Heme oxygenase-1 gene promoter polymorphisms are associated with coronary heart disease and restenosis after percutaneous coronary intervention: a meta-analysis

Numerous published studies have suggested that there is association between heme oxygenase-1 (HO-1) gene polymorphisms and coronary heart disease (CHD) or restenosis (RS) after percutaneous coronary intervention (PCI). This study aimed to clarify this association using a meta-analysis method. We used a systematic search for studies on the association of HO-1gene polymorphisms with CHD or RS in PubMed, Web of Science, the Cochrane Library, Wanfang Data and CNKI (China National Knowledge Infrastructure). We used Stata 12.0 software to perform the meta-analyses. Twenty-three studies, involving 12,130 patients with CHD or RS and 14,181 controls, were included. A statistically significant association between the HO-1(GT)n repeat length polymorphism and CHD was observed under allelic (odds ratio (OR) = 0.929, 95% confidence interval (CI) = 0.881-0.978, p= 0.005), recessive (OR = 0.858, 95%CI = 0.780-0.945, p= 0.002), and co-dominant (OR = 0.843, 95%CI = 0.754-0.942, p= 0.003) models. Moreover, we also found a statistically significant association between the HO-1(GT)n repeat length polymorphism and RS under allelic (OR = 0.718, 95%CI = 0.541-0.953, p= 0.022) and co-dominant (OR = 0.522, 95%CI = 0.306-0.889, p=0.017) models. We found a significant association of the HO-1T(−413)A single-nucleotide polymorphism (SNP) with CHD under allelic (OR = 0.915, 95%CI = 0.842-0.995, p= 0.038), recessive (OR = 0.869, 95%CI = 0.760-0.994, p= 0.041), and co-dominant (OR = 0.792, 95%CI = 0.663-0.946, p=0.010) models. Our study indicates that both the HO-1(GT)n repeat length polymorphism and the T(−413)A SNP are associated with decreased risk of CHD. The (GT)n repeat length polymorphism was associated with RS following PCI.


INTRODUCTION
Coronary heart disease (CHD) is a multifactorial disorder resulting from the interaction between environmental and genetic factors [1]. Many genes that associate with CHD have been identified in recent years. In the treatment of CHD, percutaneous coronary intervention (PCI) is the main therapy. However, restenosis (RS) following coronary stenting is a disadvantage of this therapy [2]. Current studies suggest that there are associations between genetic factors and the development of CHD or RS after PCI [3,4].
Current studies have documented the interaction between HO-1 gene polymorphisms and CHD or RS after PCI. HO-1 is a subtype of heme oxygenase (HO) which plays a key regulatory role in the synthesis and catabolism of bilirubin [5]. HO will be increased significantly when the body responds to oxidative stress. During the degradation of heme to biliverdin, HO plays an important role as a rate-limiting enzyme [6]. Recently, two loci of HO-1 gene have been identified to be associated with CHD or RS in different population [7,8]. One is the (GT) n dinucleotide repeat length polymorphism, the other is the T(-413)A (rs2071746). Both loci are located in the
Based on these observations, two meta-analyses [27][28] related to this topic have been published. Qiao et al. [27] reported a positive correlation between genetic polymorphisms of HO-1 gene and CHD or RS after PCI. However, the meta-analysis from Rong et al. [28] do not support this results. Thus, the association of HO-1 gene polymorphisms with CHD or RS remains unclear. To clarify these associations, we performed an updated metaanalysis.

Study characteristics
There were 176 potentially relevant papers acquired from PubMed, Web of Science, the Cochrane Library, Wanfang and CNKI databases. Of these, we excluded 143 documents because of irrelevance to the aim of our study after reading the title and abstract. The remaining 33 documents were full-text reviewed, and 3 studies were excluded due to reported associations with diabetes [29][30][31]. Four studies were not case-control studies [32][33][34][35], 3 studies were excluded because the variable number tandem repeat was different [36][37][38], 2 studies were excluded because there was no genotype data or it was a review [39][40]. Furthermore, 2 papers were excluded because of their lack of relation to CHD risk but rather to cardiovascular disease prognosis [41][42]. Finally, our meta-analysis included 19 eligible studies [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26]. Table  1 and Table 2 listed the characteristics of each study. Finally, a total of 13 studies of the (GT)n repeat length polymorphism and 4 studies of the T(-413)A SNP with CHD were included. Six studies of the (GT)n repeat length polymorphism with RS were included. Because 4 papers included 2 studies, there were 23 studies included in final analysis.

HO-1(GT)n repeat length polymorphism and CHD
First, we investigated the relation between HO-1(GT)n repeat length polymorphism and CHD. No significant heterogeneity was identified by H-test and I 2 test in any of the genetic models (Table 3), therefore, the fixed-effect model was used. Significant statistical association was found between HO-1(GT)n repeat length polymorphism and CHD risk under an allelic contrast (S vs. L, OR = 0.929, 95%CI = 0.881-0.978, P = 0.005), the recessive genetic model (SS vs. SL+LL, OR = 0.858, 95% CI = 0.780-0.945, P = 0.002), and the co-dominant genetic model (SS vs. LL, OR = 0.843, 95% CI = 0.754-0.942, P = 0.003). Comparing to SL+LL and LL genotypes carriers, the CHD risk was significantly decreased among the SS genotype patients (Figure 1-5). However, we did not find significant association in the dominant genetic model ( Table 4).
In addition, we conducted subgroup analysis according to quality assessment. The fixed-effects model was used in all of the genetic models. Significantly decreased risk of CHD was found among individuals with the SS genotype compared to patients with L allele (SL + LL and LL genotypes) in the good-quality subgroup (S vs. L, OR = 0.951, 95% CI = 0.863-0.971, P = 0.003; SS vs. SL+LL, OR = 0.830, 95% CI = 0.746-0.924, P = 0.001; SS vs. LL, OR = 0.822, 95% CI = 0.726-0.930, P = 0.002). However, this association was not found in the poor-quality reports (Table 4).

HO-1(GT)n repeat length polymorphism and RS
In 6 independent studies, drug-eluting stents were utilized. These studies examined the main baseline characteristics and identified no significant difference. First, significant heterogeneity was found in the contrast models, and therefore, the random-effects model was used in this meta-analysis. In the overall population, we found that patients with S allele had a decreased RS risk after PCI compared with the L allele carriers (S vs. L, OR = 0.718, 95% CI = 0.541-0.953, P = 0.022; SS vs. LL, OR = 0.522, 95% CI = 0.306-0.889, P = 0.017) ( Figure  6 Table 4).

HO-1 T(-413)A SNP and CHD risk
There were 4 studies that investigated the relationship between the HO-1T(-413)A SNP and CHD. No significant heterogeneity was found in the contrast models, and so the fixed-effect model was used in this of meta-analysis. Meta-analysis suggested that there was a significant association between the HO-1 T(-413)A polymorphism and CHD under the allele contrast (A vs. T, OR = 0.915, 95% CI = 0.842-0.995, P = 0.038), the        Table 4).

Sensitivity analysis
We performed the sensitivity analysis to examine the influence of each study on the pooled ORs by deleting each study one at a time in each genetic model. The pooled ORs showed no significant change, suggesting the results are stable (Figure 8).

Publication bias
In the present study, we utilized Egger's test and funnel plots to evaluate the publication bias of all contrast models. By Egger's test and funnel plots, we did not found publication biases for both the (GT)n repeat length polymorphism and T(-413)A SNP (Table 5, Figure 9).

DISCUSSION
The present study indicates that HO-1 gene polymorphisms are associated with CHD independently. There are three isoforms of heme oxygenase in human, including HO-1, HO-2 and HO-3. HO-1 is up-regulated by oxidative stress and its own substrate heme [43] and may be modulated by fragile histidine triad gene (FHIT) [44]. Animal experiments and clinical trials have confirmed that the HO-1 enzyme is expressed in various tissues and cells, including asatherosclerotic lesions and vascular smooth muscle cells [43]. Therefore, HO-1 is considered to provide protective functions against asatherosclerotic lesions formation [43] and cellular proliferation [45].
In the present study, a significant association of the (GT)n SS genotype or S allele with decreased the risk of CHD and RS after PCI was observed. In the subgroup analysis, the Asian population showed a positive association in the all genetic models, while the study conducted with the Caucasian population showed no significant association in any of the genetic models. This can be explained by the high prevalence of S allele in Asian subjects. In addition, because of difference in life styles, ethnicity, region, and other factors, there are large differences in gene distribution. Nevertheless, more studies have been included in the present meta-analysis, and all the included studies were of high quality according to the methodological quality assessment. No significant heterogeneity was identified, and supplementary analysis, including subgroup and sensitivity analysis, were performed to strengthen our conclusions.
The present study also shows an association between genetic factors and the risk of stenting RS. We found that the HO-1T(-413)A SNP was associated with decreased risk for CHD. However, this significant association should be interpreted cautiously. First, CHD or RS after PCI are complex diseases with multifactorial etiology, including gene and environmental factors. Only one SNP is not sufficient to provide the appropriate explanation of genetic risk for CHD or RS after PCI.
Gene-gene or gene-environment interaction factors may influence the risk of a subject for CHD or RS. Second, some potentially confounding factors should be discussed. Primary sources of heterogeneity include the following: the condition of the population included in this study, the main characteristics of the stents following PCI and the treatment compliance of the patients. In addition, the number of included studies for the HO-1T(-413)A SNP is small, and so we did not perform further subgroup analysis in the present study. Even so, the conclusion still give us some information on the pathogenesis of CHD and RS risk factors. Indeed, HO-1 is involved in the occurrence of restenosis by inhibiting vascular smooth muscle cells, attenuating vascular remodeling, and other mechanisms [48,49]. Although in our meta-analysis, we found that S allele carriers have decreased risk for RS after PCI compare with L allele carriers and that the HO-1 T(-413)A SNP was associated with decreased risk of CHD, the importance of HO-1 in human RS following coronary stenting has not been fully defined.
Several limitations of our study need to be considered. First, the number of included studies for HO-1T(-413)A SNP is small, and so we did not perform further subgroup analyses in the present study. Second, limiting the included studies to those published in English and Chinese might have missed some eligible studies in other languages. In addition, it is possible that the results included in the present meta-analysis are affected by miscounting the genotypes or misclassification of CHD and RS.
Hence, our results suggested that carrying the S allele of the (GT)n locus or the A allele of the T(-413) A locus in the HO-1 gene promoter decreased the risk of CHD. We also found that carrying shorter (GT)n repeats (S or SS genotype) but not the T(-413)A SNP was associated with decreased risk of RS after PCI. These effects appeared more significant in Asian populations.

Identification of eligible studies
We carried out a systematic search in PubMed, Web of Science, the Cochrane Library, Wanfang Data and CNKI (China National Knowledge Infrastructure), with the last search updated on February 1, 2016. The following terms were used: "heme oxygenase 1" or "HO-1" or "HMOX-1" and "polymorphism" and "coronary artery disease" or "cardiac heart disease" or "myocardial infarction" or "MI" or "angina pectoris" or "arteriosclerosis" or "coronary disease" or "acute coronary syndrome" or "coronary stenosis" or "restenosis" or "stent-restenosis". We included literature on relevant studies carried out in human subjects published in English and Chinese. CHD was defined as confirmed myocardial infarction, typical angina pectoris (by the World Health Organization criteria), or a history of PCI or as diagnosed by angiography. The controls were defined as in-patients, outpatients, or members of the general population who were without typical angina pectoris or electrocardiographic abnormality and without coronary artery stenosis of more than 20% upon coronary angiography [28].

Inclusion criteria
The studies in our meta-analysis met the following inclusion criteria: (1) case-control or cohort studies; (2) investigation of the association between the HO-1 gene polymorphisms and coronary artery disease or coronary restenosis; (3) inclusion of all patients, which were using drug-eluting stents(DES) and had 6 months follow-up angiography, after stenting (Restenosis, was defined as angiographic restenosis, diameter stenosis of >50%, and clinical restenosis, target vessel revascularization during the follow-up.);(4) studies focusing on humans; and (5) unabridged genotype data could be acquired to calculate the odds ratios (ORs) and 95% confidence intervals (CIs).

Exclusion criteria
We excluded papers according to the following criteria: (1) studies with no genotype data; (2) commentaries, reviews and editorials; (3) family-based studies of pedigrees; and (4) repeated studies using the same population data or duplicated data.

Data extraction
Data collection from the eligible studies were conducted independently by two investigators (Zhang and Zheng). The following contents were collected: name of the first author, year of publication, ethnicity or geographic location of the study subjects, the characteristics of cases and controls, genotyping methods, number of cases and controls, the criteria for cases and controls, genotype frequency in cases and controls for HO-1 genotypes, Hardy-Weinberg equilibrium, and type of stents. Two investigators checked the extracted data and reached a consensus on all the data. If a disagreement existed, a third investigator (Xie) would adjudicate the disagreement.

Quality assessment
To determine the methodological quality of each study, we used the Newcastle-Ottawa scale (NOS), which uses a ''star'' rating system to judge the quality of all included studies. The NOS ranges between zero (worst) and nine stars (best). Studies with a score equal to or higher than seven were considered to be of good quality. A score equal to or higher than four and less than seven was regarded as being poor quality. Two investigators (Zheng and Zhang) independently assessed the quality of included studies, and the results were reviewed by a third investigator (Xie). Disagreement was resolved by discussion.

Statistics analysis
Stata 12.0 software (StataCorp, College Station, TX, USA) was used for statistical analysis in our metaanalysis. The Hardy-Weinberg equilibrium (HWE) was calculated for each study using the Chi-square test in control groups, and P < 0.05 was considered a significant deviation from the HWE. Odds ratios and 95% confidence intervals were calculated to assess the strength of the associations of the polymorphism and susceptibility to CHD or RS risk. The statistically significant level was determined by Z-test, and significance was set at p<0.05. Heterogeneity was assessed using the H test (significance level of P< 0.1) and the I 2 test (greater than 50% as evidence of significant inconsistency). Pooled effect sizes were determined using a fixed-effects model (the Mantel-Haenszel method) when heterogeneity was negligible (I 2 <50%) or a random-effects model (the DerSimonian and Kacker method) when significant heterogeneity was present (I 2 ≥50%). We also performed a sensitivity analysis to evaluate the effect of each study on the combined ORs by omitting each study in turn. Finally, we utilized Egger's tests to assess the potential publication bias.