Variations of risk factors for ischemic stroke and its subtypes in Chinese patients in Taiwan

Chinese have a higher stroke incidence and a different distribution of ischemic stroke (IS) subtypes as compared with Caucasians. Herein we aimed to investigate the prevalence and associations of major risk factors in IS and its subtypes in Chinese patients. From 2006 to 2011, we included 4953 acute IS patients consecutively recruited in National Taiwan University Hospital Stroke Registry (mean age 68 years; male 59%). For each risk factor, we accessed the proportion in all IS patients, and calculated odds ratios for each main IS subtype versus other subtypes. Multiple logistic regression models were used to adjust for confounders, and to examine the associations of risk factors with IS subtypes. Compared with other ischemic subtypes, large artery atherosclerotic and lacunar strokes were associated with hypertension, diabetes, and hyperlipidaemia, while cardioembolic strokes were associated with ischemic heart disease. Furthermore, the associations with hypertension and diabetes became stronger in lacunar strokes after adjusting for confounders, but not in other ischemic subtypes. Here we report the variable effects of risk factors on different IS subtypes in Chinese patients in Taiwan. Our findings could help shed light on different mechanisms of IS subtypes and provide targets to make more effective strategies for IS prevention.


Diagnosis of ischemic stroke and its subtypes. The diagnosis of IS was based on the definition from
World Health Organization, and confirmed by brain CT/MRI 11 . IS patients were classified into five etiological subtypes according to the TOAST criteria (risk-factor dependent)-large artery atherosclerosis (LAA: infarct lesions greater than 1.5 cm on CT/MRI with more than 50% stenosis of an intracranial or extracranial artery by duplex imaging or CT angiography (CTA)/MR angiography (MRA)/angiography, and excluding cardiogenic embolism), LI (lacunar syndrome with infarct lesion less than 1.5 cm on brain CT/MRI, and excluding cardiac embolism and large artery stenosis more than 50% ), cardioembolism (CE: at least one cardiac source of emboli such as atrial fibrillation being identified, and excluding large artery atherosclerosis), other determined, and undetermined etiology. We also classified these IS patients into four anatomical subtypes of the Oxfordshire Community Stroke Project (OCSP classification)-LI, total and partial anterior circulation infarct (TACI and PACI), and posterior circulation infarct (POCI) 12,13 . The OCSP classification system is a risk-factor free scheme based on clinical features of stroke syndrome and brain image findings, being free of etiological assumptions of risk factors to reduce the possibility of classification bias. Patients with previous stroke and its complications, intracerebral/subarachnoid/subdural/epidural haemorrhage, tumor, non-cerebrovascular causes or no brain imaging were excluded.
Major risk factor definitions. Major risk factors were defined as follows: hypertension (history, antihypertensive treatment or blood pressure ≥ 140/90 mmHg 7 days after stroke), diabetes (history, anti-diabetic treatment, or fasting plasma glucose ≥ 126 mg/dl), AF (history or electrocardiographic evidence), IHD (history or electrocardiographic evidence), hyperlipidemia (history, anti-hyperlipidemic medication, hypercholesterolemia or hypertriglyceridemia ≥ 200 mg/dl), current or previous smoking, and alcohol intake (habitual drinking more than once per week). We also recorded previous stroke and history of transient ischemic attack (TIA).

Statistical analysis.
We used analysis of variance (ANOVA) to compare mean age, and chi-square test to compare proportions of male sex and risk factors among IS subtypes. We divided age into five equal-sized categories as it was not assumed to contribute a linear change, and calculated proportions of risk factors in each IS subtype. For each risk factor, we computed crude odds ratio (OR) with 95% confidence interval (CI) for LAA, LI, and CE versus other subtypes in the TOAST classification, and used logistic regression to adjust for age and sex and to examine if there was any interaction, obtaining adjusted ORs (OR I). Then we developed a second logistic regression model to adjust for age, sex, and all major risk factors, and used stepwise selection to incorporate the significant 2-way interactions among variables (p < 0.001), obtaining adjusted ORs (OR II). Also, we categorized OCSP subtypes into LI and non-lacunar infarct (non-LI), doing the same analyses for LI versus non-LI.
For TOAST IS subtypes, we did not include AF in the second logistic regression to avoid classification bias as IS patients with AF were most likely assigned to CE subtype. If there was a significant interaction among variables, we conducted further subgroup analyses to exam risk factor associations in different groups. All the statistical hypothesis tests were two-sided, and p values less than 0.05 were regarded as significant. Statistical analyses were performed with R 14 .

Results
Characteristics of overall IS patients. We included 4953 acute IS Chinese patients into current analyses, and all had brain CT and electrocardiography (EKG). For cerebrovascular evaluation, around 60% of these patients had extracranial and intracranial CTA/MRA/angiography, while others had Carotid duplex imaging and transcranial color-coded ultrasonography. Cardiac sonography and 24-h EKG monitoring were performed in patients had cardiac disease, electrocardiography abnormality, suspected embolic stroke on brain image, sudden onset or young age of stroke, or unclear cause of stroke. The mean age was 68.1 (± 13.8) years; 59% were men. In our study, male patients had younger onset of stroke (mean age: men 66.5 ± 13.8 years, women 70.4 ± 13.6 years, p < 0.001), more smoking and alcohol intake, but less diabetes, AF and hyperlipidemia than female patients. The clinical characteristics and risk factor distributions were shown in Table 1.
Risk factors for the TOAST subtypes. LI was the most common subtype in the TOAST classification (27.5%), followed by CE (24.3%), undetermined (22.1%), LAA (21.0%), and other determined etiology (5.0%) ( Table 2). There were significant age and sex variations among subtypes (p < 0.001)-CE patients were the eldest (mean 72.5 ± 13.2 years), while LAA had the highest proportion of men (67%). Different distributions were significant in all vascular risk factors studied among IS subtypes (p < 0.001).
CE strokes were more associated with age and IHD in comparison with other subtypes (Fig. 2), while having lower prevalence of male sex, diabetes, hyperlipidemia and smoking. There was a strongly significant interaction between IHD and age (p < 0.001). While incorporating this interaction, the fully adjusted OR for IHD was 6.74 (95% CI 4.16-10.87, p < 0.001).
For LI versus other IS subtypes, unadjusted analyses showed significant associations with hypertension, diabetes, and hyperlipidemia ( Fig. 3), with significant interactions between age and hypertension, and between

Risk factors for the OCSP subtypes.
In the OCSP classification, ANOVA showed significant different distributions in age, male sex and all major risk factors studied in different subtypes except TIA (Supplementary Table S1). The mean age was older in TACI (69.6 ± 15.0 years), while younger in POCI (66.7 ± 13.9 years). POCI had the highest proportion of men (63%), whereas TACI possessed the lowest (51%). In comparison to Non-LI, LI was significantly associated with hypertension (OR II 4.58, 95% CI 3.16-6.75, p < 0.001), diabetes (OR II 2.40, 95% CI 1.71-3.37, p < 0.001) and hyperlipidemia (OR II 1.55, 95% CI 1.34-1.79, p < 0.001), while having lower prevalence of AF (OR II 0.10, 95% CI 0.07-0.13, p < 0.001) and IHD (OR II 0.42, 95% CI 0.33-0.54, p < 0.001) in fully adjusted analysis (Fig. 4). There were significant interactions between age and hypertension, and between hypertension and diabetes (p < 0.001). The unadjusted and adjusted results for LI versus Non-LI in the OCSP were in line with those in the TOAST classification.

Discussion
Our results showed that the proportions of hypertension, diabetes and hyperlipidaemia in our IS patients based in Taiwan were significantly higher, while AF, IHD, smoking and alcohol intake were lower than our previous meta-analyses for Caucasian patients 9 . As for the risk factor associations with IS subtypes, LAA and LI were more associated with hypertension, diabetes, and hyperlipidemia (varied in size), while CE strokes were more associated with IHD as compared with other subtypes. In addition, the associations with hypertension and Hypertension and diabetes were independent risk factors for LAA and lacunar strokes in our previous metaanalyses and other report 9,15 . Current study further revealed that hyperlipidemia was also a significant risk factor for both LAA and lacunar strokes in Chinese patients after adjusting for confounding factors. As compared with those in mainland China, our study showed a higher prevalence of these risk factors in IS patients based in Taiwan 9,16 . This finding was also noted in the international REACH registry-a stepwise increase of hypertension,   18 . When the life expectancy and economics continue improving with more high-fat diet and sedentary life, atherosclerotic diseases such as diabetes and IHD increase in the third stage (e.g. Latin America) 18 . Taiwan is probably also at this stage because of earlier economic development and more influence from western countries. The epidemiological data in Taiwan could be useful to predict the disease evolution in China and other developing countries since they are experiencing rapid economic and lifestyle changes along www.nature.com/scientificreports/ with aging populations. As the prevalence of hypertension, diabetes, and hyperlipidemia have increased in Chinese populations in recent decades, more effective strategies through various education and intervention projects are of vital importance to reduce these risk factors and associated cerebrovascular diseases ahead 2,19-21 . LI was the most common IS subtype in Chinese stroke patients 5 . In our analyses, there were significant interactions between age and hypertension, and between hypertension and diabetes. The associations with hypertension and diabetes became stronger for LI after incorporating the significant interactions in fully adjusted analyses in in both risk-dependent TOAST and risk-free OCSP classifications, but not in LAA or CE subtypes, or western patients 9 . These findings suggested hypertension and diabetes are strong and important risk factors for LI in Chinese patients. Our findings were different from western research, which showed no difference of hypertension www.nature.com/scientificreports/ and diabetes between LI and non-LI 22 , but in line with others studies in eastern Asians 23,24 . The Hisayama study in Japan also reported a greater impact of blood pressure on LI than other IS subtypes 23 . In Taiwan, the prevalence of hypertension and diabetes in IS patients had increased in the past 10 years, while the proportion www.nature.com/scientificreports/ of taking anti-hypertensive or anti-diabetic medication before stroke decreased 19 . Similar findings were also noted in mainland China, where only a low proportion of patients achieved the target goals of control 16 . This may be responsible (at least in part) for a higher proportion of LI in Chinese IS patients and a higher incidence of stroke in Chinese populations. Difference in genetics among ethnic groups, variations of risk factor control, and different methods of case ascertainment may account for these disparities [6][7][8][9]19,20 .
As for CE subtype, it had a totally different risk factor distribution. Our results showed CE strokes were more associated with IHD, while having lower prevalence of male sex, diabetes, hyperlipidemia, and smoking. In our analyses for TOAST IS subtypes, we did not include AF in the second logistic regression to avoid classification bias. Even though the classification system does not define subtypes explicitly based on risk factors, there is a tendency for clinicians to bias their assessments according to the strong risk factors present. For example, IS patients with AF are usually assigned to CE subtype, irrespective of other investigation results. In our study, the proportion of CE strokes was higher than previous meta-analyses for the whole Chinese patients or earlier studies, which was probably related to aging population and increasing detection of AF by echocardiography and 24-h holter monitoring 5,9,25 . The incidence of AF and related strokes are expected to increase ahead because of longer life expectancy and improved survival of IHD, but anticoagulation is still not adequate, especially in Asians 25,26 . Proper holter monitor and adequate anticoagulation may reduce the burden from CE strokes.
Our study documented various influences of risk factors on IS subtypes in Chinese populations, who had a higher incidence of stroke, a different distribution of IS subtypes, especially a higher proportion of lacunar infarct than Caucasians. Furthermore, our results showed that hypertension and diabetes were strong and important risk factors for lacunar stroke in Chinese patient. It had several strengths. First, NTUH Stroke Registry was a well-established registry, prospectively recruiting consecutive acute stroke patients with comprehensive records, without selection of age, sex or socioeconomic status as the Taiwan National Health Insurance provided affordable and compulsive medical care, covering 99% people in Taiwan. This study benefited from a relatively large number of patients, and detailed information of risk factors. Second, the inclusion of acute stroke patients was based on a standard stroke definition, having timely Neurologist and brain imaging evaluation, and in-depth investigations. Third, we used multiple logistic regressions to adjust for possible confounders, and incorporated significant interactions among variables. There were some limitations. NTUH registry was a large hospital-based stroke registry in Taiwan, which might not be fully representative of all Chinese IS patients. Also, not all acute IS patients received cardiac echography and 24-h EKG monitoring, and only few patients had prolonged EKG monitoring up to 72 h or longer. Finally, risk factors were collected from medical records before stroke, and from the included patients and their families. Whilst most patients had previous medical records, we could not totally exclude recall bias in a few patients who had no available data before stroke.
Accumulating evidence has suggested that the distributions of IS subtypes and risk factors vary among ethnicities and geographical areas, and there are different associations of risk factors with ischemic subtypes 2,5,9,15,[22][23][24] . While genetics may play a certain role on ischemic subtypes, it is still of vital importance to target the modifiable risk factors through well-designed education and intervention projects. Herein we report various influences of risk factors on main IS subtypes in Chinese IS patients based in Taiwan. As IS accounts for the majority of heavy stroke burden in Chinese populations and around the world, our findings could help shed light on various impacts of risk factors on ischemic subtypes, reveal the potential targets to improve further, and reduce the stroke burden in the years ahead.

Data availability
The datasets for this study are available from the corresponding author on reasonable request.