Mendelian randomization study and meta‐analysis exploring the causality of age at menarche and the risk of intracerebral hemorrhage and ischemic stroke

Abstract Background The relationship between the age at menarche (AAM) and the risk of intracerebral hemorrhage (ICH) and ischemic stroke (IS) is still up for debate. The purpose of this study was to investigate potential causal connections between them. Methods Genome‐wide association analysis (GWAS) of AAM conducted by the MRC‐IEU consortium was utilized for association analyses of ICH and IS by two‐sample Mendelian randomization (MR) study. AAM data of the within‐family GWAS consortium were used as replication phase data to verify the causal relationship between each other. Inverse variance weighting (IVW) method was the primary method used in this MR study. For additional proof, the weighted median estimation, MR‐Egger regression, MR‐PRESSO test, and MR‐Robust Adjusted Profile Score evaluation were performed. The Cochran's Q test and the MR‐PRESSO global test were used, respectively, to examine the sensitivity and pleiotropy. Random effects meta‐analysis was utilized to analyze the causal data from the two consortiums to further explore the causality between AAM and ICH, IS. Results We found that the AAM was causally linked with the risk of ICH (OR = 0.48, 95% CI: 0.28–0.80, p = 0.006). On the contrary, the causal effect from AAM to IS (OR = 0.98, 95% CI: 0.91–1.06, p = 0.64) has not been confirmed. For all subtypes of ICH, we found that nonlobar intracerebral hemorrhage (NLICH, OR = 0.41, 95% CI: 0.23–0.75, p = 0.004) but not lobar intracerebral hemorrhage (LICH, OR = 0.65, 95% CI: 0.34–1.24, p = 0.19) was associated with AAM without surprise. Similarly, we used the within‐family GWAS consortium data to explore causality and found that AAM may reduce the risk of ICH (OR = 0.78, 95% CI: 0.72–0.86, p = 9.5 × 10−8) and NLICH (OR = 0.68, 95% CI: 0.61–0.75, p = 3.4 × 10−13) by IVW methods, but is not related to IS (OR = 0.97, 95% CI: 0.93–1.02, p = 0.26). These findings are further supported by the meta‐analysis. Both Cochran's Q test and the MR‐PRESSO global test failed to detect the presence of sensitivity. Conclusion AAM and ICH, particularly NLICH, are causally related, but not LICH, IS, or its subtypes in European population.


| INTRODUC TI ON
Age at menarche (AAM) refers to the age at which puberty women first exhibit endometrial bleeding brought on by the regulation of estrogen and progesterone, which signals the start of women's menstrual cycle. More and more research has shown that women's AAM is gradually earlier as a result of lifestyle changes. 1,2 Critical illnesses like obesity, 1 type II diabetes, 2 cardiovascular disease, 3,4 and cancers of the reproductive system like breast cancer and ovarian cancer 5,6 in the female population have also been linked to early AAM.
Premature AAM is closely linked to the effects of estrogen and progesterone and may be influenced by them as a potential cause of many female diseases.
Estrogen and progesterone may have important effects on stroke, such as female pregnancy, 7,8 birth rate, 8,9 menopause, 9 and external use of these two hormones. 9 The second largest cause of death and disability worldwide is stroke, which is also the third leading cause of death and disability. 10 Stroke poses a severe hazard to people's health all over the world with a prevalence of 101 million and a death toll of 6.55 million. 10 Intracerebral hemorrhage (ICH) and ischemic stroke (IS) are the two main subtypes of stroke but differ markedly in their pathogenesis (vascular rupture/vascular embolism). There are also significant gender differences in the incidence and mortality rates of the two diseases, and the effects of estrogen and progesterone may be important in accounting for this difference. 9,11,12 Age at menarche is the time when the body is first affected by estrogen and progesterone, and it affects the length of a woman's exposure to estrogen and progesterone and her reproductive life span. It is the earliest indication of the effects of hormones compared with pregnancy and menopause and provides the best predictor of future prevention and control. Therefore, exploring the interrelationship between AAM and ICH/IS has significant clinical indications.
A multicenter case-control study showed that women with first menstruation younger than 13 years old had a higher risk of developing IS (OR = 1.64, 95% CI: 1.11-2.17). 13 Conversely, a meta-analysis of 75 studies found that women with shorter reproductive cycles had a higher risk of stroke (OR = 1.31, 95% CI: 1.25-1.36), which indicates that the earlier the AAM, the lower the risk of stroke. 14 Furthermore, according to a cohort study of 1.2 million adults in the UK, the association between AAM and stroke was U-shaped, with the risk of stroke rising with age between ≥17 and ≤10. 15 Taken together, the relationship between AAM and IS risk is controversial (IS accounts for more than 80% of strokes). Also, the relationship between AAM and the risk of ICH is unclear.
Mendelian randomization (MR) is frequently used to assess the effects of confounding factors in epidemiological etiology on inferences. 16 MR can be used to directly analyze the causal relationship between exposure and outcome through genetic variables, eradicating the influence of confounding factors and reversing causality on the outcomes with the aid of GWAS data. 17 On the above background, we want to use the novel method of MR to explore the causal relationship between AAM and the risk of IS and ICH from a genetic perspective, and meta-analysis was used to integrate the results of the two different consortiums to obtain a robust causal relationship.

| Study design
Aiming to elucidate the relationship between AAM and the stroke subtypes of IS and ICH, we designed this main MR investigation.
Additionally, several sensitivity analysis techniques were performed to guarantee the reliability of the findings (Figure 1). Data from another GWAS database were then used to validate the causal relationship between AAM and the risk of ICH and IS. Finally, a meta-analysis was conducted to integrate the causal relationships explored by different GWAS data between AAM and ICH, IS, and their subtypes to obtain a robust causal relationship ( Figure 2). The public database was used to obtain all the study's data. 18 All the study subjects provided their informed consent, and the design of the investigations passed an ethical assessment. Therefore, further ethical evaluation of this work is not necessary.

| Sources of exposure and outcome
The GWAS database run by the MRC-IEU consortium served as the initial source of exposure data (AAM) for this investigation in 2018. 19 There were 243,944 individuals overall in this study who were of

| Two-Sample Mendelian randomization
Instrumental variables, namely single-nucleotide polymorphisms (SNPs), are extracted from GWAS data of exposure variables. The significant effect of extracting instrumental variables is defined as the p-value of SNP less than 5 × 10 −8 . On the contrary, it cannot be ignored that the existence of linkage disequilibrium will affect the analysis quality. Therefore, we defined the criteria for removing F I G U R E 1 Three core assumptions of AAM with ICH and IS in this two-sample MR study. AAM, age at menarche; CES, cardioembolic stroke; ICH, intracerebral hemorrhage; IS, ischemic stroke; LAS, large artery stroke; LICH, lobar intracerebral hemorrhage; NLICH, nonlobar intracerebral hemorrhage; SVS, small vessel stroke.

F I G U R E 2
Research design of this study. In the initial stage, the MRC-IEU Consortium was used to obtain GWAS data of AAM as exposure data, and the MEGASTROKE consortium obtained GWAS data of IS and ICH as outcome data to analyze the causal relationship between AAM and IS, ICH, and their subtypes. In the replication phase, the WFG Consortium was used to obtain GWAS data for AAM as exposure data, and the MEGASTROKE consortium obtained GWAS data for IS and ICH as outcome data to analyze the causal relationship between AAM and IS, ICH, and their subtypes. Finally, the causal data obtained by MRC-IEU and WFG consortium were meta-analyzed to obtain more robust causal relationships. CES, cardioembolic stroke; ICH, intracerebral hemorrhage; IS, ischemic stroke; ISG, International Stroke Genetics; LAS, large artery stroke; LICH, lobar intracerebral hemorrhage; NLICH, nonlobar intracerebral hemorrhage; SVS, small vessel stroke; WFG, within-family GWAS.
linkage disequilibrium as r 2 < 0.01 and kb as 10,000. All included SNPs and excluded SNPs are shown in Tables S1 and S2.
Currently, the IVW approach has been chosen as the primary analysis method in this research. 21 This technique can be used to determine whether exposure and outcome have a consistent and reliable causal relationship. But for it to produce consistent findings, all genetic variations must be treated as instrumental variables. In addition, a few more techniques are utilized to support the IVW method and offer more proof for this study's causality assessment.
Even with 50% invalid instrumental variables, the weighted median estimate (WME) approach can still produce trustworthy findings. 22 Although directional pleiotropic bias can be controlled via MR-Egger regression, its assessment effect is not very strong. 23 25 has been adopted in this study.
A series of sensitivity analyses are employed to determine whether the causality assessed by the above methods has heterogeneity, directional pleiotropy, etc. To ascertain whether each instrumental variable is heterogeneous or not, the heterogeneity test (also known as Cochran's Q test) is used. To mitigate the effect of heterogeneity on the results, we must use a random effect model.
The p-value of the global test in MR-PRESSO and the intercept MR-Egger regression was used to determine the directionality of pleiotropy. 23,24 Both the MR-Egger regression approach and the IVW method did not perform the actual evaluation, known as the no measurement error (NOME) assessment, but instead simply took into account the association between SNPs and exposure. Therefore, it must be >90% for I2GX to evaluate the NOME of MR-Egger regression. 26 F-statistics, a tool to identify the presence of weak instrumental factors and assess the efficacy of research, is used to estimate the NOME in the IVW approach. 27 Table S1 is used to display these findings.

| Meta-analysis
Finally, we performed a random effects meta-analysis of the results (OR, 95% UL, 95% LL) obtained from the MRC-IEU and within-family GWAS data to obtain a combined analysis of causality. The metaanalysis was carried out in STATA software version 14.0, and the statistical criterion for meta-analysis results was set at 0.05; similarly, the statistical criterion for the test of heterogeneity was also 0.05.

| Statistical analysis
All the statistical analyses in the MR study were conducted using R studio and R 4.0.3, including the "Mendelian randomization" package and the "TwoSampleMR" package. AAM and stroke subtypes' statistical significance was assessed using the stringent statistical approach of Bonferroni correction. 28

| The AAM was related to the risk of ICH, especially NLICH and was validated by the results of the within-family GWAS consortium data and metaanalysis
In the initial phase of the MR analysis employing MRC-IEU data, AAM and the risk of ICH (OR = 0.48, 95% CI: 0.28-0.80, p = 0.006) were strongly associated, as shown in Table 2  0.58-0.95, p < 0.05) studies investigated a causative association between AAM and the risk of ICH. In the meta-analysis, the causal link between AAM and the risk of ICH and NLICH was also established (as shown in Figure 3). These support the reliability of the findings from our MR study.

| No Causal effect of AAM on the risk of IS and its subtypes was revealed
The initial phase involved using a random effect model to examine whether AAM affected the IS and its subtypes in light of the presence of heterogeneity (Table 3). For IS (Figure 4) by IVW method were discovered using data from the within-family GWAS consortium in the replication phase. Furthermore, a causal link between AAM and IS and its subtypes was not established in the meta-analysis of data from the MRC-IEU collaboration and withinfamily GWAS consortium (Table 3 and Figure 4).

| The results of the sensitivity analysis further attest to the reliability of these findings
As can be shown in

| DISCUSS ION
To our knowledge, this is the first MR study to explore the potential causal relationship between AAM and ICH/IS and its subtypes. A two-stage MR analysis of a sizable sample size of GWAS was conducted to establish a more trustworthy causal link between AAM and ICH, IS, and their subtypes. We then used meta-analysis to combine the causal effects of the two stages. Our MR investigation offers proof that AAM has a direct causal relationship with the chance of developing ICH and its subtype NLICH. However, we were unable to identify a causal connection between AAM, LICH, IS, or the majority of its subtypes. These serve as a reference for comprehending the influence and probable processes of hereditary variables connected to AAM on cerebrovascular illness.
There are a growing number of epidemiological studies examining how gender affects the incidence, prevalence, and subtypes of IS and ICH. 13,14,15,29 Gender differences in ICH were investigated in 2017 in a prospective study with 2212 patients, and results revealed that women were more likely predisposed with LICH than males. 29 Our conclusion that AAM may be causally linked to F I G U R E 3 Causal relationships between AAM and ICH with its subtypes by IVW, WME, MR.RAPS, MR-PRESSO methods in the initial phase, replication phase, and meta-analysis. A causal relationship between AAM and NLICH was found in the IVW, WME, MR-PRESSO, and MR.RAPS methods after the meta-analysis of causality in the initial and replication phases. In addition, the causal relationship between AAM and ICH was also confirmed in the meta-analysis of the IVW, MR-PRESSO, and MR.RAPS methods. AAM, age at menarche; ICH, intracerebral hemorrhage; IVW, inverse-variance-weighted; NLICH, nonlobar intracerebral hemorrhage; SNP(n), SNP number; WFG, withinfamily GWAS; WME, weighted median estimation.

TA B L E 3
Causality of AAM with ICH and IS with its subtypes by IVW method (within-family GWAS consortium).  14,15,30,31 Early AAM may cause early initiation of estrogen exposure, elevated blood levels of estradiol, and potentially compromised vascular elasticity and coagulation. 30,32 In some cases, extending the time that people are exposed to estrogen may lower their risk of developing endometrial cancer. The body, however, is exposed to estrogen too early if the AAM is premature, leading to a significant estrogen accumulation impact. 31 This is comparable to the effects of oral contraceptives 33  AMM may influence obesity and hypertension, in addition to the direct effects previously indicated, to affect the risk of ICH, particularly NLICH. According to one study, late AAM may somewhat lower the risk of hypertension. This observational study showed that 1 year of late AAM would reduce the incidence of hypertension by roughly 3%. 36 They have looked into the basic mechanisms that connect them. Adolescent obesity is frequently present in early AAM patients. 37 Adipose tissue buildup boosts the activity of the enzyme aromatase, which transforms testosterone into estrogen and causes an increase in estrogen levels in the female population. 31 The buildup F I G U R E 4 Causal relationships between AAM and IS with its subtypes by IVW, WME, MR.RAPS, MR-PRESSO methods in the initial phase, replication phase, and meta-analysis. Meta-analysis of causality in the initial and replication phases did not confirm a causal relationship between AAM and IS. AAM: age at menarche. ICH: intracerebral hemorrhage, IVW, inverse-variance-weighted; NLICH, nonlobar intracerebral hemorrhage; SNP(n), SNP number; WFG, within-family GWAS; WME, weighted median estimation.

AAM (MRC-IEU
of estrogen will make hypertension more likely and consequently leading to high risk of ICH. 38 AAM may also prematurely trigger NLICH through heightened sympathetic nervous system activity, the renin-angiotensin-aldosterone system, increased peripheral vascular resistance, or gonadotropin effects on blood pressure. 4 The most frequent risk factor for ICH is hypertension, which frequently causes ICH in the putamen, basal ganglia, and other areas due to rapid changes in blood pressure and the area where the bleeding is occurring is an NLICH area. In light of this, the low risk of late AAMinduced hypertension may contribute in some ways to the low risk of ICH and particularly NLICH.
Additionally, early AAM is linked to high BMI in adults, which has a variety of effects on the body in adults. 39 High prepubertal BMI in females results in significantly higher plasma total estradiol levels, which would also have an estrogenic compounding effect and raise the risk of ICH in adulthood. 40 Furthermore, earlier observational studies support our findings. 9 Obesity has been linked to insulin resistance, renin-angiotensin system activation, increased adipocytederived resistance, and increased leptin effects, but not LICH, which has been suggested to increase the risk of profound ICH but not LICH. [41][42][43] These are all strong arguments in favor of drawing a connection between AAM and ICH, NLICH.
In addition, we found no causal relationship between AAM and IS and its subtypes, unlike previous observational studies. This  NLICH risk in the female population, but further experimental and clinical research is required to establish the underlying mechanisms.

AUTH O R CO NTR I B UTI O N S
The research was planned, and the manuscript's organization was chosen by Le Zhang and Xuelun Zou. References were chosen, and writing assistance was provided by Xuelun Zou, Leiyun Wang, and Sai Wang. Data from GWAS were gathered by Leiyun Wang. These Mendelian randomization study's findings were examined with assistance from Xuelun Zou and Leiyun Wang. The article was revised and polished with help from Le Zhang. The article's submission and all of its writers' contributions were both accepted.

ACK N OWLED G M ENTS
This project was supported by the National Science and Technology Basic Resources Survey (2018FY100900).

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors do not declare any potential conflict of interest relevant to this article.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in IEU-GWAS at https://gwas.mrcieu.ac.uk/datas ets/?trait__icont ains=Body%20mas s%20ind ex.