Prevalence of Depression among Epileptic Patients and its Association with Drug Therapy: A Systematic Review and Meta-Analysis

Background Despite the high prevalence of epilepsy in sub-Saharan Africa and the established relationship between depression and epilepsy, the extent of comorbid epilepsy and depression in the region is still poorly understood. This systematic review and meta-analysis aims to address this gap in the literature by determining the pooled prevalence of depression among epileptic patients in sub-Saharan Africa. Methods A systematic desk review and electronic web-based search of PubMed, Google Scholar, EMBASE, PsycINFO and the World Health Organization’s Hinari portal (which includes the SCOPUS, African Index Medicus, and African Journals Online databases) identified peer-reviewed research studies and doctoral dissertations on the prevalence of depression among patients with epilepsy using pre-defined quality and inclusion criteria. Relevant data were extracted and descriptive summaries of the studies presented in tabular form. The I2 test was used to assess heterogeneity across studies. A random effects model was used to estimate the pooled prevalence of comorbidity at a 95% confidence interval (CI). Funnel plot asymmetry and Egger’s tests were used to check for publication bias. The final effect size was determined by applying a trim and fill analysis in a random-effects model. Results Our search identified 167 studies, of which 16 articles were eligible for inclusion the final analysis. The pooled estimate of prevalence of depression among patients with epilepsy was 32.71 (95% CI: 25.50 - 39.91). Regional sub-group analysis found that the pooled prevalence in East Africa was 34.52 (95% CI: 23.53 - 45.51) and 29.69 (95% CI: 22.7 - 36.68) in Southern and West Africa. The odds of depression among epileptic patients receiving polytherapy were 2.65 higher than in those receiving monotherapy (95% CI: 1.49 - 4.71, I2=79.1%, p < 0.05). Conclusion Our findings indicate high comorbidity in sub-Saharan Africa and suggests that it may be more prevalent there than elsewhere. Comorbidity is statistically associated with polytherapy. Given the high levels of epilepsy in the region, more attention should be paid to incorporating depression screening and treatment into existing epilepsy programs and to revising treatment guidelines on comorbid depression to reduce polytherapy.


Introduction
Epilepsy is one of the most common neurological disorders globally and in sub-Saharan Africa where it affects approximately 10 million people annually(1). Untreated, epilepsy can cause traumatic injuries, impair physical functioning, and reduce social engagement. This in turn can result in significant psychological stress and premature death.
People with epilepsy (PWE) are more vulnerable to psychiatric illnesses. Rates of psychiatric illness are 9% higher among PWE that in the general population and rates of depression, 22% higher. Depression is the most common psychiatric disorder in PWE (2) and major depressive episodes are one of the most common diagnoses among PWE. Social stigma, feelings of frustration and low self-esteem due to the danger and unpredictability of the illness, and the psychotropic effects of antiepileptic drugs (AEDs) have all been posited as reasons for the strong association between the two illnesses (3). There is also evidence that the association between depression and epilepsy may be bi-directional with people who are depressed having higher risk for developing epilepsy perhaps due to higher rates of head injuries and substance abuse (2) Having a depression-related comorbidity is associated with in poorer quality of life and increased suicidal ideation for epileptic patients (4,5). Greater severity of comorbid depression with epilepsy is associated with significantly reduced overall seizure recovery, higher seizure severity, and increased cognitive, emotional, and physical illness (6,7). Clinicians may also find managing anti-depression treatment particularly challenging for their patients with epilepsy due to concerns about drug interactions, the side effects of polytherapy, and fears of lowering seizure thresholds (8,9).
In addition to complicating treatment and reducing the quality of life, depression in PWE also places a strain on health systems, particularly in low-income countries because patients with untreated depression tend to use significantly more health resources. For example, Cramer et al's study of health care utilization among comorbid patients found that epileptic patients with mild to moderate depression had a two-fold increase in medical visits, and those with severe depression, a four-fold increase, compared to those who were not depressed (6,10). Moreover, studies suggest that depressed patients might be less adherent to epilepsy treatment than their non-depressed counterparts (11)and may respond relatively poorly to drug treatment (12).
Several factors have been found to be associated with increased risk of depression in the epileptic population. In sub-Saharan Africa, lower educational status, lower monthly income, frequency of seizure, the side effects of AEDs, and difficulties adhering to AEDs have all been found to be prominent risk factors for depression (13)(14)(15)(16)(17). Polytherapy has also been reported as an important factor but the association between depression and polypharmacy exhibits significant variation across studies; some finding a significant positive association (13)(14)(15)18) and others none (16).
Studying the psychiatric comorbidities of epilepsy in sub-Saharan Africa is important because of the high prevalence of epilepsy in the region. This elevated prevalence is thought to be due to the endemicity of bacterial and parasitic infections that affect the central nervous system and to poor labor and delivery and perinatal care practices that could result in head trauma in infants and young children (19,20) It is also reasonable to expect that the prevalence of depression comorbidity and its negative health and socio-economic effects would be more pronounced in the sub-Saharan African region where social stigma surrounding epilepsy is more pronounced, and the availability of adequate treatment lacking. However, studying the extent of depression and epilepsy comorbidity in sub-Saharan Africa is complicated by the fact that only about 20% of PWE in low-and middle-income countries receive treatment(1) and by poor estimates of the underlying population prevalence of both diseases. While there is relatively consistent evidence of high prevalence of comorbidity globally, most systematic reviews to date have included either no African studies (21,22) or only one or two of studies from the continent (23) while the African literature on the subject has been characterized by considerable variability, inconsistency, and inconclusive findings.
Better information on the extent of this comorbidity is important for reducing inappropriate treatment and for improving suicide prevention in this vulnerable population (24). This systematic review and meta-analysis therefore, aims to synthesize evidence on the prevalence of depression among epileptic adults and adolescents and its association with drug therapy in Sub-Saharan Africa.

Search approach and appraisal of studies
Articles reviewed in this meta-analysis were accessed through electronic web-based database searches, desk reviews of the grey literature, and reference list reviews using the Preferred Reporting Items of Systematic Reviews and Meta-Analysis (PRISMA) checklist guidelines (25). The researchers used the following key terms for the database searches: "depression" AND "epilepsy" OR "co-morbid depression" AND "epilepsy" OR "mental illness" AND "sub-Saharan Africa". Searches were conducted from December 1, 2017 to January 30, 2018.

Inclusion and exclusion criteria
All English-language, full-text articles on observational studies (case-control or cross sectional) conducted in the sub-Saharan Africa region from 2005 to 2017, with adults or adolescents that were published in peer-reviewed journals or grey literature, who use internationally accepted scales to measure epilepsy and depression, who define epilepsy and depression according to internationally accepted definition and that concerned comorbid depression and epilepsy were eligible for inclusion.

Data abstraction and quality assessment
After initial screening, two reviewers (FW and GD) downloaded abstracts to assess them for inclusion. The relevance of the articles was evaluated based on their topic, objectives, and methodology as listed in the abstract. The abstracts were also assessed for agreement with the inclusion criteria. When it was unclear whether an abstract was relevant, it was included for retrieval. At this stage articles deemed irrelevant or out of the scope of the study were excluded and the full text of the remainder downloaded for a detailed review. Two reviewers (GD and FW) then assessed the quality of potentially eligible articles using the Newcastle-Ottawa Scale (NOS) criteria (26). Discrepancies in quality assessment scores were resolved with a third reviewer (HM), whenever appropriate. The average of two independent reviewers' score was used to determine whether the articles should be included. Articles whose NOS quality scores were less than six, or that had methodological flaws, incomplete reporting of results, or for which full text was not available, were excluded from the final analysis. Study researchers made two separate attempts to contact article authors whenever additional study information was needed; for example, when patient outcome data were incomplete.

Outcome of interest
The outcome of interest was the pooled prevalence of depression among epileptic patients in sub-Saharan Africa. The prevalence is measured as the number of comorbid study subjects divided by the number of patients in a study multiplied by 100. We also estimate the association (as measured by crude odds ratios) between comorbidity and polypharmacy as a secondary outcome.

Data analysis
Information on the study characteristics (time frame, study location, study design, sample size, number of comorbid patients, and the age-range of patients) was extracted from each study using a Microsoft Excel spreadsheet template. These data were then transferred to Stata version 11 software to describe the pooled prevalence of depression among epileptic patients and to identify the significant association between the outcome variable and factors. The heterogeneity of study outcomes was assessed using the I 2 test (27) and the Cochran's Qstatistic. Because the test statistic indicated significant heterogeneity among studies (I 2 >70%, p<0.05) a random effects model was used to estimate the pooled prevalence of comorbidity at a 95% confidence interval (CI) and a geographic subgroup analysis was conducted. We used funnel plot asymmetry ( Figure 1) and Egger's and Begg-Mazumdar Rank correlation tests to check for publication bias (28). Because the results of these tests suggested the possible existence of significant publication bias, the final effect size was determined by applying trim and fill analysis in the random-effects model (29). To confirm results, two researchers independently carried out the main statistical analysis and results were cross-checked for consistency.

Identification and description of studies
The database search and desk review yielded a total of 167 articles ( Figure 2). Of these, 133 articles were found in PubMed, Hinari and Google Scholar and the remaining 34 were found from the desk review. After reviewing the titles and abstracts we excluded 143 articles due to irrelevance. The full text of the remaining 24 articles were downloaded and assessed for quality and relevance. An article from Sierra Leone was exclude because the outcome was not clearly stated (30); three articles from Zambia were excluded due to duplication (31), lack of clarity regarding outcome (32) or irrelevance (33); three articles from Nigeria (34), Togo (35) and Benin (35) were excluded due to low quality scores; and one article from Ethiopia was excluded because the full text was not available (36). The remaining 16 studies were included in the analysis (Figure 2). The 16 articles reported on six cross-sectional studies from Ethiopia (14-17, 37, 38), one cross-sectional study from Kenya (13), five from Nigeria (39)(40)(41)(42)(43), two from Rwanda (44,45), one article from Sudan (46) and one from Zambia (47).

Quality Assessment
Most studies were cross-sectional-two (42, 46) used a case-control design; and all recruited participants from in-patient or out-patient clinical settings.
Most studies had moderate sample sizes; only two had small samples of less than 100 participants (40,41). Reported response rates were high (>90%) but more than half of the studies did not report a response rate and only one discussed the characteristics of nonresponders. All studies used standardized methods for measuring depression with the Beck's Depression Inventory (BDI) and the Hospital Anxiety and Depression Scale (HADS) being the tools most frequently used.
Half of the studies examined the factors associated with depression in PWE; three studies examined the factors related to PWE having any psychotic disorder or with them have a lower quality of life.

Publication Bias
Both funnel plots of precision asymmetry and the Egger's test of the intercept indicated the presence of publication bias in the studies. Visual examination of the funnel plot showed it to be asymmetric and Egger's test of the intercept (B0) was 0.54 (95% CI: 0.2 -0.87 p<0.05). To mitigate against publication bias we applied a trim and fill analysis in the random effects model; however, the prevalence estimates did not differ significantly between the initial model and the trim and fill model.

Prevalence of depression in PWE
Depression prevalence ranged from 49.3% in an Ethiopian study (16) to 6.5%, reported in study conducted in Rwanda (44).
Because the I 2 test for heterogeneity indicated significant difference between the studies (I 2 =98 %, p<0.05) and because theoretically we expected that the study settings and socioeconomic contexts might differ radically across these studies we fitted a DerSimonian and Laird random effect model to estimate the pooled prevalence of depression (48,49). In the model each individual study is given a weight based its reported effect size and sample size (50). The studies with the largest weight were Michel (44), Kiko (13), and Tegegne et al (15) with respective weights of 6.5%, 6.43%, and 6.4%. Smaller weights were given for Mosaku,
four of the studies reviewed explicitly intended to find factors associated with depression in epilepsy patients (14)(15)(16)39).

Key Findings
This systemic review and meta-analysis attempted to estimate the pooled prevalence of depression and its association with polytherapy among epilepsy patients in sub-Saharan Africa. We found a very high prevalence of depression (32.71%) among PWE that was roughly 10 percentage points larger than that reported in recent meta-analyses; which reviewed studies that were almost exclusively conducted in Europe, the Americas, or Asia (21)(22)(23). In the one recent meta-analysis that included sub-Saharan Africa, the region was also found to have the highest regional prevalence of comorbidity although that pooled prevalence, 25.6% was lower than that found in our study (23).
While we expected to find significant comorbidity due to the well-established bi-directional relationship between the depression and epilepsy (2,51), the high prevalence of depression reported in the studies under review suggest that the psychological toll of the epilepsy may be particularly severe in sub-Saharan African settings where the illness is more poorly managed, treatment options limited, epilepsy-related stigma high, and the social and economic costs of illness particularly acute. This is because the relationship between depression and epilepsy is not only neurological but also triggered and shaped by socioeconomic factors (52). In settings where epilepsy is thought to be a potentially contagious spiritual curse, PWE face profound social isolation and limited prospects for education, employment, or marriage (20) leaving them physically, emotionally, and economically vulnerable. In addition, endemic poverty in many communities means that PWE must manage the stress of affording treatment while maintaining their livelihoods in households that are often food insecure and financially precarious. The argument that high social stigma, poor treatment, and low socio-economic status in sub-Saharan Africa are important drivers of its higher rates of comorbidity is supported by the findings in the individual studies reviewed that these factors are significantly associated with depression among PWE.
A secondary objective of this study was to determine the effect of polytherapy on the prevalence of depression in PWE. We found that polytherapy doubled the odds of depression (OR= 2.65; 95% CI: 1. 49 -4.71) compared with monotherapy. Polytheraphy not only increases pill burden, making adherence, and therefore control of seizures, more difficult, but also increases the risk of adverse drug reactions and the severity of those reactions (53,54). In addition patients taking multiple medications have higher risk of drug-to-drug interactions and may more prone to medication errors (53,54). Our findings suggest that greater emphasis should be placed on assessing pill burden, tolerability, and potential drug interactions as well as on providing appropriate health education about drug regimens when designing epilepsy treatment guidelines and provider training curricula.
Our regional sub-group analysis found higher prevalence of depression in East Africa than in  (55) and most of these providers are concentrated in the capital city. Non-mental health professionals are often assigned to provide mental health services in Ethiopia, but studies indicate that they lack the knowledge required to deliver comprehensive mental health care (56). Ethiopia's mental health system also faces many other challenges that complicate treatment of epilepsy such as delayed and inadequate supportive supervisions for trainees, lack of funding for supportive supervision and mentoring, and interrupted drug supplies (55). Ethiopia's overall poverty compared to Zambia and Nigeria may also contribute to the difference in regional prevalence rates, since as discussed above, poor socio-economic status is associated with depression in PWE.

Study Limitations
There are several limitations in this review. Firstly, we were only able to review Englishlanguage studies, which may have caused us to omit important studies from Francophone and Lusophone Africa. Secondly, as noted above, many of the studies reviewed were published in small regional journals or were PhD dissertations making it difficult to gauge the extent of peer review. Moreover, due to the absence of data, crude odds ratios were used to estimate factors related with comorbidity, which prevented us excluding confounding factors. The findings of this meta-analysis would be best interpreted keeping these analytical limitations and the limitations of the original studies in mind.

Conclusion
This meta-analysis found that the prevalence of comorbid depression with epilepsy in sub-Saharan Africa was high, and of greater magnitude than that reported in other geographic regions. We also find that comorbidity is and significantly associated with polytherapy. Given the high overall prevalence of epilepsy in sub-Saharan Africa, more attention should be paid to increasing health education on epilepsy to reduce stigma; to incorporating depression screening and treatment into existing epilepsy programs; and to revising treatment guidelines on co-morbid depression to reduce polytherapy. Future research in sub-Saharan Africa should focus on identifying appropriate medication regimens for patients with comorbid depression.

Availability of data and material
Data are available upon request to the corresponding author.

Competing interests
The authors have declared that there is no competing interest.

Author contributions
FW conceived and designed the research protocol. GD and FW conducted the literature review, data extraction, and data analysis tasks, and led the interpretation and drafting of the manuscript. HM and CT assisted with quality control. SB assisted in data interpretation and manuscript preparation and language editing. All the authors (GD, FW, HM, CT, and SB) were involved in revising the manuscript and editing the manuscript.

Funding
No funding was obtained for this study