Atherosclerosis xxx (2017) 1e7

Contents lists available at ScienceDirect

Atherosclerosis
journal homepage: www.elsevier.com/locate/atherosclerosis

Total cholesterol and stroke mortality in middle-aged and elderly

adults: A prospective cohort study
Sang-Wook Yi a, b, *, Dae-Hee Shin c, Hyeyun Kim d, Jee-Jeon Yi e, Heechoul Ohrr f
a
Department of Preventive Medicine and Public Health, Catholic Kwandong University College of Medicine, Gangneung, 25601, Republic of Korea
b
Institute for Clinical and Translational Research, Catholic Kwandong University International St. Mary's Hospital, Incheon, 22711, Republic of Korea
c
Cardiovascular center, Catholic Kwandong University College of Medicine, International St. Mary's Hospital, Incheon, 22711, Republic of Korea
d
Department of Neurology, Catholic Kwandong University College of Medicine, International St. Mary's Hospital, Incheon, 22711, Republic of Korea
e
Institute for Occupational and Environmental Health, Catholic Kwandong University, Gangneung, 25601, Republic of Korea
f
Department of Preventive Medicine, Yonsei University College of Medicine, Seoul, 03722, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: Background and aims: The association between cholesterol and stroke has been inconsistent. This study
Received 28 September 2017 aimed to examine the association between total cholesterol (TC) and mortality from total stroke and
Received in revised form stroke subtypes.
28 November 2017
Methods: 503,340 Korean adults aged 40e80 years without a history of heart disease or stroke partic-
Accepted 5 December 2017
Available online xxx
ipated in routine health examinations in 2002 and 2003, and were followed up until 2013. Adjusted
hazard ratios (HRs) for stroke (I60-I69) mortality were calculated.
Results: Nonlinear associations for total stroke (U-curve) and hemorrhagic stroke (L-curve), especially
Keywords:
Blood cholesterol
intracerebral hemorrhage (ICH), but a linear association for ischemic stroke, were found. In the range
Intracerebral hemorrhage <200 mg/dL, TC was inversely associated with stroke mortality (HR per 39 mg/dL [1 mmol/L]
Ischemic stroke increase ¼ 0.88 [95% CI ¼ 0.80e0.95]), mainly due to hemorrhagic stroke (HR ¼ 0.78 [0.68e0.90]),
Cohort studies especially ICH (HR ¼ 0.72 [0.62e0.85]). In the upper range (200e349 mg/dL), TC was positively asso-
Subarachnoid hemorrhage ciated with stroke mortality (HR ¼ 1.09 [1.01e1.16]); ICH and subarachnoid hemorrhage mortality
showed no inverse association. The associations were generally similar in middle-aged (40e64 years)
and elderly (65 years) adults and, in the upper range, each 1 mmol/L (39 mg/dL) higher TC was
associated with 11% higher mortality from stroke (95% CI ¼ 2%e21%) in the elderly. Both middle-aged
(39%) and elderly (23%) adults had higher ischemic stroke mortality associated with TC 240 mg/dL,
compare to <200 mg/dL.
Conclusions: TC level around 200 mg/dL was associated with the lowest risk of overall stroke in the
elderly and middle-aged adults. No stroke subtype including ICH, was inversely associated with TC in the
range 200 mg/dL.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction European guidelines, statin therapy is recommended for preven-

Stroke imposes huge health and economic burdens worldwide
[1,2]. The associations between cholesterol and stroke have been
inconsistent. Randomized controlled trials have provided evidence
that lowering cholesterol, particularly by statins, reduces the risk of
stroke, except fatal stroke [3e6]. Accordingly, in the recent
* Corresponding author. Department of Preventive Medicine and Public Health,
College of Medicine, Catholic Kwandong University, Bumil-ro 579, Gangneung,
Gangwon-do, 25601, Republic of Korea.
E-mail address: flyhigh@cku.ac.kr (S.-W. Yi).
tion of stroke [7]. Reviews and meta-analyses of prospective cohort
studies, however, found no consistent graded association between
cholesterol and stroke [8e10]. Furthermore, concerns persist that
lowering cholesterol level may increase the risk of hemorrhagic
stroke [11e13]. The fact that prospective cohort studies consistently
showed inverse associations of cholesterol with hemorrhagic
stroke, especially intracerebral hemorrhage (ICH) [10,14,15], fueled
these concerns. These discrepancies between studies on the asso-
ciation of cholesterol with overall stroke and stroke subtypes in-
crease the complexity of decision making in the clinical and public
health settings, such as whether or when statins can be used in
persons with, or at risk for, ICH, or for primary prevention for stroke

https://doi.org/10.1016/j.atherosclerosis.2017.12.003
0021-9150/© 2017 Elsevier B.V. All rights reserved.

Please cite this article in press as: S.-W. Yi, et al., Total cholesterol and stroke mortality in middle-aged and elderly adults: A prospective cohort
study, Atherosclerosis (2017), https://doi.org/10.1016/j.atherosclerosis.2017.12.003
2 S.-W. Yi et al. / Atherosclerosis xxx (2017) 1e7
in older adults.
Through a large prospective cohort study, we aimed to examine
the association between total cholesterol (TC) and stroke mortality,
and to elucidate whether inconsistencies between previous studies
may be resolved. TC remains an integral part of risk prediction and
prevention models for cardiovascular disease [16e18], and of public
health initiatives, such as Healthy People 2020 [1,19], Therefore,
precise estimates of association of TC below 140 mg/dL could help
inform decision-making for stroke prevention and management, in
the coming era of mean TC levels below 180 mg/dL, in populations
with much higher TC levels decades ago [1,19].
2. Materials and methods
2.1. Data availability
Data are accessible to researchers by the National Health In-
surance Service (NHIS) of Korea, when their study protocol is
reviewed and approved by the NHIS [20].
2.2. Study population and follow-up
The NHIS provides compulsory health insurance to 97% of the
Korean population [21]. The study cohort (n ¼ 514,795) comprised a
random sample of 10% of the 5.15 million NHIS beneficiaries aged
40e79 years in 2002, who underwent health examinations during
2002e2003. Of these subjects, 11,455 were excluded due to missing
information on TC (n ¼ 1710), fasting glucose, systolic blood pres-
sure, and body mass index (BMI), due to a history of heart disease
and stroke (n ¼ 9693), or due to an extremely high BMI (50 kg/m2,
n ¼ 52). For the remaining 503,340 subjects, follow-up on stroke
deaths through December 31, 2013 was conducted via record
linkage with national death records. The International Classification
of Diseases-10th Revision was used to define death from total
stroke (I60-I69), and the subtypes of stroke were classified into
hemorrhagic stroke (I60-I62), subarachnoid hemorrhage (SAH,
I60), intracerebral hemorrhage (ICH, I61-I62), and ischemic stroke
(I63). In Korea, hospitals have routinely used computed tomogra-
phy and/or magnetic resonance imaging for stroke diagnosis since
the late 1990s; 89% of hospital admissions used those imaging
techniques for stroke according to a nationwide survey in 2000
[22]. The NHIS can provide data without specific informed consent
from participants according to the Korean law [20]. This study was
approved by the Institutional Review Board of the Catholic Kwan-
dong University, Republic of Korea. The NHIS provided authors with
access to the anonymized data.
2.3. Data collection
TC and glucose were assayed using fasting serum samples by
enzymatic methods. Systolic blood pressure (SBP) was measured in
a seated position using a standard mercury sphygmomanometer.
Weight and height were measured to the nearest kilogram and
centimeter, respectively [21]. BMI was calculated by weight in ki-
lograms divided by the square of height in meters (kg/m2). Smoking
history, alcohol use, and physical activity were assessed via a
questionnaire. Health examination and data collection followed a
standard protocol, the Health Examination Practice Guide, officially
registered by the Ministry of Health and Welfare. The Korean As-
sociation of Quality Assurance for Clinical Laboratory supervised
external quality assessment in clinical chemistry, such as TC mea-
surements, for participating hospitals, and assessments of the
quality of assays were regularly performed [23].
Please cite this article in press as: S.-W. Yi, et al., Total cholesterol and stroke mortality in middle-aged and elderly adults: A prospective cohort
study, Atherosclerosis (2017), https://doi.org/10.1016/j.atherosclerosis.2017.12.003
2.4. Statistical analysis
For analysis, TC levels were categorized into 8 groups (<140,
140e159, 160e179, 180e199 [reference], 200e219, 220e239,
240e259, and 260 mg/dL). The category with the lowest mortality
was used as reference. Log risk was regressed on TC as a continuous
variable within the range <200 mg/dL (termed “lower range”),
200e349 mg/dL (“upper range”), or <350 mg/dL (“full range”),
yielding HRs per 39 mg/dL (1 mmol/L) increase in TC in each range.
Analysis using a restricted cubic spline transformation of TC with 3
knots (150, 190, and 230 mg/dL) was also performed.
HRs for stroke mortality were calculated using Cox proportional
hazards models stratified by age (years) at baseline (40e44, 45e54,
55e64, 65e74, and 75e80) after adjustment for age at baseline
(continuous variable; within each age group), sex (when appli-
cable), smoking status (current smoker, former smoker, never
smoker, and missing information [n ¼ 21,282]), alcohol use (fre-
quency; monthly or less, 2 days/month to 2 days/week, 3e7 days/
week, and missing information [n ¼ 9458]), physical activity (at
least once a week; yes, and no), beneficiary income status (deciles;
below 4 [low income], 4e7, 8e10 [high income]), BMI (continuous
variable), SBP (continuous variable), and fasting glucose (contin-
uous variable). In a sensitivity analysis, lipid-lowering medication
use (yes [n ¼ 7742], and no) at baseline was further adjusted for.
The nonlinear associations of TC with stroke mortality were
assessed with a likelihood ratio test, in which we compared the
model with only the linear term to that with both the linear and the
cubic spline terms. Subgroup analyses with varying categories of TC
served as sensitivity analyses.
All p values were 2-sided. All analyses used SAS version 9.4 (SAS
Institute Inc., Cary, NC, USA).
3. Results
During 5,223,381 person-years of follow-up of 503,340 people
(45.7% women), 3383 individuals died from stroke: 1184 from
hemorrhagic stroke; 354 from SAH; 830 from ICH; and 1044 from
ischemic stroke. At baseline the mean age was 52.9 ± 9.7 years and
the mean TC level was 200.4 ± 38.7 mg/dL (Table 1), and 14.5% of
subjects had TC levels 240 mg/dL or higher. Persons with higher TC
values were more likely to be women and current smokers, and
were less likely to exhibit frequent alcohol use than those with
lower levels. Higher TC levels were associated with older age and
higher levels of systolic blood pressure, fasting glucose, and BMI
(Table 1).
3.1. TC and stroke mortality
The associations between TC and total stroke mortality were
nonlinear. Total stroke mortality showed U-curve associations with
a nadir at 200e219 mg/dL (Fig. 1). Mortality from hemorrhagic
stroke, especially ICH, was inversely associated with TC in the lower
range <200 mg/dL, but not in the upper range (L-curve). Ischemic
stroke was positively associated with TC, particularly in the upper
range. After adjustment for other confounders, the associations
generally did not substantially change (Supplementary Table1).
Further adjustment for lipid-lowering medication use at baseline
yielded no material change (Supplementary Table2).
In the restricted cubic spline analysis, the patterns of associa-
tions and TC values associated with the lowest stroke mortality
were generally the same as in the categorical analysis of TC, and the
nonlinear associations were statistically confirmed for total stroke
and each subtype, except for SAH and ischemic stroke (Fig. 2).
Assuming a linear association below 200 mg/dL, TC was
inversely associated with stroke mortality (Fig. 3; HR per 39 mg/dL
 S.-W. Yi et al. / Atherosclerosis xxx (2017) 1e7 3

Table 1
Characteristics of participants according to total cholesterol categories.

Variables Characteristics Total Desirable Borderline High

<200 mg/dL 200-239 mg/dL 240 mg/dL

N ¼ 503,340 n ¼ 261,458 n ¼ 169,099 n ¼ 72,783

Total cholesterol mg/dLa 200.4 ± 38.7 171.8 ± 20.0 217.1 ± 11.1 264.8 ± 30.4
Age years 52.9 ± 9.7 52.3 ± 9.8 53.2 ± 9.5 54.3 ± 9.4
Systolic blood pressure mmHg 127.1 ± 18.2 125.5 ± 18.0 128.1 ± 18.1 130.5 ± 18.6
Fasting serum glucose mg/dLb 98.3 ± 34.6 96.4 ± 32.3 98.8 ± 32.4 104.0 ± 45.1
Body mass index kg/m2 24.0 ± 3.0 23.6 ± 2.9 24.3 ± 2.9 24.7 ± 2.9
Sex Women 229,865 (45.7) 114,877 (43.9) 78,235 (46.3) 36,753 (50.5)
Men 273,475 (54.3) 146,581 (56.1) 90,864 (53.7) 36,030 (49.5)
Smoking status Never smoker 322,696 (64.1) 166,609 (63.7) 108,638 (64.2) 47,449 (65.2)
Past smoker 42,543 (8.5) 21,856 (8.4) 14,583 (8.6) 6104 (8.4)
Current smoker 116,820 (23.2) 62,389 (23.9) 38,478 (22.8) 15,953 (21.9)
Missing 21,281 (4.2) 10,604 (4.1) 7400 (4.4) 3277 (4.5)
Alcohol use frequency, days 1/month 278,394 (55.3) 142,621 (54.5) 93,756 (55.4) 42,017 (57.7)
2/month-2/week 158,360 (31.5) 83,214 (31.8) 53,640 (31.7) 21,506 (29.5)
3-7/week 57,128 (11.3) 30,888 (11.8) 18,522 (11.0) 7718 (10.6)
Missing 9458 (1.9) 4735 (1.8) 3181 (1.9) 1542 (2.1)
Physical activity 1 times/week 206,821 (41.1) 106,429 (40.7) 70,730 (41.8) 29,662 (40.8)
Income status, decile <4 (low-income) 115,882 (23.0) 60,274 (23.1) 38,564 (22.8) 17,044 (23.4)
4e7 163,984 (32.6) 86,673 (33.1) 54,044 (32.0) 23,267 (32.0)
>7 (high-income) 223,474 (44.4) 114,511 (43.8) 76,491 (45.2) 32,472 (44.6)
Total cholesterol, mg/dL <140 19,167 (3.8) 19,167 (7.3) 0 (0.0) 0 (0.0)
140e159 46,068 (9.2) 46,068 (17.6) 0 (0.0) 0 (0.0)
160e179 86,874 (17.3) 86,874 (33.2) 0 (0.0) 0 (0.0)
180e199 109,349 (21.7) 109,349 (41.8) 0 (0.0) 0 (0.0)
200e219 99,108 (19.7) 0 (0.0) 99,108 (58.6) 0 (0.0)
220e239 69,991 (13.9) 0 (0.0) 69,991 (41.4) 0 (0.0)
240e259 39,913 (7.9) 0 (0.0) 0 (0.0) 39,913 (54.8)
260 32,870 (6.5) 0 (0.0) 0 (0.0) 32,870 (45.2)

Data are expressed as mean ± SD or n (%).

p values, which were calculated with the Chi-square test and one-way ANOVA between cholesterol groups, were <0.001 for each variable.
a
To convert total cholesterol from mg/dL to mmol/L, multiply by 0.0259.
b
To convert glucose from mg/dL to mmol/L, multiply by 0.0555.

Total stroke (I60-I69) Hemorrhagic stroke (I60-I62) Subarachnoid hemorrhage (I60)

2.6 2.6 2.6
HR (95% CI)

HR (95% CI)

HR (95% CI)

2.2 2.2 2.2

1.8 1.8 1.8

1.4 1.4 1.4

1.0 1.0 1.0

0.6 0.6 0.6

120 140 160 180 200 220 240 260 280 120 140 160 180 200 220 240 260 280 120 140 160 180 200 220 240 260 280

Total cholesterol (mg/dL) Total cholesterol (mg/dL) Total cholesterol (mg/dL)

Intracerebral hemorrhage (I61-I62) Ischemic stroke (I63)
2.6 2.6
HR (95% CI)

HR (95% CI)

2.2 2.2

1.8 1.8

1.4 1.4

1.0 1.0

0.6 0.6
120 140 160 180 200 220 240 260 280 120 140 160 180 200 220 240 260 280

Total cholesterol (mg/dL) Total cholesterol (mg/dL)

Fig. 1. Age- and sex-adjusted HRs* across 8 categories of total cholesterol for stroke mortality.
CI, confidence interval; HR, hazard ratio. *Cox proportional hazard models stratified by baseline age were used.
To convert TC from mg/dL to mmol/L, multiply by 0.0259.

Please cite this article in press as: S.-W. Yi, et al., Total cholesterol and stroke mortality in middle-aged and elderly adults: A prospective cohort
study, Atherosclerosis (2017), https://doi.org/10.1016/j.atherosclerosis.2017.12.003
4 S.-W. Yi et al. / Atherosclerosis xxx (2017) 1e7

Total stroke Hemorrhagic stroke SAH

pnon-linearity <0.001 pnon-linearity = 0.004 pnon-linearity = 0.938

ICH Ischemic stroke

pnon-linearity <0.001 pnon-linearity = 0.280

Fig. 2. HRs* for stroke mortality using restrictive cubic spline analysis.
CI, confidence interval; HR, hazard ratio; ICH, intracerebral hemorrhage; SAH, subarachnoid hemorrhage. *Cox proportional hazards models stratified by baseline age, after
adjustment for age at baseline, smoking status, alcohol use, physical activity, body mass index, systolic blood pressure, and fasting glucose levels were used in persons with total
cholesterol values < 350 mg/dL (n ¼ 502,369). To convert cholesterol from mg/dL to mmol/L, multiply by 0.0259.

TC < 350 mg/dL TC < 200 mg/dL TC of 200-349 mg/dL

Stroke subtypes Deaths 0.7 1.0 1.3 Deaths 0.7 1.0 1.3 Deaths 0.7 1.0 1.3

Total stroke 3,375 1,761 1,614

Hemorrhagic stroke 1,181 640 541

SAH 352 186 166

ICH 829 454 375

Ischemic stroke 1,041 514 527

0.5 0.8 1.1 1.4 0.5 0.8 1.1 1.4 0.5 0.8 1.1 1.4
Hazard ratio Hazard ratio Hazard ratio
p value HR (95% CI) p value HR (95% CI) p value HR (95% CI)
Total stroke 0.311 0.98 (0.95-1.02) 0.002 0.88 (0.80-0.95) 0.022 1.09 (1.01-1.16)

Hemorrhagic stroke 0.003 0.91 (0.86-0.97) <0.001 0.78 (0.68-0.90) 0.312 1.06 (0.94-1.20)

SAH 0.272 0.94 (0.84-1.05) 0.905 1.02 (0.77-1.35) 0.661 1.05 (0.84-1.31)

ICH 0.012 0.91 (0.85-0.98) <0.001 0.72 (0.62-0.85) 0.306 1.08 (0.93-1.25)

Ischemic stroke 0.050 1.06 (1.00-1.13) 0.724 0.97 (0.83-1.14) 0.154 1.09 (0.97-1.24)

Fig. 3. HRs* per each 39 mg/dL (1 mmol/L) increase in total cholesterol (TC), according to TC range.
CI, confidence interval; HR, hazard ratio; ICH, intracerebral hemorrhage; SAH, subarachnoid hemorrhage. *Cox proportional hazard models stratified by baseline age, after
adjustment for the same variables as in Fig. 2.
To convert glucose from mg/dL to mmol/L, multiply by 0.0259.

[1 mmol/L] higher level ¼ 0.88), mainly due to strong inverse as-
sociations with hemorrhagic stroke (HR ¼ 0.78), especially ICH
(HR ¼ 0.72). Assuming a linear association in the upper range, TC
was positively associated with stroke mortality. No subtypes of
stroke had an inverse association in the upper range. In the full
range, assuming linear relationships, TC was positively associated
with mortality from ischemic stroke (HR ¼ 1.06, p ¼ 0.0496).
3.2. Subgroup analysis according to sex, age, and blood pressure
The elderly (aged 65 years) and hypertensive (SBP
140 mmHg) persons had similar estimated relative risks for stroke
mortality than middle-aged (40e64 years old) and non-
hypertensive persons (SBP <140 mmHg) both in the lower and
upper range (Fig. 4). Women had lower HRs for total stroke

Please cite this article in press as: S.-W. Yi, et al., Total cholesterol and stroke mortality in middle-aged and elderly adults: A prospective cohort
study, Atherosclerosis (2017), https://doi.org/10.1016/j.atherosclerosis.2017.12.003
 S.-W. Yi et al. / Atherosclerosis xxx (2017) 1e7 5

(A) Total cohort (B) Age group

All participants ≥ 65 yrs
< 65 yrs

pnon-linearity < 0.001 pnon-linearity (≥ 65 yrs) = 0.005

pnon-linearity (<65 yrs) = 0.023

(C) Sex group (D) SBP group

Men ≥ 140 mmHg
Women < 140 mmHg

pnon-linearity (men) = 0.002 pnon-linearity (≥ 140 mmHg) = 0.025

pnon-linearity (women) = 0.023 pnon-linearity (< 140 mmHg) = 0.007

Fig. 4. HRs* for stroke mortality using restrictive cubic spline analysis according to age, sex, and systolic blood pressure.
HR, hazard ratio; SBP, systolic blood pressure. *Cox proportional hazards models stratified by baseline age, after adjustment for age at baseline, sex (when applicable), smoking
status, alcohol use, physical activity, body mass index, SBP, and fasting glucose levels in persons with total cholesterol values < 350 mg/dL (n ¼ 502,369).
To convert cholesterol from mg/dL to mmol/L, multiply by 0.0259.

mortality in the upper range of cholesterol, but with overlapping
confidence intervals.
For the subtypes of stroke, HRs were generally similar in both
the lower and upper ranges in each age (Supplementary Fig.1-3),
each sex (Supplementary Fig.4-6), and each systolic blood pressure
groups (Supplementary Fig. 7), except perhaps for hemorrhagic
stroke mortality, for which men than women (HR, 0.71 versus 0.92,
pheterogeneity ¼ 0.074) and middle-aged than elderly persons (HR,
0.70 versus 0.90, pheterogeneity ¼ 0.078) seemed to have stronger, yet
statistically not significant, inverse associations. In the elderly, the
HRs per each 39 mg/dL [1 mmol/L] higher TC were 1.11 (1.02e1.21)
for total stroke mortality in the upper range (Supplementary Fig. 3).
3.3. Associations across standard classification of TC
Compared to the desirable level of <200 mg/dL, high TC levels
(240 mg/dL) were not associated with greater stroke mortality
(Supplementary Table 3), while both high TC (HR ¼ 1.17, 95%
CI ¼ [1.06e1.30]) and desirable levels (HR ¼ 1.10, [1.02e1.19]) were
associated with greater stroke mortality, compared to borderline
level. For subtypes of stroke, borderline levels were associated with
lower mortality for hemorrhagic stroke (HR ¼ 0.81) compared to
the desirable level, while high levels were associated with greater
ischemic stroke mortality (HR ¼ 1.28).
In both middle-aged (HR ¼ 1.39) and elderly (HR ¼ 1.23) adults,
high levels of TC, compared to desirable level, were associated with
greater ischemic stroke mortality (Supplementary Table 4).
4. Discussion
4.1. Hemorrhagic stroke
Our study found that TC had a nonlinear association (an L-curve)
with hemorrhagic stroke, especially ICH, with an inverse graded
association in the lower range, but not in the upper range. A sys-
tematic review explored nonlinear associations for cholesterol, and
detected a potential nonlinear association between TC and hem-
orrhagic stroke, but it finally concluded an inverse association be-
tween TC and hemorrhagic stroke [15]. Previous large prospective
cohort studies, including Multiple Risk Factor Intervention Trial
observational study and Prospective Study Collaboration, generally
have reported inverse relationships between TC and hemorrhagic
stroke [10,14,15,24]. Meanwhile, meta-analyses of randomized
controlled trials concluded that cholesterol-lowering statin therapy
did not increase the risk of ICH [13,25]. The current study further
demonstrated that total cholesterol down to 200 mg/dL were not
associated with increased risk of ICH.
For the inverse association in the lower range, few studies

Please cite this article in press as: S.-W. Yi, et al., Total cholesterol and stroke mortality in middle-aged and elderly adults: A prospective cohort
study, Atherosclerosis (2017), https://doi.org/10.1016/j.atherosclerosis.2017.12.003
6 S.-W. Yi et al. / Atherosclerosis xxx (2017) 1e7
examined the association after exclusion of the first 5 years of cases,
and still showed inverse associations [15]. In the current study,
when the first 5 years of follow-up were excluded, the inverse as-
sociations were maintained for hemorrhagic stroke (HR per
1 mmol/L increase in TC ¼ 0.919; 95% CI, 0.848e0.997) and ICH
(HR ¼ 0.896; 0.812e0.988). Several putative mechanisms have
been proposed. An autopsy study suggested that low blood
cholesterol is associated with angionecrosis in cerebral hemor-
rhages, which contributes to fragility of the vascular wall [26].
Coupled with the reduction of platelet aggregability associated
with low cholesterol [27], these mechanisms could increase the risk
of microaneurysm rupture and slower repair after hemorrhages.
Underlying diseases at baseline, which may both lower blood
cholesterol and increase the risk of hemorrhagic stroke such as
cancers and inflammatory diseases might partially explain this
inverse association.
SAH was generally not associated with TC in the current study.
In a recent systematic review, “a meta-analysis limited to pro-
spective studies with major limitations only in an analysis by sex
and exclusion of certain SAH types”, showed a statistically non-
significant association [28]. Our study provides further evidence
that TC is not associated with the risk of SAH.
4.2. Ischemic stroke
The association between cholesterol and ischemic stroke is
controversial. A few [10,14,29,30], but not all [31], studies have
suggested that cholesterol is associated with a higher risk of
ischemic or non-hemorrhagic stroke in young and middle-aged
adults. However, many studies did not show consistent strong
positive associations [31e35], especially for ischemic stroke mor-
tality [36], in the elderly [10,30]. Our study found a modest, but
statistically significant, positive association between TC and
ischemic stroke mortality, and the associations, especially in the
upper range, were very similar between men and women, and
between middle-aged and elderly persons. Further, both middle-
aged and elderly persons had higher ischemic stroke mortality
associated with high TC range (240 mg/dL), compared to desirable
TC level (<200 mg/dL). Hypercholesterolemia should not be
neglected in the prevention of ischemic stroke [37].
4.3. Total stroke
Previous reviews and meta-analyses of cohort studies have
generally concluded that there is no consistent positive association
between cholesterol and overall stroke, especially stroke mortality
[8e10,38]. In the current study, when assuming a linear association
in the full range of TC, TC was not positively associated with stroke
mortality. However, when a nonlinear association was taken into
consideration, our study found a negative association in the lower
range and a positive association in the upper range. Additionally, in
the upper range, no stroke subtypes including ICH had negative
associations with TC.
The associations of TC were quite different between stroke
subtypes, namely ICH, SAH, and ischemic stroke. Among ischemic
stroke subtypes, atherothrombotic stroke, but not lacunar or
embolic stroke, was suggested to be associated with cholesterol
[39]. When these findings combined, for overall stroke, the pattern
of associations and the TC range associated with the lowest risk
may differ across regional and ethnic populations or different time
periods in a population, since the distribution of stroke subtypes
may vary by time period, region, and ethnicity [40,41]. Randomized
trials of statins have shown that statin therapy reduces the risk of
nonfatal stroke, but not fatal stroke [3e6]. It might be partly
explained by varying stroke subtypes in nonfatal and fatal stroke:
Please cite this article in press as: S.-W. Yi, et al., Total cholesterol and stroke mortality in middle-aged and elderly adults: A prospective cohort
study, Atherosclerosis (2017), https://doi.org/10.1016/j.atherosclerosis.2017.12.003
nonfatal stroke may predominantly consist of ischemic stroke
including transient ischemic attack, while the proportion of hem-
orrhagic stroke is relatively high in fatal stroke.
The patterns of associations for stroke and its subtypes were
generally similar between the middle-aged and elderly, as well as
non-hypertensive and hypertensive individuals, which may be in
discordance with a collaborative study [10]. However, that collab-
orative study used data from 61 prospective studies with different
time periods, regions, and ethnicities [10].
Overall, our study provides evidence that TC level around
200 mg/dL, perhaps as low as to 180 mg/dL (equivalent to around
110 mg/dL of low-density lipoprotein-cholesterol in Koreans)
considering the shape of associations, were associated with a low
risk of stroke with no excess risk of ICH.
4.4. Strengths and limitations of the study
The prospective design, a large number of participants, and the
complete follow-up for mortality are main strengths of the study.
The fact that our study population was composed of ethnically
homogeneous individuals living in a similar environment covered
by the same health care system is another major strength [42].
Another strength is that our study estimated the risk associated
with TC levels down to below 140 mg/dL. There are several limi-
tations. First, lipid-lowering medication use during follow-up was
unaccounted for in the analysis. Considering the effect of lipid-
lowering medication, especially statins, the impact of high
cholesterol on stroke mortality may be underestimated to some
degree. Second, the relative risk based on a single measurement of
TC may underestimate the true association, due to a regression
dilution effect [10]. Third, information on other lipid measures was
unavailable such as low-density lipoprotein- and high-density li-
poprotein-cholesterols. It is a limitation, especially since guidelines
for the management of dyslipidemias are more focused on these
subfractions of cholesterol. Fourth, the fact that the study popula-
tion was homogeneously Korean may affect the generalizability of
our findings. Some results, such as the magnitude of relative risk
associated with TC for stroke mortality and the TC range associated
with the lowest stroke mortality, may need further assessment in
other populations with varying TC levels, healthcare utilization,
distribution of stroke subtypes, and environmental and individual
risk factors.
4.5. Conclusions
Various nonlinear and linear associations were found between
TC and mortality from stroke and its subtypes. In the U-curve as-
sociations for overall stroke with a nadir at approximately 200 mg/
dL, the negative association in the lower range <200 mg/dL is
mostly accounted for by hemorrhagic stroke, especially ICH (an L-
curve). Ischemic stroke had a weak but positive association for TC.
No stroke subtype including ICH, was inversely associated with TC
in the upper range 200 mg/dL. The discordance between previous
studies could be partly resolved by these nonlinear associations.
The associations for stroke mortality were generally similar be-
tween middle-aged (40e64 years) and elderly (65 years) adults
and, in the upper range, increments in TC were clearly associated
with higher stroke mortality in the elderly.
Conflict of interest
The authors declared they do not have anything to disclose
regarding conflict of interest with respect to this manuscript.
 S.-W. Yi et al. / Atherosclerosis xxx (2017) 1e7
Financial support
Health examinations were funded and managed by the NHIS of
Korea. This work was supported by research funds from Catholic
Kwandong University (201705850001). Catholic Kwandong Uni-
versity had no role in study design, data collection, the writing of
the manuscript, or the decision to submit it for publication.
Author contributions
SWY conceived the study concept and design, and acquired data.
SWY analyzed the data and wrote the first draft. SWY, DHS, HK, JJY,
and HO interpreted data and contributed to critical revision of the
manuscript. All authors have read and approved of the final sub-
mitted version of the manuscript. SWY is the study guarantor.
Acknowledgements
The authors thank the staff at the Big Data Steering Department
at the NHIS for providing the data and support.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.atherosclerosis.2017.12.003.
References
[1] E.J. Benjamin, M.J. Blaha, S.E. Chiuve, et al., Heart disease and stroke Statistics-
2017 update: a report from the american heart association, Circulation 135
(2017) e146ee603.
[2] GBD DALYs Hale Collaborators, Global, regional, and national disability-
adjusted life-years (DALYs) for 315 diseases and injuries and healthy life ex-
pectancy (HALE), 1990-2015: a systematic analysis for the Global Burden of
Disease Study 2015, Lancet 388 (2016) 1603e1658.
[3] R. Chou, T. Dana, I. Blazina, et al., Statins for prevention of cardiovascular
disease in adults: evidence report and systematic review for the US preventive
services task force, JAMA 316 (2016) 2008e2024.
[4] R. De Caterina, M. Scarano, R. Marfisi, et al., Cholesterol-lowering in-
terventions and stroke: insights from a meta-analysis of randomized
controlled trials, J. Am. Coll. Cardiol. 55 (2010) 198e211.
[5] P. Amarenco, J. Labreuche, Lipid management in the prevention of stroke:
review and updated meta-analysis of statins for stroke prevention, Lancet
Neurol. 8 (2009) 453e463.
[6] C. O'Regan, P. Wu, P. Arora, et al., Statin therapy in stroke prevention: a meta-
analysis involving 121,000 patients, Am. J. Med. 121 (2008) 24e33.
[7] A.L. Catapano, I. Graham, G. De Backer, et al., 2016 ESC/EAS guidelines for the
management of dyslipidaemias, Eur. Heart J. 37 (2016) 2999e3058.
[8] T. Hamazaki, H. Okuyama, Y. Ogushi, et al., Towards a paradigm shift in
cholesterol treatment. A Re-examination of the cholesterol issue in Japan,
Ann. Nutr. Metab. 66 (Suppl 4) (2015) 1e116.
[9] S.A. Peters, Y. Singhateh, D. Mackay, et al., Total cholesterol as a risk factor for
coronary heart disease and stroke in women compared with men: a sys-
tematic review and meta-analysis, Atherosclerosis 248 (2016) 123e131.
[10] Prospective Studies Collaboration, S. Lewington, G. Whitlock, et al., Blood
cholesterol and vascular mortality by age, sex, and blood pressure: a meta-
analysis of individual data from 61 prospective studies with 55,000 vascular
deaths, Lancet 370 (2007) 1829e1839.
[11] P. Amarenco, J. Bogousslavsky, A. Callahan 3rd, et al., High-dose atorvastatin
after stroke or transient ischemic attack, N. Engl. J. Med. 355 (2006) 549e559.
[12] R. Collins, J. Armitage, S. Parish, et al., Effects of cholesterol-lowering with
simvastatin on stroke and other major vascular events in 20536 people with
cerebrovascular disease or other high-risk conditions, Lancet 363 (2004)
757e767.
[13] J.S. McKinney, W.J. Kostis, Statin therapy and the risk of intracerebral hem-
orrhage: a meta-analysis of 31 randomized controlled trials, Stroke 43 (2012)
2149e2156.
[14] J.D. Neaton, H. Blackburn, D. Jacobs, et al., Serum cholesterol level and mor-
tality findings for men screened in the multiple risk factor intervention trial.
Multiple risk factor intervention trial research group, Arch. Intern Med. 152
Please cite this article in press as: S.-W. Yi, et al., Total cholesterol and stroke mortality in middle-aged and elderly adults: A prospective cohort
study, Atherosclerosis (2017), https://doi.org/10.1016/j.atherosclerosis.2017.12.003
7
(1992) 1490e1500.
[15] X. Wang, Y. Dong, X. Qi, et al., Cholesterol levels and risk of hemorrhagic
stroke: a systematic review and meta-analysis, Stroke 44 (2013) 1833e1839.
[16] R.M. Conroy, K. Pyorala, A.P. Fitzgerald, et al., Estimation of ten-year risk of
fatal cardiovascular disease in Europe: the SCORE project, Eur. Heart J. 24
(2003) 987e1003.
[17] R.B. D'Agostino Sr., R.S. Vasan, M.J. Pencina, et al., General cardiovascular risk
profile for use in primary care: the framingham heart study, Circulation 117
(2008) 743e753.
[18] J. Hippisley-Cox, C. Coupland, P. Brindle, Development and validation of
QRISK3 risk prediction algorithms to estimate future risk of cardiovascular
disease: prospective cohort study, BMJ 357 (2017) j2099.
[19] US Department of Health and Human Services, Healthy People 2020. 2020
Topics & Objective: Heart Disease and Stroke., In.
[20] S.C. Seong, Y.Y. Kim, >S.K. Park, et al., Cohort profile: the National Health In-
surance Service-National Health Screening Cohort (NHIS-HEALS) in Korea,
BMJ Open 7 (2017), e016640.
[21] S.W. Yi, H. Ohrr, S.A. Shin, et al., Sex-age-specific association of body mass
index with all-cause mortality among 12.8 million Korean adults: a pro-
spective cohort study, Int. J. Epidemiol. 44 (2015) 1696e1705.
[22] J.W. Park, S.Y. Lee, S.Y. Kim, et al., BMI and stroke risk in Korean women, Obes.
(Silver Spring) 16 (2008) 396e401.
[23] W.K. Min, K.D. Kim, D.J. Kim, et al., Annual report on external quality
assessment in clinical chemistry in Korea, J. Laboratory Med. Qual. Assur. 25
(2003) (2002) 1e14.
[24] K. Yano, D.M. Reed, C.J. MacLean, Serum cholesterol and hemorrhagic stroke in
the honolulu heart program, Stroke 20 (1989) 1460e1465.
[25] D.G. Hackam, M. Woodward, L.K. Newby, et al., Statins and intracerebral
hemorrhage: collaborative systematic review and meta-analysis, Circulation
124 (2011) 2233e2242.
[26] M. Konishi, H. Iso, Y. Komachi, et al., Associations of serum total cholesterol,
different types of stroke, and stenosis distribution of cerebral arteries. The
Akita Pathology Study, Stroke 24 (1993) 954e964.
[27] N. Tandon, J.T. Harmon, D. Rodbard, et al., Thrombin receptors define
responsiveness of cholesterol-modified platelets, J. Biol. Chem. 258 (1983)
11840e11845.
[28] J.V. Lindbohm, J. Kaprio, M. Korja, Cholesterol as a risk factor for subarachnoid
hemorrhage: a systematic review, PLoS One 11 (2016) e0152568.
[29] S. Ebrahim, J. Sung, Y.M. Song, et al., Serum cholesterol, haemorrhagic stroke,
ischaemic stroke, and myocardial infarction: Korean national health system
prospective cohort study, BMJ 333 (2006) 22.
[30] R.B. Horenstein, D.E. Smith, L. Mosca, Cholesterol predicts stroke mortality in
the Women's Pooling Project, Stroke 33 (2002) 1863e1868.
[31] J.S. Choi, Y.M. Song, J. Sung, Serum total cholesterol and mortality in middle-
aged Korean women, Atherosclerosis 192 (2007) 445e447.
[32] R. Cui, H. Iso, H. Toyoshima, et al., Serum total cholesterol levels and risk of
mortality from stroke and coronary heart disease in Japanese: the JACC study,
Atherosclerosis 194 (2007) 415e420.
[33] E. Shahar, L.E. Chambless, W.D. Rosamond, et al., Plasma lipid profile and
incident ischemic stroke: the Atherosclerosis Risk in Communities (ARIC)
study, Stroke 34 (2003) 623e631.
[34] Y. Zhang, J. Tuomilehto, P. Jousilahti, et al., Total and high-density lipoprotein
cholesterol and stroke risk, Stroke 43 (2012) 1768e1774.
[35] M.J. O'Donnell, D. Xavier, L. Liu, et al., Risk factors for ischaemic and intrace-
rebral haemorrhagic stroke in 22 countries (the INTERSTROKE study): a case-
control study, Lancet 376 (2010) 112e123.
[36] X. Zhang, A. Patel, H. Horibe, et al., Cholesterol, coronary heart disease, and
stroke in the Asia Pacific region, Int. J. Epidemiol. 32 (2003) 563e572.
[37] P.U. Heuschmann, J. Kircher, T. Nowe, et al., Control of main risk factors after
ischaemic stroke across Europe: data from the stroke-specific module of the
EUROASPIRE III survey, Eur. J. Prev. Cardiol. 22 (2015) 1354e1362.
[38] J.F. Meschia, C. Bushnell, B. Boden-Albala, et al., Guidelines for the primary
prevention of stroke: a statement for healthcare professionals from the
American Heart Association/American Stroke Association, Stroke 45 (2014)
3754e3832.
[39] T. Tanaka, T. Okamura, Blood cholesterol level and risk of stroke in
community-based or worksite cohort studies: a review of Japanese cohort
studies in the past 20 years, Keio J. Med. 61 (2012) 79e88.
[40] GBD Mortality Causes of Death Collaborators, Global, regional, and national
life expectancy, all-cause mortality, and cause-specific mortality for 249
causes of death, 1980-2015: a systematic analysis for the Global Burden of
Disease Study 2015, Lancet 388 (2016) 1459e1544.
[41] B.J. Kim, J.S. Kim, Ischemic stroke subtype classification: an asian viewpoint,
J. Stroke 16 (2014) 8e17.
[42] D.T. Ko, D.A. Alter, H. Guo, et al., High-density lipoprotein cholesterol and
cause-specific mortality in individuals without previous cardiovascular con-
ditions: the CANHEART study, J. Am. Coll. Cardiol. 68 (2016) 2073e2083.