two recent studies are in concordance with our results showing a link between CM and epilepsy. In a study in children aged between six and nine, which took place in Kenya, Carter et al. (2004) observed that the prevalence of epilepsy in children with a previous history of CM (9.2%) (OR: 4.4; 95% CI: 1.4–13.7) and in those with previous history of malaria-related convulsions (11.5%) (OR: 6.1; 95% CI: 2.0–18.3) was higher than in children who had not had malaria (2.2%). In their study, Carter et al. (2004) did not estimate a multivariate model. We also found a significant OR for convulsions alone in the univariate analysis and nonsignificant in the multivariate analysis. This was due to a colinearity of this variable with the variable “history of febrile convulsions.” In fact, all patients with convulsions alone had had febrile convulsions.

Furthermore, most of the children with cerebral malaria have convulsions so the group with coma and no convulsions was small. With one person each in the epilepsy and no-epilepsy groups having coma without convulsions, no conclusion can be made about this condition and the risk of subsequent epilepsy.

In a second longitudinal study done in Mali, where a cohort of children aged from six months to 14 years were followed after a severe or a simple episode of malaria, the incidence rate of epilepsy during a cerebral malaria, compared to a symptomatic noncerebral malaria, was evaluated for the first time (Ngoungou et al., 2006). The annual incidence was 17 new cases per 1,000 children in the cerebral malaria group versus 1.8 new cases per 1,000 children in the noncerebral malaria group, i.e., a relative risk of 9.4 times higher (95% CI: 1.3–80.3) of developing a sequel of epilepsy in children after cerebral malaria was found. In the same study, the occurrence of seizures in some children stopped one year after the treatment, indicating a rather benign form of epilepsy.

Multiple seizures associated with acute malarial encephalopathy are frequent in children (Crawley et al., 1996). During the normal course of malaria, an isolated seizure could simply be due to fever (Akpede and Sykes, 1993; Obi et al., 1994). Febrile seizures in malaria are often not simple seizures (i.e., episodes of short duration, secondary generalization, single seizures etc.). They can also be characterized by their high frequency and their occurrence at a later age than in other febrile illnesses (Wattanagoon et al., 1994). Therefore the observation of the above mentioned type of seizures should lead to the suspicion that they could be associated with the onset of cerebral malaria. In addition, tonic–clonic status epilepticus can occur in the severe stage of malaria leading to the development of neurological sequela (Crawley et al., 1996; Mbonda et al., 1996). Even if, it is unclear if these convulsions are due because of a genetic propensity or caused by exogenous factors of malaria (Versteeg et al., 2003), it can therefore be hypothesized that convulsions during CM have a major responsibility in the occurrence of these epilepsies. Studies elsewhere have shown that febrile convulsions play a major role as a risk factor for the subsequent occurrence of epilepsy, specifically temporal lobe epilepsy (Maher and McLachlan, 1995; Berg, 2003).

The pathophysiology of CM-associated epilepsy has not been studied and the physiopathological mechanisms leading to CM are not yet completely clear. It has been suggested that the removal of parasitized red blood cells (PRBCs) from peripheral circulation and their sequestration within the vital organs, particularly the brain, contribute to the pathogenesis of CM by causing hypoxia and subsequent brain ischemia (Newton and Krishna, 1998). However, this sequestration is not sufficient to explain all the neurological features of severe malaria. The capillary obstruction, even partial, does not explain the frequent reversibility of the cerebral deficit or the low incidence of neurological sequela. Intracranial hypertension with a cerebral oedema could also be a contributing mechanism for cerebral malaria and related death (Newton and Krishna, 1998). This intracranial hypertension could be primarily caused by the sequestration of PRBCs, the circulatory phenomena linked to acidosis, the nitric oxide and other vasoactive mediators (Newton and Krishna, 1998). Additional complex mechanisms are implicated in the physiopathological process of CM. Local metabolic modifications are probably secondary to the microcirculation disturbances and associated with cytokines and other substances that induce the cerebral vascular endothelium to produce neuroactive mediators (WHO, 2000). The CT scan is usually normal but it can reveal vascular lesions or a cerebral oedema, which is more frequently found in children than in adults and occurs late in the evolution of the phenomenon (Newton and Krishna, 1998). Structural lesions such as severe vasculopathy, generalized hemorrhage, Durk malarial astroglial granuloma formed by astroglial reaction were reported in the acute encephalopathic phase (Toro and Roman, 1978). It is conceivable that these lesions could act as epileptogenic foci in survivors.

In the tropical context, the inclusion of people in general population would have not allowed us to gather complete biological and clinical data and indeed would be very difficult to conduct in the African context. We recognize that a study relying on case notes and interviews in African hospitals is likely to suffer from constraints and may be biased in ways that are difficult to identify. However, we believe that the results of this study, which included a large number of patients, are important for public health in Africa.

The FJE and HPO are two hospitals run by the “Caisse Nationale de Securité Sociale (CNSS)” essentially treating civil servants and patients ensured by the social security system. Therefore, as the costs are not an issue in these groups, subjects are well willing to come in the hospitals. CHL is a large hospital in Libreville, which in particular, welcomes people with a low economic status. Therefore we believe that a sample of people with epilepsy, drawn from these three hospitals, should be a good reflect of the situation of the patients in the population of Libreville. In this urban population, most of the families are used to seek care in the hospital for their children, even for common signs such as fever, which is potentially an alert for a malarial episode. Demographic characteristics of the sample (age, sex ratio, type of seizures) were similar to the data existing in other literature about epilepsy in Africa (Preux and Druet-Cabanac, 2005).

In our study, a family history of epilepsy increases the risk of developing epilepsy, a fact confirmed by other studies carried out in Africa (Rwiza et al., 1992; Matuja et al., 2001). Also underscored is the fact that febrile convulsion in childhood increases the risk for developing epilepsy (Annegers et al., 1979). In this study, this risk was increased by nine-fold, slightly lower than the result observed in Nigeria (OR: 11) (Ogunniyi et al., 1987). People with an underlying propensity towards seizures are more likely to have convulsions during a febrile illness. It is then possible that a prolonged postictal state or a decreased level-of-consciousness had resulted to an erroneous diagnosis of CM in few cases. As we used a strict definition based on the presence of unarousable coma, we believe this bias unlikely in our study. Taylor et al. (2004) realized autopsies in 31 cases with PF parasitemia documented in sophisticated laboratories and found that 23% died from other associate causes. This raises a possible problem of misclassification. The numbers of subjects with a medical history of meningitis or measles were low, leading to a low power to find a statistically significant risk for meningitis in the univariate analysis. It is possible also that subjects who had seizures during the measles infection were labeled as severe measles, but this could simply reflect their underlying propensity to have seizures.

The fact that SCD was an independent protective factor for epilepsy was a surprising finding in our results. As a result of a selection bias, SCD patients could be more prevalent in controls because of the high frequency of infections in this group. In fact, infections are one of the most frequent complications of SCD. There is the possibility that, given someone is ill, he is more likely to be hospitalized if he has SCD than if he does not. Finally, CM could also be in the causal pathway of SCD relationship to epilepsy (SCD reduces CM, CM increases epilepsy, so SCD decreases epilepsy). SCD could rather lead to a vascular damage in the brain, which could be a risk factor for subsequent epilepsy. Recent findings suggest that vasculopathy and focal hypoperfusion may be factors in the development of SCD-associated seizures (Prengler et al., 2005). But, to our knowledge, no study has been conducted to specifically quantify a possible relationship between SCD and epilepsy, taking into account the confusing role of malaria. Gabon could be a country where this type of research could be done since the prevalence of homozygous SCD is very high and 24% of the general population carried out the S-trait hemoglobin profile (Koko et al., 1998).

In conclusion, CM is a cause of epilepsy in tropical regions. Long-term monitoring is needed in patients who suffered from a CM episode in their childhood, in particular in the tropical context, where epilepsy is still badly accepted and the patient frequently stigmatized.


Acknowledgments:  This study was the beneficiary of a study grant from la Ligue Française Contre l'Epilepsie (Laboratoire Novartis Pharma)” and received further sponsorship from “l'Agence Universitaire de la Francophonie” and “Conseil Régional du Limousin.” We would also like to thank the doctors—Lucien Mwanyombet, Gertrude Minso, Adrien Mougougou, Christelle Chaviot, and Luc Nzamba—for their help and support in the realization of this study, and Pr. Jorge Seixas for helpful comments.