Introduction

Asthma is an inflammatory disorder of the airways characterized by reversible airway obstruction and bronchial hyper-responsiveness [1]. To date, several studies have revealed that the complex interaction of cell and pro-inflammatory mediators are responsible for the pathogenesis of asthma. Among the inflammatory mediators, sources of oxidant injury are endogenous reactive oxygen species generated by cellular metabolism and by the inhalation of atmospheric pollutants [2]. These factors, which contribute to the severity and symptom exacerbation of asthma, are countered by both enzymatic and non enzymatic antioxidants, including vitamins C and E and glutathione (GSH), a major protective antioxidant in the lungs that also plays a fundamental role in the regulation of inflammatory responses [3]. Glutathione S-transferases (GSTs) catalyze the conjugation of electrophilic compounds to GSH. Their catalytic activities make these enzymes crucial to the detoxification of a wide range of endogenous and exogenous compounds [4]. Polymorphisms of genes encoding GST enzymes influence their functionality in the lungs and other organs, and this can confer genetic susceptibility to oxidative stress and, consequently, to asthma [5, 6].

For this reason, the potential role for GST polymorphisms to modulate disease and adverse responses has been the subject of numerous molecular epidemiologic studies [7–9]. The most extensively studied GST polymorphisms occurs in three isozymes: GSTM1, GSTP1, and GSTT1. Gene polymorphisms affect the activity of these enzymes: gene deletion polymorphisms occurs in the GSTM1 and GSTT1 gene, whereas a missense substitution (I105V, rs1695) characterizes GSTP1 [10].

Considering their biochemical function, the association between GST genes and asthma development risk has been thoroughly examined; published data have reported contrasting results [11, 12]. Specifically, studies have assessed the effects of GSTM1 and GSTT1 null genotypes and GSTP1 Val105 allele. Studies on GSTM1, GSTT1 and asthma have confirmed or rejected that these genes have been associated with an increased susceptibility to asthma development in children and adults. Also, for the GSTP1 gene, the Val allele reportedly protects against asthma in adults [13–15]. Conversely, in other studies, the Val105 allele seems to increase the risk of asthma in children [16, 17]. Others authors have not noted evidence of associations between GSTP1*Ile105Val polymorphism and asthma [18, 19]. All the mentioned studies have reflected upon the meta-analytical level, albeit with significant potential for improvement. Concerning the association between GST polymorphisms and asthma, only two meta-analyses have been published, with conflicting outcomes [2, 20]. Careful examination of the studies included therein reveals, however, that other eligible case–control studies have not been taken into account; additional studies have appeared, thereby increasing the amount of the available data. Furthermore, the previous meta-analyses did not analyze the effects of important confounding variables, such as ethnicity and urbanization.

The aim of our study is to perform a meta-analysis that includes genetic association studies on GST and asthma risk, analyzing the effect of ethnicity, population age, and urbanization in order to explain the heterogeneity of the results.

Materials and methods

Literature search

Eligible articles were identified through a search of the Medline and Hugenet databases through February of 2012 using combination of the following keywords: “glutathione S-transferase”, “GST”, “glutathione S-transferase M1”, “GSTM1”, “glutathione S-transferase P1”, “GSTP1”, “glutathione S-transferase T1”, “GSTT1”, “asthma”, and “asthma development”. In addition, we checked all the references of relevant reviews and eligible articles that our search retrieved. No restrictions were placed on language or type of report, and conference abstracts were also included. When multiple reports were available for a single study, only the most recent article, or the article with the largest sample size, was included. Authors of studies were contacted for further information whenever the data required by the meta-analysis were not fully reported in the article.

Meta-analysis

Case–control, cohort, and cross sectional studies with any sample size examining the association between asthma and GSTM1 null genotype, GSTT1 null genotype and GSTP1 Ile105Val polymorphism were considered eligible for this meta-analysis. For each study we collected the following data: journal name, year of publication, characteristics of the included studies (in relation to the source of cases and controls), method of ascertainment of the diagnosis (inclusion and exclusion criteria), asthma characteristics (atopic or non- atopic), demographic characteristics of the population being studied, prevalence of meaningful risk factors (smoking habits, and family history of asthma). This was done both in cases and controls. We also collected data on geographical location and ethnicity. On the basis of the information reported by the authors, we divided geographical location in urban areas or non urban areas by population estimate in accordance with demographia world urban areas: seventh annual edition [http://www.demographia.com/db-worldua.pdf].

We excluded from our analysis familiar, only-case, and occupational asthma studies. For each study frequencies of GSTP1*I105V genotypes and frequencies of GSTM1 and GSTT1 positive/null genotypes in cases and controls and crude odds ratios (ORs) were collected. A wide range of studies evaluated GSTM1 and GSTT1 as presence/absence of gene deletion, so meta-analyses of these polymorphisms were performed using a single OR (null vs. present). Results for GSTP1 Ile105Val were reported as two ORs: a dominant OR (Ile/Val + Val/Val vs. Ile/Ile) and a recessive OR (Val/Val vs. Ile/Ile + Ile/Val). Regarding GSTP1 Ile105Val polymorphisms, a sensitivity analysis was performed excluding studies where allele frequencies exhibited a significant deviation from the Hardy–Weinberg equilibrium (HWE). Statistical significance was defined as p < 0.05. The Q test and the I 2 statistic were used to investigate respectively the presence of between-study heterogeneity and the proportion of variation across studies due to heterogeneity rather than random error [21]. Possible causes of heterogeneity were investigated by subgroup analyses based on ethnicity, urbanization, and population age. For ethnicity, we considered four ethnic classes: Asian, European, Mediterranean African, and Turkish. For urbanization, the variables were defined as metropolis (>1 million inhabitants), city (500 thousand to 1 million inhabitants), and country-side (<500 thousand inhabitants). In the analyses, we used as a subgroup variable the combination of these categories, or, rather, we used a variable with four categories: 1= “Cases: Adults and Controls: Adults”, 2= “Cases: Children and Controls: Children–Adults”, 3= “Cases: Children and Controls: Adults”, 4= “Cases: Children and Controls: Children”. All statistical analyses were performed using STATA 10.1 (Stata Corp, College Station, TX, USA); the meta-analyses were performed using a fixed or, where appropriate, random effects model to estimate pooled OR. Publication bias was evaluated with the funnel plot and the Begg and Mazumdar Rank correlation test and Egger’s regression method. Then with Duvall and Tweedie’s trim and fill procedure, we tested how the effect size would shift, when the apparent bias was removed. Temporal effect was also estimated with a cumulative meta-analysis.

Results

Systemic review and meta-analysis

In Fig. 1, we reported the flow-chart that represents the inclusion and exclusion criteria of studies. From our initial research we identified 2,483 articles. Among those, after screening for titles and abstracts, we identified 95 studies; after cross-checking references we identified 20 additional papers so we examined and retrieved 115 papers for more detailed evaluation. 43 eligible articles were considered; we excluded occupational asthma reports, multiple reports of the same studies, family-based studies, other outcomes reports, editorials and reviews. We contacted the authors of 14 papers with no genotype data available and seven of those authors gave us the genotype data.

Fig. 1
figure 1

The flow chart of the studies included in the meta-analysis

Full size image

We included in the meta-analysis data collected from cohorts (n = 3), cross-sectional studies (n = 3), and case–control studies (n = 37). For the cohorts and cross-sectional studies we contacted authors to better define a control group (defined as group of individual without any respiratory disease), in order to equate to a case–control study and to collect data on GST gene polymorphism distribution. Association with asthma was evaluated in 35 studies on GSTM1, 31 on GSTT1, and 28 on GSTP1. In Table 1, other characteristics of the population based studies included in the meta-analyses are summarized, while genotype counts by disease outcome are reported in Supplemental Table 1 (Online Resource 1).

Table 1 Characteristics of the studies evaluating the effects of GST genes on asthma risk
Full size table

Meta-analyses on GSTM1 positive/null genotype

For the evaluation of the association of GSTM1 and asthma, 35 published studies were considered, a total of 6,661 affected and 17,220 non-affected individuals were included. The meta-analysis showed an increased risk of asthma associated with the GSTM1 null genotype (pooled OR = 1.12, 95 %CI = 0.99–1.26, p = 0.072; Fig. 2). This outcome, however, appears to be insignificant and large between-study heterogeneity was observed (Q = 109.17, p < 0.001; I 2 = 68.9 %). A funnel plot was performed to evaluate the publication bias of literature on GSTM1 and asthma. The shape of the funnel plot seemed asymmetrical (Supplemental Fig. 1; Online Resource 2), but the Begg and Mazumdar Rank correlation test (p = 0.551) and Egger’s regression method (bias = 1.95, t = 1.06, p = 0.299) highlighted no publication bias. Moreover, a temporal effect has been recognized in the study outcomes (Supplemental Fig. 2; Online Resource 3).

Fig. 2
figure 2

Forest plot (random-effects model) of the association between GSTM1 null genotype and asthma

Full size image

To explain the large heterogeneity in the general meta-analysis, stratification analyses were conducted. Firstly we considered ethnicity as a stratification variable. 16 studies were carried out on European populations, ten studies on Asian populations, four studies on Turkish population, and three studies evaluated the Mediterranean African population. Two studies with particular ethnic origins (a study on Iranian population [22] and a study on a mixed European/African-American population [23]) were not reported. No statistical association was observed in each ethnic group (Europe: pooled OR = 1.08, 95 %CI = 0.98–1.19; Asia: pooled OR = 1.07, 95 %CI = 0.84–1.36; Turkey: pooled OR = 1.10, 95 %CI = 0.61–1.99; Mediterranean Africa: pooled OR = 0.84, 95 %CI = 0.28-–2.52) and a high heterogeneity within each group was observed, with the exception of the European subgroup in which there was a consistency (Europe: I 2 = 24.6 %, p = 0.176; Asia: I 2 = 76.2 %, p < 0.001; Turkey: I 2 = 72.4 %, p < 0.013; Mediterranean Africa: I 2 = 88.9, p < 0.001). In the stratification analysis by urbanization, we considered 18 studies on metropolis locations, and 15 studies on country-side locations. Two studies were not reported: a study conducted in a city [24], and a study without information about the sampling location [22]. The stratification analysis by urbanization did not reveal any significant associations (metropolis: pooled OR = 1.12, 95 %CI = 0.93–1.35; country-side: pooled OR = 1.07, 95 %CI = 0.91–1.25) and highlighted the presence of a high heterogeneity within each group (metropolis: I 2 = 70.2, p < 0.001; country-side: I 2 = 66.3, p < 0.001). In the age-stratification analysis we considered 14 “Cases: Adults and Controls: Adults” studies, 11 “Cases: Children and Controls: Children” studies, four “Cases: Children and Controls: Adults” studies, and two “Cases: Children and Controls: Children–Adults” studies; we did not report four studies without age information. The OR estimated for each subgroup was not statistically significant (“Cases: Adults and Controls: Adults”: pooled OR = 1.08, 95 %CI = 0.91–1.27; “Cases: Children and Controls: Children”: pooled OR = 1.07, 95 %CI = 0.86–1.32; “Cases: Children and Controls: Adults”: pooled OR = 1.10, 95 %CI = 0.71–1.71; “Cases: Children and Controls: Children–Adults”: pooled OR = 1.45, 95 %CI = 0.89–2.37) and high heterogeneity was observed in each subgroup, with the exception of “Cases: Children and Controls: Children–Adults” (“Cases: Adults and Controls: Adults”: I 2 = 62.4 %, p = 0.001; “Cases: Children and Controls: Children”: I 2 = 66.9 %, p = 0.001; “Cases: Children and Controls: Adults”: I 2 = 73.6 %, p = 0.010; “Cases: Children and Controls: Children–Adults”: I 2 = 0.0 %, p = 0.368).

Meta-analyses on GSTT1 positive/null genotype

A total of 31 published studies on the association of GSTT1 and asthma was taken into consideration, including a total of 5,454 cases and 14,513 controls. The meta-analysis for GSTT1 shows an OR of 1.33 (95 % CI 1.10–1.60; p = 0.003) with an increased risk of asthma associated with the GSTT1 null genotype (Fig. 3). This estimate appears to be significant, but a large between-study heterogeneity was observed (Q = 118.58, p < 0.001; I 2 = 74.7 %). The funnel plot analysis revealed publication bias of literature on GSTT1 and asthma (Supplemental Fig. 3; Online Resource 4). The Begg and Mazumdar Rank correlation test (p = 0.003) and Egger’s regression method (bias = 8.76, t = 3.27 p = 0.003) consistently confirmed the visual inspection of the funnel plot. After the trim and fill procedure, there was a relevant change in overall summary estimate (OR); the procedure suggested that almost 8 studies had to be included to convert the combined p value to a non-significant value (p = 0.946) and this brought the pooled OR of 1.33–1.01. Conversely, no temporal effect seemed to be present in the study outcomes (Supplemental Fig. 4; Online Resource 5).

Fig. 3
figure 3

Forest plot (random-effects model) of the association between GSTT1 null genotype and asthma

Full size image

In the stratification analysis by ethnicity, 18 studies were carried out on European populations, four studies on Asian populations, four studies on Turkish populations, and three studies on Mediterranean African populations. The studies conducted by Saadat et al. [22] on the Iranian population and by Passos-Lima et al. [23] on European/African-American population were not reported. No significant association was observed in the ethnic groups (Europe: pooled OR = 1.18, 95 %CI = 0.98–1.43; Asia: pooled OR = 2.12, 95 %CI = 0.62–7.26; Turkey: pooled OR = 1.41, 95 %CI = 0.93–2.16; Mediterranean Africa: pooled OR = 1.73, 95 %CI = 0.60–4.96) and the studies were characterized by high heterogeneity within the subgroups, with the exception of the Turkish group (Europe: I 2 = 66.0 %, p < 0.001; Asia: I 2 = 92.5 %, p < 0.001; Turkey: I 2 = 37.2 %, p = 0.189; Mediterranean Africa: I 2 = 86.3 %, p = 0.001).

For the stratification analysis by urbanization, we considered the same studies of GSTM1. The analysis revealed significant association with high heterogeneity for “metropolis” (pooled OR = 1.65, 95 %CI = 1.12–2.42; I 2 = 85.0 %, p < 0.001) and no association with low heterogeneity for “country-side” (pooled OR = 1.05, 95 %CI = 0.91–1.21; I 2 = 27.2 %, p = 0.156).

Regarding the population age-stratification analysis, we considered 13 “Cases: Adults and Controls: Adults” studies, nine “Cases: Children and Controls: Children” studies, three “Cases: Children and Controls: Adults” studies, and two “Cases: Children and Controls: Children–Adults” studies; we did not report four studies without age information. Significant associations were observed for “Cases: Children and Controls: Children–Adults” and “Cases: Children and Controls: Adults” subgroups (“Cases: Adults and Controls: Adults”: pooled OR = 1.24, 95 %CI = 0.99–1.54; “Cases: Children and Controls: Children”: pooled OR = 1.30, 95 %CI = 0.90–1.86; “Cases: Children and Controls: Adults”: pooled OR = 0.76, 95 %CI = 0.52–1.12; “Cases: Children and Controls: Children–Adults”: pooled OR = 2.80, 95 %CI = 1.09–7.21), but only in the “Cases: Children and Controls: Adults” subgroup was no significant heterogeneity observed (“Cases: Adults and Controls: Adults”: I 2 = 64.2 %, p = 0.001; “Cases: Children and Controls: Children”: I 2 = 67.0 %, p = 0.683; “Cases: Children and Controls: Adults”: I 2 = 74.7 %, p < 0.001; “Cases: Children and Controls: Children–Adults”: I 2 = 68.6 %, p = 0.074).

Meta-analyses on GSTP1 Ile105Val polymorphism

A total of 28 published studies with 5,559 affected and 9,199 non-affected individuals was available for the meta-analysis of GSTP1*Ile105Val polymorphism. For this variant, two genetic models were considered: dominant and recessive. Regarding the general meta-analysis, no significant association and high heterogeneity were found both for the dominant (pooled OR: 0.93, 95 %CI: 0.82–1.06; Q = 59.44, p < 0.001, I 2 = 54.6 %) and for the recessive (pooled OR = 0.92, 95 %CI = 0.73–1.15, Q = 60.90, p < 0.001, I 2 = 57.3 %) genetic models (Figs. 4, 5). The funnel plots did not reveal publication bias in literature on GSTP1 and asthma in both considered genetic models (Supplemental Figs. 5 and 6; Online Resources 6, 7), and no temporal effects were observed (Supplemental Fig. 7 and 8; Online Resources 8, 9). The Begg and Mazumdar Rank correlation test (p dominant  = 0.213, p recessive = 0.868) and Egger’s regression model (dominant: bias = 0.34, t = −1.90, p = 0.07; recessive: bias = 0.04, t = −0.29, p = 0.777) also confirmed this assessment.

Fig. 4
figure 4

Forest plot (random-effects model) of the association between GSTP1*I105V polymorphism (dominant genetic model) and asthma

Full size image
Fig. 5
figure 5

Forest plot (random-effects model) of the association between GSTP1*I105V polymorphism (recessive genetic model) and asthma

Full size image

In the stratification analysis by ethnicity, 13 studies were carried out on European populations, five studies on Asian populations, four studies on Turkish population, and four studies on the Mediterranean African populations. The single study conducted by Munoz et al. [19] on Mexican population was not reported. For the dominant model, the subgroup analysis by ethnicity highlighted no significant association and high heterogeneity for each subgroup, with the exception of the Asian and Turkish subgroups where the non-significant association were linked to low heterogeneity across studies pooled (Europe: pooled OR = 0.96, 95 %CI = 0.82–1.14, I 2 = 52.3 %, p = 0.014; Asia: pooled OR = 0.93, 95 %CI = 0.78–1.11, I 2 = 10.4 %, p = 0.349; Turkey: pooled OR = 1.11, 95 %CI = 0.77–1.59, I 2 = 55.7 %, p = 0.079; Mediterranean Africa: pooled OR = 0.59, 95 %CI = 0.27–1.29, I 2 = 82.4 %, p = 0.001). For the recessive model, a similar outcome was observed (Europe: pooled OR = 1.28, 95 %CI = 0.88–1.20, I 2 = 0.0 %, p = 0.457; Asia: pooled OR = 1.12, 95 %CI = 0.69–1.82, I 2 = 0.0 %, p = 0.578; Turkey: pooled OR = 0.89, 95 %CI = 0.32–2.44, I 2 = 83.4 %, p < 0.001; Mediterranean Africa: pooled OR = 0.52, 95 %CI = 0.17–1.61, I 2 = 83.8 %, p < 0.001). For the urbanization stratification, only one study was conducted in a city [25] and, thereby, it was not reported. 20 studies were carried out on metropolis locations and 15 studies were carried out on country-side locations. For the dominant model, the subgroup analysis by urbanization highlighted no associations and high heterogeneity within each group (metropolis: pooled OR = 0.84, 95 %CI = 0.69–1.02, I 2 = 52.7, p = 0.009; country-side: pooled OR = 1.05, 95 %CI = 0.88–1.25, I 2 = 55.2, p = 0.011). The same results were observed in urbanization stratification for the recessive genetic model (metropolis: pooled OR = 0.81, 95 %CI = 0.52–1.24, I 2 = 59.0, p = 0.003; country-side: pooled OR = 1.03, 95 %CI = 0.77–1.38, I 2 = 56.4, p = 0.003). In the stratification analysis by population age, we considered 12 “Cases: Adults and Controls: Adults” studies and 13 “Cases: Children and Cases: Children” studies. three studies were not reported because no age information was available. This stratification analysis showed no association and high heterogeneity within each group for both dominant (“Cases: Adults and Controls: Adults”: pooled OR = 0.98, 95 %CI = 0.84–1.14, I 2 = 63.3 %;p = 0.001; “Cases: Children and Controls: Children”: pooled OR = 0.92, 95 %CI = 0.76–1.13, I 2 = 41.9 %; p = 0.062) and recessive (“Cases: Adults and Controls: Adults”: pooled OR = 0.87, 95 %CI = 0.61–1.24, I 2 = 55.8 %; p = 0.009; “Cases: Children and Controls: Children”: pooled OR = 0.92, 95 %CI = 0.65–1.30, I 2 = 41.9 %; p = 0.001) models.

Discussion

GSTs are involved in many detoxification processes, and it has been hypothesized that genetic alterations of GST enzymes may affect the ability of the airways to handle toxic substances and may influence airway inflammation [12].

A number of studies has investigated the role of functional GST polymorphisms in asthma susceptibility both in children and adults, with contrasting results. To the best of our knowledge, two previous meta-analyses have attempted to clarify the association between GST genes and asthma, but these studies showed conflicting outcomes [2, 20]. Saadat and Ansari-Lari [20] stated that GSTM1 and GSTT1 null genotypes are significantly associated with asthma phenotypes, especially for non-smokers and adults. Conversely, Minelli et al. [2] reported no significant association between GSTM1, GSTP1, and GSTT1 and asthma, with high heterogeneity among the studies, even though they were stratified by asthma diagnosis.

The aim of our study was to perform a meta-analysis that included independent genetic association studies on GSTM1, GSTP1, and GSTT1, evaluating also the effect of potential confounding variables (i.e. ethnicity, population age, and urbanization).

Our results highlighted that no significant associations with asthma susceptibility were observed for GSTM1 and GSTP1 gene polymorphisms whereas significant outcomes were detected for GSTT1 positive/null genotype. However, high between-study heterogeneity was identified in all the general analyses. This heterogeneity among the genetic association studies does not ensure that these results are reliable, though [26]. Moreover, the asymmetrical distribution of the studies on GSTT1 genes in the funnel plot suggests the presence of a publication bias. In order to verify whether the heterogeneity among the studies was caused by the confounding effects of ethnicity, population age, and urbanization, we performed different stratification analyses. For GSTM1 and GSTP1 in which no associations with high heterogeneity were observed in the general analysis, the stratification revealed, in some cases, no associations with low heterogeneity, leaving most of the stratification analyses with high heterogeneity within the subgroups. Regarding GSTT1, in which a positive association was observed, a significant association with low heterogeneity was observed only in the “children versus adults” studies, whereas, in most of the other subgroups the non-significant associations were related to high heterogeneity. Therefore, the uncertainty about both the significant and non-significant results of our meta-analysis was confirmed, even after stratification analyses. Furthermore, for GSTT1 we obtained a non-significant association correcting the effect of publication bias through the trim and fill procedure. This situation is probably due to disease pathogenesis. Indeed, asthma expression is a complex process within genetics, and environment strongly impact it [27].

Our stratification analysis has tried to address confounding effects of environmental (urbanization), demographic (population age) and genetic (ethnicity) factors. Indeed, as previously reported by other authors, the inhalation of hazardous environmental factors may increase the risk for disease development or worsening of symptoms [28]. Given the multiplicity of exposures, there are many possible effects from a single pollutant, as well as from its interactions [29]. Furthermore, genetic predisposition may alter the ability of the airways to protect itself against these inhaled toxic substances from the environment [27] and genetic differences in the population structure of antioxidant enzymes may influence the genetic association between GSTs and asthma susceptibility [30]. Although our approach seems to explain a portion of this complex picture in a few cases, further studies on interactions of GST genes with environmental oxidative exposures and with other antioxidant genes are required to explain the role of GST enzymes in asthma pathogenesis.