Independent ischemic stroke risk factors in older Americans: a systematic review

Serologic or diagnostic risk factors

In recent years, investigating serologic biomarkers has positively impacted our understanding of the pathophysiology of ischemic stroke. We identified nine studies that investigated either serological or diagnostic risk factors for incident ischemic stroke (Table 1). The results were mixed. One study [11] assessed the relationship between advanced glycation endproducts (specifically carboxymethyl-lysine [CML]), and incident ischemic stroke. An adjusted cubic spline analysis was consistent with a linear relationship between CML and ischemic stroke. A multivariable hazard ratio (mHR) of 1.14 (95% CI, 1.01-1.28; p= 0.027) per 0.99 pmol/l of serum CML was reported. The authors concluded that CML is a marker of the vascular effects of advanced glycation end products, thus predisposing an individual to stroke.

Using a similar study design, associations between an established biomarker of inflammation, serum C-reactive protein levels (CRP) and carotid intima-media thickness (CIMT), a marker of atherosclerosis, were associated with incidence of ischemic stroke. Associations were examined after stratification by quartile of serum CRP levels [12]. Using the first quartile of CRP as the referent, the mHR for ischemic stroke by second, third, and fourth quartiles were 1.19 (95% CI, 0.92-1.53), 1.05 (95% CI, 0.81-1.37), and 1.60 (95% CI, 1.23-2.08), respectively. While CRP levels in the fourth, or or highest, quartile were an independent risk factor for ischemic stroke, the magnitude of the risk was attenuated when CIMT was added as a covariate to the model. The authors concluded that CRP is a risk factor for ischemic stroke when controlling for atherosclerosis markers, such as CIMT.

Serum total homocysteine (tHcy) [13], a known cardiovascular risk factor, was investigated in relation to ischemic stroke. A tHcy ≥ 15 umol/L was a significant and independent predictor of ischemic stroke risk, with a mHR of 2.01 (95% CI, 1.00-4.05).

Given interest in the role of estrogens and androgens in cardiovascular and cerebrovascular disease, two studies [14,15] investigated endogenous blood sex hormone or related protein levels and risk of ischemic stroke, yielding similar results. One study [14] investigated men of Japanese ancestry who were living in the United States, observing that serum levels of albumin, sex hormone binding globulin, and testosterone were not associated with risk of ischemic stroke. However, estradiol was associated with an over 2-fold increased risk for ischemic stroke, after adjusting for serum testosterone, hypertension, and diabetes (mHR 2.2, 95% CI, 1.5-3.4; p<0.001). The second study [15] investigated serum levels of testosterone and dihydrotestosterone (DHT) in association with ischemic stroke in men. Testosterone levels were again not associated with ischemic stroke risk. However, DHT levels that were either higher or lower than the reference value (50-75 ng/dL) were associated with ischemic stroke. Specifically, low (<25 ng/dL) DHT levels were associated with a mHR of 2.20 (95% CI, 1.18-4.12) and high (>100 ng/dL) DHT levels were associated with mHR of 1.89 (95% CI, 0.55-6.54). This overall U-shaped association of DHT with ischemic stroke risk, adjusting for covariates (e.g., age; gender; ever smoked), was significant (p=0.006).

As obesity is a known risk factor for stroke, adipose tissue-related metabolites and hormones were examined in relation to ischemic stroke in three studies [16–18]. One study [16] assessed serum free fatty acid (FFA) levels and found no association between FFA levels and incidence of ischemic stroke. A second case-control study [17] enrolled post-menopausal women who had experienced an ischemic stroke and assessed the association with three plasma-based adipokines: adiponectin, leptin, and resistin. After adjusting for body mass index (BMI) and other ischemic stroke risk factors, resistin was the only adipokine found to predict the odds of ischemic stroke (mOR, 1.61; 95% CI, 1.22-2.13; p<0.001). The third study [18] of both men and women investigated the relationship between leptin levels and risk of ischemic stroke and found no association. It should be noted that sex hormone metabolism changes in women and men as they age. Testosterone levels decline in men, however, most notably, in postmenopausal women, with cessation of ovarian function, the primary site of estrogen synthesis is adipose tissue [19–21]. Androstenedione is aromatized to estrone, which while having low activity in the presence of estradiol, becomes primary when estradiol production is diminished [19–21].

There was one study that evaluated a non-serological biomarker [22]. This study investigated subjects who had received baseline M-Mode echocardiography to assess the relationship between echocardiographic measures and risk of ischemic stroke. After adjusting for conventional risk factors, left ventricular (LV) mass was associated with an increased HR for ischemic stroke (mHR=3.36, 95% CI, 1.96-5.74). However, no specific pattern of LV hypertrophy conferred higher risk for ischemic stroke. Additionally, left atrial dimensions and the presence of mitral annulus calcification were not associated with ischemic stroke risk.

Conventional

Risk factors (Table 2) for ischemic stroke that are deemed ‘conventional’, as identified in the Framingham Study [5], were evaluated. Nine studies were identified that measured these conventional risk factors in adults >65 years. These risk factors included alcohol intake [23,24], type 2 diabetes [25], myocardial infarction [26,27], blood pressure/pulse pressure [28,29], and physical functioning [30,31].

Genetics

Table 3 describes results from studies that investigated presence of the APOEε2, ε3, and ε4 alleles in association with risk of ischemic stroke [32]. The presence of APOEε2 allele was associated with a lower incidence of stroke in individuals 71-79 years old. No protective effect was seen in those over 80 years old, nor was an association found between APOEε3 or APOEε4 and ischemic stroke in any age group. The APOEε2 allele has been shown to be protective for later-life outcomes such as Alzheimer’s disease and related dementias as well, which are also associated with vascular risk factors, including stroke. As aforementioned, APOEε4 status was an effect modifier in the analyses associating alcohol intake with stroke risk.

Two studies [33,34] analyzed single nucleotide polymorphisms (SNPs) in the phosphodiesterase 4D and AC008828-2 genes and risk of ischemic stroke. Both the phosphodiesterase 4D and AC008828-2 genes have been studied extensively with research showing their positive association with ischemic stroke risk [35]. Among women [33], 4 phosphodiesterase 4D SNPs were associated with both higher and lower ischemic stroke risk, after stratification by hypertension status. Notably, among women without hypertension, the SNPs rs9 (HR=0.48; 95% CI, 0.26-0.91), rs42 (HR=1.73; 95% CI, 1.10-2.70), rs219 (HR=1.73; 95% CI, 1.13-2.64), and rs220 (HR=1.56; 95% CI, 1.05-2.32) of the phosphodiesterase 4D gene were associated with ischemic stroke. The second study [34] evaluated the relationship between SNPs in the coagulation factor XII gene (F12), peak thrombin generation (pTG) level, and ischemic stroke risk. pTG was associated with higher ischemic stroke risk (HR=1.09; 95% CI, 1.01-1.17; p=.03) and 2 F12 SNPs were associated with greater pTG. However, these latter SNPs were not independently associated with an increased incidence of ischemic stroke.

Psychosocial

Cognition

While it is well-known that stroke contributes to vascular cognitive impairments and disorders, it is less clear whether cognitive function predicts stroke. In a study [39] measuring cognition using the Mini Mental Status Exam (MMSE), a measure of global cognitive function, those scoring > 21 (range 0-30), denoting better cognitive function, were half as likely to have an ischemic stroke at seven year follow-up (HR=0.49; 95% CI, 0.35-0.69) compared to those scoring < 20. All participants had no history of stroke at baseline and were followed for seven years, with follow-up visits at 2, 5 and 7 years. Using MMSE has a continuous score, global cognitive function predicted ischemic stroke, irrespective of baseline age group (65-74; 75-84; >85 years). For every 1-point increase in MMSE there was a 5% reduction in ischemic stroke risk across all age groups (Table 5). While crude, the MMSE has been used in numerous epidemiological studies of elderly and has high construct validity [40],

In a similar study of participants > 65 years old [41], the relationship between cognitive function and incident ischemic stroke was evaluated. Another test of global cognitive function, the modified Mini Mental State Examination (3MS), was used. 3MS score was dichotomized to define those with high cognitive function (scores > 96 of 100) and low cognitive function (scores <96 of 100). Those with high cognitive function experienced a lower risk of ischemic stroke (OR=0.68; 95% CI, 0.52-0.88; p=.005). A secondary analysis was conducted after stratification by baseline quartile of stroke risk factor burden defined via systolic blood pressure, blood glucose level, presence of Type-2 diabetes, walking the previous day, and history of heart disease.

The association between cognitive function and ischemic stroke was evaluated in older Americans, >71 years [42]. The majority of the sample was female (66%) and White (95%), with a sub-sample of African Americans (5%). Seventy-nine percent of the sample scored in the normal range (7-9 correct answers) on the Pfeiffer’s Short Portable Mental Status Questionnaire (SPMSQ), 18% showed moderate impairment (4-6 correct answers) and 3% showed severe impairment (0-3 correct answers). Stroke incidence was lowest among those with normal cognitive functioning (12.1 per 1,000 persons), intermediate with moderate impairment (16.3 per 1,000 person), and highest in those with severe impairment (30.9 per 1,000 persons). Results were similar with control for age and sex.

Antibiotic use as a risk factor

Infectious diseases, both bacterial and viral, are associated with higher cardiovascular and cerebrovascular disease risk. One study meeting our criteria [43] tested the hypothesis that short-term antibiotic use targeted at Chlamydia pneumoniae, decreases ischemic stroke risk. C. pneumoniae is associated with lung infections via airborne routes of transmission. Using a healthcare database sample of individuals over the age of 65 with an insurance claim for ischemic stroke, individuals with and without a pharmacy claim for an antibiotic prescription were compared. In a study of almost 200,000 claims, incident stroke was experienced by 7335 individuals over a mean follow-up period of 1855 ± 554 days (median=1159). In univariate analyses, general antibiotic use was not found to be associated with stroke risk. However, 2 of the 6 antibiotics prescribed were associated with a 20-30% higher stroke risk (cephalosporins (HR=1.17; 95% CI, 1.10-1.25) and quinolones (HR=1.30; 95% CI, 1.21-1.40). After multivariate adjustment for stroke risk factors such as diabetes, hypertension, hyperlipidemia, coronary artery disease, chronic obstructive pulmonary disease, and atrial fibrillation, only quinolones remained associated with stroke risk (HR=1.17; 95% CI, 1.09-1.26). The researchers concluded that antibiotic use does not reduce risk of stroke, although a higher risk may be evident among those using specific antibiotics. Given the potential chronic nature of C. pneumoniae infection, intermittent use of quinolones over time may be problematic.

Serologic or diagnostic risk factors

Serologic biomarkers have greatly expanded our understanding of the pathophysiology of ischemic stroke [44]. Reliable serologic diagnostic markers to predict ischemic stroke risk are of great interest given the relative ease by which they can be obtained. The studies included in this review found that several serologic markers, including CML, CRP, tHcy, estradiol, DHT and resistin, are associated with an increased risk of ischemic stroke. However, other serologic risk factors for ischemic stroke demonstrated in studies involving younger populations (<65 years) were not found to be significant, including serum albumin, SHBG, testosterone, leptin, and adiponectin. The imaging biomarker, M-Mode echocardiography, was also not associated [45]. While these serologic biomarkers may be informative for clinicians assessing ischemic stroke risk, the risk associated with each of these markers was relatively low. The physiologic basis underlying these associations requires further exploration, especially within older age groups, to identify more precise risk biomarkers.

Conventional

Equivocal results for conventional risk factors among older adults were associated with ischemic stroke. While the studies reviewed provide insight, there are limitations to be considered. Up-to-date studies are needed, as many of the studies [23,24,27–31] reviewed were published between 1992 and 2005. Trends in both risk factors and outcomes due to, for example, secular trends, changes in diagnostic criteria, and diagnostic advances may influence observed associations. Moreover, as common in epidemiologic studies, subjective and/or retrospective measures of exposures may be biased. While still commonly used, some self-reported conventional risk factors are being tracked using new technologies (e.g., Ecological Momentary Assessment). This may help to advance exposure assessments. For example, both studies reviewed that examined alcohol use [23,24] were based on participants’ self-report of prior alcohol consumption that occurred more than a year before the interview (i.e., “how much alcohol did you drink two years ago?”). Another study [24] relied on participants’ self-report of diabetes and ischemic stroke and did not corroborate the responses with medical records. Furthermore, one of these studies [i.e., 31] did not report a risk estimate, making it difficult to evaluate and compare results. Notwithstanding, these published reports significantly add to our knowledge of stroke risk, and if they are replicated in accordance with our criteria, will provide more confidence in the observed association.

Genetic

Four studies were identified that assessed genetic predisposition as a risk factor for ischemic stroke based on candidate genes and previously published SNPs. Studies continue to examine the role of other SNPs and genetic markers with ischemic stroke. The four studies identified here found that four phosphodiesterase 4D SNPs were associated with ischemic stroke after stratification by hypertensive status, pTG was associated with increased risk of ischemic stroke and an association between two F12 SNPs were associated with pTG. The APOEε2 allele was associated with a lower incidence of stroke in individuals 71-79 years old, but was not associated with stroke among individuals over the age of 80 years old.

Psychosocial

The extant literature regarding psychosocial risk factors for stroke in older adults was reviewed and suggested multiple risk and protective factors. Our search results led to publications on depressive symptoms and purpose in life. Overall the findings suggest that the presence of depressive symptoms in older adults are associated with stroke incidence, though comorbid risk factors reduced this association; and that purpose in life may be protective against future stroke.

Cognition

There are clear weaknesses in the studies examining cognition in association with future ischemic stroke in older populations. For example, the scales used, such as the MMSE and 3MS, have weak psychometric properties and small sample sizes comprise the published studies reviewed. However, taken together, some studies provide evidence that higher global cognition is associated with lower risk for incident ischemic stroke. This points to the potential importance of cognitive and brain reserve in relation to both cognition and cerebrovascular events.

Search strategy

A comprehensive search of PubMed, EMBASE and PsycINFO data bases was conducted by two researchers, to include articles published between the date of origin of each database through December 23, 2018. The search terms used were [“stroke” or “ischemic stroke” or “cerebrovascular accident” or “brain stem infarction” or “infarction, anterior cerebral artery” or “infarction, middle cerebral artery” or “infarction, basilar artery” or “infarction, brainstem” or “brain volume” or “thalamic stroke” or “putaminal stroke” or “cerebrovascular disorders” or “cerebral infarction” or “cerebellar infarction”] and [“risk factors” or “hypertension” or “diabetes mellitus” or “smoking” or “atrial fibrillation” or “dyslipidemia” or “previous myocardial infarction/ coronary artery disease” or “family history of myocardial infarction/ coronary artery disease” or “previous stroke/transient ischemic attack” or “family history of stroke/ transient ischemic attack” or “illicit drug use” or “alcohol” or “obesity” or “sleep apnea”] and [“Aged, 70 and Over” or “aging population” or “elderly” or “geriatric”]. Boolean logic was used to help maximize the number of relevant studies. We applied search parameters to limit the results to full-text. We applied search parameters to limit the results to full-text.

Eligibility criteria

Selected for inclusion in this systematic review were: studies involving human subjects and case-control or prospective studies. These studies had to include ischemic stroke as an outcome, risk factor measurements at age >65 years, and had to be comprised of a study sample with mean age > 70 years. In addition, all reports had to be published in English. Degree of risk or protection associated with any given factor had to be presented using an estimate of Relative Risk (RR) an Odds Ratio (OR). Studies were excluded if they lacked primary data (e.g., editorials, review articles, or protocol papers), were not published in a peer-reviewed journal, were prevalence studies, focused on hemorrhagic stroke or transient ischemic attacks (TIA), examined risk factors for recurrent stroke, or did not specify the type of stroke outcome. Assessment of publication eligibility was performed independently in a blinded (to each other) standardized manner by two reviewers (JS; JM). A third reviewer (CC) resolved any disagreements.

Quality assessment

We used the Newcastle-Ottawa quality assessment scale [10] for assessing the quality of studies in meta-analyses. The Newcastle-Ottawa scale rates the quality of categories pertaining to observational prospective longitudinal studies. Categories include participant selection, comparability, and outcome. Two reviewers independently assessed each study using the assessment scale (see Table 1). Any disagreements were resolved by consensus with the fifth author (JM). All authors had 100% agreement with the final results reported in Table 2. Inter-class correlation coefficients (ICC) were calculated to measure inter-rater reliability among reviewers using SPSS (IBM SPSS, Version 20, IBM Corp.)