A Comprehensive Systematic Review and Meta-analysis on the Risk Factors of Stroke in Iranian Population

1Non-Communicable Diseases Research Center, Fasa University of Medical Sciences, Fasa, Iran 2Health Policy Research Center, Institute of Health, Shiraz University of Medical Sciences, Shiraz, Iran 3Clinical Neurology Research Center, Shiraz University of Medical Sciences, Shiraz, Iran 4Department of Biochemistry, School of Medicine, Kerman University of Medical Sciences, Kerman, Iran 5Student Research Committee, School of Medicine, Kerman University of Medical Sciences, Kerman, Iran 6Department of Neurology, Ghaem Hospital, Mashhad University of Medical Sciences, Mashhad, Iran 7Department of Clinical Neurological Science, University Hospital, London Health Science Center, University of Western Ontario, London, Ontario, Canada 8Department of Epidemiology and Biostatistics, Western University, London, Ontario, Canada


Introduction
Stroke, with its main types including ischemic and hemorrhagic, is known as the second leading cause of death and the fourth leading cause of mortality worldwide; thus, approximately 5.7 million deaths are recorded annually due to this medical condition, accounting for 9.7% of all-cause mortality across the world. 1,2 It also occurs in 1-3 cases per thousand people each year in developed countries, 3 and it is growing in Iran, affecting 372 people per 100 000 population. 4 The reports in this respect have also demonstrated that age-standardized stroke mortality have totally decreased in the past two decades. However, the absolute number of individuals experiencing a stroke every year, as well as stroke survivors, related deaths, and the overall global burden of this condition (i.e., the disability-adjusted life years: DALYs lost) have been great and increasing, especially in low-and middle-income countries (LMICs) over the past decades. 5 The LMICs also account for more than 85% of strokes correlated with mortality and morbidity, leading to the highest disease burden in these countries. In addition, geographic variations can add to the epidemiology of the stroke, as the highest incidence and prevalence rates are being reported in the same nations. 6 Determining potential stroke risk factors (SRFs) within each geographical and ethnic region is accordingly of Risk factors of Stroke in Iran utmost importance to adjust stroke prevention programs based on existing needs and facilities. 7,8 Previously, it had been reported that there could be regional variations in the significance of risk factors, which had in turn resulted in worldwide discrepancies observed in incidence rate and subtypes of stroke. 9 Recently, the INTERSTROKE, a large-scale case-control study of SRFs in 32 countries, has reported 10 potentially modifiable risk factors for either ischemic or hemorrhagic stroke, associated with 90% population attributable risks of stroke in each major geographical region of the world as well as in specific age and gender groups. 10,11 The regional variations of SRFs are thus of great importance for risk stratification and lifestyle modification programs in each region. In this way, analysis of SRFs should be conducted in different regional and ethnic groups in order to specify stroke prevention programs and increase their success rates. 2 Iran, as a populous LMIC located in the Middle East with considerably diverse ethnic groups and social classes, largely reflects the core of such countries. In addition, the relatively younger age of stroke in Iran (approximately one decade earlier in Iran than that in Western countries) can significantly increase stroke DALYs. 12 Thus, providing better evidence on SRFs in the Iranian population as an LMIC is particularly relevant. Although several studies have previously addressed the analysis of SRFs in the Iranian population, there are variations in methodologies and sample sizes, restricting their implications to develop a nationwide stroke prevention program based on a risk stratification model. [13][14][15][16][17][18][19] With this purpose in mind and given the unavailability of structured studies, pooling literature data, the present systematic review and metaanalysis was undertaken to determine SRFs in the Iranian population.

Materials and Methods
This meta-analysis was designed and performed according to the guidelines in the preferred reporting items for systematic reviews and meta-analyses (PRISMA).

Search Strategy
As the main strategy, two authors independently searched international and national databases including PubMed, Embase, Web of Science, Scopus, Scientific Information Database (SID), Magiran, and IranMedex for relevant articles published until March 2018. The searches were also conducted using the following MeSH terms and text words as well as their Persian equivalents: ["stroke" OR "brain infarction" OR "ischemic stroke" OR "intracerebral hemorrhagic' OR "intracranial bleed" OR "hemorrhagic stroke" OR "subarachnoid hemorrhage" OR "cerebrovascular disorders" OR "cerebrovascular disease"] AND ["all related risk factors" (such as demographic, anthropometric, vascular, diabetes mellitus, hypertension, cigarette smoking, opioid dependence, and hyperlipidemia factors)] AND ["Iran" OR "Iranian"].
In order to augment the sensitivity of the search strategy, the reference lists of the included articles were manually screened out to detect any additional relevant studies and even Google Scholar as a gray literature was hand-scanned. The searches were restricted to observational studies published in the English or Persian literature without any limitations in their publication date.

Definitions
Most of the investigated SRFs had similar definitions in national and international publications. However, the authors used the definitions of SRFs across selected articles as included here: hypertension (HTN) HTN represented systolic blood pressure (SBP) ≥140 mm Hg and/or diastolic blood pressure (DBP) ≥90 mm Hg or taking anti-hypertensive medications. 20 Diabetes mellitus (DM) was defined as fasting blood glucose (FBG) ≥ 7 mmol/L, 2 h-PG ≥ 11.1 mmol/L or current intake of antidiabetic medicines. 20 Cigarette smoking (CS) represented current or past use of cigarettes or any other type of pipes or smoking instruments, 20 and family history of cardiovascular diseases (CVDs) was described as any prior diagnosis of CVDs in first-degree relatives. 20 Also, body mass index (BMI) was weight divided by height squared (kg/m 2 ) and waist circumference (WC) was measured in a standing position at the point midway between the lowest rib and the iliac crest in both genders 21 . Opioid dependence (OD) was defined according to the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision (DSM-IV-TR) (304.00, opium dependence): continued for at least one year (via inhalation or oral ingestion). 22 Furthermore, total-cholesterol (TC), highdensity lipoprotein-cholesterol (HDL-C), low-density lipoprotein-cholesterol (LDL-C), triglycerides (TG), and FBG levels in serum were further considered.

Inclusion and Exclusion Criteria
Studies meeting the eligibility criteria including observational human studies with cross-sectional, casecontrol, or cohort designs as well as articles investigating associations between related risk factors and stroke were selected for this study. Moreover, studies published until March 2018, without any limitations on their sample size, were found eligible. We used articles reporting sufficient data to extract the number of events for calculating odds ratios (ORs) or weighted mean differences (WMDs) with 95% confidence intervals (CIs) for the related risk factors in patients with first-ever stroke, compared with control groups without any strokes or related CVDs. However, in-vitro studies, animal experiments, case reports, review articles, and seminar abstracts with no full-texts, together with studies that had not obtained the least required scores of the quality assessment process were excluded. Data Extraction and Quality Assessment Two independent authors (RT and BK) extracted the data from the retrieved articles and imported them into standard forms in the Microsoft Excel spreadsheets. Accordingly, the extracted information included first author's name, date published, study setting, participants' demographic characteristics, sample size (either in stroke patients or control group), study design, stroke type, risk factors, and required data to estimate ORs or WMDs and 95% CIs for risk factors if they were categorical or continuous variables, respectively, in stroke patients and control groups. The quality of the given studies was also assessed by two independent authors (RT and BK) according to a comprehensive quality assessment using the Newcastle-Ottawa Scale, comprised of eight and six items for cohort/ case-control and cross-sectional studies, respectively. These items had covered three aspects including selection, comparability, and exposure. The selected study obtaining 4-6 points could be considered as an article of moderate quality and the one with more than 6 points could be assigned as high quality.
In case of disagreements in the domains of data extraction and quality assessment, RT and BK discussed to reach consensus or resolved the disagreement through talks with a third author (A-BH).

Data Synthesis and Statistical Analysis
All the statistical analyses were performed using the Stata (version 12.0) (Stata Corp., College Station, TX) software package. The association between the related risk factors and the risk of stroke was also estimated. To assess the overall pooled effect sizes (ESs), unadjusted ORs or WMDs were respectively employed with 95% CIs for continuous or categorical factors. The Cochran Q test and the I-squared (I 2 ) test were further utilized to measure heterogeneity across the included studies. According to the heterogeneity findings, whenever I 2 ≥ 50% and P < 0.05, a random-effect model using the DerSimonian and Laird method could be applied to combine the studies, a fixedeffect model could be also implemented using the Mantel and Haenszel method provided that I 2 < 50% and P > 0.05. The sensitivity analysis was additionally conducted to examine the impact of each study on the overall pooled ESs after excluding each one using the leave-oneout method. Subgroup analyses were also completed to identify the possible sources of heterogeneity regarding the following moderator variables: type of controls (healthy vs. hospital-based control), type of stroke (all vs. ischemic vs. hemorrhagic stroke), study design (cross-sectional vs. cohort vs. case-control study), and matching status (nonmatching vs. gender or age matching). The Begg's and the Egger's tests were consequently utilized to evaluate the possibility of publication bias. The two-tailed P ˂ 0.05 was also considered as statistically significant.

Characteristics of Selected Studies
In the first step searching the online databases, 573 articles related to SRFs were identified. After removing the duplicates and the irrelevant ones, 15 studies (namely, 10 case-control, three cohort, and two cross-sectional studies) were selected for the current meta-analysis. The flowchart details of the systematic process of study identification and selection are summarized in Figure 1. The sample size in the selected studies included 13 342 participants (10 009 ischemic, 255 hemorrhagic, and 3078 all stroke) ranging from 72 to 5398 cases. The selected articles had been published from 2003 up to March 2018. The characteristics of the observational studies included are presented in Table 1.

Main Outcomes
Forest plots reporting the meta-analysis of the included studies on SRFs consisting of demographic, traditional vascular, and life-style factors are illustrated in Figures  2 and 3.
The pooled findings showed that 13 factors influenced the risk of stroke in the Iranian population. As compared with the control group, the risk of stroke was correlated with the mean age (WMD = 5.49 years; 95% CI, 4.  (Table 2).

Risk factors of Stroke in Iran
Subgroup and Sensitivity Analyses Subgroup analyses were conducted to identify the effects of moderator variables, i.e. type of control, stroke, study design, and matching status on heterogeneity statistics. The reduction of heterogeneity was demonstrated in some strata of the subgroups as shown in Table 2.
In subgroup analyses for gender, a significant association was observed between male gender and increased risk of stroke in studies with all types of stroke (OR = 1.71, 95% CI: 1. 26 16, 40.30), compared with other strata. There was also a significant association in terms of the increased risk of stroke and LDL-C levels (WMD = 41.63 mg/dL; 95%    After excluding several studies from the sensitivity analysis, the results revealed that the relationship between gender, CS, FBG, TC, LDL-C, and low levels of HDL-C and the risk of stroke had changed significantly upon respectively excluding the articles by Afshari et al, 28 Hamzehee-Moghadam et al, 29 Parizadeh et al, 20 Savadi-Oskouei et al, 31 Moshayedi et al, 21 and Maghbooli et al. 24 The association between CS, FBG, and HDL-C levels and the risk of stroke also changed significantly after removing more than one study.

Risk factors of Stroke in Iran
For the remaining factors, there was no significant change between the pre-and post-sensitivity WMDs or ORs. The lower and the higher pooled WMDs or ORs in the sensitivity analyses are presented in Table 3.
Publication Bias and Quality Assessment The Begg's and the Egger's tests showed that publication bias had affected only four factors in the studies on the Iranian population, which included mean age (P Begg's test = 0.17, P Egger's test < 0.01), FBG levels (P Bg = 0.11, P Ee = 0.01), WC (P Bg = 0.11, P Ee = 0.04), and low levels of HDL-C (P Bg = 0.09, P Ee = 0.02). Non-parametric methods (i.e., Duval and Tweedie' Trim and Fill method) were further utilized to compute the findings of the removed observational studies for two factors; however, the pooled ESs for the related factors did not change significantly after using the method.
Likewise, quality assessment using the Newcastle-Ottawa Scale established that 86.67% of the selected primary studies obtained at least two stars for selection, 100% of these articles were assigned with at least one star for comparability, and 93.34% of the primary studies included were given at least two stars for exposure category. The detailed quality assessment results are summarized in Table 4.

Discussion
Stroke risk stratification depends on various regional and ethnic factors that are taken into account as integral parts of resource assessment as well as bases for developing targeted stroke prevention programs. Currently, there is lack of related literature on the Iranian population regarding the risk factors of either ischemic or hemorrhagic strokes. Thus, the present comprehensive meta-analysis of observational studies on the Iranian population as an example of other LMICs was conducted to explore SRFs among Iranians. Overall, the included observational studies covered demographics (age and gender), traditional VRFs (HTN, DM, dyslipidemia, and family history of CVDs), and lifestyle factors (CS, OD, WC, and BMI). Notably, the majority of these predisposing factors were preventable. This study also provided essential information and a global insight into all modifiable and non-modifiable SRFs in the Iranian population and added to the information of all previous epidemiological studies in this field. 18,[32][33][34][35][36] The meta-analysis findings also revealed that age, but not gender, was associated with the risk of stroke. HTN, SBP and DBP, DM, and FBG were accordingly correlated with increased risk of stroke among the traditional VRFs. Other possible risk factors including CS, OD, and WC were further confirmed as independent stroke determinants in addition to high circulating levels of TG, LDL-C, TC, and low HDL-C.

Demographics Factors
In this regard, a significant association was observed for mean age, whereas gender did not influence the risk of stroke. Stroke distribution was also correlated with age and gender. 37 The highlighted impact of aging was in line with a previous systematic study from Iran, 17 underlining a slight predominance for females in older age in contrast to a small prevalence for males in younger age. 17 It should be noted that a large cohort study from northwestern Iran had reported a similar incidence in women and men. 38 Nevertheless, in subgroup analyses, male gender was found to be significantly associated with stroke. However, these findings should be cautiously explored and further studies are needed to evaluate the rate of age-specific stroke in Iran because the number of articles in age-matched groups is not sufficient (only one study was age-matched for gender while five studies were not). Considering the fact that several studies from Middle Eastern countries such as Iraq, * For case-control studies, there are four items within the selection: 1) indicates cases adequately defined; 2) representativeness of the series of cases; 3) community controls; 4) controls have a group of stroke-free population), one item (with two sub-items A and B) within the comparability: A study controls for important factor and B study controls for additional factor(s), and three items within the exposure: 1) ascertainment of exposure by records; 2) same method of ascertainment used for cases and controls; and 3) same non-response rate for both groups. For cohort studies are four items within the selection: 1) shows exposed cohort truly representative in the studied community; 2) non-exposed cohort drawn from the same community as the exposed cohort; 3) ascertainment of exposure by records; 4) outcome of interest was not present at start of study, one item (with two sub-items A and B) within the comparability: A cohorts comparable on the basis of important factor and B cohorts comparable for any additional factor, and three items within the outcome: 1) assessment of outcome by record linkage or independent blinding; 2), follow-up long enough for outcomes to occur; and 3) follow up adequacy of cohorts. For cross-sectional studies are three items within the selection: 1) shows samples truly representative; 2) same characteristics between respondents and non-respondents and enough response rate; 3) exposure ascertainment by validated measurement tool, one item (with two sub-items A and B) within the comparability: A study controls for important factor and B study controls for additional factor, and two items within the outcome: 1) assessment of outcome by record linkage or independent blinding; 2) the statistical test is clearly described and appropriately presented.
Saudi Arabia, and Palestine have shown significant female predominance, 39 determining gender-related predisposing factors requires in-depth investigations to shed light on the causes of this discrepancy between Iran and other nations in the Middle East.

Traditional VRFs
To examine the metabolic profile of the patients, the crucial role of diet should be taken into account. Diet can be thus involved in numerous interconnected mechanisms underlying stroke and related to balance in blood pressure, blood lipids, thrombosis and coagulation, oxidative stress, systemic inflammation (SI), endothelial function, glucose/ insulin homeostasis, gut microbiome, and body weight. 40 HTN, DM, and dyslipidemia (including high levels of TG, TC, LDL-C, and low levels of HDL-C), as three comorbidities that could very often co-exist in patients, were thus regarded as SRFs in this study. It seems that these populations are susceptible to higher TG and lower HDL-C levels. Nevertheless, after additional analysis based on gender and age matching, the findings revealed that TC and LDL-C levels were also significantly associated as SRFs. A meta-analysis by Wang et al further concluded that a higher level of HDL-C was correlated with a lower risk of hemorrhagic stroke. 41 Of note, overall dyslipidemia is one of the important risk factors for coronary artery diseases (CADs) and stroke, which can be affected and changed by lifestyle, diet, and medications. 42 In this respect, recent studies have suggested that inadequate physical activity is correlated with decreased HDL-C and increased TG levels. 43,44 Also, HTN and DM have been repeatedly among the strongest SRFs. [45][46][47] DM also augments the risk of stroke by four times. 48 The INTERSTROKE study had similarly listed HTN and DM among the ten factors accounting for over 90% of SRFs, 10 which is consistent with the results of a nationwide German study focusing on stroke onset below 55 years of age. In that study, HTN, dyslipidemia, and DM were mentioned among the top eight risk factors and the similarity with attributed risks in the older population was emphasized. 49 A meta-analysis of the Chinese population correspondingly underlined the significance of HTN and DM as leading SRFs while hyperlipidemia was mildly correlated. 47 A review of the Middle East studies on stroke patients further revealed that the prevalence rate of HTN was 62.1% followed by dyslipidemia with 36.8% and DM with 33.1%. 18 Although many risk factors are common across demographic groups, a universal weight cannot be applied due to variations in importance and prevalence of the given factors. 10,47 Thus, screening and prevention should be designated based on local priorities. However, a holistic approach is crucial as risk factors are often intertwined. Furthermore, LMICs should assess what part of resources, i.e., human resources, healthcare facilities, diagnostic and laboratory Risk factors of Stroke in Iran services, medications, and access to transportation, is more important to be improved with regard to stroke care. 50 In this study, family history of CVDs was not significantly correlated with stroke in the Iranian population. This might be due to the limitation in the included studies that had been considered suitable for evaluating the relationship between family history and stroke. It should be noted that family history is a well-established SRF 51 especially at a younger age. 33,52 In this respect, a national cohort study in the United States showed that a positive family history of stroke was associated with physiological and behavioral factors such as HTN and CS, implying that family history could be also partially attributed to modifiable factors. 53 Unfortunately, the presence of such an association could not be assessed based on the existing data in this study. Accordingly, encouraging family members of stroke patients to pursue a stricter healthy lifestyle and to adhere to routine monitoring seems very reasonable principally in LMICs, wherein prevention should be highly prioritized due to limited resources.

Lifestyle Factors
Extensive research has revealed the role of CS in stroke. 10,46,49 A meta-analysis covering studies for almost 50 years on approximately four million individuals also confirmed CS as an independent SRF with a similar risk in both men and women, confirming that both genders could benefit from CS cessation. 54 Moreover, non-smokers, especially women, could be subject to a greater risk of stroke, ischemic heart disease, and chronic obstructive pulmonary disease (COPD) subsequent to second-hand smoke exposure. 55 Accordingly, Austin et al 56 suggested that the link between COPD and stroke was not only correlated with shared traditional risk factors including aging and smoking, but also the COPD-related SI and oxidative stress that could aggravate cerebral vascular dysfunction and platelet hyperactivity.
Regarding dependence, most studies have focused on alcohol abuse, while it is less prevalent in Iran and very likely to be underreported in surveys due to its illegality. As a result, no assumptions could be offered on alcohol. On the other hand, OD is a more common public health concern in Iran. It is noteworthy that substance abuse is illegal in Iran and it is socially stigmatized. Therefore, the possibility of underreporting cannot be ignored. The findings revealed that total OD (regardless of its type) was a SRF. Several studies on the Iranian population have also confirmed OD as a risk factor for stroke and CVDs, 13,57-59 possibly by enhancing plasma fibrinogen activity and chance of atherosclerosis. 59 The impact of obesity and BMI on development of stroke has been similarly supported in large-scale studies. 49 A meta-analysis by Guo et al 60 concluded that young adults with overweight or obesity were prone to an elevated risk of stroke probably independent of other cardioVRFs. A meta-analysis of prospective studies with two million participants by Strazzullo et al further found that overweight and obesity were linked with increased risk of ischemic stroke, at least in part, independently from lifestyle, age, and other VRFs. 61 The findings of the present study were in favor of the adverse impact of WC on stroke but not BMI. Interestingly, it has been demonstrated that by excluding the effect of BMI, abdominal obesity per se, in terms of WC and waist-to-height ratio, turns out to be a common risk factor in both genders. 62 Further studies are thus needed to assess the role of WC as a SRF in comparison with other anthropometric factors such as BMI. Unfortunately, the infrastructure to facilitate physical activity is rather underdeveloped in LMICs like Iran, even though the costs can outweigh the damage from diseases related to excess weight and sedentary life.

Strengths and Limitations
The present study was the first attempt as the most comprehensive systematic review and meta-analysis to determine SRF in the Iranian population. However, there were several limitations. Firstly, the number of studies evaluated for risk factors affecting stroke in the Iranian population was small. Secondly, some articles were excluded due to their methodological limitations. Thirdly, there was heterogeneity among the included primary studies regarding their outcomes. Fourthly, the subgroup analysis did not adjust for confounding variables. Therefore, the results are possibly affected by confounding factors and should be interpreted with caution. Accordingly, the relationship between each risk factor and stroke was estimated based on a comparison of the distribution of risk factors between case and control groups (univariate analysis) because the studies with the adjusted findings were not adequate.

Conclusion
Apparently, stroke seems to be strongly associated with lifestyle and environmental age-related factors in Iran. Among the risk factors mentioned, HTN, SBP and DBP, DM, and FBG were related to increased risk of stroke among traditional VRFs. Other possible risk factors including CS, OD, and WC were further established as independent stroke determinants as well as high circulating levels of TG, LDL-C, TC, and low HDL-C. Considering recent lifestyle alterations in the Iranian society, stroke continues to take its toll on a more extensive scale, especially with a shift towards the younger population if prompt public awareness and strict preventive policies remain neglected.