Association between MGMT Promoter Methylation and Non-Small Cell Lung Cancer: A Meta-Analysis

Background O6-methylguanine-DNA methyltransferase (MGMT) is one of most important DNA repair enzyme against common carcinogens such as alkylate and tobacco. Aberrant promoter methylation of the gene is frequently observed in non-small cell lung cancer (NSCLC). However, the importance of epigenetic inactivation of the gene in NSCLC published in the literature showed inconsistence. We quantified the association between MGMT promoter methylation and NSCLC using a meta-analysis method. Methods We systematically reviewed studies of MGMT promoter methylation and NSCLC in PubMed, EMBASE, Ovid, ISI Web of Science, Elsevier and CNKI databases and quantified the association between MGMT promoter methylation and NSCLC using meta-analysis method. Odds ratio (OR) and corresponding 95% confidence interval (CI) were calculated to evaluate the strength of association. Potential sources of heterogeneity were assessed by subgroup analysis and meta-regression. Results A total of 18 studies from 2001 to 2011, with 1, 160 tumor tissues and 970 controls, were involved in the meta-analysis. The frequencies of MGMT promote methylation ranged from 1.5% to 70.0% (median, 26.1%) in NSCLC tissue and 0.0% to 55.0% (median, 2.4%) in non-cancerous control, respectively. The summary of OR was 4.43 (95% CI: 2.85, 6.89) in the random-effects model. With stratification by potential source of heterogeneity, the OR was 20.45 (95% CI: 5.83, 71.73) in heterogeneous control subgroup, while it was 4.16 (95% CI: 3.02, 5.72) in the autologous control subgroup. The OR was 5.31 (95% CI: 3.00, 9.41) in MSP subgroup and 3.06 (95% CI: 1.75, 5.33) in Q-MSP subgroup. Conclusion This meta-analysis identified a strong association between methylation of MGMT gene and NSCLC. Prospective studies should be required to confirm the results in the future.


Introduction
Lung cancer is the leading cause of global cancer deaths in recent decades [1]. Human lung cancer contains two histological types, small-cell lung cancer (SCLC) and non-small lung cancer (NSCLC). The latter comprises the majority of lung cancer and has an increasing incidence and mortality in the last two decades worldwide. DNA methylation is an epigenetic modification of the genome and methylation associated with silencing can affect genes expression in cellular pathways [2].
The epigenetic alterations are early and frequent events occurred in carcinogenesis. O 6 -methylguanine-DNA methyltransferase (MGMT) is a DNA damage reversal protein against DNA adduct formation of carcinogens [3,4]. It can protect cells from the carcinogenic effects of alkylating agents by removing adducts from the O 6 position of guanine [5]. Therefore, the repair capacity of the MGMT protein helps decrease the probability so that the damaged guanine becomes a mutagenic site. Methylation of MGMT gene promoter has been associated with loss or decrease of MGMT expression in tumor tissues of various cancers, including lung tumors [6][7][8].
Taken together, methylation of MGMT gene has been considered as potentially useful candidate biomarker for early detection of lung cancer. Many studies had also shown that methylation of the gene can be found in clinical samples, such as tissues, serum and bronchoalveolar lavage fluid (BALF) of NSCLC [9][10][11].
The purpose of this study was to understand the difference of the prevalence of aberrant promoter methylation of MGMT in NSCLC tissue from control. We conducted a meta-analysis using available data in the literature on the basis of MGMT promoter methylation and NSCLC to better identify the association between MGMT promoter methylation and NSCLC.

Selection criteria and study search
We searched the electronic databases online including Pubmed, EMBASE, Ovid, ISI Web of Science, Elsevier and CNKI database, using the search terms "MGMT", "NSCLC" and "methylation". The following search strategy was performed in Pubmed "NSCLC" (MESH), "methylation" and "MGMT or O 6methylguanine-DNA methyl-transferase" to collect eligible articles. The search was limited to articles published in English and Chinese. Similar searches were performed in other databases. The search was updated until June 1, 2013.
Studies selected had to meet the following criteria: (a) The study was about MGMT methylation and NSCLC. (b) The authors offered a measure of the association either as an effect estimate with 95% CI and OR, or sufficient data in the original article to calculate it. (c) MGTM methylation status was examined using methylation-specific PCR (MSP) or real-time quantitative MSP (Q-MSP). (d) Specimens of NSCLC were surgically respected primary tumor sample and the styles of control were composed of plasma or non-cancerous lung tissues (NLT) including autologous control and heterogeneous control.
Studies were excluded according the following criteria: (a) The study did not provide control information. (b) The articles were repetitively reported, (c) The articles researched chromate lung cancer, brain metastases of lung cancers and malignant pleural mesothelioma [7,12,13]. Firstly, we evaluated whether a study met the inclusion criteria by title and abstract of initial searching articles. Then all the potentially relevant articles were evaluated by accessing full-text paper. The selection procedure of studies was illustrated in statement flow chart ( Figure 1).

Abstraction of Data and Quality Assessment
All the data were independently abstracted by three authors (Changmei Gu, Cheng Lu, and Hao Shi) with the use of standardized data-extraction forms. For each study, the following characteristics were extracted: first author's name, year of publication, the study country, ethnicity, age, sample size, type of control, the method for methylating detection, and status of MGMT methylation were collected using standard data extraction forms. If there were disagreements, we discussed with Meixia Lu and Jiachun Lu repeatedly, until a consensus. To ensure the transparent and complete reporting of this meta-analysis, we designed and checked selected studies according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [14] statement.

Statistical Methods
The ORs and 95% CIs were abstracted or calculated to evaluate the strength of the association between MGMT promoter methylation and NSCLC risk. Summary of OR was acquired from all studies or calculated with the data in selected studies.
We, then, investigated between-study heterogeneity by using Cochran's Q test with a significance level of P value less than 0.1 and heterogeneity was also examined using I 2 statistic [15], which is a quantitative measure of inconsistency across studies. If I 2 > 50% or P < 0.1 is considered as a measure of severe heterogeneity, then the random-effects model was used to calculate summary OR according to the Der-Simonian Laird method; otherwise, the fixed-effects model (Matel-Haenszel method) was applied [16]. τ 2 was used to determine how much heterogeneity was explained by subgroup differences. The meta-regression was performed to explore the source of heterogeneity based on ethnicity (Asian and Caucasian), publication year, style of control (autogenous or heterogeneous), method (MSP or Q-MSP), and sample size. Subgroup analyses were performed according to ethnicity (Asian and Caucasian), control style (autogenous or heterogeneous) and method of methylation detection (MSP or Q-MSP) in consideration of the source of heterogeneity. Sensitivity analyses were also performed to assess the contributions of each study on the overall result by omitting one study at a time. The funnel plot asymmetry was used to evaluate the evidence of publication bias. Peter's test was applied to quantitatively evaluate the evidence for publication bias [17]. Fail-safe number (N fs ) adopted by Rosenthal [28] is considered to be a useful indicator of leaning [18]. The metatrim method was used to re-estimate the effect size when there was possible bias. All P values are two-tailed with a significant level at 0.05. In forest plot displayed in Figure 2, the size of the box for each study was inversely proportional to the variance of the log relative risk, and the horizontal lines represent 95% CI.

Study Characteristics
The electronic search found that 145 potentially relevant studies were initially identified. These studies were further screened based on inclusion and exclusion criteria. A total of 18 studies (2001-2011) were included in the analysis ( Figure  1). In these studies, 6 studies were conducted in Asia (1 in Japan and 5 in China) and the remaining 11 were in the USA and 1 in the Netherlands.
Among the 18 retrieved studies, 14 studies used methylation-specific polymerase chain reaction (MSP) and 4 studies used real-time quantitative MSP (Q-MSP) to explore MGMT methylation in NSCLC tissue and control. There were two control styles, including autogenous control (the tissues from the patients themselves) and heterogeneous control (i.e Plasma, tissue, and bronchoalveolar lavage fluid from other individuals). The main characteristics of these studies were presented in Table 1.
There are 1, 160 NSCLC tumor tissues and 970 controls. The frequencies of MGMT promote methylation ranged from 1.5% to 70.0% (median 26.1%) in NSCLC tissue and 0.0% to 55.0% (median 2.4%) in non-cancerous control, respectively, which indicated the methylation frequency in cancer tissue was much higher than that in the control group. Under the randomeffects model, the pooled OR of MGMT methylation in NSCLC tissue was 4.43 (95% CI: 2.85, 6.89), in comparison with control group (Figure 2).

Meta-regression and subgroup analyses
Considering the existence of heterogeneity in the metaanalysis (I 2 =32.0%, P= 0. 095), the meta-regression was performed for finding the source of heterogeneity. The restricted maximum likelihood modification was used to estimate between study variances. As the restriction to access raw data, we assumed the source of heterogeneity may appear from the year of publication, ethnicity, type of control, detection method and sample size. The results showed that the sources of the heterogeneity were control style (P = 0.073) and detection method (P = 0.094). Other factors such as sample size, year of publication, and ethnicity could not explain the heterogeneity (Table 2).   With these observations, we performed subgroup analyses based on control style and detection method. The OR in the heterogeneous control subgroup was 20.45 (95% CI: 5.83, 71.73; fixed-effects model), while it was 4.16 (95% CI: 3.02,5.72; fixed-effects model) in the autologous control subgroup. I 2 changed to 16.6% and 34.0% in those subgroups, respectively, compared with 32% of the total. Stratification by control style showed that the OR in the heterogeneous control subgroup was higher than that in the autologous control subgroup. After stratification by detection method, the OR was 5.31 (95% CI: 3.00, 9.41; random-effects model) in MSP subgroup and 3.06 (95% CI: 1.75, 5.33; fixed-effects model) in Q-MSP subgroup (Table 3).

Sensitivity analysis
To determine the effects of omitting a single study at a time on the overall effect, the sensitivity analysis was performed. Omission of a single study changed the overall OR from 3.85 (95% CI: 2.66, 5.56) to 4.87 (95% CI: 2.96, 8.02) using the random-effects model, which demonstrated that no single sensitivity analysis existed (Figure 3).

Publication bias
A funnel plot of methylation status of tumor versus control showed that there were two studies exceed the 95% confidence limits as shown in Figure 4. However, no publication bias was detected by using Peter's test (P = 0.624).
We also used a fail-safe number (N fs ) to assess the efficacy of meta-analysis (Z = 45.63, N fs0. 05 = 756.12, N fs0. 01 = 365.52), which indicated a slight publication bias in the meta-analysis.

Discussion
Aberrant methylation of promoter regions in DNA repair genes is a key event in the formation and progression of cancer. It was concluded that genes methylation was potentially a new generation of cancer biomarker [34]. Epigenetic change of CpG islands in the gene promoter region is an important reason for gene dysfunction [35], which can lead to the transcription of the gene down or stops [36].
MGMT promoter methylation is very common in the primary NSCLC, which has been reported in some studies. However, the results about the association between the status of MGMT methylation and NSCLC were inconsistent. The OR values fluctuated from 0.65 [28] to 114.98 [24], so we performed this meta-analysis to identify the association between MGMT promoter methylation and NSCLC.
A total of 18 studies including 1, 160 tumor tissues and 970 controls were involved in the meta-analysis. The frequencies of MGMT promote methylation ranged from 1.5% to 70.0% (median; 26.1%) in NSCLC tissue and 0.0% to 55.0% (median; 2.4%) in non-cancerous control, respectively. The findings indicated that the methylation frequency in cancer tissue was much higher than that in the control group. MGMT methylation had an increased risk in tumor tissue (OR = 4.43; 95% CI: 2.85, 6.89) in comparison with non-cancerous samples including plasma, tissue, and bronchoalveolar lavage fluid. This finding was consistent with other studies [10,25]. Furthermore, subgroup analysis of control style showed that an OR was 20.45 (95% CI: 5.83, 71.73; fixed-effects model) in the heterogeneous control subgroup verse 4.16 (95% CI: 3.02, 5.72; random-effects model) in the autologous tissues subgroup, indicating that the OR of methylation in heterogeneous control group was higher than that in the autologous control group. It suggested that MGMT gene promoter methylation was a frequent event in NSCLC, but it rarely happened in the non-tumor group. Stratification analysis by different methods of methylation detection showed that the OR for the MSP subgroup was higher than that for Q-MSP subgroup. Sensitivity analysis was performed to determine the effects when omitting a single study at a time on the overall effect. The analytic results demonstrated that no single study could affect the summarized OR. In addition, the shape of the funnel plot did not show any evidence of funnel plot asymmetry and no publication bias was detected.
The current meta-analysis has some limitations. Firstly, selection bias is inevitable due to the strategy restricted to articles published in English or Chinese. Secondly, although the total sample sizes in the meta-analysis were not more than 1, 000, the results maybe have still no vigorous power. Thirdly, we did not study the methylation status in histological subtypes, smoking and different clinical stages because of limitation of insufficient raw materials.
In conclusion, MGMT gene is an important DNA repair gene on maintaining the integrity of the genome. Aberrant MGMT promoter methylation may be associated with the occurrence and development of NSCLC. This meta-analysis identified a strong association between MGMT promoter methylation and NSCLC. Prospective studies should be required to confirm the results in the future.