Jee SH, Suh I, Kim IS, Appel LJ. Smoking and Atherosclerotic Cardiovascular Disease in Men With Low Levels of Serum CholesterolThe Korea Medical Insurance Corporation Study. JAMA. 1999;282(22):2149-2155. doi:10.1001/jama.282.22.2149
Author Affiliations: Department of Epidemiology and Disease Control, Graduate School of Health Science and Management, Yonsei University, Seoul, Republic of Korea (Dr Jee); Department of Preventive Medicine and Public Health, Yonsei University College of Medicine, Seoul, Republic of Korea (Drs Suh and Kim); and Welch Center for Prevention, Epidemiology and Clinical Research, Johns Hopkins Medical Institutions, Baltimore, Md (Dr Appel).
Context Few studies have examined the interactive effects of smoking and serum
cholesterol level on morbidity and mortality from cardiovascular dieseases.
In East Asia, where the prevalence of smoking is among the highest in the
world, morbidity and mortality from ischemic heart disease (IHD) is rapidly
Objectives To determine whether cigarette smoking is an independent risk factor
for atherosclerotic cardiovascular disease (ASCVD) in the Republic of Korea
(South Korea), a population that has relatively low levels of serum cholesterol,
and to determine whether serum cholesterol levels modify the risk relationship
between smoking and ASCVD.
Design Prospective cohort study with a follow-up period of 6 years (1993-1998).
Setting and Subjects A total of 106,745 Korean men aged 35 to 59 years who received health
insurance from the Korea Medical Insurance Corporation and who had biennial
medical evaluations in 1990 and 1992.
Main Outcome Measures Hospital admissions and deaths from IHD, cerebrovascular disease (CVD),
and total ASCVD.
Results At baseline, 61,389 (58%) were current cigarette smokers and 64,482
(60%) had a total cholesterol level of less than 5.17 mmol/L (200 mg/dL).
Between 1993 and 1998, 1006 IHD events (176 per 100,000 person-years), 1364
CVD events (238 per 100,000 person-years), and 716 other ASCVD events (125
per 100,000 person-years) occurred. In multivariate Cox proportional hazard
models controlling for age, hypertension, hypercholesterolemia, and diabetes,
current smoking increased the risk of IHD (risk ratio [RR], 2.2; 95% confidence
interval [CI], 1.8-2.8), CVD (RR, 1.6; 95% CI, 1.4-1.8), and total ASCVD (RR,
1.6; 95% CI, 1.5-1.8). For each outcome, there were significant dose-response
relationships with amount and duration of smoking. Throughout the range of
serum cholesterol levels, current smoking significantly increased the risk
of IHD and CVD. In the lowest quartile of serum cholesterol levels (<4.42
mmol/L [171 mg/dL]), the RR from current smoking was 3.3 (95% CI, 1.7-6.2)
for IHD and 1.6 (95% CI, 1.2-2.3) for CVD. There was no evidence of an interaction
between smoking and serum cholesterol (P for interaction
= .75, .87, and .92 for IHD, CVD, and total ASCVD, respectively).
Conclusions This study demonstrates that in Korea smoking is a major independent
risk factor for IHD, CVD, and ASCVD and that a low cholesterol level confers
no protective benefit against smoking-related ASCVD.
Morbidity and mortality from ischemic heart disease (IHD) is rapidly
escalating in economically developing countries, including those of East Asia.
In the Republic of Korea (South Korea), atherosclerotic cardiovascular disease
(ASCVD) became the leading cause of death in the early 1990s, accounting for
approximately 30% of all registered deaths.1,2
Of particular concern is the emerging epidemic of IHD in this country. Between
1981 and 1994, age-adjusted IHD mortality increased from 1.8 deaths per 100,000
person-years to 11.2 deaths per 100,000 person-years in Korean men and from
1.0 deaths per 100,000 person-years to 6.8 deaths per 100,000 person-years
in Korean women.1 The escalating burden of
IHD in Korea may be related to adverse changes in traditional ASCVD risk factors,
particularly cigarette smoking.3
Cigarette smoking is widely recognized as a major risk factor for IHD
in Western countries.4- 9
Nonetheless, few studies have examined the relationship between cigarette
smoking and IHD in Korea and other East Asian countries10- 12
where the prevalence of smoking is among the highest in the world (72% in
Korean men,13 50% in Chinese men,11
and 58% in Japanese men14). Despite the extremely
high prevalence of smoking, populations in these countries tend to be leaner
and have lower levels of serum cholesterol and perhaps less atherosclerosis
than their Western counterparts. Such differences have led to the belief that
smoking is a less important risk factor for IHD in East Asian populations.
Because smoking mediates its effects, in part, through adverse changes in
platelet adhesion and hemostasis, ie, smoking acts as a triggering factor
for IHD events in the presence of atherosclerosis,15
there is also a biological basis for exploring interactions between smoking
and serum cholesterol.
Surprisingly, few studies have examined the interactive effects of smoking
and serum cholesterol on IHD morbidity and mortality. Of these studies, virtually
all were conducted in Western countries where the effects of serum cholesterol
and smoking appear synergistic, at least in populations with high serum cholesterol
levels.16,17 In contrast, smoking
was not an IHD risk factor in several observational studies conducted in populations
with low serum cholesterol levels. In the Puerto Rico Heart Health Program,
where the mean total cholesterol level was 5.22 mmol/L (202 mg/dL), smoking
did not predict the occurrence of IHD.18 Likewise,
in a widely cited study that compared IHD risk factors in men of Japanese
ancestry living in Hawaii with men living in Japan, smoking was significantly
associated with IHD in Hawaii but not Japan.19
In another prospective study of Japanese adults, smoking was significantly
associated with coronary heart disease in persons with total cholesterol levels
of 4.65 mmol/L (180 mg/dL) or higher but not in persons with cholesterol levels
of less than 4.65 mmol/L (<180 mg/dL).10
Even though many of these studies had limited power to detect an association
between smoking and ASCVD at low levels of serum cholesterol as well as limited
power to detect an interaction between smoking and cholesterol, such results
have contributed to the belief that smoking is not an important ASCVD risk
factor in populations with low serum cholesterol levels.15,19
In this setting, we examined prospectively the main and interaction
effects of smoking and serum cholesterol on ASCVD in Korean men, a population
with relatively low levels of serum cholesterol.
The Korea Medical Insurance Corporation (KMIC) provides health insurance
to civil service workers, teachers, and their dependents in the Republic of
Korea (South Korea). Of the entire Korean population (approximately 43 million
in 1990), 4,603,361 (11%) were insured by KMIC, including 1,213,594 workers
and their 3,389,767 dependents. All insured workers are required to participate
in biennial medical examinations performed by KMIC. In 1990 and 1992, 95%
and 94% completed biennial examinations.
The KMIC Study is a prospective cohort study designed to assess risk
factors for ASCVD in Korean men and women.3
Because only 0.6% of the women smoked, we restricted our analyses to men.
The KMIC Study cohort consists of 112,105 men, aged 35 to 59 years, who attended
both the 1990 and 1992 examinations; the cohort was a 25% systematic sample
with a random start, drawn from insured members ordered on the national identification
number. Of the 112,105 men, we excluded 5360 smokers (4.8%) with incomplete
data on the amount and/or duration of smoking. Hence, the final sample size
was 106,745 (Table 1).
The KMIC biennial examinations are conducted in a standardized fashion
by medical staff at local hospitals. In 1990, examinations were conducted
in 416 hospitals. A questionnaire was administered to each participant 3 to
4 days before the examination. In the 1992 questionnaire, participants were
asked to describe their smoking habits, including the number of cigarettes
smoked per day and the duration of cigarette smoking in years, along with
other health habits, including alcohol consumption. Completed questionnaires
were reviewed and edited by trained staff.
The 1990 and 1992 biennial medical examinations included measurements
of weight, height, and blood pressure. Blood pressure was measured in the
seated position by a registered nurse or blood pressure technician using a
standard mercury sphygmomanometer or automatic manometer. In the case of manual
manometers, systolic blood pressure and diastolic blood pressure were measured
as the first and fifth Korotkoff sounds, respectively. One measurement was
taken. A fasting serum specimen was drawn and analyzed for total cholesterol
and serum glucose. Each hospital that participated in the examination had
internal and external quality control procedures directed by the Korean Association
of Laboratory Quality Control.
Using data collected in the 1992 examination, participants were classified
as "current" smokers if they smoked currently for at least 1 year, "nonsmokers"
if they never smoked, and "ex-smokers" if they smoked but quit. Current smokers
were further classified by the average number of cigarettes smoked per day
(1-9, 10-19, and ≥20 cigarettes per day) and duration of smoking (1-19,
20-29, and ≥30 years). Body mass index was calculated as weight in kilograms
divided by height in meters squared. Hypertension was defined as a systolic
blood pressure of at least 140 mm Hg or a diastolic blood presure of at least
90 mm Hg; stages of hypertension were further classified according to the
Sixth Report of the Joint National Committee on the Detection, Evaluation,
and Treatment of High Blood Pressure (JNC VI) criteria.20
Using National Cholesterol Education Program guidelines,21
serum total cholesterol was classified as desirable (serum cholesterol level,
<5.17 mmol/L [200 mg/dL]), borderline-high (serum cholesterol level, 5.17-6.20
mmol/L [200-239 mg/dL]), and high (serum cholesterol level, ≥6.21 mmol/L
[240 mg/dL]). Using diagnostic criteria from the National Diabetes Data Group,22 diabetes was defined by a fasting serum glucose level
of 6.99 mmol/L (126 mg/dL) or higher. Categories of average alcohol consumption
were 0 g/d, 1 to 20 g/d, and more than 20 g/d, as used in previous studies.11 In our analyses, blood pressure, total cholesterol,
fasting serum glucose, and body mass index were the average of measurements
collected in 1990 and 1992.
The principal outcome variables were morbidity and mortality from (1)
IHD alone (International Classification of Diseases, Ninth
Revision [ICD-9] codes 410-414), (2) cerebrovascular disease (CVD)
alone (ICD-9 codes 430-438), and (3) total ASCVD.
The latter category included hypertensive disease (ICD-9 codes 401-405), IHD (ICD-9 codes 410-414),
hemorrhagic stroke (ICD-9 codes 430-432), thrombotic
stroke (ICD-9 codes 433-434), other stroke (ICD-9 codes 435-438), other heart disease likely related
to ASCVD (ICD-9 codes 426-429), sudden death (ICD-9 code 798), and other vascular disease (ICD-9 codes 440-444). For those individuals with more than 1 event,
we used just the first event in our analyses. The follow-up period, which
lasted 6 years, was January 1993 to December 1998.
Outcomes were ascertained from diagnoses on hospital discharge summaries
and from causes of death on death certificates. In Korea, professionally trained
and certified medical chart recorders abstract charts and assign discharge
diagnoses in a standardized fashion using World Health Organization codes
for common diseases such as stroke and myocardial infarction. Likewise, these
recorders complete death certificates using information provided by physicians.
Computerized searches of death certificate data from the National Statistical
Office in Korea were performed for each of the KMIC enrollees; hence, in terms
of mortality, follow-up is likely to be 100% complete. For morbidity, which
is defined exclusively by hospital discharge diagnoses, follow-up is likely
to be quite high, if not close to 100%, because hospitals cannot receive payments
until the bill with discharge diagnoses is submitted to KMIC. For morbidity,
the duration of follow-up was 6 years for those persons without an ASCVD event;
for those experiencing an ASCVD event, follow-up time was the interval between
January 1, 1993, and the date of the event; for those who retired (21.8%)
or those who switched medical insurance because of a job change (0.1%), follow-up
time was the interval between January 1, 1993, and the date of retirement
or job change, respectively. Just 4.4% were unavailable for follow-up.
In bivariate analyses, we examined the relationship between smoking
status and traditional ASCVD risk factors, adjusting for age using the age
distribution of the 1990 Korean Census population. In these bivariate analyses,
we tested for trends across categories of amount of current smoking, using
nonsmokers as the reference. For ASCVD risk factors with a continuous distribution,
we used simple linear regression and entered an ordinal variable for the categories
of current smoking ("0" for nonsmokers, "1" for 1-9 cigarettes per day, "2"
for 10-19 cigarettes per day, and "3" for ≥20 cigarettes per day). For
dichotomous variables, we used the method of Mantel-Haenszel.23
In these models, ex-smokers were excluded.
Cox proportional hazards models were used to assess the independent
effects of smoking (both current smokers and ex-smokers) on IHD, CVD, and
ASCVD events, controlling for age and traditional risk factors (hypertension,
hypercholesterolemia, and diabetes). For the CVD analyses, we assessed the
impact of smoking with and without adjustment for alcohol consumption. To
determine whether smoking was an independent risk factor throughout the range
of total cholesterol, we performed separate Cox proportional hazard models
within each quartile of serum cholesterol. After excluding ex-smokers, tests
of interaction were performed in Cox proportional hazard models adding terms
for current smoking (yes/no), cholesterol status (<5.17 mmol/L [200 mg/dL]
vs ≥5.17 mmol/L [200 mg/dL]), and a corresponding interaction term.24 To calculate the population attributable risk from
cigarette smoking and other ASCVD risk factors, we used the formula devised
by Levin25; separate population attributable
risk analyses were performed in those with a total cholesterol level of less
than 5.17 mmol/L [200 mg/dL]. In all analyses, a 2-sided α level of
.05 was considered statistically significant.
The mean (SD) age of study participants was 45 (6.7) years. Among the
106,745 men, 61,389 (58%) were current smokers and 22,739 (21%) were ex-smokers;
23,022 (22%) had stage 1, 5128 (4.8%) stage 2, and 1721 (1.6%) stage 3 hypertension.
With respect to total cholesterol, 64,482 (60%) had a total cholesterol level
of less than 5.17 mmol/L (200 mg/dL), 32,973 (31%) had a borderline level
of 5.17-6.20 mmol/L (200-239 mg/dL), and 9286 (8.7%) had a level of 6.21 mmol/L
(240 mg/dL) or higher. Among current smokers, 62% smoked for more than 20
years; 27%, 41%, and 32% of current smokers smoked 1 to 9, 10 to 19, and 20
or more cigarettes per day, respectively. Characteristics of nonsmokers, ex-smokers,
and current smokers are presented in Table
1. After adjustment for age, current smokers compared with nonsmokers
had significantly lower diastolic blood pressure (P
for trend = .002), consumed more alcohol (P for trend
= .04), and had a lower prevalence of hypertension (P
for trend = .008).
During 6 years of follow-up (572,645 person-years), 3086 men (2.9%)
were either hospitalized or died of ASCVD. Of these, 1006 IHD events (176/100,000
person-years), 1364 CVD events (238/100,000 person-years), and 716 other ASCVD
events (125/100,000 person-years) occurred; these events included 131 deaths
(23/100,000 person years) from IHD, 270 deaths (47/100,000 person-years) from
CVD, and 168 deaths (29/100,000 person-years) from other ASCVD. These 569
deaths from ASCVD were 18% of all deaths that occurred during follow-up.
The independent effects of smoking on IHD, CVD, and total ASCVD were
examined in Cox proportional hazards models that simultaneously controlled
for age, hypertension, hypercholesterolemia, and diabetes (Table 2). Compared with nonsmokers, the risk ratio (RR) for IHD
was 2.2 in smokers (P<.001) and 2.1 in ex-smokers
(P<.001). For CVD, the corresponding RRs were
1.6 in smokers (P<.001) and 1.1 in ex-smokers
(P = .42). Compared with nonsmokers, the RRs for
any ASCVD event were 1.6 (P<.001) in smokers and
1.3 (P<.001) in ex-smokers. In these models, hypertension
status was associated with cardiovascular events in a progressive, dose-response
fashion. Borderline and high total cholesterol levels were associated with
IHD and total ASCVD; high cholesterol but not borderline cholesterol was associated
with CVD alone. In other analyses, current smoking was significantly associated
with the occurrence of hemorrhagic stroke (RR = 1.5, P
= .001) and thrombotic stroke (RR = 1.9, P<.001).
After additional adjustment for alcohol intake, the corresponding RRs were
1.5 (P = .001) and 2.0 (P<.001).
After adjustment for age, hypertension, hypercholesterolemia, and diabetes,
the RRs for IHD alone, CVD alone, and total ASCVD were determined in groups
classified by amount of smoking (Figure 1) and duration of smoking (Figure
2). For both IHD and CVD, the risk of events increased progressively
with higher levels of daily consumption (P for trend,
<.01 and <.05, respectively) and with increased duration of smoking
(P for trend, <.01 and <.05, respectively).
To assess the relationship between smoking and ASCVD events by level
of total cholesterol, we divided the cohort into quartiles of total cholesterol
(<4.42 mmol/L [171 mg/dL], 4.42-4.91 mmol/L [171-190 mg/dL], 4.92-5.51
mmol/L [191-213 mg/dL], and >5.51 mmol/L [213 mg/dL]). In these quartiles,
the number of IHD events was 156, 209, 249, and 392, respectively, while the
number of CVD events was 291, 293, 341, and 439, respectively. In each quartile
of serum cholesterol, current smoking was significantly associated with both
IHD and CVD (Figure 3). The RR from
current smoking in the lowest quartile of serum cholesterol was 3.3 (95% confidence
interval [CI], 1.7-6.2) for IHD and 1.6 (95% CI, 1.2-2.3) for CVD. In Cox
proportional hazards models that tested for interactions between current smoking
(yes/no) and total cholesterol level (≥5.17 mmol/L [200 mg/dL] vs <5.17
mmol/L [200 mg/dL]), there was no evidence of an interaction (P for interaction = .75, .87, and .92 for IHD, CVD, and total ASCVD,
For current smoking and other traditional risk factors, we estimated
the population attributable risks for IHD alone, CVD alone, and total ASCVD
using risk factor prevalence estimates from this study (Table 3). For IHD, current smoking accounted for approximately 41%
of events and hypertension for 21% of events. For CVD, corresponding estimates
were 26% from smoking and 35% from hypertension. Among men with a total cholesterol
level of less than 5.17 mmol/L (200 mg/dL), current smoking accounted for
41% of IHD events and for 29% of CVD events.
In this large prospective, observational study of Korean men with relatively
low levels of serum cholesterol, we documented that current cigarette smoking
was a strong, independent risk factor for IHD, CVD, and total ASCVD events.
These risk relationships were present throughout the range of total serum
cholesterol, including the lowest quartile (<4.42 mmol/L [171 mg/dL]).
Furthermore, in the context of a high prevalence of cigarette smoking, our
data indicate that this risk factor accounts for a substantial fraction of
ASCVD events in Korean men (ie, >40% of IHD events), even among those individuals
with low levels of serum cholesterol.
Previous observation studies conducted in East Asian countries and in
other countries with populations that have low levels of serum cholesterol
yielded inconsistent results. In 1 recent observational study conducted in
China, cigarette smoking was significantly associated with IHD mortality.11 However, the IHD analyses were only adjusted for
age and alcohol intake, not for traditional ASCVD risk factors. In an initial
and subsequent report from the Nippon, Honolulu, and San Francisco (NI-HON-SAN)
Study, cigarette smoking was a significant, independent risk factor for coronary
heart disease in Japanese men living in Hawaii but not in those living in
Japan, where serum cholesterol levels were relatively low.12,19
In the Hisayama, Japan, cohort study, smoking was significantly associated
with coronary heart disease in persons with a total cholesterol level of 4.65
mmol/L (180 mg/dL) or higher but not in persons with a cholesterol level of
less than 4.65 mmol/L (180 mg/dL).10 In an
autopsy series from the same study, smoking was not a risk factor for coronary
atherosclerosis.26 Finally, in the Puerto Rico
Heart Health Program, where the mean total cholesterol level was 5.22 mmol/L
(202 mg/dL), smoking did not predict the occurrence of IHD.18
In the context of these equivocal results, our study provides unambiguous
evidence that current cigarette smoking is a risk factor for IHD and CVD,
even among persons with low serum cholesterol levels.
The KMIC Study also demonstrated that cigarette smoking was a significant
risk factor for cerebrovascular events and stroke subtypes, both thrombotic
and hemorrhagic. In this respect, our data are consistent with observational
data from the United States,8,27
but not with recent data from China.11 In the
latter study, cigarette smoking did not predict the occurrence of stroke mortality
in analyses adjusting for age and alcohol use. This inconsistency may have
resulted from differences in study power, ie, 197 events in the Yuan study
vs 1364 events in the KMIC Study.
In our analyses, ex-smokers had a similar risk of IHD as current smokers.
In contrast, the risk of CVD in ex-smokers was not significantly different
from that of nonsmokers. This pattern of findings suggests that the high risk
of CVD subsides after cessation of smoking, while the risk of IHD persists
for a longer, albeit unknown, period of time. Unfortunately, the KMIC data
do not have information on quit dates in ex-smokers.
Among the strengths of the KMIC Study are high follow-up rates, repeated
measures of several exposures (blood pressure, total serum cholesterol, fasting
serum glucose, weight, and height), and its large, national sample. Other
aspects of the study population deserve comment. The participants in the KMIC
Study tend to be middle-class, employed individuals, who may be healthier
than the general population in Korea. Although incidence rates of disease
likely differ by socioeconomic status, there is little reason to suspect that
risk factor-disease relationships should markedly differ. Second, because
smoking remains uncommon among Korean women (just 0.6%),3
we restricted our analyses to men. Finally, most KMIC enrollees were middle-aged.
Hence, patterns of disease may change as the cohort ages. In particular, proportionate
morbidity and mortality from ASCVD should increase with time. Overall, results
from this cohort study should be generalizable to other populations in East
Potential limitations of our study include the relatively brief duration
of follow-up, inclusion of individuals with prevalent ASCVD in the cohort,
and reliance on diagnoses from discharge summaries and death certificates.
Although the duration of follow-up in our analyses is just 6 years, the large
size of the cohort (>100,000 participants) provided sufficient statistical
power, even in subgroup and dose-response analyses. The inclusion of persons
with antecedent ASCVD events could potentially lead to biased estimates. The
bias, however, is likely to be conservative, because such individuals tend
to adopt desirable habits after clinical ASCVD events; hence, the category
of nonsmokers may be differentially enriched with persons who stopped smoking
after a myocardial infarction or stroke. The impact of prevalent ASCVD is
also diminished because individuals who experienced ASCVD events between January
1990 and December 1992, the years of baseline data collection, were excluded.
Reliance on diagnoses from hospitalizations and death certificates may introduce
random and systematic errors. Random error would tend to diminish the study's
power to detect associations. Systematic error could alter the distribution
of events and perhaps risk factor-disease relationships if the errors were
related to exposure status. However, the consistency of our findings, ie,
significant relationships of current smoking within the broad category of
ASCVD events as well as the component categories of IHD and CVD events, suggest
that major systematic errors related to the coding of ASCVD events was unlikely.
In conclusion, current cigarette smoking is a major, independent risk
factor for IHD, CVD, and total ASCVD in Korea, an East Asian country with
a population that has a high prevalence of smoking and relatively low levels
of serum cholesterol. A low cholesterol level confers no protective benefit
against the harmful effects of smoking on ASCVD.