Académique Documents
Professionnel Documents
Culture Documents
E-Mail karger@karger.com
Edinburgh, EH16 4SB (UK)
www.karger.com/ned
E-Mail cathie.sudlow @ ed.ac.uk
In a systematic review, we have reported that Chinese Major Risk Factor Definitions
populations had a higher stroke incidence and a 2-fold Vascular risk factor information was obtained from the pa-
tients, family, and medical records. Hypertension was defined
proportion of ICH among all strokes, as compared with as history along with antihypertensive treatment or blood pres-
white populations [5]. The reasons for the prevalence of sure ≥140/90 mm Hg 7 days after stroke; diabetes was defined
this trend are not clear but could include differences in as history along with anti-diabetic treatment or fasting plasma
genetic and environmental exposures, and differences in glucose >126 mg/dL. Atrial fibrillation (AF) and ischemic heart
risk factor associations between ICH and ischemic stroke disease (IHD) were defined as history or electrocardiographic
evidence. Hyperlipidemia was recorded if patients had history
(IS) [6–8]. Our recent meta-analysis of major risk factors along with anti-hyperlipidemic treatment, hypercholesterol-
found that hypertension and alcohol intake were signifi- emia, or hypertriglyceridemia (≥200 mg/dL). Smoking included
cantly more frequent in ICH than IS in Chinese stroke current or previous smoking history. Alcohol intake was de-
patients, but not in white stroke patients [9]. However, fined as habitual drinking more than once per week. We also
these findings could be confounded by age, gender, and recorded previous stroke and history of transient ischemic at-
tack (TIA).
other risk factors.
To explore this further, we performed individual Statistical Analyses
Chinese patient data analyses using data from the For ICH vs. IS, we used Student t test to compare the mean age,
National Taiwan University Hospital (NTUH) Stroke and Pearson’s chi-square test to compare the proportions of male
Registry, comparing major risk factors between ICH gender, risk factors and 1-month case fatality between groups.
Since it was not assumed to contribute a linear change in odds of
and IS, and adjusting for potential confounders. We ICH vs. IS, we divided age into 5 equal-sized categories. First, we
aimed to test the hypothesis that risk factor prevalence calculated crude ORs with 95% CIs for each risk factor. Then we
and associations vary between ICH and IS in Chinese fitted a logistic regression model for the relevant risk factor, adjust-
stroke patients. As Chinese populations are aging rap- ing for age and gender, obtaining adjusted ORs (OR I). We devel-
idly and make up around one-fifth of the world’s popu- oped a second logistic regression model to adjust simultaneously
for all other major risk factors along with age and gender. In the
lation, a better understanding of the reasons for a high- second model, we used stepwise selection and analysis of variance
er incidence and proportion of ICH in Chinese could to incorporate the strongly significant 2-way interactions among
lead to more efficient stroke-prevention strategies, variables (p < 0.001), and to choose the best fitting model. From
thereby reducing the burden of the impending stroke [5, the parameters of this model, we obtained further adjusted ORs
10]. (OR II).
In addition, we conducted subgroup analyses to examine risk
factor associations with ICH vs. IS in different subgroups where
there was a strongly significant interaction. All statistical hypoth-
Methods esis tests were 2-sided, and p values <0.05 were considered signifi-
cant. Statistical analyses were performed with R statistical software
Subjects (http://www.R-project.org/) [16].
The NTUH Stroke Registry is a large hospital-based registry
initiated in 1995, prospectively recruiting consecutive acute
stroke patients [11–13]. Patients who had acute stroke within 10
days were included. A neurologist assessed each patient upon ar- Results
rival, recorded demographic information and clinical features,
reviewed medical records, arranged timely brain imaging, and Characteristics of ICH and IS Patients
discussed the imaging findings with a neuroradiologist. The In-
stitutional Review Board of NTUH approved the stroke registry, From January 1, 2006 to December 31, 2011, we in-
and informed consent was given by the patients. In this study, we cluded 1,373 spontaneous ICH patients and 4,953 adult
conducted analyses for first-ever or recurrent spontaneous ICH IS patients for analyses. For patients with more than one
and IS adult patients recruited between January 2006 and stroke during the study period, analyses were based only
December 2011. on the first stroke episode. All patients had brain CT im-
Diagnosis of Stroke and Its Types aging and 50% had additional brain MR and MR angiog-
The diagnosis of stroke was based on clinical features: acute raphy.
neurologic dysfunction of vascular origin lasting for more than ICH patients were younger than IS patients (mean age
24 h [14], along with brain CT or MRI findings. Strokes were clas- 61.4 vs. 68.1 years, p < 0.001), while there was no signifi-
sified as IS, ICH, or subarachnoid hemorrhage (SAH) according to cant difference in gender between groups (males 61.9% in
standard definitions [15]. Patients with traumatic ICH, postin-
farct hemorrhagic transformation, subdural/epidural/SAH, tumor ICH and 59.1% in IS, p = 0.064). Patients with ICH had a
bleeding, no brain imaging, or non-cerebrovascular causes were much higher 1-month case fatality than those with IS
excluded. (19.3 vs. 5.5%, p < 0.001).
132.239.1.231 - 6/12/2017 4:10:19 AM
Univ. of California San Diego
ICH, intracerebral hemorrhage; IS, ischemic stroke; NTUH, National Taiwan University Hospital; AF, atrial fibrillation; TIA,
transient ischemic attack. ** Based on Student t test; * based on Pearson’s chi-square test.
Risk Factor Comparisons for ICH vs. IS Subgroup Analysis of Hypertension in Different
The clinical characteristics and risk factor distribu- Age Groups
tions for ICH and IS are shown in Table 1. In unadjusted We found a strongly significant interaction between
comparisons, ICH patients had a significantly lower prev- age and hypertension (p < 0.001), implying a significant
alence of diabetes, AF, IHD, hyperlipidemia, smoking, difference in the association across different age groups
previous stroke and TIA than IS patients. Alcohol intake for hypertension, but not other risk factors. We, therefore,
was slightly more common in ICH vs. IS (crude OR 1.19, examined the association of hypertension with ICH vs. IS
95% CI 1.00–1.40), while hypertension did not differ sig- among 5 equal-sized age groups (18–55, 56–64, 65–72, 73–
nificantly between these 2 types (Fig. 1). 80, 81–106 years). Compared with older stroke patients,
After adjusting for age and gender, the association of hypertension had a stronger association with ICH vs. IS in
hypertension with ICH vs. IS became significantly younger stroke patients (Fig. 2). The fully adjusted results
stronger (OR I 1.25, 95% CI 1.07–1.45), while the re- showed that hypertension was significantly associated
sults for other risk factors remained similar except that with ICH vs. IS in patients aged <73 years (18–55 years OR
associations with alcohol and previous stroke became II 2.33, 95% CI 1.79–3.05; 56–64 years OR II 2.33, 95% CI
nonsignificant. After further adjusting for all risk fac- 1.58–3.48; 65–72 years OR II 1.84, 95% CI 1.22–2.87) but
tors but before including significant interactions, hy- not in those aged ≥73 years (73–80 years OR II 0.81, 95%
pertension and alcohol intake were significantly more CI 0.56–1.18; 81–106 years OR II 0.96, 95% CI 0.63–1.50).
associated with ICH than IS (hypertension OR II with-
out interactions 1.60, 95% CI 1.42–1.95; alcohol OR II
without interactions 1.45, 95% CI 1.16–1.79). For hy- Discussion
pertension, the association became yet stronger when
strongly significant interactions were included (hyper- In our study, we showed that ICH patients had a much
tension OR II with interactions 2.23, 95% CI 1.74–2.87; higher case fatality and a younger age of onset than IS pa-
Fig. 1). Fully adjusted analyses yielded similar but tients, but no significant difference in gender. In terms of
somewhat stronger negative associations than in unad- risk factor comparisons between ICH and IS, unadjusted
justed analyses for AF, hyperlipidemia, and smoking, and adjusted analyses were qualitatively similar, though
while results were changed only in a very negligible they varied to a certain extent in terms of the size of esti-
manner following adjustment for diabetes, TIA, and mates and significance. Both hypertension and alcohol
previous stroke. intake had stronger associations with ICH than IS after
132.239.1.231 - 6/12/2017 4:10:19 AM
Univ. of California San Diego
adjusting for confounders, whereas diabetes, AF, IHD, tients are limited. Our results are in line with our meta-
hyperlipidemia, smoking, and TIA were less frequent in analyses after adjusting for confounding factors – hyper-
ICH vs. IS. Furthermore, the association of hypertension tension and alcohol intake had stronger associations with
with ICH vs. IS differed across different age groups, being ICH than IS in Chinese [9]. They may be responsible – at
more marked in younger stroke patients. Although least in part – for a higher incidence and proportion of
Chinese are reported to have a much higher ICH inci- ICH in Chinese populations.
dence than white populations, the reasons are still not Few large-scale studies or systematic reviews have di-
clear [17, 18]. Risk factor analyses for ICH vs. IS using rectly compared risk factors between ICH and IS. An ear-
robust methods based on a large number of Chinese pa- lier systematic review in mainly white populations re-
132.239.1.231 - 6/12/2017 4:10:19 AM
Univ. of California San Diego
ported that increased age, male gender, hypertension, and consolidate this finding that hypertension is a stronger
alcohol intake were risk factors for ICH, whereas hyper- risk factor for ICH vs. IS in Chinese patients [9, 22].
cholesterolemia was associated with a lower risk of ICH The reasons for a higher incidence of total strokes and
[19]. Recently, INTERSTROKE study conducted risk fac- ICH, and the greater association with hypertension for ICH
tor analyses for all strokes, IS, and ICH among 3,000 vs. IS in Chinese populations are not fully understood. The
stroke cases and controls from 22 countries (around one preference of high salt intake, inadequate treatment of hy-
third from Southeast Asia) [8]. Hypertension, smoking, pertension, and a higher proportion of deep ICH in Chinese
waist-to-hip ratio, diet, and alcohol intake were signifi- (predominantly associated with hypertension) may all con-
cant risk factors for ICH. Hypertension was more strong- tribute to these findings [23–25]. In the INTERMAP study
ly associated with ICH than IS, whereas the association among British, American, Japanese and Chinese popula-
with smoking appeared stronger with IS than ICH. How- tions, the Chinese were found to have the highest daily so-
ever, risk factor associations for ICH and IS were not pro- dium intake, which was known to lead to high blood pres-
vided solely in Chinese populations. sure and increased risk of stroke [23]. Also, the proportion
Hypertension is one of the most important risk factors of awareness, treatment, and control rates are relatively low-
for ICH, and its association with ICH probably differs er among the Chinese people. In mainland China, only 45%
among ethnic groups [20, 21]. Research from Asia Pacific were aware of high blood pressure, 28% took antihyperten-
Cohort study suggested a stronger association of increased sive medication, and 8% achieved control target (<140/90
blood pressure with hemorrhagic stroke among eastern mm Hg) [24]. In Taiwan, despite improvements in medical
Asians than White populations [20, 21]. However, these care after the implementation of the National Health Insur-
studies were conducted for all hemorrhagic strokes, in- ance scheme, the proportions of people with hypertension
cluding ICH and SAH together, while their causes were receiving treatment or achieving good control are still low,
somewhat different. No separate analysis for ICH was re- particularly in stroke patients less than 65 years old [26].
ported. Our recent systematic review and meta-analyses This may be one of the important reasons accounting for the
showed a higher incidence of stroke along with a higher stronger association with ICH vs. IS in younger stroke pa-
proportion of ICH in Chinese, and hypertension had a tients, and highlight the need to enhance education and
stronger association with ICH vs. IS as compared with treatment in this group.
white populations [5, 9]. Current analyses from individu- With regard to lipids, increased total cholesterol is as-
al patient data after controlling for confounders further sociated with the risk of IS but not hemorrhagic stroke
132.239.1.231 - 6/12/2017 4:10:19 AM
Univ. of California San Diego
References
1 Feigin VL, Lawes CM, Bennett DA, Barker- 4 Feigin VL, Krishnamurthi RV, Parmar P, Nor- angiopathy and intracerebral haemor-
Collo SL, Parag V: Worldwide stroke rving B, Mensah GA, Bennett DA, et al; GBD rhage: systematic review and meta-analysis.
incidence and early case fatality reported 2013 Stroke Panel Experts Group: Update on J Neurol Neurosurg Psychiatry 2012; 83:
in 56 population-based studies: a syste- the global burden of ischemic and hemorrhag- 275–281.
matic review. Lancet Neurol 2009; 8: 355– ic stroke in 1990–2013: the GBD 2013 study. 8 O’Donnell MJ, Xavier D, Liu L, Zhang H,
369. Neuroepidemiology 2015;45:161–176. Chin SL, Rao-Melacini P, et al: Risk factors for
2 van Asch CJ, Luitse MJ, Rinkel GJ, van der 5 Tsai CF, Thomas B, Sudlow CL: Epidemiolo- ischaemic and intracerebral haemorrhagic
Tweel I, Algra A, Klijn CJ: Incidence, case fa- gy of stroke and its subtypes in Chinese vs stroke in 22 countries (the INTERSTROKE
tality, and functional outcome of intracere- white populations: a systematic review. Neu- study): a case-control study. Lancet 2010;376:
bral haemorrhage over time, according to age, rology 2013;81:264–272. 112–123.
sex, and ethnic origin: a systematic review and 6 Donnan GA, Hankey GJ, Davis SM: Intracere- 9 Tsai CF, Anderson N, Thomas B, Sudlow CL:
meta-analysis. Lancet Neurol 2010; 9: 167– bral haemorrhage: a need for more data and Comparing risk factor profiles between intra-
176. new research directions. Lancet Neurol 2010;9: cerebral hemorrhage and ischemic stroke in
3 Qureshi AI, Mendelow AD, Hanley DF: In- 133–134. Chinese and white Populations: systematic
tracerebral haemorrhage. Lancet 2009; 373: 7 Samarasekera N, Smith C, Al-Shahi Salman review and meta-analysis. PLoS One 2016;
1632–1644. R: The association between cerebral amyloid 11:e0151743.
132.239.1.231 - 6/12/2017 4:10:19 AM
Univ. of California San Diego