PAI-1 -675 4G/5G Polymorphism in Association with Diabetes and Diabetic Complications Susceptibility: a Meta-Analysis Study

A meta-analysis was performed to assess the association between the PAI-1 -675 4G/5G polymorphism and susceptibility to diabetes mellitus (DM), diabetic nephropathy (DN), diabetic retinopathy (DR) and diabetic coronary artery disease (CAD). A literature-based search was conducted to identify all relevant studies. The fixed or random effect pooled measure was calculated mainly at the allele level to determine heterogeneity bias among studies. Further stratified analyses and sensitivity analyses were also performed. Publication bias was examined by the modified Begg’s and Egger’s test. Twenty published articles with twenty-seven outcomes were included in the meta-analysis: 6 studies with a total of 1,333 cases and 3,011 controls were analyzed for the PAI-1 -675 4G/5G polymorphism with diabetes risk, 7 studies with 1,060 cases and 1,139 controls for DN risk, 10 studies with 1,327 cases and 1,557 controls for DR and 4 studies with 610 cases and 1,042 controls for diabetic CAD risk respectively. Using allelic comparison (4G vs. 5G), the PAI-1 -675 4G/5G polymorphism was observed to have no significant association with diabetes (REM OR 1.07, 95% CI 0.96, 1.20), DN (REM OR 1.10, 95% CI 0.98, 1.25), DR (REM OR 1.09, 95% CI 0.97, 1.22) or diabetic CAD risk (REM OR 1.07, 95% CI 0.81, 1.42), and similar results were obtained in the dominant, recessive and co-dominant models. Our meta-analyses suggest that the PAI-1 -675 4G/5G polymorphism might not be a risk factor for DM, DN, DR or diabetic CAD risk in the populations investigated. This conclusion warrants confirmation by further studies.


Introduction
The plasminogen activator inhibitor-1 (PAI-1) belongs to the serine protease inhibitor superfamily, and plays a key role in the regulation of extracellular matrix degradation [1]. Studies have indicated that the increase of PAI-1 level was related to the incidence of diabetes mellitus (DM) and its complications, such as an increased risk of diabetic nephropathy (DN), diabetic retinopathy (DR) and diabetic coronary artery disease (CAD), etc [2][3][4][5].
Numerous single nucleotide polymorphisms (SNPs) have been observed in the PAI-1 gene. Some increase the t ½ of PAI-1 while others inactivate it or slow down its secretion into the plasma. It has been found that the most-studied 4G/5G polymorphism (rs1799889), characterized by a single guanosine nucleotide insertion/deletion variation at -675 bp of the PAI-1 promoter, increases PAI-1 concentration or its activity in the plasma of humans without changing its t ½ [6]. This polymorphism is a cause of high plasma PAI-1 level in 4G/4G allele carriers suggesting that the PAI-1 4G/5G polymorphism is a genetic risk factor for diabetes [7]. Recently many studies reported the association between the 4G/5G polymorphism of PAI-1 gene and the risk of diabetes and diabetic complications [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27], mainly focusing on DN, DR and diabetic CAD. Despite strong functional evidence for the relevance of several studies, the results for the association with diabetes and its complications showed significant betweenstudy variations and were inconclusive.
Considering a single study may lack the power to provide a reliable conclusion, we performed a meta-analysis on these eligible case-control studies, to investigate the precise relationship between the PAI-1 -675 4G/5G polymorphism and susceptibility to DM, DN, DR and diabetic CAD, which would have a much greater possibility of reaching reasonably strong conclusions.

Methods
This meta-analysis is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) statement [28].

Search strategy
The comprehensive literature searches were carried out independently by two investigators (Kuanfeng Xu and Xiaoyun Liu) using the electronic data-bases PubMed, Embase and ScienceDirect without language restrictions, and updated on July 11, 2013. The PAI-1 -675 4G/5G polymorphism was investigated by combinations of the following search terms: 'plasminogen activator inhibitor-1 or PAI-1', 'polymorphism, variant or mutation' and 'diabetes or diabetic complications'. We used the PubMed option 'Related Articles' for each study to retrieve additional potentially relevant articles, and also handsearched the included articles to identify any other relevant citations. No restriction was set on the source of control participants (general population, clinic or hospital).

Inclusion and exclusion criteria
The inclusion and exclusion criteria were as follows: (1) each case-control study were published as an original study designed to evaluate the association; (2) numbers in case and control groups were reported for each allele or genotype; (3) case-control studies had sufficient published data to estimate an odds ratio (OR) with 95% confidence intervals (CI) or provide raw data that allowed us to calculate them; (4) only the most recent or complete study was used if the data were duplicated and had been published in several publications; (5) studies were excluded if the genotype distribution of the controls deviated from Hardy-Weinberg equilibrium (HWE); (6) the following were excluded: animal studies, review articles, abstracts, editorials, reports with incomplete data, studies based on pedigree data or prospective studies, etc.

Data extraction
Data were independently extracted by two investigators who discussed disagreements and reached a consensus on all of the items. Information extracted from each study was considered as follows: name of first author, year of publication, ethnic origin of the studied population, available number of participants in case and control groups, genotype and allele frequency by case/control status, and OR (95% CI). Not all papers reported the necessary statistics directly, so in some instances we transformed and estimated an OR from the reported data [29]. We did not define any minimum number of patients for a study to be included in our meta-analysis. In addition, all participants of the included studies provided informed consent and the studies were approved by the ethics committees of the participating institutions.

Statistical analysis
The association of the PAI-1 -675 4G/5G polymorphism with diabetes and diabetic complications was estimated by calculating the pooled OR and 95% CI in the allelic and genotypic (dominant, recessive and co-dominant model) comparisions respectively. The significance of the OR was determined by the Z test (P <0.05 was considered statistically significant). HWE of the genotype distribution of controls was tested by a goodness-of-fit χ 2 analysis. The distribution was considered to be deviated from HWE at P < 0.05.
We employed the DerSimonian and Laird random effect model (REM) to bring the individual effect size estimates together, and quantified between-study heterogeneity by inconsistency index (I 2 ) statistic. The I 2 statistic is defined as the percentage of the observed between-study variability that is due to heterogeneity rather than chance, with high values suggesting more possible existence of heterogeneity. Potential heterogeneity between results of individual studies or in subgroups respectively by ethnicity, study design, source of controls, genotyping method, diagnostic criterion, and sample size was explored using χ 2 test. Sensitivity analysis was conducted to evaluate the key studies with a substantial impact on between-study heterogeneity. Influence analysis was performed to assess the stability of the results, with a single study in the meta-analysis being deleted each time to reflect the influence of the individual data set on the pooled OR.
Publication bias was assessed by the modified Begg's and Egger's test [30]. The significance of the intercept was determined by the t test suggested by Egger, with p <0.10 considered representative of statistically significant publication bias. All statistical analyses were conducted using RevMan 5.0 (The Nordic Cochrane Centre, The Cochrane Collaboration) & STATA 11.0 (Stata, College Station, TX, USA). All tests were two-sided.

Quantitative syntheses
Our meta-analysis revealed no association between the PAI-1 -675 4G/5G polymorphism and diabetes risk, either by allelic comparision (     Results of pooled analyses are summarised in detail in Table 4 & Figure 3. In addition, our meta-analysis showed no significant association between the PAI-

Sensitivity analyses and influence analyses
As shown in Table 2-5, low or high heterogeneity in most of the inherited models was observed among studies in the overall population except for the association with diabetes risk. To identify the studies with the greatest impact on the overall between-study heterogeneity, sensitivity analyses and stratified analyses were conducted. On the association with DN risk, the heterogeneity in the recessive model was not significantly decreased after either sensitivity analysis or stratified analysis by ethnicity. On the association with DR risk, low heterogeneity existed in all genetic models. Sensitivity analysis indicated that the heterogeneity was also effectively decreased in all genetic models when excluded the study from Nagi DK et al [9]. Moreover, when the data were stratified by ethnicity, the heterogeneity in all genetic models was significantly decreased or eliminated in populations of Asian and European descent. On the association with diabetic CAD risk, the study from Mansfield MW et al [8] were mainly responsible for the observed heterogeneity, but sensitivity analysis suggested that the significant heterogeneity still existed in all genetic models when excluded this study.  To assess the degree to which each individual study affected the overall OR estimates, influence analysis was conducted by repeating the meta-analysis sequentially excluding one study at a time. No single study excessively influenced the analyses (data not shown).

Publication bias
Funnel plots and Egger's test were performed to assess the publication bias of the literature. As expected, symmetrical funnel plots were obtained in diabetes and its complications tested in all genetic models. And Egger's test further confirmed no publication bias for any of the polymorphisms examined, indicating that our results are statistically robust, as shown in Table S1.

Discussion
Elevated concentrations of PAI-1 have been observed consistently in blood from patients with T2D or insulin resistance [2,3]. The PAI-1 -675 4G/5G polymorphism, a single guanosine insertion/deletion, has been identified to contain an additional binding site for a DNA binding protein that may play a pivotal role as a repressor during transcription and exert the greatest impact on plasma PAI-1 concentration [6,31]. The information suggested that this polymorphism might be a genetic risk factor for diabetes. However, our meta-analysis results indicated that the PAI-1 -675 4G/5G polymorphism had no association with T2D in all genetic models. Further studies also indicated that the elevation of PAI-1 concentration correlates with complications of diabetes, including DN, DR and diabetic CAD risk [32][33][34]. Also studies on the PAI-1 -675 4G/5G polymorphism with the risk of diabetic complications were reported, which showed significant between-study variations and were inconclusive. Unfortunately, in our results of meta-analysis this polymorphism had no association with risk of diabetic complications in overall populations in all genetic models, given that the underlying studies were carried out in different populations, we also performed by random effects model, in which the results indicated no association in either European or Asian descendent.
Heterogeneity is potentially a significant problem when interpreting the results of any meta-analysis of genetic association studies [35]. To determine the amount of heterogeneity that existed among these variants, we did an χ 2based Q test. Our meta-analysis showed no significant between-study heterogeneity except for diabetic CAD risk in all genetic models and DN risk in the recessive model. Many of the variables that varied between the various studies might be responsible for this observed heterogeneity, including the source of the controls, sex bias, ethnicity, etc. Initial inspection of the data did not immediately identify any likely candidate variable or study that was significantly impacting on our overall results. Then we conducted sensitivity analyses by repeating the meta-analysis with one study excluded at each time [36]. The results showed that none of the individual study dramatically influenced the heterogeneity or pooled ORs.
The results of the present meta-analysis should also be interpreted within the context of its limitations. First, other potential environment-gene interactions or gene-gene interactions may well be contributors to the observed diseaseeffect unconformity except for ethnicity, but we had insufficient data to perform an evaluation of such interactions. Second, our meta-analysis is based on unadjusted estimates because of a lack of original data. Third, though the Egger's test gave no publication bias, sample size is still a limitation of our metaanalysis [37], especially on the association with diabetic CAD risk.
In conclusion, our results indicate that the PAI-1 -675 4G/5G polymorphism might not be a risk factor for DM, DN, DR or diabetic CAD risk in the populations investigated. Maybe this association is not robust and could be due to chance, and additional larger studies that allow stratification for other genegene and gene-environment should also be conducted in future analyses.