Associations between MTHFR gene polymorphisms and the risk of intracranial hemorrhage: Evidence from a meta‐analysis

Abstract Introduction Previously, a number of genetic epidemiological studies have evaluated associations between MTHFR gene polymorphisms and the risk of intracranial hemorrhage (ICH), with controversial results. Accordingly, we carried out this meta‐analysis to more conclusively evaluate associations between MTHFR gene polymorphisms and the risk of ICH. Methods MEDLINE, EMBASE, Wanfang, VIP, and CNKI were searched comprehensively, and thirty‐one genetic association studies were finally selected to be included in this meta‐analysis. Results Eight literatures (963 cases and 2,244 controls) assessed relationship between MTHFR rs1801131 (A1298C) polymorphism and the risk of ICH, and thirty‐one literatures (3,679 cases and 9,067 controls) assessed relationship between MTHFR rs1801133 (C677T) polymorphism and the risk of ICH. We found that AA genotype of rs1801131 polymorphism was significantly associated with a decreased risk of intraventricular hemorrhage (IH) compared with AC/CC genotypes (OR = 0.63; p = .003), AC genotype was significantly associated with an increased risk of IH compared with AA/CC genotypes (OR = 1.55; p = .005), and A allele was significantly associated with a decreased risk of IH compared with C allele (OR = 0.75; p = .02). Additionally, CC genotype of rs1801133 polymorphism was significantly associated with a decreased risk of cerebral hemorrhage (CH) compared with CT/TT genotypes (OR = 0.75; p = .04), TT genotype was significantly associated with an increased risk of CH compared with CC/CT genotypes (OR = 1.27; p = .02), and C allele was significantly associated with a decreased risk of CH compared with T allele (OR = 0.85; p = .007). Conclusions This meta‐analysis shows that rs1801131 polymorphism may influence the risk of IH, while rs1801133 polymorphism may influence the risk of CH.


| INTRODUC TI ON
Intracranial hemorrhage (ICH), an acute cerebral vascular disorder characterized by nontraumatic bleeding into brain parenchyma, has high death and disability rates (Collaborators, 2015;Global Burden of Disease Study, 2013). ICH can be divided into subarachnoid hemorrhage (SAH), intraventricular hemorrhage (IH), cerebral hemorrhage (CH), and hemorrhagic stroke (HS). SAH refers to spontaneous bleeding into subarachnoid space, IH refers to spontaneous bleeding into ventricular system, CH refers to spontaneous bleeding into brain parenchyma, and HS refers to loss of blood supply to brain parenchyma caused by rupture of blood vessels (may result in spontaneous bleeding into subarachnoid space, subdural space, or brain parenchyma). Although the incidence of ICH is found to be relatively lower than ischemic cerebral vascular disorder, the associated mortality and disability rates of ICH are actually much higher than that of ischemic cerebral vascular disorder (Caceres & Goldstein, 2012;Freeman & Aguilar, 2012). So far, the disease mechanisms of ICH are still not clear. Nevertheless, the remarkable differences observed in incidences of ICH among different ethnic groups by previous epidemiological investigations indicated that genetic architecture may play a vital role in its development (Backhaus et al., 2015;Krishnan et al., 2018).
Methylenetetrahydrofolate reductase (MTHFR) regulates the metabolism of folate and homocysteine (Moll & Varga, 2015;Zaric et al., 2019). Previous experimental studies have demonstrated that deficiency of MTHFR could lead to hyperhomocysteinemia and it also give rise to an elevated risk of multiple cardiovascular and cerebrovascular disorders Santilli, Davì, & Patrono, 2016;Wei et al., 2017). Accordingly, it is rational to speculate that functional polymorphisms of MTHFR gene, which may lead to disruption of homocysteine metabolism and give rise to hyperhomocysteinemia, may also influence the risk of a wide variety of cardiovascular and cerebrovascular disorders. Past experimental studies have found that the C/T variation of rs1801133 and A/C variation of rs1801131 are both associated with alteration of amino acid sequence and reduction in MTHFR enzymatic activity (Li, Dai, Zheng, Liu, & Huang, 2015;Trimmer, 2013). Therefore, it is possible that individuals carrying the mutant allele of MTHFR rs1801131 or rs1801133 polymorphism may be more prone to develop hyperhomocysteinemia and its associated vascular disorders.
Previously, many investigators across the world have extensively tried to examine the associations of MTHFR rs1801131 and rs1801133 polymorphisms with the risk of ICH, yet with inconsistent results. Accordingly, a meta-analysis was carried out by us to robustly estimate the associations between those two MTHFR polymorphisms and the risk of ICH.

| MATERIAL S AND ME THODS
This meta-analysis was conducted in accordance with the PRISMA guideline (Moher, Liberati, & Tetzlaff, 2009

| Data extraction and quality assessment
The following data items were extracted from eligible publications: (a)  (Stang, 2010), and these with a score of 7-9 were considered to be of good quality.
Two authors carried out data extraction and quality assessment in parallel. A thorough discussion until a consensus is reached would be initiated if discrepant results were found between two authors.

| Statistical analyses
All statistical analyses in this meta-analysis were performed with the Cochrane Review Manager software. Associations between MTHFR gene polymorphisms and the risk of ICH were explored by using odds ratio and its 95% confidence interval. The statistically signifi- DerSimonian-Laird method, which is also known as the random effect model, to pool the results of eligible studies if I 2 is larger than 50%. Otherwise, the authors would use Mantel-Haenszel method, which is also known as the fixed effect model, to pool the results of eligible studies. Meanwhile, the authors also conduct subgroup analyses by ethnic background and disease subtypes. Sensitivity analyses were conducted by deleting one eligible study each time, and then recalculating the pooled results based on the rest of eligible studies. Publication biases were evaluated with funnel plots and Egger's tests. Meta-regression analysis was further performed to determine whether study quality of eligible studies can change the effect size.

| Characteristics of included studies
Seventy-eight publications were identified by us using our searching strategy. We screened forty publications for eligibility of inclusion after excluding thirty-four unrelated publications and four repeated reports. Seven reviews and two case series were further excluded.
Totally, thirty-one studies met the selection criteria and were finally enrolled for pooled analyses (Figure 1). Data items extracted from eligible studies were summarized in Table 1. In brief, the eligible studies were published between 1998 and 2018, and they all used healthy blood donors as controls. The NOS score of eligible studies ranged from 7 to 8, which indicated that all eligible studies were of relatively good quality.  Table 2).  Table 2).

| Sensitivity analyses
We tested stability of quantitative analysis results by deleting one eligible study each time, and then recalculating the pooled results based on the rest of studies. The overall trends of associations remained unchanged in sensitivity analyses, which indicated that our quantitative analysis results were quite reliable and stable (Relevant datasets can be found at https://osf.io, username: j7ffofnjwi@163. com, password: j7ffofnjwi @).

| Publication biases
We estimated potential publication biases in this meta-analysis with funnel plots and Egger's tests. Funnel plots were found to be overall symmetrical ( Figure S1). We further calculated p values for

| Meta-regression analysis
Meta-regression analysis was performed to determine whether study quality of eligible studies can change the effect size, and the findings of meta-regression analysis indicated that the effect sizes of pooled analyses were not altered by study quality (see Table 3).

| D ISCUSS I ON
This meta-analysis, robustly estimated associations between gene polymorphisms in MTHFR and the risk of ICH. It is so far the very first meta-analysis regarding rs1801131 (A1298C) polymorphism and ICH, and it is also so far the most complete meta-analysis regarding rs1801133 (C677T) polymorphism and ICH. The pooled analyses results showed that rs1801131 (A1298C) polymorphism

TA B L E 3
Meta-regression analysis for study quality was significantly associated with the risk of IH, whereas rs1801133 (C677T) polymorphism was significantly associated with the risk of CH.
The following points need be taken into consideration when interpreting our pooled analyses. First, previous experimental studies have found that those two investigated MTHFR polymorphisms may alter enzymatic function of MTHFR and raise plasma homocysteine levels, which may attribute to a higher risk of ICH Santilli et al., 2016;Trimmer, 2013;Zaric et al., 2019). Nevertheless, further experimental studies still need to explore the exact molec- studies that were conducted in the Chinese population, and two recent publications in the Chinese population were not covered by Hu and colleagues (Jiang, Sheng, & Luo, 2018;Shao, Meng, & Wu, 2016).
In contrast to those two previous meta-analyses, we did not observe positive associations with ICH in overall population and Asians.
Positive results were only detected for rs1801133 (C677T) polymorphism in the CH subgroup in our integrated analyses, but not in other types of ICH. Considering that our integrated analyses were derived from more eligible studies, our observations should be more statistically robust. Nevertheless, further studies with larger sample sizes are still warranted to confirm our findings. Actually, if we set type I error at 5%, set the power at 80%, and set the estimated between group difference at 10%, a sample size of around 950 would be sufficient to detect the differences between two groups. In the current meta-analysis, only the pooled sample sizes of SAH and IH subgroups for rs1801131 as well as IH subgroup for rs1801133 did not reach these criteria, so the findings observed in those subgroups should be considered as exploratory and future studies still need to test the robustness of those pooled results.
Despite being the so far most comprehensive meta-analysis on this topic, it should be acknowledged that the present study also has two major limitations. Firstly, our quantitative analysis results were derived from unadjusted pooling of previous publications due to lack of access to raw data of eligible studies. Without genotypic data according to age, sex or comorbid status, it is impossible for us to conduct adjusted analyses accordingly. Although almost all meta-analyses shared the same limitation, it should be acknowledged that lack of further adjustment for baseline characteristics such as age, sex or comorbid status may somehow influence reliability of our findings (Chen, Chang, & Chen, 2018;Falcone & Woo, 2017).
Secondly, environmental factors such as diets or exercise levels may also influence associations between MTHFR polymorphisms and the risk of ICH. However, most of previous publications failed to consider environmental factors, so it is impossible for us to explore genetic-environmental interactions in a meta-analysis based on previous publications (Ma et al., 2015).
In conclusion, this meta-analysis shows that rs1801131 (A1298C) polymorphism may affect the risk of IH, whereas rs1801133 (C677T) polymorphism may affect the risk of CH. Nevertheless, further studies with larger sample sizes are still needed to confirm our findings.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R S ' CO NTR I B UTI O N S
Fenghui Wang and Youbin Jing conceived and designed this metaanalysis. Fenghui Wang and Zhendong Xu searched literatures.
Haiyan Jiao and Aixiang Wang analyzed data. Fenghui Wang and Youbin Jing wrote the manuscript. All authors have approved the final manuscript as submitted.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.1840.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.