This article is mentioned in:

Introduction

Migraine is a common, chronic, recurrent andneurovascular disorder which is associated with digestive systemand autonomic nervous system symptoms (1). The two main clinical types are MA(migraine with aura) and MO (migraine without aura). In the generalpopulation, 10% of males and 24% of females suffer from migraine(2,3). A review (4) reported that the global incidence ofadult migraine is over 10%. It has been confirmed that migraine isassociated with other diseases. Research has shown that MA mayincrease the risk of cardiovascular disease (CVD), myocardialinfarction and ischemic stroke in female patients (5). The mortality rate of patients with MAwho have CVD and stroke is higher than that of those who do notsuffer from migraine (6).

However, the pathophysiology of migraine remainsunclear. Previously, the vascular hypothesis (7) proposed that migraine was caused byintracranial and extracranial vascular dysfunction. However, thishypothesis did not address neurogenic changes and did not explainthe typical migraine (MA) and common migraine (MO) phenomena.Neurogenic inflammation may be a key mechanism in stimulating thetrigeminal system and causing the headache.

Tumor necrosis factor (TNF) is a pro-inflammatorymolecule and a polypeptide effector of the inflammatory reactionwhich also appears to play a role in migraine. TNF-α activates thetranscription of calcitonin gene-related peptide (CGRP) and plays akey role in migraine pathophysiology (8). A study (9) revealed that levels of CGRP in theexternal jugular vein are significantly increased during amigraine. The TNF-α gene is located on chromosome 6p21, in theclass III area of the major histocompatibility complex (MHC)(10). The -308G/A polymorphism inTNF-α is associated with certain autoimmune, neoplastic andinfectious diseases (11).Polymorphisms in the TNF-α gene (-308G/A) have been confirmed toincrease the production of TNF-αin vitro.

Migraine is correlated with genetic susceptibility.The correlation between the -308G/A polymorphism in the TNF-α geneand migraine has been widely evaluated. Several studies havereported that TNF-α polymorphisms at -308 may be a risk factor formigraine in the Asian population (12–15),but one study did not agree (16)and the results of another were unclear (17). The reason may be the small numberof samples and different genetic region between these studies. Tofurther study the correlation between the TNF-α -308G/Apolymorphism and migraine risk, we performed a meta-analysis in anAsian population.

Materials and methods

Search strategy

A systematic literature search in HuGENet, Pubmed,EMBASE and Google scholar was carried out to identify originalstudies concerning the correlation betweeen the TNF-α -308G/Apolymorphism and migraine on Sep 10, 2011. The search key wordswere as follows: ‘migraine’, ‘headache’ and ‘variant orpolymorphism or SNP’ and ‘tumor necrosis factor or tumor necrosisfactor-α or TNF-a’ and ‘rs1800629’. We also searched studiesselected from the references of the retrieved studies.

The included articles had to meet the followingcriteria: i) evaluation of the TNF-α -308 G/A polymorphism andmigraine risk; ii) case-control study; iii) sufficient dataconcerning gene frequency.

Data extraction

Two independent investigators (L.G. and Y.Y.)extracted the data and reached a consensus in all cases. Theinformation quoted from each study included: first author, year ofpublication, country or region in which the study was performed,sample size, diagnostic criteria for migraine, selection method ofcontrols, genotyping method, genotype distribution, gene frequency,clinical type of migraine and migraine risk factor. For therepeated studies, we only used the data from the latest and mostcomprehensive research.

Statistical analysis

The Hardy-Weinberg equilibrium (HWE) was used totest the genotype distribution in the controls of each study usingPearson’s Square (P≥0.05) (18).Between-study heterogeneity was evaluated using Cochran’s Qstatistic (19) and theinconsistency index (I2) (20). We set P<0.05 for the Q-test andI2 >50% as the threshold of heterogeneity (20). Random effects models were used whenheterogeneity existed, otherwise fixed effects models wereselected. Logistic regression analysis was used to evaluate theassociation between the TNF-α -308 G/A polymorphism and migrainerisk. As the frequency of the AA genotype in the majority of thestudies was zero, we could not use common methods to select thegenetic model. Therefore, we used three models (AA+GA vs. GG, A vs.G and GA vs. GG) to assess the association between the TNF-α-308G/A polymorphism and migraine risk. Begg’s and Egger’s testswere used to assess possible publication bias. Sensitivity analysiswas performed on all studies, including those that deviated fromHWE. All analyses were performed using STATA 11.1 (Stata, CollegeStation, TX, USA).

Results

Study inclusion and characteristics

Eleven studies concerning the association betweenthe TNF-α -308G/A polymorphism and migraine susceptibility wereretrieved from HuGENet, Pubmed, EMBASE and Google scholar. However,five studies (21–25) were performed in Caucasianpopulations and were excluded from our study. Finally, six studies(12–17) were included in our meta-analysis.Of these included studies, three were performed in Turkey (12,14,16)and the others were carried out in India (15), South Korea (17) and Iran (13). All the included studies werecase-control designed, comprising 1,206 cases and 1,141controls.

Of the included studies, five (13–17)selected migraine patients based on International Headache Society(HIS) diagnosis and one (12)based on the International Classification of Headache Disorders-II(ICHD-II). All controls were healthy, however, the sources ofcontrols in the studies varied (TableI).

Table I. Characteristics of all eligiblestudies in this meta-analysis.

Table I.

Characteristics of all eligiblestudies in this meta-analysis.

			Sample size		Diagnostic criteria
Author (ref.)	Year	Country	Cases	Controls	Cases	Controls	Genotypingmethods	Source ofcontrols	P-value for HWE
Yilmazet al(12)	2010	Turkey	67	96	ICHD-II	Hospital workers,students of the university, family members of patients	RFLP-PCR	Population, hospital,family members of patients	0.5845
Ateset al(14)	2011	Turkey	203	202	HIS	Healthy hospitalworkers with no previous or current history of migraine who livedin Tokat, Turkey	ARMS-PCR	Hospital	0.228
Ghoshet al(15)	2010	India	216	216	HIS	Healthy staff membersand the general population, age- and gender-matched	RFLP-PCR	Population	0.5507
Mazaheriet al(13)	2006	Iran	221	183	HIS	Medical andnon-medical staffs, matched for age and geographic area	SSP-PCR	Hospital andpopulation	0.0006
Herkenet al(16)	2005	Turkey	60	62		Same ethnic origin aspatients	RFLP-PCR	Population	1.0000
Leeet al(17)	2007	South Korea	439	382	HIS	Korean females	PCR	Population	0.1608

[i] ICHD-II,International Classification of Headache Disorders-II; HIS,International Headache Society; HWE, Hardy-Weinberg equilibrium;RFLP, restriction fragment length polymorphism; ARMS, amplificationrefractory mutation system; SSP, single specific primer; PCR,polymerase chain reaction.

In addition, among the included studies, oneconcerned only MO (12) andanother only MA (13), three(15–17) clarified the clinical type ofmigraine and one (14) did notclarify the clinical type.

Meta-analysis database

In the controls, the prevalence rate of AAhomozygosity in the -308G/A variant was 0.52% and the GAdistribution was 13.5%. For clinical types, the prevalence rates ofAA were 0.6% and 0.7% in the control subjects of MA and MO,respectively. The respective prevalence rates of GA were 11.2% and11.9%. The genotype distribution of the included studies and theP-values for HWE testing are shown inTables I andII.

Table II. Distribution of TNF-α genotype andalleles between the cases and controls.

Table II.

Distribution of TNF-α genotype andalleles between the cases and controls.

			Cases					Controls
Author (ref.)	Year	Disease	GG	GA	AA	G	A	GG	GA	AA	G	A
Ateset al(14)	2011	Migraine	125	78	0	328	78	162	40	0	364	40
Ghoshet al(15)	2010	Migraine	175	41	0	391	41	191	24	1	406	26
		MA	65	19	0	149	19	191	24	1	406	26
		MO	110	22	0	242	22	191	24	1	406	26
Herkenet al(16)	2005	Migraine	54	5	1	113	7	53	9	0	115	9
		MA	36	4	0	76	4	53	9	0	115	9
		MO	18	1	1	37	3	53	9	0	115	9
Leeet al(17)	2007	Migraine	377	61	1	815	63	338	41	3	717	47
		MA	54	11	0	119	11	338	41	3	717	47
		MO	282	44	1	608	46	338	41	3	717	47
Yilmazet al(12)	2010	MO	37	23	7	97	37	79	16	1	174	18
Mazaherietal (13)	2006	MA	51	163	7	265	177	94	86	3	274	92

[i] TNF, tumornecrosis factor; MA, migraine with aura; MO, migraine withoutaura.

Main results, subgroup analyses

Six studies concerning the correlation between theTNF-α -308G/A polymorphism and migraine are shown inTable I. Yilmazet al (12), Mazaheriet al (13), Ateset al (14) and Ghoshet al (15) revealed a significant associationbetween the TNF-α -308G/A polymorphism and migraine risk. However,Hasanet al (16) revealedno significant association between the polymorphism and migraineand the results of Leeet al (17) were uncertain. Of these six studies,one (13) deviated from HWE(P=0.0006). After excluding this study, the samples contained 985cases and 958 controls.

The results of the meta-analysis are shown inTable III. The overall ORs and 95%CIs were calculated based on the data of the five included studies.After computing in Stata, random effects models were used and asignificant association between the TNF-α -308G/A polymorphism andmigraine risk was revealed in the A vs. G, GA vs. GG and dominant(AA+GA vs. GG) models. The ORs were 1.735 (95% CI, 1.129–2.666) forA vs. G, 1.781 (95% CI, 1.166–2.718) for GA vs. GG, 1.821 (95% CI,1.153–2.874) for AA+GA vs. GG. There was significant between-studyheterogeneity (I2=69.1% for A vs. G, 60.8% for GA vs. GGand 67.8% for AA+GA vs. GG).

Table III. Summary of comparative results.

Table III.

Summary of comparative results.

		A vs. G		GA vs. GG		AA+GA vs. GG
Variables	N	OR (95% CI)	I2(%)	OR (95% CI)	I2(%)	OR (95% CI)	I2(%)
Total	5	1.735(1.129–2.666)	69.10	1.781(1.166–2.718)	60.8	1.821(1.153–2.874)	67.8
Subgroupanalysis
MA	3	1.516(0.986–2.331)	20.3	1.728(1.095–2.726)	35.2	1.651(1.049–2.598)	31.7
MO	4	1.654(0.916–2.985)	69.80	1.557(1.124–2.156)	49.20	1.650(0.917–2.969)	63.10

[i] MA, migrainewith aura; MO, migraine without aura; OR, odds ratio; CI,confidence interval; I², inconsistency index.

The meta-analysis consisted of three case-controlstudies concerning MA (15–17)and four case-control studies concerning MO (12,15–17).The summary result of the subgroup analysis revealed a significantcorrelation between the TNF-α -308G/A polymorphism and MA risk. TheORs were 1.728 (95% CI, 1.095–2.726) for GA vs. GG and 1.651 (95%CI, 1.049–2.598) for AA+GA vs. GG, and no significant between-studyheterogeneity was found. However, the correlation between the TNF-α-308G/A polymorphism and MO risk was not significant; the ORs were1.654 (95% CI, 0.916–2.985) for A vs. G and 1.650 (95% CI,0.917–2.969) for AA+GA vs. GG, and there was significantbetween-study heterogeneity. The results are shown inTable III.

Begg’s and Egger’s tests were performed to checkpublication bias, however, the result showed that publication biaswas not significant (data not shown).

Discussion

TNF-α is a pro-inflammatory molecule and apolypeptide effector of the inflammatory reaction. It activates thetranscription of CGRP and plays a key role in migrainepathophysiology. The TNF-α -308G/A polymorphism has been confirmedto be correlated with certain neuropsychiatric disorders and anumber of studies have been performed to confirm the hypothesisthat the TNF-α -308G/A polymorphism is associated with migrainerisk; however, the results have been conflicting. Therefore, weconducted this meta-analysis.

In this meta-analysis, the results showed that theTNF-α -308G/A polymorphism was significantly correlated withmigraine risk in several comparisons. Heterogeneity is anunavoidable problem. In our meta-analysis, heterogeneity may be dueto a mixed population, with patients of different ethnicities andfrom different geographic regions. Other factors, includingdiagnostic criteria, genotyping methods and selection methods ofcontrols may also lead to heterogeneity.

The subgroup analysis was based on MA and MO and theresults revealed that the TNF-α -308G/A polymorphism was associatedwith MA risk, but not with MO. It is possible that the geneticsusceptibility to the two clinical types are different. Russell andcolleagures (26,27) have confirmed that the genetic basesof MA and MO are markedly different: MA is more dependent ongenetic factors and MO is determined by genetic and environmentaleffects. In addition, migraine is a complex disease which is alsocorrelated with psychological factors. The small sample size in ourstudy may be another reason for the results of the subgroupanalysis.

The limitations of our study should be addressed.First, the diagnostic criteria for migraine in the included studieswere not the same; for example, one study based the diagnosis onthe International Classification of Headache Disorders-II(ICHD-II), so the use of this article may have led to selectionbias. Second, all of the included studies in our meta-analysis werein English, so certain studies in other languages may have beenmissed and we were unable to provide a more accurately powerfulresult. Third, we lacked an unified source of controls. Thecontrols were selected from three sources: hospital-based, healthypopulation and family members of the migraine patients.

In conclusion, our meta-analysis revealed that the-308G/A polymorphism in the TNF gene is associated with migrainerisk in the Asian population. However, we did not further researchthe gene-to-gene and gene-to-environment interactions of TNF-α-308G/A and migraine. The sample size in the present study wassmall, therefore, larger studies with thousands of subjects shouldbe performed.

Acknowledgements

This study was supported by theSignificant Scientific Research Foundation of the Guangxi HealthDepartment (grant no. 200933), the Science and Technology Projectof Traditional Chinese Medicine, Guangxi (grant no. 200911LX203)and the Scientific Research of the Provincial Education Department,Guangxi (grant no. GZKZ1107).

References