Association Between MTHFR Polymorphisms and the Risk of Essential Hypertension: An Updated Meta-analysis

Background: Since the 1990s, there have been a lot of research on single-nucleotide polymorphism (SNP) and different diseases, including many studies on 5,10-methylenetetrahydrofolate reductase (MTHFR) polymorphism and essential hypertension (EH). Nevertheless, their conclusions were controversial. So far, six previous meta-analyses discussed the internal relationship between the MTHFR polymorphism and EH, respectively. However, they did not evaluate the credibility of the positive associations. To build on previous meta-analyses, we updated the literature by including previously included papers as well as nine new articles, improved the inclusion criteria by also considering the quality of the papers, and applied new statistical techniques to assess the observed associations. Objectives: This study aims to explore the degree of risk correlation between two MTHFR polymorphisms and EH. Methods: PubMed, EMBASE, the Cochrane Library, CNKI, and Wan Fang electronic databases were searched to identify relevant studies. We evaluated the relation between the MTHFR C677T (rs1801133) and A1298C (rs1801131) polymorphisms and EH by calculating the odds ratios (OR) as well as 95% confidence intervals (CI). Here we used subgroup analysis, sensitivity analysis, cumulative meta-analysis, assessment of publication bias, meta-regression meta, False-positive report probability (FPRP), Bayesian false discovery probability (BFDP), and Venice criterion. Results: Overall, harboring the variant of MTHFR C677T was associated with an increased risk of EH in the overall populations, East Asians, Southeast Asians, South Asians, Caucasians/Europeans, and Africans. After the sensitivity analysis, positive results were found only in the overall population (TT vs. CC: OR = 1.14, 95% CI: 1.00–1.30, P h = 0.032, I 2 = 39.8%; TT + TC vs. CC: OR = 1.15, 95% CI: 1.01–1.29, P h = 0.040, I 2 = 38.1%; T vs. C: OR = 1.14, 95% CI: 1.04–1.25, P h = 0.005, I 2 = 50.2%) and Asian population (TC vs. CC: OR = 1.14, 95% CI: 1.01–1.28, P h = 0.265, I 2 = 16.8%; TT + TC vs. CC: OR = 1.17, 95% CI: 1.04–1.30, P h = 0.105, I 2 = 32.9%; T vs. C: OR = 1.10, 95% CI: 1.02–1.19, P h = 0.018, I 2 = 48.6%). However, after further statistical assessment by FPRP, BFDP, and Venice criteria, the positive associations reported here could be deemed to be false-positives and present only weak evidence for a causal relationship. In addition, when we performed pooled analysis and sensitivity analysis on MTHFR A1298C; all the results were negative. Conclusion: The positive relationships between MTHFR C677T and A1298C polymorphisms with the susceptibility to present with hypertension were not robust enough to withstand statistical interrogation by FPRP, BFDP, and Venice criteria. Therefore, these SNPs are probably not important in EH etiology.


INTRODUCTION
Essential hypertension (EH) is a common disease in the world and a threat to the health of people. If blood pressure is not well controlled, it can lead to serious consequences, such as coronary heart disease, enlarged heart, heart failure, cerebral hemorrhage, optic papillary edema, and renal insufficiency. At the same time, chronic high blood pressure could lead to clinical symptoms such as dizziness, headache, chest pain, tinnitus, vomiting, palpitation, and blurred vision (O'Donnell et al., 1998). EH is also called primary hypertension because its etiology is not clear, and there is, so far, no complete understanding thereof (Bähr et al., 2003). It is the result of the interaction of various environmental factors and genetic factors. The former mainly includes diet, smoking, mental stress, and lifestyle. The latter includes contributors such as obesity, family history, and genetic variations or polymorphisms (Singh et al., 2016). In recent years, many genes related to hypertension have been reported, and relevant evidence suggested that genetic factors accounted for about 25-65% of the proportion in those with hypertension (Krzesinski and Saint-Remy, 2012). Therefore, genetic factors are critical to explore as a possible internal cause of this disease. In Europe, Newton-Cheh et al. conducted a genome-wide association study (GWAS) which used blood pressure as a continuous trait (Newton-Cheh et al., 2009). Through a two-stage meta-analysis, three loci associated with systolic blood pressure (MTHFR, CYP17A1, and PLCD3) and five diastolic blood pressure loci (FGF5, C10orf107, SH2B3, CYP1A2, and ZNF652) were identified at the genome-wide level. This is of great significance for the genetic research of EH (Rafiq et al., 2010).
Studies have shown that hyperhomocysteinemia (hHcy) is an important risk factor for hypertension. 5,10-Methylenetetrahydrofolate reductase (MTHFR) plays an important role in folate metabolism-methionine cycle. Together with methionine synthase reductase (MTRR), it maintains the normal metabolism of folate and participates in the maintenance of normal homocysteine (Hcy) levels in the body. Specific MTHFR gene mutations can lead to a decrease in the activity of key enzymes and disorders in folate metabolism, which increased the demand for folate to maintain Hcy methylation to methionine (Met) and ultimately lead to the increase of Hcy level and hHcy (Ren et al., 2018). The human MTHFR gene is located in autosomal 1p36.3. There are a variety of mutation types and multiple mutation sites in the MTHFR gene, of which C677T (rs1801133) and A1298C (rs1801131) are two common gene sites. The 677C-T gene mutation is located in exon 4 of the catalytic activity of the N-terminal region, where cytosine is replaced by thymine, and the corresponding protein is changed from alanine (Ala) to valine (Val). The 1298A-C mutation is located in the C-terminal regulatory region of exon 7, where adenosine mutates to cytosine, causing the glutamate (Glu) encoding to be replaced by alanine (Ala) (Weisberg et al., 2001;Chen et al., 2009;Nursal et al., 2018). Globally, the ethnic and geographic distribution of these two loci are significantly different. In 2010, approximately 1.39 billion people suffered from EH, especially in low-and middle-income countries (Kearney et al., 2005;Yang and Gu, 2016;Gheorghe et al., 2018;Mills et al., 2020). Therefore, EH was a major risk factor contributing to the global burden of disease (Falaschetti et al., 2014). Although many studies have elucidated the pathogenesis, the associations between the MTHFR SNPs and blood pressure were unclear (Doris, 2002;Staessen et al., 2003). Studies have shown that EH was related to the inheritance of a variety of susceptibility genes.

Search Strategy
Databases, including PubMed, EMBASE, the Cochrane Library, CNKI, and Wan Fang databases, were searched for evidence between the carrier status of the MTHFR gene variants and susceptibility to EH. The retrieval strategy was as follows: (MTHFR C677T OR rs1801133 OR Ala222Val) AND (MTHFR A1298C OR rs1801131 OR Glu429Ala) AND (polymorphism OR mutation OR variant OR genotype) AND (essential hypertension OR hypertension OR EH OR blood pressure). Two researchers who were familiar with the literature search process and meta-analysis methods conducted a literature search to identify all possible original studies. The search deadline was December 2020.

Selection Criteria
The inclusion criteria were as follows: (1) case-control study, (2) complete study of genotype data, and (3) study exploring the relationship between MTHFR polymorphisms and EH.
The exclusion criteria were as follows: (1) unrelated diseases or other diseases associated with hypertension or secondary

Data Extraction
According to the abovementioned inclusion and exclusion criteria, two researchers searched independently the abovementioned databases and finally checked them to ensure a comprehensive search and information extraction of all relevant studies. If there was a difference in opinion, it would be discussed with a third researcher. The extraction will be based on the following information: last name of the first author, publication year, country, geographical location, ethnicity of the study subjects, source of the case group, source of the control group, matching, diagnostic criteria for EH, genotype data in cases and controls, sample size, genotype examination, and adjustments.

Quality Assessment
All studies were independently evaluated and verified by two investigators. According to previously published articles (Thakkinstian et al., 2011;Xue et al., 2014) together with the characteristics of this study, a quality assessment table was designed. The maximum score a study could achieve based on the quality assessment criteria was 20. We regarded studies achieving a score of 12 and higher as good-quality research. Those scoring below 12 were not eligible for inclusion based on the low quality thereof. The specific evaluation criteria are shown in Table 1.

Credibility Analysis
The reliability of the statistically significant association was evaluated by false-positive reporting probability (FPRP), Bayesian false discovery probability (BFDP) test, and Venice criterion (Ioannidis et al., 2008). When FPRP <0.2 and BFDP <0.8, based on a predetermined prior probability of 0.001, the results of the genetic association were valid, indicating that the association might be true (Wakefield, 2007). The Venice criterion evaluated the credibility of cumulative results in terms of validity of evidence, reproducibility of studies, and bias control. The evaluation indicators were as follows: (1) evidence validity (n: genotype sample size): A: n ≥ 1,000; B: 100 ≤ n < 1,000; C: n < 100; (2) reproducibility: A: I 2 <25%; B: I 2 >25% and I 2 < 50%; C: I 2 >50% or higher; and (3) bias control: A: no bias; B: no obvious bias but data is missing; C: obvious bias. A, B, and C represent strong, moderate, and weak cumulative epidemiological evidence, respectively (Wakefield, 2007;Ioannidis et al., 2008).

Statistical Analysis
Chi-square goodness-of-fit test was used to perform Hardy-Weinberg equilibrium (HWE) on the included control group and to select the studies conforming to HWE. We calculated the combined odds ratio (OR) and 95% confidence intervals (CI) to determine the association between MTHFR gene polymorphisms and EH. Evaluation of heterogeneity was by Cochran Q test and I 2 test. When P >0.10 and/or I 2 ≤50%, a fixed-effect model was selected to combine the effect sizes. When P ≤0.10 and/or I 2 >50%, a random-effects model was selected   (Higgins et al., 2003). Allelic model, dominant model, recessive model, heterozygote model, and homozygote genetic model were used for the evaluation (Fu et al., 2019). We could identify the sources of heterogeneity in the following ways: (1) metaregression was performed on the factors that might lead to heterogeneity in the study itself; (2) according to ethnicity, geographical distribution, and genetic testing quality, a subgroup analysis was conducted; and (3) we conducted a sensitivity analysis by excluding individual studies one by one and judged the publication bias by Begg's funnel plot and Egger test. If there was publication bias, we would use the trim-and-fill method to correct it (Zhang et al., 2018). Cumulative metaanalysis of included literature was performed to dynamically observe the results (Lau et al., 1992). STATA12.0 software was used for statistical analysis, and P <0.05 was considered to be statistically significant.

Description of the Included Studies
Based on STREGA and PRISMA (Swartz, 2011), we carried out this research. According to the retrieval method, we searched PubMed, Embase, the Cochrane Library, CNKI, and Wan Fang electronic databases and obtained 930 articles. According to strict inclusion and exclusion criteria, 66 original literatures were finally included in this study, including 63 articles on C677T polymorphism and 11 articles on A1298C polymorphism (see Figure 1 for the detailed flow chart of the included literature).

Basic Characteristics and Quality Evaluation of Literature
A total of 66 articles were included, of which 63 were about MTHFR C677T and EH and 11 were about MTHFR A1298C. The results of the quality assessment were 25 high-quality studies and 41 low-quality studies. There were 58 studies in the control group that complied with HWE. The specific features of the included literature are shown in Table 2. The detailed genotype distribution is shown in Table 3 and Table 4.

MTHFR C677T Polymorphism
In Table 5, we summarized the association between MTHFR C677T and EH. In the overall population, EH was found to be at a high risk across all genetic models (TT vs.

MTHFR A1298C Polymorphism
The association between MTHFR A1298C and EH is shown in

Heterogeneity and Sensitivity Analyses
In Table 5, we showed the very obvious heterogeneity. A metaregression analysis was performed based on geographic origin, ethnicity, control group origin, gene quality control, matching, sample size, HWE, and literature quality assessment. Ethnicity was the source of heterogeneity in the final results (TC vs. CC: P 0.035; TT + TC vs. CC: P 0.042).
A sensitivity analysis was performed to rule out individual studies one by one, and the results were found to be stable. When we restricted the high-quality and HWE studies, there was no significant change in A1298C genotype. However, the results of the C677T genotype changed significantly in the overall population (TC vs.

Publication Bias
Begg's funnel plot and Egger test were used to evaluate publication bias. For MTHFR C677T polymorphism, there was publication bias in four genetic models (TT vs. CC: P 0.009; TT + TC vs. CC: P 0.045; TT vs. TC + CC: P 0.007; T vs. C: P 0.005) (Figure 2). In order to further clarify the publication bias, we applied the trim-and-fill method, and the results of the overall population did not change (Figure 3). For MTHFR A1298C polymorphism, the shape of the funnel plot was uniform and symmetrical, indicating that there was no significant publication bias under all genetic models (P). 3 | (Continued) The genotype distribution and HWE of MTHFR C677T polymorphism in this meta-analysis. The bold value is C/T T/T A/C C/C genotype data is incomplete, and HWE evaluation cannot be performed.

Cumulative Meta-analysis Results
According to year, the literature was included sequentially to evaluate the stability of cumulative effect size. The results showed that the correlation degree tended to be stable ( Figure 4).

Credibility of the Positive Results
After FPRP, BFDP, and Venice criterion evaluation, results that met both FPRP <0.2, BFDP <0.8, and above moderate epidemiological evidence were only in the heterozygote (TC vs. CC) and recessive model (TT vs. CT + CC). Table 7 shows the details. This was different from the results of the sensitivity analyses that limited high-quality and HWE studies.

DISCUSSION
In 1996, a group of researchers first investigated the association between MTHFR C677T polymorphism and EH in Japan (Nishio et al., 1996). Then, another group of researchers studied MTHFR C677T and A1298C polymorphisms for the first time in the Czech Republic (Kahleová et al., 2002). Thereafter, many scholars worldwide partook in the research, but the results were inconsistent and controversial. Even though several meta-analyses (Qian et al., 2007;Niu et al., 2012;Yang B. et al., 2014;Yang K.-M. et al., 2014;Wu et al., 2014;Fu et al., 2019) appeared to investigate the association between the MTHFR SNPs, methodological problems, especially the inclusion of pregnancy-related hypertension and lack of quality assessment of the studies included, hampered the acceptance of their findings. Therefore, it is necessary to conduct a new, more reliable, and more comprehensive meta-analysis. After data collection for this meta-analysis, HWE compliance checks of the control group, and literature quality score assignments, the effect sizes were combined using five genetic models. Ultimately, the C677T polymorphism was found to be a risk genotype in the overall population, East Asians, Southeast Asians, South Asians, Caucasians/Europeans (allelic model, homozygote model, dominant model, and recessive model), and Africans. It was possible that both were linked in the study population from African, South Asia region, and Southeast Asia region (Table 5). However, we are of the opinion that, based on a lack of studies on populations representative of Africa [one report (Ghogomu et al., 2016)], South Asia [two reports (Markan et al., 2007;Dwivedi and Sinha, 2017)], and South-East Asia [two reports (Wei et al., 2015;Candrasatria et al., 2020)], these results should be interpreted with caution until more studies on these underrepresented groups accumulate (Higgins, 2008). Then, we conducted a sensitivity analysis to choose high-quality studies and studies whose populations adhered to the assumptions of HWE. The results showed that C677T polymorphism was significantly associated with EH only in the overall population (allele, dominant, and homozygote model) and the Asian population (dominant, allele, and heterozygote model). When we used the FPRP, BFDP test, and the Venice criterion for evaluation, all significant associations were less-credible positive results, indicating that the association could not be noteworthy and could therefore be regarded as weak-level evidence. These tests are necessary to avoid random error and confounding bias, but sometimes they can alter the results of the epidemiological studies (Wakefield, 2007;Ioannidis et al., 2008;Zhao et al., 2018;Tian, 2019). At the same time, among the low-quality studies, there were four HWD studies, 28 unadjusted studies, 34 gene quality uncontrolled studies, and 34 no matching studies, which may be the main source of bias. In addition, we did not find evidence that the A1298C polymorphism was associated with EH. The South Asia [one report (Markan et al., 2007)] and Southeast Asia [one report (Wei et al., 2015)] results were also significantly less reliable because of the small number of studies (Tables 5-7). In conclusion, our findings are comparable to previous reports which observed that the MTHFR variant SNP at position 677 conveys a risk for hypertension; however, compared with the previous metaanalyses, the FPRP, BFDP, and Venice criterion indicated that this relationship is probably not causal. Some observational studies have shown that genetic changes are inseparable from the impact of environmental factors such as diet, exercise, smoking, and drinking. A method to explore the connection between genes and environmental factors is to study people in different regions and cultures (Anand, 2005). Hcy is an independent risk for EH (Zhong et al., 2017). Alam et al. found that the MTHFR 677 TT genotype was associated with hHcy in Indians (Alam et al., 2008). However, no association between MTHFR C677T polymorphism and hHcy was found in long-term resident Indians in the UK (Chambers et al., 2000). Further research suggested that the difference may be caused by different environmental factors such as diet and exercise. We could find a significant heterogeneity between the articles included in the final results ( Table 5 and Table 6). Heterogeneity exists naturally in meta-analysis, so sources of heterogeneity should be sought as far as possible. First, after excluding each study one by one, we believe that the included studies are within the confidence interval of the total effect size. We performed a sensitivity analysis which, in certain instances, can be useful to identify the source of heterogeneity. Therefore, we conducted a meta-regression analysis and found that ethnicity was the source of heterogeneity only in the dominant model and the heterozygote model. In addition, publication bias was also observed (Figure 2 suggests that small sample studies lead to publication bias). In order to identify and correct publication bias caused by the asymmetry of a funnel figure (Rothstein et al., 2005), we applied trim-and-fill method to remove the small sample study and (to) include missing studies. The results showed that there were no obvious changes before and after the cut to fill the consolidation effect (before the cut to fill the consolidation effect was statistically significant, while after the cut to fill the consolidation effect was statistically significant) ( Figure 3); publication bias may not exist (Zhang and Zhong, 2009). In molecular epidemiological studies, small-sample research is very likely to have random errors and bias, resulting in unreliable final results (Ioannidis et al., 2008). Because the experimental design of small-sample research is not strict, its research quality is often low. Cumulative meta-analysis was carried out according to year. The results showed that, with the extension of year, the effect size did not change much, and the final results were gradually stable   (Figure 4). When we carried out a strict and high-quality design for small-sample research, the results were close to the real level. Compared with previously published meta-analyses, this study has several advantages. Firstly, we have the largest sample size to date. A total of 66 case-control studies were included. For MTHFR C677T polymorphism, 63 studies were collected, including 11,556 case groups and 12,523 control groups. For MTHFR A1298C polymorphism, there were 11 studies with 2,504 case groups and 2,979 control groups. Secondly, we analyzed the sensitivity of the combined results by excluding individual studies one by one and limiting high-quality and HWE studies. Thirdly, meta-regression, cumulative meta-analysis, trim-and-fill method, and geographic origin subgroup analysis were used to verify the qualified combined effect size. Fourthly, we used the FPRP and BFDP tests that were first introduced to judge the credibility of the positive results. The Venice criterion was used to assess the cumulative level of epidemiological evidence for genetic association. The GWAS on Europeans was an association study for the detection of multiple SNPs. The method was regression analysis. An assumption of regression analysis is the independent distribution of data. However, the fact is that many individuals may have distant relationships, which will lead to false-positive results in an association analysis (O'Donnell and Nabel, 2008;Sul et al., 2018). However, our study conducted a credibility analysis of positive results obtained by traditional methods.
However, there are also shortcomings to our meta-analysis that should be kept in mind. Firstly, only Chinese and English studies were included in this study; studies in other languages were not included. There is a certain degree of heterogeneity. Secondly, this study did not include the data of family genetic aggregation, smoking, body weight, body mass index, gender, age, and medication history. Therefore, the results may be affected by confounding factors. Thirdly, the sources of the control group were not uniform, which may have classification bias. Fourthly, there were few studies on MTHFR C677T polymorphism in Africa, South Asia, Southeast Asia, and West Asia. Therefore, heterogeneity may exist. More research should be done in these areas. Fifthly, with regard to the MTHFR A1298C polymorphism, since there were only two studies in West Asia, it is still uncertain whether this SNP is associated with EH.

CONCLUSION
In summary, this study provided a comprehensive analysis of MTHFR polymorphisms on the risk of EH in different populations worldwide. The results showed that MTHFR C677T gene polymorphism was associated with increased EH risks in overall population, East Asian, and Caucasian. However, after FPRP, BFDP, and Venice Criteria tests, the abovementioned associations became less reliable. These results should be treated with caution because they could be falsepositives.
The final conclusion highlighted weak epidemiological credibility for a higher risk of EH of the MTHFR 677T allele. Therefore, more investigations are needed to provide researchers with conclusive evidence of whether MTHFR C677T is a genetic predictor of EH. Similarly, there was no significant association between the MTHFR A1298C polymorphism and EH. Therefore, MTHFR polymorphisms could not contribute to the development of hypertension. In the future, more studies are needed to repeatedly verify the current conclusions.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
In this study, HM conducted the research, collected data, analyzed data, and wrote papers. SH carried out the research design, data collection, statistical analysis, and revision of the paper. YY was responsible for sorting data and checking data. YX and LF designed the research and performed statistical analysis and revision of the paper. At the same time, YX also completed the generation of Figures 1-4 and Tables 5-7. HM and SH were responsible for making Tables 1-4 and Supplementary forms and filling in relevant information. XH was involved in designing the research, verifying data, and revising papers. All authors participated in the revision and uploading of the manuscript.