Hyperhomocysteinemia and Ischemic Stroke: A Potential Dose-Response Association—A Systematic Review and Meta-analysis

Background and Purpose  Previous studies suggest an association between increased homocysteine (Hcy) and risk of ischemic stroke. Yet, it remains unknown whether a dose-response association exists between Hcy levels and risk of ischemic stroke. Methods  Systematic literature searches were performed in PubMed, Embase, Scopus, and Web of Science. Inclusion criteria were studies investigating ischemic stroke risk in an adult population with measured Hcy levels. We computed odds ratios (ORs) for a 5 µmol/L increase in Hcy levels using a random effects meta-analysis. Results  In total, 108 studies met the inclusion criteria of which 22 were rated as high-quality studies, and 20 studies included a dose-response analysis. Hcy levels were analyzed either as a continuous or categorical variable. The majority of the studies found an increased risk of ischemic stroke when comparing the highest-to-lowest Hcy strata. A graded association was observed over the Hcy strata, indicating a dose-response association, with the most apparent effect when Hcy levels exceeded approximately 15 µmol/L. No studies explored a potential nonlinear association between Hcy levels and ischemic stroke. Six studies were included in a meta-analysis, showing an OR of 1.43 (95% confidence interval [CI]: 1.28–1.61) per 5 µmol/L increase in Hcy levels. Conclusion  This review and meta-analysis indicate a dose-response association between Hcy levels and ischemic stroke. An evident increase in effect measures was observed when Hcy levels exceeded 15 µmol/L, indicating a nonlinear association between ischemic stroke and Hcy levels. This nonlinear association warrants further study. This study is registered with clinical trial ( https://www.crd.york.ac.uk/prospero/ ; unique identifier: CRD42019130371).


Introduction
The comprehension that elevated homocysteine (Hcy) in plasma might predispose to arterial or venous throm-boembolism emerged more than 40 years ago, when patients with homocysteinuria were observed to have a high risk of early vascular disease. 1 This led to extensive research regarding the role of Hcy in cardiovascular disease (CVD) and whether elevated Hcy is a modifiable risk factor.
Elevation of Hcy levels may be caused by several factors, including deficiency of vitamin B6, folate, and/or vitamin B12, due to insufficient intake or absorption, renal insufficiency, several drugs, 2 lifestyle factors, such as smoking and alcohol intake, or genetic factors. 3 Among fasting individuals, normal Hcy levels commonly range from 5 to 15 µmol/L. 4 Animal studies have demonstrated that elevated Hcy levels leads to complex changes within the blood vessel wall, with increased oxidative stress, proinflammatory effects, and endothelial dysfunction, indicating that an association between increased Hcy and CVD is biologically plausible. [5][6][7] Several studies have investigated the potential association between elevated Hcy concentration and risk of CVD, including stroke, but results are inconsistent. [8][9][10] The Norwegian Vitamin Trial indicated that treatment with folic acid and vitamin B combination therapy effectively lowered Hcy levels by 28%, but no effect was found on the incidence of ischemic stroke. 11 The Vitamin Intervention for Stroke Prevention trail demonstrated similar results, with no significant reduction in the risk of stroke among patients treated with B-vitamin combinations. 12 In contrast, the China Stroke Primary Prevention Trial reported a 24% risk reduction for ischemic stroke in the group that received folic acid treatment. 13 While, a Cochrane review from 2017 found a small reduction in risk of stroke in patients treated with B12, folate and B6 vitamins compared with patients receiving placebo. 14 Systematic reviews of observational studies have reported a dose-response related association between Hcy levels and the risk of stroke, independent of other cardiovascular risk factors. 15 The most recent literature investigating the doseresponse relationship between stroke and Hcy levels was performed in 2002; however, this review did not differentiate between ischemic stroke and hemorrhagic stroke. 15 Therefore, we performed a systematic review and metaanalysis to assess the dose-response association between Hcy levels and the risk of ischemic stroke.

Methods
The present systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 16 The protocol was published in the Prospero database (ID: CRD42019130371).

Literature Search
Searches in PubMed, Embase, Scopus, and Web of Science were performed on May 8, 2020. Where possible, filters were applied to remove nonhuman studies, and non-English language publications. No limits were set with regard to publication year. Free-text and the Medical Subject Headings (MeSH) terms or Emtree-preferred terms were used. Search combinations included terms related to the following search categories: Hcy, thromboembolism, biomarker, and adult human population. The complete search combinations used in PubMed is provided hereinafter. Similar search combinations were used for searches in the remaining three databases.

PubMed
Search (((((((("Homocysteine" [Mesh] OR "Hyperhomocysteinemia" [Mesh] OR homocyst Ã OR hyperhomocyst Ã ))) AND (("Embolism and Thrombosis" [Mesh] OR "Myocardial Infarction" [Mesh] OR "Acute Coronary Syndrome" [Mesh] OR "Brain Ischemia" [Mesh] OR "Stroke" [Mesh] OR "Deep vein thrombosis" OR "pulmonary embolism" OR "lung embolism" OR thrombosis OR embolism OR thromboembolism OR stroke OR "acute stroke" OR "brain infarction" OR "cerebral infarction" OR "brain ischemia" OR "cerebral ischemia" OR "ischemic stroke" OR "intracranial embolism" OR "intracranial thrombosis" OR apoplexy OR "cerebrovascular accident" OR "cerebral stroke" OR "myocardial infarction" OR "myocardial infarct" OR "heart infarct" OR "heart infarction" OR "acute coronary syndrome" OR "acute myocardial infarction" OR "brain embolism" OR "cardiovascular stroke" OR "heart attack" OR "acute myocardial infarct" OR "acute heart infarction"))) AND (("Biomarkers" [Mesh] OR "Blood"[Mesh] OR blood OR serum OR plasma OR biomarker OR "biological marker" OR "blood level" OR "blood levels"))) AND ( Initially, all thromboembolic events were included as outcome, as seen in our search combinations, resulting in a large number of eligible articles (►Fig. 1). Therefore, our inclusion/exclusion criteria were revised after screening of abstracts to include only articles with ischemic stroke as outcome for this review. This choice was based on results found in the 2017 Cochrane review 14 which showed a potential association between Hcy and ischemic stroke in randomized trials of B-vitamin treatment.
Our inclusion criteria were studies investigating ischemic stroke events in patients with measured plasma Hcy providing original data on adult human populations. Both interventional and observational studies were included, including randomized trials, cohort, cross-sectional, and case-control studies. The accepted endpoint was acute ischemic stroke, including all subtypes. The exclusion criteria were as follows: nonoriginal literature, reviews, meta-analyses, guidelines, case studies, conference abstracts, and letters/editorials/ comments without original data; missing information on Hcy concentration or studies including hyper-Hcy as a binary variable; endpoint of transient cerebral ischemia and arteriosclerotic lesions without sign of thrombosis, and silent brain infarction; and language other than English.
First, 100 abstracts were randomly selected and screened independently by the three authors for either exclusion or inclusion to full-text screening. Any disagreement was solved by consensus. Screening of the remaining abstracts was performed by MH. Similarly, 50 randomly selected articles proceeding to full text screening were read in full by all authors, and any disagreement was solved by consensus. Remaining papers were screened by MH for inclusion or exclusion, and in case of doubt, all three authors discussed the study in question.

Data Extraction and Quality Assessment
Data extraction from the included articles was performed by M.H. and verified by A.M.H. and J.F.H.A. Study quality was assessed by all three authors using the Study Quality Assessment Tools for Observational Cohort and Cross-Sectional studies and for Case-Control studies, National Heart, Lung, and Blood Institute, National Institute of Health, the United States. 17 Each study was rated good, fair or poor according to the estimated risk of bias. Disagreement between authors was solved by consensus.

Data Synthesis
We performed a meta-analysis of odds ratios (ORs) in which Hcy was included as a continuous linear variable. [18][19][20][21][22][23] Hcy increments was standardized to 5 µmol/L. A random effects model was used to obtain a summary OR as a measure of the relative risk. Forest plots and funnel plots were used to visualize the data.

Results
In total, 108 original articles were included in the review. Of these, 22 articles rated good, 47 fair and 39 poor. Screening of abstracts and inclusion of articles are shown in ►Fig. 1. Articles rated good were grouped according to study design of which 18 were case-control studies and 4 were observational cohort studies, as presented in ►Tables 1-3. Only articles rated good are reported on and discussed in this

Studies with Effect Measures Based on Homocysteine Strata
Eleven studies stratified Hcy levels into several strata, estimating the risk of ischemic stroke in the lowest stratum compared with the higher strata (►Table 2). 18,19,[21][22][23][105][106][107][108]111,112 Nine out of 11 studies found an increased risk of ischemic stroke when comparing patients in the highest versus lowest Hcy level strata. 18,19,[21][22][23][105][106][107][108] Two studies found no association between risk of ischemic stroke and Hcy level 111,112 and one study reported an association for cardioembolic stroke, but not with large-vessel atherosclerotic stroke. 107 Two studies estimated effect measures using a dichotomous Hcy; one study found an association, 113 whereas one did not. 9

Studies with Effect Measures Based Homocysteine Increments
Ten studies included Hcy as a continuous variable and presented effect measures based on various increments of Hcy (►Table 2). [18][19][20][21][22][23]105,108,109,114 All studies found an association between increasing Hcy levels and odds of ischemic stroke, despite variations in the Hcy level increments that were employed.
Six studies included analysis of ischemic stroke subclasses. 23,105-109 Among these, four studies showed an association between small-vessel disease and/or large-vessel disease, 23,106,108,109 and three studies demonstrated an association with cardioembolic stroke. 105,107,109 Two studies did not include effect measures illustrating the dose-response association between ischemic stroke and Hcy levels. 110,115 Overall, in studies comparing Hcy strata, effect measures were clearly elevated when Hcy level reached 15 µmol/L and above (►Table 2).

Cohort Studies
Four cohort studies were rated as good, (►Table 3). [116][117][118][119] Mean follow-up time ranged from 18 months to 9.9 years. Two studies included patients with CVD, with subanalyses for ischemic stroke. 117, 119 Zee et al did not find an association when comparing quintiles of Hcy levels in the population, 119 whereas Petri et al found increased risk of ischemic stroke  with increasing Hcy levels. 117 Bostom et al included elderly patients with stroke and found an association for both nonhemorrhagic stroke and atherothrombotic brain infarction, when comparing the highest quartile of Hcy to the lowest quartile. 118 Shi et al investigated recurrence of ischemic stroke as outcome, with enrolment at admission for first ever stroke. Blood sampling was performed at 3 days and 3 months after enrollment. An association between the risk of recurrent ischemic stroke (within 12-36 months) and Hcy levels was found, when comparing the highest and lowest Hcy quartiles in blood samples performed 3 months after the enrollment. 116

Meta-analysis of Dose-Response Association
Eleven studies included Hcy as a continuous variable of which six were included in the meta-analysis. [18][19][20][21][22][23] We normalized ORs to increments of 5 µmol/L in Hcy. 21,22 The remaining five studies performed log transformation of Hcy levels prior to statistical analysis and were therefore not included in the meta-analysis (►Table 2). 106,108,109,114,117 The studies included in the meta-analysis reported similar results and included similar numbers of patients. All adjusted for age, sex, main CVD risk factors (diabetes, hypertension, hypercholesterolemia, smoking, and body mass index [BMI]), with an exception of Eikelboom et al that did not adjust for BMI. Eikelboom et al was the only study that adjusted for renal insufficiency (►Tables 1 and 3). 23 We performed a random effect analysis, resulting in an OR of 1.43 (95% confidence interval [CI]: 1.28-1.61; . A funnel plot for the metaanalysis is provided in ►Fig. 3, as the resulting funnel plot was severely asymmetric. Results reported in studies rated fair and further supported the observations reported in studies rated good (►Table 5).

Discussion
The present study indicates a dose-response association between Hcy levels and the risk of ischemic stroke. It was apparent that risk estimates reported in studies were notably higher when reaching Hcy levels above 15 µmol/L, indicating a possible nonlinear association between Hcy and ischemic stroke. Both studies rated good and fair supported this observation.
Studies have shown that Hcy levels increase in patients within 1-week poststroke. 27,39,45 This could explain some of the differences observed between cases and controls in the case-control studies where blood sampling was performed in cases during hospitalization for stroke. 27,74 Moreover, Hcy levels have been shown to increase in critically ill patients. 120 In this review, no difference in means was observed when comparing Hcy levels in blood samples obtained during the acute phases 23,107,108,114,116 and convalescence phases of ischemic stroke 20,22,106,109 ; but based on the aforementioned previous studies, timing of blood sampling should be considered when evaluating Hcy as an exposure.
Additionally, several studies performed a logarithmic transformation of Hcy levels prior to statistical analysis, indicating that they initially observed a nonlinear association with ischemic stroke risk, but without exploring this further.
To assess publication bias, we performed a funnel plot of our meta-analysis (►Fig. 3). Generally, at least 10 studies with varying sample sizes should be included for the test to have the power to distinguish chance from true asymmetry. Even though only six studies were included in the metaanalysis, we included the funnel plot as it was severely asymmetric. This could indicate publication bias, but it  may also result from an overestimation of the effect of Hcy because of small study populations. 123 Several large randomized controlled trials of Hcy-lowering treatment report mainly no effect on risk of CVD. [12][13][14]124 These studies were excluded in the inclusion process of this review, as Hcy levels were not reported in the studies.

Strengths and Limitations
One of the strengths of the present systematic review was the strict requirement of outcome definition required for inclusion. Numerous studies investigating the relationship between Hcy and stroke do not differentiate between hemorrhagic or ischemic stroke in their outcome variable which could lead to a reduction of estimates toward the null. 10,125,126 Second, this review only included multivariable adjusted risk estimates.
Some limitations have to be considered as well. First, statistical analyses were not standardized across studies and a meta-analysis of the dose-response relationship was only based on six studies. Second, the strategies for choosing control groups varied between hospital-and community-based controls; this could lead some studies to include a healthier control group compared with others. Third, we were not able to take into account differences in laboratory methods for measuring Hcy levels, and in turn, differences in reference intervals between studies. Forth, Hcy levels are influenced by a vast array of environmental and genetic factors, but most studies only adjusted effect measures for the main known cardiovascular risk factors, age, sex, hypertension, diabetes, cholesterol, smoking status, and BMI. However, six studies adjusted for renal function 23,106,109,113,115,117 and additional nutritional factors, such as folate and vitamin B12 levels, were adjusted in six studies. 21-23,106,108, 109 We were not able to further assess the possible differential impact of the etiology of elevated Hcy levels and the association with ischemic stroke. Furthermore, lipid-lowering medication, such as fibrates, commonly prescribed for patients in risk of CVD, might influence the Hcy levels. 127 Any potential influence of lipid-lowering drugs on the association between Hcy and ischemic stroke was not assessed.

Conclusion
The present review and meta-analysis indicate that a nonlinear association could exist between Hcy levels and the risk of ischemic stroke. This implies that the risk of ischemic stroke increases when Hcy exceeds a certain level. Identifying this cut-off point would be of strong clinical interest, as it could help distinguish which patients could benefit of Hcylowering treatment.

Funding
This study was supported by grants from Redordati Rare Diseases. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or manuscript preparation.

Conflict of Interest
M.H. has no conflicts of interest. A.M.H. has no conflicts of interest regarding the present paper but has the following general conflicts for interest: has received speaker's fees from CSL Behring, Bayer, Boehringer-Ingelheim, Bristol-Myers Squibb, and Astellas, and unrestricted research support from Octapharma and CSL Behring. J.F.H.A. has no conflicts of interest regarding the present paper but has the following conflicts of interest outside the present work: received a speaker's fee within the last 36 months from Siemens Healthineers, Denmark, and a speaker's fee within the last 36 months from Teva Denmark A/S.