Gene 518 (2013) 405–411

Contents lists available at SciVerse ScienceDirect

Gene

j ourna l homepage: www.e lsev ie r .com/ locate /gene

Null genotypes of GSTM1 and GSTT1 contribute to increased risk of diabetes mellitus:A meta-analysis

Jingwen Zhang a, Hu Liu b, Hongyi Yan c, Guoliang Huang a,⁎, Bin Wang d,⁎⁎a Department of Endocrinology, Fujian Institute of Endocrinology, Union Hospital of Fujian Medical University, 29 Xinqu and Road, Fuzhou, Fujian 350001, Chinab Graduate School, Second Military Medical University, Shanghai 200433, Chinac Department of Endocrinology, People's Liberation Army 210 Hospital, Shenyang 116021, Chinad Department of Internal Medicine, Maternal and Children Health's Hospital, Weifang 261011, China

Abbreviations: DM, diabetes mellitus; GSTs, GlutatGlutathione S-Transferase M1; GSTT1, Glutathione S-T95%CI, 95% confidence interval.⁎ Correspondence to: G. Huang, Department of Endo

Endocrinology, Union Hospital of Fujian Medical UnFuzhou, Fujian 350001, China. Tel./fax: +86 8335789⁎⁎ Correspondence to: B. Wang, Department of InternaHealth Hospital of Weifang, 76 Young Road, Weifang 26

E-mail addresses: huangguoliang17@yahoo.cn (G. H(B. Wang).

0378-1119/$ – see front matter © 2013 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.gene.2012.12.086

a b s t r a c t

a r t i c l e i n f o

Article history:Accepted 19 December 2012Available online 4 January 2013

Keywords:Diabetes mellitusGene polymorphismGlutathione S-transferaseMeta-analysis

Diabetes mellitus (DM) is a common disease which results from various causes including genetic and environ-mental factors. Glutathione S-TransferaseM1 (GSTM1) and Glutathione S-Transferase T1 (GSTT1) genes are poly-morphic in human and the null genotypes lead to the absence of enzyme function. Many studies assessed theassociations between GSTM1/GSTT1 null genotypes and DM risk but reported conflicting results. In order to geta more precise estimate of the associations of GSTM1/GSTT1 null genotypes with DM risk, we performed thismeta-analysis. Published literature from PubMed, Embase and China Biology Medicine (CBM) databases wassearched for eligible studies. Pooled odds ratios (OR) and corresponding 95% confidence intervals (95%CI)were calculated using a fixed- or random-effectsmodel. 11 publications (a total of 2577 cases and 4572 controls)were finally included into this meta-analysis. Meta-analyses indicated that null genotypes of GSTM1/GSTT1 anddual null genotype of GSTM1–GSTT1 were all associated with increased risk of DM (GSTM1: OR random-effects=1.60, 95%CI 1.10–2.34, POR=0.014; GSTT1: OR random-effects=1.47, 95%CI 1.12–1.92, POR=0.005; GSTM1–GSTT1:OR fixed-effects=1.83, 95%CI 1.30–2.59, POR=0.001). Subgroup by ethnicity suggested significant associationsbetween null genotypes of GSTM1 and GSTT1 and DM risk among Asians (GSTM1: OR random-effects=1.77, 95%CI1.24–2.53, POR=0.002;GSTT1: OR random-effects=1.58, 95%CI 1.09–2.27, POR=0.015). Thismeta-analysis suggestsnull genotypes of GSTM1/GSTT1 and dual null genotype of GSTM1–GSTT1 are all associated with increased risk ofDM, and null genotypes of GSTM1/GSTT1 and dual null genotype of GSTM1–GSTT1 are potential biomarkers ofDM.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Diabetes mellitus (DM), which is estimated that 347 (314–382)million adults are suffering from (Danaei et al., 2011), has becomean important cause of mortality and morbidity worldwide, throughboth direct clinical sequelae and increased mortality from cardiovas-cular and kidney diseases (Danaei et al., 2006; Khaw et al., 2004;Lawes et al., 2004; Nakagami, 2004). DM results from body's ineffec-tive use of insulin, which is determined by several different genes andenvironmental factors. Causes of the DM are both various and com-plex, and one of these causes is oxidative stress, arising as a result

hione S-Transferases; GSTM1,ransferase T1; OR, odds ratio;

crinology, Fujian Institute ofiversity, 29 Xinqu and Road,6 8423.l Medicine, Maternal and Child1011, China.uang), bin.wangdr@gmail.com

rights reserved.

of an imbalance between free radicals and antioxidant defenses(West, 2000). As β-cells are very sensitive to cytotoxic stress becauseof their little expression of the antioxidant enzymes, they are suscep-tible to the oxidative stress attack, and the dysfunction of β-cells afteroxidative stress attack may further result in the development of DM(Tiedge et al., 1997).

Glutathione S-Transferases (GSTs) are the most important family ofphase II isoenzymes known to detoxify a variety of electrophilic com-pounds, including carcinogens, chemotherapeutic drugs, environmentaltoxins, and DNA products generated by reactive oxygen species damageto intracellular molecules, chiefly by conjugating them with glutathione(Hayes et al., 2005). GSTs play a major role in cellular antimutagen andantioxidant defense mechanisms (Baiocco et al., 2006). GlutathioneS-Transferase M1 (GSTM1) and Glutathione S-Transferase T1 (GSTT1)genes are polymorphic in human and the null genotypes result in the ab-sence of enzyme function, contributing to interindividual differences inresponse to xenobiotics (Binkova et al., 2007). In recent yearsmany stud-ies have assessed the associations between DM and GSTM1 and/or GSTT1polymorphisms (Amer et al., 2011; Bekris et al., 2005; Bid et al., 2010;Datta et al., 2010; Hori et al., 2007; Hayek et al., 2006; Ramprasath etal., 2011; Wu et al., 2006; Wang et al., 2006; Yalin et al., 2007).

Table 1Characteristics of included studies in this meta-analysis evaluating the effects of GSTM1/GSTT1 polymorphisms on DM risk.

No. Study Year Ethnicity Null GSTM1 N (%) Null GSTT1 N (%) Dual null N (%)

Case Control Case Control Case Control

1 Wang GY 2006 Asian(China) 40(40) 44(44) 61(61) 51(51)2 Yalin S 2007 Asian(Turkish) 63(64.3) 32(32.7) 21(21.2) 22(22.4) 13(13.2) 6(6.1)3 Amer MA 2011 Africa(Egyptian) 62(62) 47(47) 35(35) 21(21) 19(19) 12(12)4 Hayek T 2006 Europe(London) Data fail to support 155(20) 422(16)5 Bid HK 2010 Asian(North India) 54(54) 73(36.5) 17(17) 28(14.0) 5(5) 8(4.0)6 Datta SK 2010 Asian(North India) 48(48) 35(35) 43(43) 37(37) 25(25) 12(12)7 Hori M 2007 Asian (Japan) 40(63.5) 216(53.2) 37(58.7) 189(46.6) 21(33.3) 97(23.9)8 Tsai JP 2011 Asian(China) 57(67.1) 121(61.1)9 Ramprasath T 2011 Asian(South India) 222(43.36) 56(20.74) 206(40.23) 48(17.80)10 Wu KH 2006 Asian(China) 3(42.86) 17(50) 4(57.14) 16(47.06)11 Bekris LM 2005 Europe(Sweden) 331(52) 266(56) 87(14) 67(14)

406 J. Zhang et al. / Gene 518 (2013) 405–411

Ramprasath T's study demonstrated significant associations betweenGSTM1/GSTT1 null genotypes and DM risk (Ramprasath et al., 2011),and similar results were also reported in other studies (Amer et al.,2011). However, some studies reported different conclusions andshowed that there were no obvious associations between GSTM/GSTT1null genotypes and DM risk (Bekris et al., 2005; Datta et al., 2010; Horiet al., 2007; Wu et al., 2006), or either GSTM1 or GSTT1 caught the asso-ciations (Bid et al., 2010; Hayek et al., 2006;Wanget al., 2006; Yalin et al.,2007). Thus, it remains unclearwhether there are significant associationsbetween GSTM1 and GSTT1 polymorphisms and DM risk.

Small genetic association studies have various designs, differentmethodology and insufficient power, and could inevitably increasethe risk that chance could be responsible for their conclusions,while combining data from all eligible studies by meta-analysis hasthe advantage of reducing random error and obtaining precise esti-mates for some potential genetic associations. Therefore, there is arole for meta-analysis in pooling these studies, particularly to clarifythe effects of GSTM1 and GSTT1 polymorphisms on DM risk. Hence,to address this controversial issue and get a more precise estimateof the associations of GSTM1/GSTT1 null genotypes with DM risk, weperformed a meta-analysis of published data from available studies.

2. Materials and methods

2.1. Search strategy and selection criteria

We collected the relative studies by conducting literature searchthrough the PubMed, Embase and China Biology Medicine (CBM) data-bases (up to May 26, 2012). The search strategy included key words:(glutathione s-transferase or GST or GSTT or GSTM or GSTM1 or GSTT1)and (polymorphism or polymorphisms or genetic polymorphism) and(diabetes mellitus or DM or diabetes). The full text of the candidate ar-ticles were further examined carefully to determine whether theyaccorded with the inclusion criteria for the meta-analysis. Thecriteria were as follows:(1) Case–control studies which evaluatedassociations between GSTM1/GSTT1 polymorphisms and DM risk;(2) Used an unrelated case–control design; (3) Had an odds ratio(OR) with 95% confidence interval (95%CI) or other available data

Table 2Main results of meta-analysis of the associations between GSTM1/GSTT1 polymorphism an

Genotype Studies (no. of cases/co

GSTM1 Total studies 10(1408/1908)Subgroup analyses by ethnicity (Asians) 8(527/594)

GSTT1 Total studies 10(2493/5184)Subgroup analyses by ethnicity (Asians) 7(389/375)

GSTM1–GSTT1 Total studies 5(461/904)Subgroup analyses by ethnicity (Asians) 4(64/123)

(* M, model of meta-analysis; R, random-effects model; F, fixed-effects model; + PH: the P

to estimate OR (95%CI). If two or more articles reported outcomesfrom the same case–control study, we selected the one with themost subjects.

2.2. Data extraction

Two investigators extracted all the data independently accordingto a pre-established form. For each study, except extracting the dataof null genotypes of GSTM1 and GSTT1, we also recorded some sub-jects including different ethnicities, year of publication, types of DM,complications, and the source of cases and controls. During the pro-cess, discrepancies in the extracted data were settled by discussionamong all reviewers.

2.3. Statistical analysis

We estimated the strength of the associations between GSTM1/GSTT1 polymorphisms and DM risk by pooled ORwith its 95%CI. Duringthe process, the significance of pooled OR depend on a P value by Z test,and it was considered to be significant if Pb0.05. To summarize thepooled OR, we applied two models, the Mantel–Haenszel's fixed effectmodel or DerSimonian Laird's random effect model according to theheterogeneity among included studies (DerSimonian and Laird, 1986;Mantel and Haenszel, 1959). In this analysis, we applied two differentmodes to quantify the heterogeneity, including Cochran's Q statistic(Cochran et al., 1954) and I2 statistic (Higgins et al., 2003). Moreover,subgroup analyses were used to investigate the possible sources of het-erogeneity due to different characteristics of included studies (TheCochrane Collaboration, 2006). In the subgroup analysis, we analyzethe data by ethnicities, in which, Asians occupied mainly proportion(8 publications), therefore we mostly operated the subgroup analysisamong Asians. The Galbraith plot was also used to evaluate the majorsources of heterogeneity by spotting the outliers (Galbraith, 1988). Ifthe heterogeneity was significant (Q statistic Pb0.05 or I2>50%), therandom effect model was conducted; otherwise, the fixed effect modelwas used (DerSimonian and Laird, 1986). Sensitivity analysis was usedto assess the stability of pooled studies by sequential omission of individ-ual studies. Eventually, we examined the publication bias by funnel plot,

d DM risk.

ntrols) Odds ratio M* Heterogeneity PEgger's test‡

OR[95% CI] P OR I2 (%) PH+

1.60[1.10,2.34] 0.014 R 83.1 b0.001 0.3851.77[1.24–3.53] 0.002 R 69.4 0.002 0.2061.47[1.12,1.92] 0.005 R 67.3 0.001 0.8031.58[1.09–2.27] 0.015 R 65.9 0.004 0.9641.83[1.30,2.59] 0.001 F 0.0 0.835 0.8951.84[1.25–2.72] 0.002 F 0.0 0.699 0.908

value of heterogeneity test; ‡ P Egger's test, the P value for Egger's test).

407J. Zhang et al. / Gene 518 (2013) 405–411

in which the standard error of logor of each studywas plotted against itslogor, and an asymmetric plot suggested possible publication bias.Furthermore, funnel plot's asymmetry was assessed by Egger's lin-ear regression test (Egger et al., 1997). Statistical analyses wereperformed with the software program STATA (version 12.0). All Pvalues were two-sided.

Fig. 1. Forest plots of pooled OR with 95%CI for associations between GSTM1/GSTT1 polymorOR and 95% CI; the box size is proportional to the meta-analysis study weight; the diamonassociation of the null genotype of GSTM1and DM risk (analysis of total studies; null vs. prof null genotype ofGSTT1 and DM risk (analysis of total studies; null vs. present; random-enotype of GSTM1/GSTT1and DM risk (analysis of total studies; dual null genotype vs. non-n

3. Results

3.1. Characteristics of included studies

Following the search strategy introduced above and the selectioncriteria, we collected 21 records, and all of them were obtained in

phisms and DM risk (The squares and horizontal lines correspond to the study-specificds represent the pooled OR and 95% CI). (A) Forest plot of pooled OR with 95% CI foresent; random-effects model). (B) Forest plot of pooled OR with 95% CI for associationffects model). (C) Forest plot of pooled OR with 95% CI for association of dual null ge-ull genotype; fixed-effects model).

Fig. 1 (continued).

408 J. Zhang et al. / Gene 518 (2013) 405–411

full-text publications. Then all the publications were further assessedfor inclusion, and we excluded 10 publications, in which two for notlisting available data for analysis (Oniki et al., 2008; Tiwari et al.,2009) and the others for controls containing DM patients and investi-gating the associations of GSTM1/GSTT1 null genotypes with DM com-plications (Cilensek et al., 2012; Doney et al., 2005; Hovnik et al.,2009; Hossaini et al., 2008; Lawes et al., 2004; Schneider et al.,2008; Tang et al., 2010; Watanabe et al., 2003). Thus, 11 articles (atotal of 2577 cases and 4572 controls) were finally involved in themeta-analysis (Table 1) (Amer et al., 2011; Bid et al., 2010; Bekris etal., 2005; Datta et al., 2010; Hayek et al., 2006; Hori et al., 2007;Khaw et al., 2004; Nakagami, 2004; Wang et al., 2006; Wu et al.,2006; Yalin et al., 2007). In the involved 11 studies, 10 articles onGSTM1 polymorphism (a total of 1408 DM cases and 1908 controls),10 studies on GSTT1 polymorphism (a total of 2493 DM cases and5184 controls), and 5 studies on GSTM1–GSTT1 interaction analysis(a total of 461 DM cases and 904 controls). Among those 11 studies,8 individual case–control studies were from Asians, and the otherswere from Africa and Europe.

3.2. Meta-analysis results

Table 2 listed the main results of this meta-analysis.Regarding the association between GSTM1 null genotype and DM

risk, we pooled data from 10 publications. As the heterogeneity wassignificant (PHb0.001, I2=83.1%), the random-effects model wasapplied to pool those studies, and the pooled result indicated a sig-nificant association between GSTM1 null genotype and DM risk(OR random-effects=1.60, 95%CI 1.10–2.34, POR=0.014) (Fig. 1-A).Subgroup analysis by ethnicity still suggested a significant relation-ship between GSTM1 null genotype and DM risk among Asians(OR random-effects=1.77, 95%CI 1.24–2.53, POR=0.002). Because ofthe less studies involved in other ethnicities (Europe: 1 publication;Africa: 1 publication), the subgroup analysis concerning other ethnic-ities apart from Asians was not conducted.

Similarly, pooling data from 10 publications concerning the asso-ciation between GSTT1 null genotype and DM risk, we also found sig-nificant between-study heterogeneity (PH=0.001, I2=67.3%), and

the random-effects model was used to pool data. The pooled resultrevealed that GSTT1 null genotype was associated with increasedrisk of DM (OR random-effects=1.47, 95%CI 1.12–1.92, POR=0.005)(Fig. 1-B). In the subgroup analysis by ethnicity, there was a signifi-cant association between GSTT1 null genotype and increased risk ofDM among Asians (OR random-effects=1.58, 95%CI 1.09–2.27, POR=0.015).

For the GSTM1–GSTT1 interaction analysis, there was no evidencein between-study heterogeneity (PH=0.835, I2=0%), thus, thefixed-effects model was used. The pooled result showed that the dualnull genotype of GSTM1–GSTT1 was also associated with an increasedDM risk (OR fixed-effects=1.83, 95%CI 1.30–2.59, POR=0.001) (Fig. 1-C).

3.3. Heterogeneity analysis and sensitivity analysis

As was displayed above, the between-study heterogeneity of GSTM1polymorphism and GSTT1 polymorphism was significant when availablestudies were pooled into meta-analyses (GSTM1: PHb0.001, I2=83.1%;GSTT1: PH=0.001, I2=67.3%), but the heterogeneity was not significantin theGSTM1–GSTT1 interaction analysis. To investigate themajor sourcesof heterogeneity, we used some methods including subgroup analysis,meta-regression analysis and Galbraith plot. As the studies involvedwere less (11 publications), the univariate analysis of meta-regressionmight be considered to be unacceptable and could cause some biases(Tompson and Higgin, 2002). For the GSTM1 polymorphism, subgroupanalysis amongAsians still indicated significant between-study heteroge-neity (PH=0.002, I2=69.4%). Galbraith plots spotted three studies (No.2,No.9 and No.11) as outliers (Fig. 2-A) and when they were removed,heterogeneity was adjusted (PH=0.305, I2=16.4%). For GSTT1 polymor-phism, there was also significant between-study heterogeneity in sub-group analysis among Asians (PH=0.016, I2=61.5%). Galbraith plotsspotted two studies (No.9 and No.11) as outliers (Fig. 2-B), and whentheywere omitted, heterogeneitywas also adjusted (PH=0.828, I2=0%).

Sensitivity analysis was used to validate the credibility of pooledoutcomes and assess the stability in meta-analysis (Saltelli, 2008).Using the sensitivity analysis method, by which the outliers spottedby Galbraith plots method were dropped sequentially, we found nochanges in pooled ORs in the analyses of both GSTM1 polymorphism

A

b/se

(b)

1/se(b)

b/se(b) Fitted values

0 8.21025

-2-2

0

2

4.99622

Wu KH

Yalin S

Datta SKAmer MA

Wang GY

Hori M

Tsai JP

Bid HK

Ramprasa

Bekris L

B

b/se

(b)

1/se(b)

b/se(b) Fitted values

0 9.57816

-2-2

0

2

5.45063

Wu KH

Yalin S

Bid HK

Amer MA

Datta SK

Wang GYHori M

Ramprasa

Bekris L

Hayek T

Fig. 2. Galbraith plots of associations between GSTM1/GSTT1 polymorphisms and DM risk. (A) Galbraith plot of the association between null genotype of GSTM1and DM risk (analysis ofpooling a total of 10 studies). (B) Galbraith plot of the association between the null genotype of GSTT1 and DM risk (analysis of pooling a total of 10 studies).

409J. Zhang et al. / Gene 518 (2013) 405–411

and GSTT1 polymorphism (All the POR values for GSTM1, GSTT1 wereless than 0.001).

3.4. Publication bias

Funnel plot analysis showed no evident asymmetry for the meta-analyses of GSTM1 polymorphism (Fig. 3-A), GSTT1 polymorphism(Fig. 3-B) and GSTM1–GSTT1 interaction analysis (Fig. 3-C). The resultsof Egger's test also indicated no significant publication bias (P valuesof Egger's test for GSTM1 polymorphism, GSTT1 polymorphism,GSTM1–GSTT1 interaction were 0.385, 0.803 and 0.895, respective-ly). Thus, there was no obvious evidence of publication bias in thismeta-analysis.

4. Discussion

GSTs are one of themajor components of phase II drug-metabolizingenzymes and antioxidant systems. GSTs catalyze the conjugation ofglutathione to a wide range of electrophiles and represent a protectivemechanism against oxidative stress, and play a major role in cellularantimutagen and antioxidant defense mechanisms (Baiocco et al.,2006). As the strong ability to protect cell from oxidative stress damage,theGSTM1 andGSTT1 genes,which are justmembers of theGSTs family,

have been widely studied, and individuals with null genotypes ofGSTM1 and GSTT1 genes may be susceptible to oxidative stress. There-fore, null genotypes ofGSTM1 andGSTT1 are considered to be associatedwith increased risk of DM for the lack of protective ability against oxida-tive stress in β-cells (Yalin et al., 2007).

In recent years, a few studies have been published to investigatethe associations between GSTM1/GSTT1 polymorphisms and DM risk,but the results of those studies are not always in accordance. For in-stance, Yalin S's study (Yalin et al., 2007) revealed that GSTM1 poly-morphism played an important role in DM development, but theGSTT1 polymorphism did not present significant relationship withDM. However, in Amer MA's study (Amer et al., 2011), the resultsshowed a highly significant increase in the frequency of both GSTM1and GSTT1 genotype in DM cases. Thus the definite effects of GSTM1and GSTT1 polymorphisms on DM development are still unclear.Small genetic association studies have various designs, differentmethodology and insufficient power, and could inevitably increasethe risk that chance could be responsible for their conclusions,while combining data from all eligible studies by meta-analysis hasthe advantage of reducing random error and obtaining precise esti-mates for some potential genetic associations. Therefore, a systematicreview and meta-analysis on the associations between GSTM1/GSTT1polymorphisms with DM risk is urgently needed. To the best of our

A Begg's funnel plot with pseudo 95% confidence limits

logo

r

s.e. of: logor0 .2 .4 .6 .8

-2

-1

0

1

2

B Begg's funnel plot with pseudo 95% confidence limits

logo

r

s.e. of: logor0 .2 .4 .6 .8

-2

-1

0

1

2

C Begg's funnel plot with pseudo 95% confidence limits

logo

r

s.e. of: logor0 .2 .4 .6

-1

0

1

2

Fig. 3. Funnel plots to explore publication bias. (A) Funnel plot of the association be-tween null genotype of GSTM1 and DM risk (analysis of total 10 studies). (B) Funnelplot of the association between the null genotype of GSTT1 and DM risk (analysis oftotal 10 studies). (C) Funnel plot of the association between the null genotype ofGSTM1–GSTT1 interaction and DM risk (analysis of total 5 studies).

410 J. Zhang et al. / Gene 518 (2013) 405–411

knowledge, this study is the first systematic review that investigatesthe associations of GSTM1 and GSTT1 polymorphisms with DM riskby meta-analysis method. In our study, 11 publications were involvedinto meta-analysis. The results indicated that GSTM1 null genotype,GSTT1 null genotype, and dual null genotype of GSTM1–GSTT1 wereall associated with increased risk of DM (all the OR values weremore than 1.0 and P values were less than 0.05). The test of publica-tion bias suggested that combined analyses were unbiased, and no ob-vious publication bias existed in our meta-analysis.

Heterogeneity is a potential problem when interpreting the resultsof all meta-analyses, and finding the sources of heterogeneity is one ofthe most important goals of meta-analysis (Ioannidis et al., 2007). Inthe study, the heterogeneity was significant when we investigated theassociations between null genotypes of GSTM1/GSTT1 and DM risk(GSTM1: PHb0.001, I2=83.1%; GSTT1: PH=0.001, I2=67.3%). On theother hand, there was no significant heterogeneity in estimating the as-sociation between GSTM1–GSTT1 interaction and DM risk. To investi-gate the major sources of heterogeneity, we brought in subgroupanalysis by ethnicity, but the heterogeneity was still significant in thesubgroup analyses.Moreover, Galbraith plotwas applied to identify out-liers, and after omitting the outliers spotted, the heterogeneity was ad-justed. Besides, we further found Ramprasath T's study (Ramprasath etal., 2011) and Bekris LM's study (Bekris et al., 2005)were both containedin the outliers in the analyses of GSTM1 null genotype and GSTT1 null ge-notype. In Ramprasath T's study, the controls contained patients suffer-ing from coronary artery disease (CAD). CAD might be associated withGSTM1/GSTT1 genotypes, and whether this association contributed tothe heterogeneity remains uncertain and needs further study. Besides,in Bekris LM's study we found that the cases were diagnosed Type 1 di-abetes (T1DM). Thus, theremight be different effects ofGSTM1/GSTT1 ge-notypes on different types of DM risk, and this difference may result inthe high heterogeneity. In addition, as the great numbers of those twostudies (Ramprasath T's study was 782, and Bekris LM's study was1123), the pooled results of all the studies were easily influenced bythose two outliers.

Some possible limitations in this meta-analysis should be acknowl-edged. Firstly, the studies involvedwere relatively less. Although the re-sults were statistically significant, some negative results that were notpublished might biased the results for its weak stability. Besides, thestudies contained less ethnicity, only Asian (8 publications), Europe (2publications) and Africa (1 publication), and it was not enough to ana-lyze different effects of GSTM1/GSTT1 genotypes on DM risk among dif-ferent ethnicity. Secondly, we did not assess the effects of GSTM1/GSTT1genotypes on the complications of DM for limited studies. In fact, thecomplications of DM had been reported to be associated with GSTM1/GSTT1 genotypes, so more analyses are further needed to identify theeffects of GSTM1/GSTT1 genotypes on risk of DM complications. Finally,gene–environment or gene–immunity interactions were not fullyaddressed in this meta-analysis for the lack of sufficient data. As is gen-erally accepted, aside from genetic factors, exposure to harmful envi-ronment and immunity imbalance were thought to be important inbringing about and promoting the DM. Further studies are needed to in-vestigate the gene–environment and gene–immunity interactions.

In summary, our meta-analysis reveal that null genotypes ofGSTM1/GSTT1 and dual null genotype of GSTM1–GSTT1 are all associatedwith increased risk of DM.Despite all that, to get amore exact conclusionon the effects of GSTM1/GSTT1 polymorphisms on DM risk, more care-fully designed case–control studies with large sample size are needed.

Conflicts of interest

None.

Author contributions

Guoliang Huang directed the whole study's design; Jingwen Zhangundertook major work of the study, including statistical operationand writing; and Hu Liu collected publications concerning the study.

Funding

None.

411J. Zhang et al. / Gene 518 (2013) 405–411

References

Amer, M.A., Ghattas, M.H., Abo-Elmatty, D.M., et al., 2011. Influence of glutathione S-transferase polymorphisms on type-2 diabetes mellitus risk. Genet. Mol. Res. 10,3722–3730.

Baiocco, P., Gourlay, L.J., Angelucci, F., et al., 2006. Probing the mechanism of GSH acti-vation in Schistosoma haematobium glutathione-S-transferase by site-directed mu-tagenesis and X-ray crystallography. J. Mol. Biol. 360, 678–689.

Bekris, L.M., Shephard, C., Peterson, M., et al., 2005. Glutathione-s-transferase M1 andT1 polymorphisms and associations with type 1 diabetes age-at-onset. Autoimmu-nity 238, 567–575.

Bid, H.K., Konwar, R., Saxena, M., et al., 2010. Association of glutathione S-transferase(GSTM1, T1 and P1) gene polymorphisms with type 2 diabetes mellitus in northIndian population. J. Postgrad. Med. 56, 176–181.

Binkova, B., Chvatalova, I., Lnenickova, Z., et al., 2007. PAH-DNAadducts in environmen-tally exposed population in relation to metabolic and DNA repair gene polymor-phisms. Mutat. Res. 620, 49–61.

Cilensek, I., Mankoc, S., Petrovic, M.G., et al., 2012. GSTT1 null genotype is a risk fac-tor for diabetic retinopathy in Caucasians with type 2 diabetes, whereas GSTM1null genotype might confer protection against retinopathy. Dis. Markers 32,93–99.

Cochran, W.G., et al., 1954. The combination of estimates from different experiments.Biometrics 10, 101–129.11.

Danaei, G., Lawes, C.M., et al., 2006. Global and regional mortality from ischaemic heartdisease and stroke attributable to higher-than-optimumblood glucose concentration:comparative risk assessment. Lancet 368, 1651–1659.

Danaei, G., Finucane, M.M., Lu, Y., et al., 2011. National, regional, and global trends infasting plasma glucose and diabetes prevalence since 1980: systematic analysisof health examination surveys and epidemiological studies with 370 country-years and 2.7 million participants. Lancet 378, 31–40.

Datta, S.K., Kumar, V., Pathak, R., et al., 2010. Association of glutathione S -transferaseM1 and T1 gene polymorphism with oxidative stress in diabetic and nondiabeticchronic kidney disease. Ren. Fail. 32, 1189–1195 (Inc.).

DerSimonian, R., Laird, N., 1986. Meta-analysis in clinical trials. Control. Clin. Trials 7,177–188.

Doney, A.S., Lee, S., Leese, G.P., 2005. Increased cardiovascular morbidity and mortalityin Type 2 diabetes is associated with the glutathione S transferase theta-null geno-type:a Go-DARTS study. Circulation 111, 2927–2934.

Egger, M., Davey, S.G., Schneider, M., et al., 1997. Bias in meta-analysis detected by asimple, graphical test. BMJ 315, 629–634.

Galbraith, R.F., 1988. A note on graphical presentation of estimated odds ratios fromseveral clinical trials. Stat. Med. 7, 889–894.

Hayek, T., Stephens, J.W., Hubbart, C.S., et al., 2006. A common variant in the glutathione Stransferase gene is associatedwith elevatedmarkers of inflammation and lipid perox-idation in subjects with diabetes mellitus. Atherosclerosis 184, 404–412.

Hayes, J.D., Flanagan, J.U., Jowsey, I.R., 2005. Glutathione transferases. Annu. Rev.Pharmacol. Toxicol. 45, 51–88.

Higgins, J.P., Thompson, S.G., Deeks, J.J., et al., 2003. Measuring inconsistency in meta-analyses. BMJ 327, 557–560.

Hori, M., Oniki, K., Ueda, K., et al., 2007. Combined glutathione S-transferase T1 andM1 pos-itive genotype afford protection against Type 2 diabetes in Japanese. Pharmacogenomics8, 1307–1314.

Hossaini, A.M., Zamrroni, I.M., Kashem, R.A., et al., 2008. Polymorphism of glutathioneS-transferases as genetic risk factors for the development of complications in type2 diabetes mellitus. ICCC-IPACCMS Annual Meeting.

Hovnik, T., Dolzan, V., Bratina, N.U., et al., 2009. Genetic polymorphisms in genesencoding antioxidant enzymes are associated with diabetic retinopathy in Type 1diabetes. Diabetes Care 32, 12.

Ioannidis, J.P., Patsopoulos, N.A., Evangelou, E., 2007. Uncertainty in heterogeneity esti-mates in meta-analyses. BMJ 335, 914–916.

Khaw, K.T., Wareham, N., Bingham, S., et al., 2004. Association of hemoglobin A1c withcardiovascular disease and mortality in adults: the European prospective investi-gation into cancer in Norfolk. Ann. Intern. Med. 141, 413–420.

Lawes, C.M., Parag, V., Bennett, D.A., et al., 2004. Blood glucose and risk of cardiovascu-lar disease in the Asia Pacific region. Diabetes Care 27, 2836–2842.

Mantel, N., Haenszel, W., 1959. Statistical aspects of the analysis of data fromretrospective studies of disease. J. Natl. Cancer Inst. 22, 719–748.

Nakagami, T., 2004. Hyperglycaemia and mortality from all causes and fromcardiovascular disease in five populations of Asian origin. Diabetologia 47, 385–394.

Oniki, K., Umemoto, Y., Nagata, R., et al., 2008. Glutathione S-transferase A1 polymor-phism as a risk factor for smoking-related type 2 diabetes among Japanese. Toxicol.Lett. 178, 143–145.

Ramprasath, T., Senthil, M.P., Prabakaran, A.D., et al., 2011. Potential risk modifications ofGSTT1, GSTM1and GSTP1 (glutathione-S-transferases) variants and their associationto CAD in patients with type-2 diabetes. Biochem. Biophys. Res. Commun. 407, 49–53.

Saltelli, A., 2008. Global Sensitivity Analysis. The Primer. John Wiley & Sons.Schneider, A., Neas, L., Herbst, M.C., et al., 2008. Endothelial dysfunction: associations

with exposure to ambient fine particles in diabetic individuals. Environ. HealthPerspect. 116, 12.

Tang, J.J., Wang, M.W., Jia, E.Z., et al., 2010. The common variant in the GSTM1andGSTT1 genes is related to markers of oxidative stress and inflammation in patientswith coronary artery disease: a case-only study. Mol. Biol. Rep. 37, 405–410.

The Cochrane Collaboration, 2006. Cochrane Handbook for Systematic Reviews of In-terventions 4.2.6 Updated September 2006.

Tiedge, M., Lortz, S., Drinkgern, J., et al., 1997. Relation between antioxidant enzymegene expression and antioxidative defense status of insulin-producing cells. Diabe-tes 46, 1733–1742.

Tiwari, A.K., Prasad, P., Kumar, K.M., et al., 2009. Oxidative stress pathway genes andchronic renal insufficiency in Asian Indians with Type 2 diabetes. J. Diabetes Com-plications 23, 102–111.

Tompson, S.G., Higgin, J.P.T., 2002. How should meta-regression analyses be undertak-en and interpreted? J. Stat. Med. 21, 1559–1573.

Wang, G.Y., Zhang, L., Li, Q.F., 2006. Genetic polymorphisms of GSTT1, GSTM1, andNQO1 genes and diabetes mellitus risk in Chinese population. Biochem. Biophys.Res. Commun. 341, 310–313.

Watanabe, I., Tomita, A., Shimizu, M., et al., 2003. A study to survey susceptible geneticfactors responsible for troglitazone-associated hepatotoxicity in Japanese patientswith type 2 diabetes mellitus. Clin. Pharmacol. Ther. 73, 5.

West, J.C., 2000. Radicals and oxidative stress in diabetes. Diabet. Med. 17, 171–180.Wu, K.H., Chang, J.G., Ho, Y.J., et al., 2006. Glutathione S-transferase M1 gene

polymorphisms are associated with cardiac iron deposition in patients with beta-thalassemia major. Hemoglobin 30, 251–256.

Yalin, S., Hatungil, R., Tamer, L., et al., 2007. Glutathione S-transferase genepolymorphisms in Turkish patients with diabetes mellitus. Cell Biochem. Funct.25, 509–513.