In this study, we pooled the results of four studies on epilepsy in eastern Africa to describe the demographic, clinical, semiological, and neuroradiological characteristics, as well as treatment of PWE.
In the two community-based studies (Malawi and Uganda), a screening of people living in the target area was conducted first, comprising more than 100,000 people. Then, screen positives were seen by a neurologist to confirm diagnosis. Lifetime epilepsy prevalence reported in Malawi (1.1% [1.0–1.3%]) and Uganda (3.3% [3.1–3.4%]) were comparable, yet slightly higher than those from a previously published meta-analysis, which found a pooled lifetime prevalence of 0.7% (0.6–0.9%) for epilepsy in eastern Africa [23]. This could be due to our screening questionnaire, which may also have captured more patients with focal onset seizures, which are often not captured at all. Another explanation could be a higher prevalence of risk factors in our cohorts, e.g. NCC. In addition, other studies previously have reported even higher lifetime prevalence estimates for epilepsy of up to 8% [24, 25].
In general, studies estimate that in LMIC around 50% of PWE have secondary epilepsy [11]. In our study, we found a large proportion of people with seizure onset after adolescence which may indicate a predominance of secondary epilepsy. Unfortunately, we were not able to assess the exact proportion of persons with secondary epilepsy in our study. On the one hand because not many patients showed remarkable findings on CT scan, and for those who did, the temporal association could not be established. On the other hand, patients with obvious traumatic brain injury or perinatal hypoxic brain injury were excluded from further work-up due to study protocol, so that (CT) data on these aetiologies are not representative and most likely vastly underrepresented.
Nevertheless, CT data of almost 950 PWE were available, which to our knowledge represents the largest and most detailed analysis in sub-Saharan Africa. The rate of remarkable CT scan findings was highest in patients with focal onset seizures. Most common findings were neurocysticercosis-associated lesions, but interestingly, an even higher proportion of post-ischaemic lesions could be observed. This finding indicates that cerebrovascular events and post-stroke epilepsy have to be considered as an emerging aetiology for epilepsy in sub-Saharan Africa.
Another frequent CT pathology was cerebral atrophy, which is a common finding in PWE, but the significance and aetiology is hard to determine [26]. The rate of this pathology was higher in patients with general onset seizures, which is also known from previous studies, where patients with idiopathic generalised and genetic seizure syndromes showed diffuse and wide-spread alterations of brain morphology [26,27,28]. Detailed MRI analyses in these patients sometimes also reveal (wide-spread) cerebral malformations such as lissencephaly with pachygyria or microgyria potentially causing both focal and generalized seizures [29]. Further imaging studies in LMIC are needed to characterize the prevalence and significance of these findings.
The majority of PWE in the four studies had generalised onset seizures according to the ILAE criteria. In this analysis, we also assessed seizures according to the classification by Winkler et al. [22], which looks at epileptic seizures in low-resource settings in a more nuanced way, according to local contexts and needs, and tries to imply information on aetiologies and epilepsy syndromes based on seizure type and occurrence without accessory diagnostic. According to that adjusted classification by Winkler et al., most PWE with generalised seizure onset had generalised seizures within a specific age range, which represents the category in which a possible genetic background can be assumed. PWE of this group are suspected to suffer from idiopathic generalised epilepsies in a large proportion of cases. Consequently, our data suggest that despite a known high prevalence of risk factors for secondary epilepsy and previously reported high prevalence of focal epilepsies [30], genetic and idiopathic epilepsy syndromes also seem to play an important role in LMIC. This finding is of note, although a proportion of 30 to 58% among PWE might represent a certain overestimation in our study due to exclusion criteria for some types of secondary epilepsy as mentioned above. Nevertheless, as genetic origins are various and investigations in such a large and diverse cohort could reveal valuable results for both low-income and high-income countries, our results support the view, that it could be the time for initiating genetic epilepsy research in Africa [31] .
Despite a certain proportion of PWE with focal onset seizures in our analysis, most seizures were accompanied by a loss of consciousness and resulted in tonic–clonic seizures. Additionally, injuries during seizures were frequently observed, with a rate of up to 56% in the community-based studies. This is much higher than studies in high-income countries suggest [32, 33] and also is a possible explanation for high epilepsy-associated mortality rates in LMIC [34].
This lack of seizure control may display a lack in adequate ASM therapy. ASM treatment concerning both type and dosage differed substantially between and within countries. In Malawi and Uganda, phenobarbital and phenytoin were almost the only ASM prescribed; they were also still very common in Tanzania. In Tanzania, however, the majority of patients was treated with carbamazepine. Valproic acid was basically not available in any of the countries. These findings display the common picture of ASM treatment status in sub-Saharan Africa [35]. These findings, however, are not only limited to LMIC but can also be observed in high income countries [36]. Treatment choice in Africa very often depends on availability and cost of treatment and phenobarbital/phenytoin are often cheaper than carbamazepine or valproic acid [17]. Even though ASM are widely available, supply chain is often not sustained which results in stock-outs of medication [17, 37]. Nevertheless, stock-out of medication was not the main reason for changing ASM, but inadequate treatment effect and side-effects. Phenobarbital, in particular, is now rarely prescribed in high-income countries due to its relatively strong side effects and has been replaced by newer ASM [38]. Yet, it is still recommended by the World Health Organization (WHO) as first-line treatment in LMIC because of its cost advantage and broad spectrum of activity [35, 39]. Nevertheless, even in low-resource settings, phenobarbital has a higher risk to be withdrawn due to side-effects than carbamazepine, phenytoin, or valproic acid [39,40,41].
We found a quite large proportion of patients who had epileptic seizures more often than monthly, especially in the community-based studies, possibly reflecting the lower rates of ASM treatment compared to the hospital-based settings. However, we also found high frequencies of seizures even in patients receiving ASM, which may be due to barriers to accessing healthcare services, supply shortages, or the cost of ASM (mainly to hospitals and sometimes to patients, although ASM are often available free of charge), but also due to underdosage of ASM, which was common in our study. A large part of PWE in our pooled analysis received ≤ 60 mg phenobarbital or ≤ 400 mg carbamazepine per day.
Interestingly, seizure control varied considerably even among the two hospital-based studies, although both showed high treatment rates and treatment with carbamazepine was available. Haydom in rural Tanzania had a high proportion of patients with seizures only yearly or even less often, and no patients with seizures more frequent than monthly, despite underdosage was present in a certain proportion of patients. In contrast, PWE in urban Dar es Salaam showed high seizure rates, comparable to the community-based settings. This could be a result of the study settings, as PWE in Haydom were primarily recruited by a doctor and invited to follow-up visits in the clinic, probably getting access to ASM for the first time in some cases, resulting in good treatment effect, whereas patients in Dar es Salaam were recruited directly from the clinic. One could suspect a high proportion of patients with regular clinic visits due to difficulties to reach seizure control. This might be also reflected by the fact that this was the study site with the highest number of PWE receiving a combination therapy and having a history of prior change in ASM treatment. More studies comparing rural and urban healthcare settings are needed to assess the possible differences and contributing factors.
Many PWE across studies also received alternative treatment for their seizures. This shows the still very prevalent disbelieves about epilepsy as supernatural or witchcraft. In many regions of Africa, PWE are still being stigmatized which makes it difficult to analyse community-based prevalence and which may be the cause of the prevalent undertreatment of PWE.
Epilepsy has a big impact on social life as shown by the fact that many PWE dropped out of school, especially in Malawi and Uganda. In contrast, > 90% of PWE went to school in urban Dar es Salaam despite inadequate seizure control. One could hypothesize that disbelieves and stigmatization is less frequent in urban areas, although alternative non-ASM treatments were equally present in this group.
We report a rather large proportion of PWE who suffer from psychiatric disorders. Studies estimate that up to 50% of all PWE suffer from mental disorders, most commonly from depression, anxiety or dementia [42,43,44,45]. This proportion, however, is likely an overestimation as phenomena during seizures may be mischaracterised as mood disorders [46]. Those studies were mainly conducted in high-income countries, and only one was from Sierra Leone. For sub-Saharan Africa, data are still scarce. In our analysis, the proportion of patients with psychiatric disorders differed substantially between studies, which may be due to different assessors, as the questionnaire was the same across countries. Interestingly, prevalence of psychiatric disorders was highest in the urban study site of Dar es Salaam. Differences in education and work as well as different degrees of stigmatization of psychiatric disorders may be a possible explanation. In the four studies, only a small proportion of patients suffered from chronic diseases other than psychiatric illnesses, which is not unusual in sub-Sahara Africa, mostly because diseases can be diagnosed with less certainty.
Strengths and limitations
This pooled collection of demographic, clinical, semiological, neuroradiological and therapeutic data on PWE in three countries of East Africa is exceptional and represents sustained work of over 10 years. We give data on epilepsy prevalence proportions derived from two large community-based studies, including screening of over 100,000 people, and describe the different characteristics and treatment of nearly1200 PWE including CT examinations in almost 950 cases, which is unique within the sub-Saharan African setting.
Consequently, this study represents an important piece of work and adds substantially to the body of literature on epilepsy globally, and specifically to the evaluation and understanding of the burden of epilepsy and the persisting treatment gaps in sub-Saharan Africa. This has great policy relevance, not only for the respective countries, i.e. Tanzania, Uganda and Malawi, but also at a global level, considering both, the burden of epilepsy, and the growing burdon of neurological disorders in LMIC in general [47, 48]. In addition, results from the present analysis may also support the WHO Intersectoral Global Action Plan on Epilepsy and other Neurological Disorders which is currently being drafted [19].
However, our analysis also suffers from several limitations.
First, examinations were performed by different personnel and translation of the questionnaire to different languages had to be conducted. This may have influenced the results. Nevertheless, studies applied the same approach to history taking and investigations for the diagnosis of epilepsy. Consequently, the data we focused on in our analyses, such as demographic factors or data on seizure onset and frequency were probably less affected.
As all scans underwent examination by the same radiologist (VR), we consider CT findings reliable and comparable among cohorts. Nevertheless, CT machines and quality of CT scans at different study sites varied considerably and are not comparable to high-income settings. Thus, CT findings may not be as accurate as desired. Additionally, MRI imaging for a more detailed assessment was basically not available.
Due to the studies’ NCC specific settings, investigations focused on evaluation of cryptogenic focal epilepsy and excluded e.g. patients with known traumatic brain injury, leading to a possible overestimation of non-focal epilepsies. A broader approach might have given a more detailed insight in epilepsy aetiologies.
In terms of seizure semiology, we did not assess focal onset signs and symptoms in detail, and therefore were not able to evaluate data with regard to possible epilepsy locations and syndromes such as e.g. temporal lobe epilepsy. Future analyses should give details on awareness during focal seizures, as well as on quality of non-motor symptoms, e.g. autonomic, sensory, cognitive, and emotional symptoms respectively. Additionally, as not all patients were assessed by neurologists, some focal (onset) signs or uncommon seizure semiologies might have been missed in our studies, contributing to a potential under-estimation of focal epilepsies. As accessory diagnostics are limited in LMIC settings, a distinctive and precise history taking might be the best chance to get a more detailed picture of possible epilepsy syndromes and aetiologies.
Another limitation of our study was the mix of hospital-based and community-based studies in our pooled analysis, which probably lead to heterogeneous patient cohorts and occurring selection bias as discussed above. Nevertheless, if one is aware of the possible problems, this heterogeneity also offers the possibility to observe and analyse the possible differences of study population characteristics. Nevertheless, stigma and only slight signs/symptoms might have influenced the recruitment process in the community-based studies [9]. We tried to constantly indicate the differences between hospital-based and community-based studies throughout our manuscript and therefore consider this pooled analysis appropriate and valuable.