The Epilepsy Genetic Association Database (epiGAD): Analysis of 165 genetic association studies, 1996–2008

Summary

We have created the Epilepsy Genetic Association Database (epiGAD, http://www.epigad.org, an online database of epilepsy genetic association studies. A systematic search using several search engines identified 165 studies. Herein we analyze the types of studies available, the sample sizes used, and the strength of the findings. Common questions examined were susceptibility to idiopathic generalized epilepsy, focal epilepsy, or febrile seizures, and pharmacogenomic approaches to drug-resistant epilepsy. Sample sizes were generally small; 80% of studies had 200 or fewer cases, although more recent studies published from 2005–2008 incorporated slightly larger sample sizes. No association was judged as “strong” using current criteria for assessing genetic associations—this is probably due to inadequate sample sizes. Sample sizes need to increase, either by research collaboration or via systematic reviews and meta-analyses. We believe epiGAD will facilitate future meta-analyses.

Purpose

Genetic association studies are case–control studies that aim to identify susceptibility genes predisposing to disease. More than 6,000 such studies are now published annually (Ioannidis et al., 2008). This approach has met with rather limited success to date in epilepsy (Tan et al., 2004; Durner et al., 2006). Initial reports of association based on plausible hypotheses, such as drug transporter variants and antiepileptic drug resistance, have not been supported by further publications (Bournissen et al., 2009).

We have, therefore, created the Epilepsy Genetic Association Database (epiGAD, http://www.epigad.org), an open-access Web-based database of epilepsy genetic association studies. EpiGAD collates data on epilepsy genetic association studies and aims to provide an overview of this field for researchers and clinicians. Herein we analyze trends, characteristics, and outcomes of epilepsy genetic association studies up to the end of 2008.

Methods

Literature search

We searched for epilepsy genetic association studies from 1985–2008. Studies were included in epiGAD and in this analysis if they fulfilled two criteria. First, they must evaluate the association between a polymorphic genetic variant in the general population and an epilepsy phenotype, comparing the frequency of the variant with a control group. We included studies examining susceptibility genes for epilepsy, as well as studies of epilepsy pharmacogenomics. Secondly, studies must be published in a peer-reviewed journal. This second criterion specifically excludes studies reported in abstracts, such as those presented at scientific conferences and meetings. Studies that were electronically published in 2008 but which appeared in print in 2009 were included. We included studies in English as well as those in other languages.

Search strategy

We used PubMed (http://www.pubmed.gov) as the primary search engine. The search strategy employed used the following MeSH terms: (“Epilepsy” [Mesh]) and (“Gene Frequency” [Mesh] or “Genetic Predisposition to Disease” [Mesh] or “Genetic Variation”[Mesh] or “Genotype” [Mesh] or “Phenotype” [Mesh] or “Polymorphism, genetic” [Mesh]) and (“association” or “associated”). The search covered from January 1, 1985 to January 3, 2009, and yielded 1,270 articles. Articles were screened using title and abstract to assess eligibility. We also used the “Related Articles” feature in PubMed and citation snowballing (Booth, 2006) to identify further studies for screening.

We repeated the search using two other search engines—Google Scholar (scholar.google.com) and HuGE Navigator (Yu et al., 2008). The advanced search function was employed for Google Scholar, using the exact phrases “epilepsy” and “genetic” or “association.” The HuGE Navigator Phenopedia was used to identify eligible studies using epilepsy as the disease phenotype.

All identified articles were screened for eligibility using the title, abstract, or full text. For duplicate publications, the most recent version was selected. Data from eligible studies were extracted using a standard form, capturing information from each study about the epilepsy phenotype or epilepsy syndrome, gene, allele, number of cases and controls, type of control (population-based, family based, or both), ethnicity, and p-value for allelic association.

Online database structure

Extracted data were then uploaded into epiGAD. Studies were further categorized into those examining epilepsy susceptibility genes and those examining epilepsy pharmacogenomics; a separate database was created for each category.

Statistical analysis

Statistical analysis was performed with SPSS version 16 (SPSS Inc, Chicago, IL, U.S.A.); p-values of <0.05 were considered statistically significant. The t-test was performed for continuous variables, and the chi-square test for categorical variables.

Results

One hundred sixty-five publications were eligible; 133 studies evaluated 77 putative susceptibility genes, and 32 studies investigated pharmacogenomic effects of 20 genes in epilepsy. A complete list is available online at http://www.epigad.org. No genome-wide association studies in epilepsy were identified.

The studies were published from 1996–2008 across 52 journals. The three journals publishing the highest number of such studies were Epilepsy Research (27 of 165, 16.3%), Epilepsia (20 of 165, 12.1%) and Neuroscience Letters (17 of 165, 10.3%). The number of published studies increased gradually from 1996–2008 (Fig. 1). Four studies (4 of 165, 2.4%) were not found using PubMed, but were instead found using another search engine. Four studies were published in a non-English language; only one of these studies was not listed on PubMed.

Figure 1.   Published epilepsy genetic association studies, 1996–2008.

The most common phenotype for the susceptibility gene studies was idiopathic generalized epilepsy (65 of 133, 48.9%). Studies on focal epilepsies (35 of 133, 26.3%) and febrile seizures (27 of 133, 20.3%) were also common. Five studies (3.8%) used susceptibility to epilepsy, without specification of focal or generalized epilepsy, as the phenotype.

For the epilepsy pharmacogenomic publications, the majority (24 of 32, 75.0%) studied antiepileptic drug-resistant patients, compared to drug-responsive subjects. Seven studies (21.9%) examined the genetics of antiepileptic drug hypersensitivity reactions, whereas one study (3.1%) examined genes involved in vigabatrin visual field constriction.

All 165 studies matched the ethnicity of cases and controls, with cases and controls derived from 25 countries across Europe, the Americas, the Middle East, and Asia. Most studies (148 of 165, 89.7%) selected population-based controls, whereas a minority (6 of 165, 3.6%) used family based controls; 11 studies (6.7%) used both types of controls.

In general, samples sizes were small across all 165 studies. The number of cases was 162.7 ± 13.7 and the number of controls was 202.8 ± 15.4 [values expressed as mean ±  standard error of the mean (SEM)]. The median number of cases and controls were 104 (range 8–1,361) and 126 (range 22–1,390), respectively. The majority (132 of 165, 80.0%) had 200 or fewer cases. Studies examining susceptibility genes tended to have higher numbers of cases and controls compared to pharmacogenomic studies (cases: 171.9 ± 16.3 vs. 124.4 ± 19.6, p = 0.17; controls: 219.1 ± 18.2 vs. 135.0 ± 20.4, p = 0.03).

We then divided all 165 studies into two groups—those published from 1996–2004 (“early studies”) and those published from 2005–2008 (“later studies”) (Table 1). Significantly more pharmacogenomic studies were published from 2005–2008, compared to the period from 1996–2004. Sample sizes for cases and controls were also significantly higher for later studies, compared to earlier studies (p < 0.01).

Table 1.   Early (1996–2004) versus later (2005–2008) studies

	Early studies 1996–2004 (n = 78)	Later studies 2005–2008 (n = 87)	p-Value
SEM, standard error of the mean.
Type of study
Pharmacogenomic, no. (%)	6 (7.7%)	26 (29.9%)	0.001
Susceptibility gene, no. (%)	72 (92.3%)	61 (70.1%)
Sample size
No. of cases, mean (SEM)	118.4 (11.2)	202.4 (23.3)	0.002
No. of controls, mean (SEM)	150.1 (11.7)	245.3 (26.3)	0.001

If we use p ≤ 0.01 as a possible cutoff for declaring a statistically significant association, 35 studies (21.2%) reported a significant association. However, under recent guidelines (Ioannidis et al., 2008), the evidence for these reported associations would be categorized as only weak to moderate; none would be categorized as strong.

Discussion

Our descriptive study surveys the field of epilepsy genetic association studies, covering 165 studies across 13 years. Strengths of our study include a systematic search strategy (Booth, 2006), use of more than one search engine, and standardized data extraction. As suggested recently (Ioannidis et al., 2006; Frodsham & Higgins, 2007), we have made this information available via an online open-access database—epiGAD. We believe our search strategy is complete, as comparison with the National Institutes of Health (NIH) Genetic Association Database (Becker et al., 2004) shows 137 studies of epilepsy compared to epiGAD’s 165 studies. In addition, we believe that the data tabulation and display in epiGAD reflects information (at-risk allele, epilepsy syndrome, sample size) that is relevant for epilepsy researchers and clinicians; in contrast, this information is not found in other online databases such as HuGE Navigator (Yu et al., 2008) or the NIH Genetic Association Database (Becker et al., 2004).

We have shown that the number of published studies has gradually increased each year, especially in the field of pharmacogenomics. These studies cover a wide range of epilepsy syndromes across different ethnicities, with appropriately matched controls. Sample sizes for cases and controls have also increased for later studies. These are, in general, encouraging developments.

However, absolute sample sizes remain small, with 80% of studies including 200 or fewer cases. It is not surprising, therefore, that many initial positive associations have been followed by negative replication studies. Furthermore, existing guidelines(Ioannidis et al., 2008) for genetic association would categorize all of these published associations as “weak” or “moderate” at best. Small sample size is likely to be the reason.

We believe our findings have implications for future epilepsy genetic association studies. Currently there is great interest in genome-wide association studies (GWAS). At this time, no GWAS in epilepsy has been reported, but the lessons from GWAS in other diseases (McCarthy et al., 2008) are that odds ratios are generally ≤1.5. Sample sizes of many hundreds, or better thousands, are needed. Success in both GWAS and gene-specific association studies depends on common variants having at least some role in epilepsy causation and pharmacogenomics, a supposition that appears reasonable but is currently unproven. Rare variants, which are not detectable by association studies, now have a small established role in common epilepsies (Heron et al., 2007; Helbig et al., 2009); work on common variants, with appropriately powered studies, is now even more important to acquire a detailed understanding of the genetic architecture of epilepsies.

Our study has some limitations. First, although we used several search methods (Booth, 2006) to identify eligible studies, we may have overlooked some studies. Secondly, we have categorized published associations using existing criteria created using expert consensus (Ioannidis et al., 2008); these are interim guidelines and are in the process of validation.

In summary, our epiGAD data show that despite more than 150 publications, epilepsy genetic association studies have not demonstrated any strong, convincing associations, probably due to sample size limitations. Our study demonstrates the need for epilepsy researchers to collaborate and pool cohorts to increase sample sizes. Alternatively, systematic reviews and meta-analyses (Bournissen et al., 2009) may be performed to improve the strength of the evidence (Ioannidis et al., 2008); recent guidelines have been published to standardize the reporting of association studies to facilitate meta-analyses (Little et al., 2009). Although heterogeneity between studies may present some difficulties (Bournissen et al., 2009), this may eventually prove tractable. We believe that by systematically searching for and identifying studies, epiGAD will help to facilitate these future meta-analyses.

Acknowledgments

This study was supported by the National Medical Research Council, Singapore (NMRC/0998/2005) and an unrestricted educational grant from UCB Pharma. We also thank the Genetics Commission of the International League Against Epilepsy (ILAE) for advice and support.

We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Disclosure

None of the authors has any conflict of interest to disclose.