Conceived and designed the experiments: SDSP MB. Analyzed the data: SDSP FBT AC. Contributed reagents/materials/analysis tools: MMT SEOM AS IB WK PMP MB. Wrote the paper: SDSP CK MMT SEOM WK MB.
The authors have declared that no competing interests exist.
We sought to evaluate the relationship between onchocerciasis prevalence and that of epilepsy using available data collected at community level.
We conducted a systematic review and meta-regression of available data.
Electronic and paper records on subject area ever produced up to February 2008.
We searched for population-based studies reporting on the prevalence of epilepsy in communities for which onchocerciasis prevalence was available or could be estimated. Two authors independently assessed eligibility and study quality and extracted data. The estimation of point prevalence of onchocerciasis was standardized across studies using appropriate correction factors. Variation in epilepsy prevalence was then analyzed as a function of onchocerciasis endemicity using random-effect logistic models.
Eight studies from west (Benin and Nigeria), central (Cameroon and Central African Republic) and east Africa (Uganda, Tanzania and Burundi) met the criteria for inclusion and analysis. Ninety-one communities with a total population of 79,270 individuals screened for epilepsy were included in the analysis. The prevalence of epilepsy ranged from 0 to 8.7% whereas that of onchocerciasis ranged from 5.2 to 100%. Variation in epilepsy prevalence was consistent with a logistic function of onchocerciasis prevalence, with epilepsy prevalence being increased, on average, by 0.4% for each 10% increase in onchocerciasis prevalence.
These results give further evidence that onchocerciasis is associated with epilepsy and that the disease burden of onchocerciasis might have to be re-estimated by taking into account this relationship.
Epilepsy is particularly common in tropical areas. One main reason is that many endemic infections have neurological consequences. In addition, the medical, social and demographic burden of epilepsy remains substantial in these countries where it is often seen as a contagious condition and where the aetiology is often undetermined. For several decades, field researchers had reported some overlapping between the geographical distributions of epilepsy and onchocerciasis, a parasitic disease caused by the filarial worm
Recent surveys, based on rapid epidemiological mapping of onchocerciasis (REMO), have revealed that the number of people infected with
The excess mortality of sighted individuals with high microfilarial loads suggests the operation of insidious and systemic involvement, neurologic involvement having been proposed as a possible candidate
The present work aims at testing the hypothesis that in communities with high onchocerciasis endemicity, the prevalence of epilepsy will clearly exceed that found in communities of low onchocerciasis endemicity. We conducted a literature review of epidemiological studies addressing the issue of the onchocerciasis-epilepsy relationship and performed a meta-regression analysis including available population-based data to quantify the influence of onchocerciasis endemicity on the prevalence of epilepsy.
We searched PubMed and ISI Web of Knowledge up to February 2008, with neither past time limit nor language restriction, to identify population-based studies reporting on the prevalence of epilepsy in communities for which onchocerciasis prevalence was available or such prevalence could be estimated. We entered the following search terms and Boolean operators, for matches under any field: epilep* AND onchocerc*, epilep* AND
Studies were selected if both epilepsy prevalence and an indicator of onchocerciasis prevalence were available or such prevalence could be calculated. Inclusion criteria for subsequent analysis were set to incorporate studies: (1) carried out following a population-based design; (2) providing information on the methods used to diagnose epilepsy and onchocerciasis and (3) in which epilepsy prevalence was assessed in the general population, i.e. both in children and adults. Study communities with a sample of less than 10 subjects were discarded.
Two authors (SDSP, MB) independently assessed eligibility and study quality, and extracted data. We recorded all basic parasitological and demographic information from each eligible study into a purpose-built database. The extracted data included demographic characteristics of the population examined (age range and sex), recruitment methods, and number and dates of previous community treatments with ivermectin. For onchocerciasis, specific information was recorded on methods used for parasitological examination. For epilepsy, details on sampling procedures, and definition of epilepsy were recorded.
To quantify the extent to which epilepsy prevalence is associated with onchocerciasis endemicity across the different studies, a meta-regression was performed. Epilepsy prevalence was defined as the outcome and onchocerciasis prevalence as the explanatory variable.
We used a logistic model to assess the relationship between prevalence of epilepsy and that of onchocerciasis. A random effect, capturing between-studies heterogeneity was subsequently incorporated into the model previously outlined. Significance of this effect was tested using the likelihood ratio test
In addition, in order to test whether the relationship between prevalence of epilepsy and that of onchocerciasis was influenced by a specific study, each study was successively omitted from the whole database and the parameters re-estimated. Parameters of the different models were estimated using the non-linear regression procedure (NLMixed) provided in the SAS v8.1 software. This procedure provides Bayes empirical estimates of the study-specific random-effect
The prevalence of onchocerciasis was considered to have been measured by a standard procedure if it has been estimated in the general population (≥5 years old) using the onchocerciasis diagnostic method used by the Onchocerciasis Control Programme in West Africa (OCP): this entails taking a skin biopsy from each iliac crest, using a 2 mm Holth-type punch, and incubating it for 24 h in normal saline before searching for the presence of
Reference (country) | Deviation from standard OCP method for evaluation of onchocerciasis prevalence | Rationale for correction/assumption | Method for correction (for reference see Supporting Information) |
Druet-Cabanac et al. |
Prevalence assessed from nodule palpation | Relationship between prevalence based on nodule palpation and prevalence of microfilaridermia | Relationship estimated through non linear regression performed on data collected in the same onchocerciasis focus |
Gbenou |
Prevalence assessed from skin snips taken at either the scapula or at the iliac crests | Relationship between prevalence based on skin snips from the scapula and prevalence based on skin snips from the iliac crests | Relationship estimated through linear regression performed on published data |
Gbenou |
Skin snips incubated for about 30 min | Relationship between the prevalence obtained at 30 min and at 24 h incubation time | Correction factor estimated from 16 villages with different endemicity levels |
Kaiser et al. |
Prevalence assessed in individuals aged 10–23 years with a time of residency in the study area between 10–19 years | Relationship between prevalence in the general population and in the 10–19 years old population | Relationship estimated through linear regression performed on data collected in a similar onchocerciasis focus |
Taylor et al. |
Prevalence assessed from 5–7 mm skin snips taken at the lower anterior portion of one leg | Distribution of microfilariae in the skin of the shin is similar to that in the calf | Relationship estimated through linear regression performed on published data |
Taylor et al. |
Skin snips incubated for about 20 min | In terms of sensitivity, a large skin snip examined at 20 min similar to a standard snip read at 30 min, then a 30 min to 24 h correction is applied | Correction factor estimated from 16 villages with different endemicity levels |
Newell et al. |
Prevalence assessed from scarification | Sensitivity to detect the presence of microfilariae with this method similar to the standard OCP method | No adjustment needed |
In addition, three studies were carried out in areas where ivermectin treatment campaigns had been performed
In the present paper, we assumed that if epilepsy were associated with onchocerciasis, it might be due not only to the presence of parasites in the cerebral tissue, but also to cicatricial lesions persisting after the disappearance of the parasites after a treatment. This is the case for epilepsy induced by other infectious diseases (e.g. malaria
Four of the 8 studies included in the present analysis referred to the definition of epilepsy proposed by the International League Against Epilepsy (ILAE) in 1993
Reference (Country/No. of study sites) | Assessment of total population | Identification of possible cases | Confirmatory examination/staff qualification | Epilepsy definition |
Gbenou |
Not specified | Door-to-door, random sample of households | Interview (non-standardized), medical examination/Neurologist | ≥2 seizures |
Kaboré et al. |
Specific census | Door-to-door, exhaustive in population ≥15 year | Not specified/Neurologist | ≥2 seizures |
Newell et al. |
Not specified | Community leaders and staff of health centre | Not specified/Health agents supervised by medical doctor | ≥4 grand mal seizures in preceding year if no AED |
≥1 grand mal seizure during preceding year if AED | ||||
Boussinesq et al. |
National census | Lists from community leaders and systematic question on epilepsy during parasitological surveys | Interview (non-standardized)/Medical doctor | ≥2 seizures during the previous 2 years |
Druet-Cabanac et al. |
Not specified | Door-to-door, exhaustive | Not specified/Health agents supervised by a medical doctor | Not specified |
Dozie et al. |
Specific census | Door-to-door, exhaustive (all households) | Interview (non-standardized), medical examination/Medical doctor trained in paediatric neurology | ≥2 seizures within the previous 2 years |
Taylor et al. |
Not specified | Mobilisation of population (self-reporting) | Standardized questionnaire, medical examination/Medical doctor | ≥2 seizures during the previous year |
≥1 seizure during the previous 5 years if AED | ||||
Kaiser et al. |
Specific census | Door-to-door, exhaustive+Self-reporting patients with residency in study area | Standardized questionnaire, medical examination/Medical doctor trained in paediatric neurology | ≥2 seizures within the previous 2 years |
Kipp et al. |
Not specified | Door-to-door, random sample of households | Not specified/Medical doctor | ≥2 grand mal seizures during the preceding year if no AED |
≥1 grand mal seizure during preceding 5 years if AED | ||||
Ovuga et al. |
Not specified | Mobilisation of population (self-reporting) | Not specified/Neurologist | ≥2 seizures during the previous year |
≥1 seizure during the previous 5 years if AED |
Study excluded: see text.
Study site excluded:
a) Site 〈〈Masongora North〉〉 from Kaiser et al.
b) Site 〈〈school〉〉 from Gbenou
AED: individuals under anti-epileptic drug treatment. CAR: Central African Republic.
From the 1752 examined abstracts and research in the
Error bars represent 95% exact confidence intervals. Solid line: predicted relationship estimated by random-effect logistic regression; dashed lines: 95% confidence interval of the model predictions.
Solid line: predicted relationship estimated by random-effect logistic regression; dashed lines: 95% confidence interval of the model predictions.
Dataset | Model | U (S.E.) | -2LL | ||
Observed data | |||||
Epilepsy prevalence vs. onchocerciasis prevalence estimates with fixed-effect model | −5.953 (0.136) | 0.022 (0.002) | Fixed to 0 | 902.8 | |
Epilepsy prevalence vs. onchocerciasis prevalence estimates with random-effect model | −6.745 (0.386) | 0.043 (0.003) | 0.831 (0.435) | 711.8 | |
Corrected data | |||||
Epilepsy prevalence vs. onchocerciasis prevalence estimates with fixed-effect model | −6.589 (0.308) | 0.029 (0.002) | Fixed to 0 | 761.2 | |
Epilepsy prevalence vs. onchocerciasis prevalence estimates with random-effect model | −7.048 (0.308) | 0.041 (0.003) | 0.308 (0.173) | 690.2 |
The logistic model is expressed as
However, as indicated by varying values of the random effect parameter
Study country | Standard error | p-value | |
Benin |
0.455 | 0.336 | 0.22 |
Burundi |
−0.327 | 0.243 | 0.22 |
Cameroon |
−0.66 | 0.226 | 0.02 |
Central African Republic |
−0.538 | 0.211 | 0.04 |
Nigeria |
0.598 | 0.242 | 0.04 |
Tanzania |
0.427 | 0.234 | 0.11 |
Uganda |
−0.533 | 0.242 | 0.06 |
Uganda |
0.663 | 0.285 | 0.05 |
Sensitivity analysis showed that the significance of the association between onchocerciasis prevalence and that of epilepsy was not affected by omission of any of the studies (results not shown).
The present study was carried out in order to evaluate whether the epilepsy prevalence in communities living in
A recent review took an approach different from ours to analyse epidemiological studies searching for a relationship between onchocerciasis and epilepsy
A major difference between the review of Druet-Cabanac et al.
The random effect included in the modelling indicates that the influence of onchocerciasis on epilepsy varied between the studies. This heterogeneity can be due to either differences in the methodology used to assess the prevalence of onchocerciasis and/or that of epilepsy, or to true biological differences modulating the epilepsy/onchocerciasis association. In this respect, there is some indication that the onchocerciasis prevalence from the Nigerian study
In the Cameroonian study site, a 6-fold increase was found in the mortality rate among individuals with epilepsy compared to that in control individuals
The assessment of epilepsy prevalence in the various studies of the present analysis may have been influenced by the occurrence of other endemic diseases known to be involved in the aetiology of epilepsy
An extensive literature comparing the performance of the various diagnostic methods of
The diversity of neuro-epidemiological methods may have influenced epilepsy prevalence obtained in the different studies. For instance, the use of community key informants for case identification may have resulted in an underestimation
This is the first time that the relationship between the prevalences of onchocerciasis and epilepsy has been quantitatively assessed using available data collected at community level to perform adequate statistical analysis. We found that in areas where onchocerciasis is endemic, epilepsy prevalence increases with onchocerciasis prevalence. This is in accordance with the results of case-control studies and supports earlier anecdotal reports of numerous researchers working in various endemic areas
As the methods used to diagnose
(0.14 MB DOC)
Relationship between arcsine transformed onchocerciasis prevalence in the general (≥5 y.o. age group) population and onchocerciasis prevalence in 10–19 y.o. subjects across 51 villages located in the Mbam valley and continuous Lekie area (Cameroon)
(0.46 MB TIF)
(0.49 MB TIF)
Evolution of onchocerciasis prevalence with annual ivermectin community treatments estimated through non linear regression (Pn = P0×qn) where Pn is the prevalence after n annual treatments, P0 is the initial prevalence and 1-q is the annual relative decrease assumed to be constant over time (q estimated as 0.926 [95%CI: 0.909–0.943]). The lines join the observations of a same study, without any assumption on the mathematical pattern of the decrease.
(0.42 MB TIF)
We are grateful to Mrs Mary Holmes for her assistance during the writing of the manuscript.