ABO blood group system and the coronary artery disease: an updated systematic review and meta-analysis

ABO blood group system, a well-known genetic risk factor, has clinically been demonstrated to be linked with thrombotic vascular diseases. However, the relationship between ABO blood group and coronary artery disease (CAD) is still controversial. We here performed an updated meta-analysis of the related studies and tried to elucidate the potential role of ABO blood group as a risk factor for CAD. All detectable case-control and cohort studies comparing the risk of CAD in different ABO blood groups were collected for this analysis through searching PubMed, Embase, and the Cochrane Library. Ultimately, 17 studies covering 225,810 participants were included. The combined results showed that the risk of CAD was significantly higher in blood group A (OR = 1.14, 95% CI = 1.03 to 1.26, p = 0.01) and lower in blood group O (OR = 0.85, 95% CI = 0.78 to 0.94, p = 0.0008). Even when studies merely about myocardial infarction (MI) were removed, the risk of CAD was still significantly higher in blood group A (OR = 1.05, 95% CI = 1.00 to 1.10, p = 0.03) and lower in blood group O (OR = 0.89, 95% CI = 0.85 to 0.93, p < 0.00001). This updated systematic review and meta-analysis indicated that both blood group A and non-O were the risk factors of CAD.

of duplicates. By screening titles and abstracts, we excluded 147 records on account of animal experiments, traditional reviews, improper or lack of comparison, or other blood group classification systems rather than ABO blood type. By browsing full-text articles, we excluded 58 records because of improper or lack of comparison, other confounding factors, irrelevant to the outcomes of this study and unavailable outcomes. At last, a total of 16 articles [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26] which met inclusion criteria were included into this systematic review. A flow-chart of study selection was generated according to the PRISMA requirements ( Fig. 1). Study characteristics. One 15 of these 16 articles contained 2 studies. All the 17 studies were published in English from 1961 to 2014. Eleven articles 11,13,14,16,17,19,[21][22][23]25,26 were case-control studies, 2 articles 12,15 were prospective cohort studies and 3 articles 18,20,24 were retrospective cohort studies. Finally, a total of 225,810 patients were included. All studies described race and characteristic of the two groups. Nine studies 12,16,17,[19][20][21][22][23][24] merely mentioned MI. Two studies 12,17 only differentiated blood group non-O from blood group O. The remaining studies described all blood types. (Basic characteristics of included studies were presented in Table 1 and blood types distribution and outcome definitions of included studies were presented in Table 2, and Newcastle-Ottawa Scale (NOS) table was shown in Table 3).
Main, subgroup and sensitivity analysis. All the 17 studies were included in this meta-analysis.
Because of unnegligible heterogeneity in them, we conducted a subgroup analysis according to the research types (case-control study, prospective or retrospective cohort study) and used random-effect model 27 . Risk of CAD was significantly increased in patients with blood group A compared with blood group non-A (odds ratio (OR) = 1.14, 95% confidence intervals (CI) = 1.03 to 1.26, p = 0.01). Subjects in blood group A had a statistical increase in CAD incidence in case-control studies (OR = 1.14, 95% CI = 1.04 to 1.26, p = 0.005) with moderate heterogeneity (I 2 = 45%), while there was no statistical difference between blood group A and non-A in cohort studies (Fig. 2). Besides, risk of CAD had no statistical significant difference in patients with blood group B, AB compared with non-B, non-AB, respectively (Figs 3 and 4). Whereas, in contrast to the result of blood group A, blood group O was proved to be a protective factor in our analysis, presenting a decrease of CAD risk (OR = 0.85, 95% CI = 0.78 to 0.94, p = 0.0008). Our analysis found that there was statistical significant difference in CAD incidence in case-control studies (OR = 0.86, 95% CI = 0.75 to 0.99, p = 0.04) in spite of high heterogeneity (I 2 = 78%), which is similar to prospective cohort studies (OR = 0.94, 95% CI = 0.89 to 0.98, p = 0.009) with no heterogeneity (I 2 = 0). However, there was no statistical significant difference in CAD incidence in retrospective cohort studies (OR = 0.58, 95% CI = 0.35 to 0.97, p = 0.04) with high heterogeneity (I 2 = 70%) (Fig. 5). In the  Publication bias. We generated a funnel plot to assess publication bias. Exploration for the funnel plot of the blood group O in CAD suggested no asymmetry. No obvious evidence of publication bias was present in the comparison of blood group O (Fig. 6).

Discussion
Previous systematic reviews and meta-analysis paid more attention to the relationship between MI and ABO blood group, but the link of ABO blood group system to CAD was rarely evaluated. Besides, almost all available studies principally focused on blood type non-O and O. Hence, the relation between ABO blood group and risk of CAD is worthy to be assessed scientifically and strictly.  Our meta-analysis involved 16 articles (17 studies) covering 225,810 individuals. It was suggested that the risk of CAD in blood group A was mildly increased compared with that in blood group non-A (OR = 1.14). Meanwhile, we investigated the relationship of blood group B, AB compared with non-B, non-AB, respectively, but failed to confirm statistical difference. Moreover, our results indicated that the risk of CAD in blood group O was significantly lower than that in non-O groups (OR = 0.85), which is similar to previous studies 8 .
To our knowledge, this is the first meta-analysis involved the relationship between the risk of CAD and blood group A and non-A. Several clinical studies have provided direct evidence with different results. Whincup et al. 28 found that the incidence of ischaemic CAD was higher in those with blood group A than that with blood group non-A (OR = 1.21, 95% CI = 1.01 to 1.46). A study from Wazirali et al. 29 suggested that blood group A was associated with a substantially increased risk of CAD, which is independent of conventional cardiovascular risk factors. Whereas, another research did not support this association and indicated that the risk of CAD in blood group A was lower than that in other blood groups 30 . As we known, meta-analyses provide advance over traditional single studies. That is a reason why we performed a meta-analysis for further evaluating the relation of blood group A to the risk of CAD. In our study, we affirmed blood group A was a risk factor, which is more convincing and reliable. Similar evidence was more robust in the analysis for MI incidence (OR = 1.24).
Our study showed a significantly reduced risk of CAD in individuals with blood group O compared with that with blood group non-O (OR = 0.85, 95% CI = 0.78 to 0.94, p = 0.0008). Evidence was more obvious when we performed an analysis concerning the relationship between ABO blood group and MI (OR = 0.81, 95% CI = 0.69 to 0.94, p = 0.007). In fact, non-O blood group as an independent risk factor was already confirmed in other systematic reviews, too 8-10 . Wu et al. 9 performed a meta-analysis with regard to the relation of ABO blood group to MI and angina in 2008. In their study, taking group O as index, group A and non-O were related to an increase in MI risk (OR = 1.29, 95% CI = 1.16 to 1.45, p < 0.00001, OR = 1.25, 95% CI = 1.14 to 1.36, p < 0.00001), while no similar effect was found in the risk of angina. Furthermore, a meta-analysis by Dentali et al. 8 found that patients with blood group non-O presented a higher prevalence of MI than that with blood group O (OR = 1.28, 95% CI = 1.17 to 1.40, p < 0.001). Takagi et al. 10 enrolled 10 studies with a total of 174,945 participants and demonstrated a 14% increase in CAD incidence in individuals with blood group non-O compared to that in blood group O (OR = 1.14, 95% CI = 1.04 to 1.25, p < 0.006). All in all, the quantitative results from these meta-analyses and our one provided plenty of evidence on the close relationship between risk of CAD and blood group non-O.
The underlying mechanism of the relationship between blood group O and CAD has been clarified. ABO antigen may affect plasma levels of vWF and coagulation factor VIII 7 , and blood group non-O has the lowest expression of O antigen and relatively higher levels of vWF and factor VIII 31 . That blood group O is a potentially important genetic risk factor for bleeding 32 , which also supports this mechanism theory. Another biologically plausible mechanism involves in glycotransferase-deficient enzyme which renders the ABO blood group to encode O phenotype, resulting in protection of subjects from MI risk 33 . The latest study reveals that serum lipid mediates the effect of ABO blood group on CAD. In fact, blood group A is one of the risk factors of CAD mainly due to higher serum total cholesterol (TC) concentration in subjects 28 . Our recent study also indicated that there is an association between blood group A and risk of CAD, and around 10.5% of the effect of blood group A on CAD is mediated by TC levels 34 . It was mentioned that there were several potential limitations in this study. Firstly, there was certain heterogeneity between various studies. Although we performed subgroup analyses, it was still different among the studies in blood testing methods and diagnostic criteria of CAD, race, life and eating habits, religious beliefs, socio-economic patterns, and concern of the disease, which might result in the heterogeneity. Secondly, we did not find unpublished studies, which may bring about publication bias.
In conclusion, this updated meta-analysis suggests that blood group A and non-O are associated with an increased risk of CAD. However, considering the heterogeneity of included studies and limited number of studies, more rigorous studies with high quality are needed to give high level of evidence to confirm this association.

Methods
This meta-analysis was performed according to the MOOSE group guidelines of observational meta-analyses 35 . Data sources and searches. Two reviewers (Zhuo Chen and Sheng-Hua Yang) searched Pubmed and the Cochrane Library from their inception to August 15, 2015 in order to identify all existing literature which assessed the association between ABO blood group and CAD. Mesh vocabulary and free text terms were used for each database with relevant key words such as blood grouping and cross-matching, ABO blood group system, blood group antigens, myocardial ischemia, myocardial infarction, acute coronary syndrome and angina pectoris. Language was limited to English. There was no limitation of country and publication date. To ensure comprehensive acquisition of studies, the reference lists of the included articles were also manually screened to identify additional eligible studies. Manual searches were also performed on other databases, including Web of Science, and Google Scholar. Furthermore, databases of ongoing trials were also searched: Clinical Trials.gov (http://clinicaltrials.gov/) and Current Controlled Trials (http://www. controlled-trials.com/). Study selection. Studies were independently identified by two reviewers (Zhuo Chen and Sheng-Hua Yang) according to inclusion criteria. Disagreements were resolved through discussion and decided by a third reviewer. Both case-control and cohort studies were included if they met all the following criteria: 1) patients with CAD or even MI; 2) separate data for patients with or without CAD were provided; 3) diseases were objectively diagnosed in line with the diagnosis level at the time; 4) a clear extractable ABO blood group typing. Patients included were regardless of age and race.
Data extraction and quality assessment. The retrieved papers were subjected to a rigorous extraction by two authors (Zhuo Chen and Sheng-Hua Yang) independently according to a predesigned form. Disagreements were resolved by consensus or consulted from the third author (Hao Xu). We did not try to contact authors to obtain unpublished data. The methodological quality of studies was assessed using the NOS checklist for observational studies 36 . We rated cohort studies a maximum of 4 stars for selection, 2 stars for comparability, and 3 stars for outcome assessment. The maximum score of case-control studies for selection, comparability, and exposure assessment was 4, 2, 3, respectively, too. The highest score is 9, and more stars meant better quality.
Data analysis and synthesis. Revman 5.2 software (The Cochrane Collaboration, Oxford, UK) was used for data analyses. We presented dichotomous data as OR and its 95% CI. Data were assessed by both random and fixed effect models, but only the random effect analyses were reported if the heterogeneity was significant evaluated by the I 2 statistic which assessed the appropriateness of pooling all studies 27 . A funnel plot was used to assess publication bias.