Cholesterol and Mortality

116 Pages·2015·1.52 MB·English

Ann Nutr Metab 2015;66(suppl 4):1–116 Published online: April 30, 2015
DOI: 10.1159/000381654
Introduction
High cholesterol levels are recognized as a major cause less support from pharmaceutical companies than re-
of atherosclerosis. However, for more than half a century searchers overseas do? Not at all. Because Japanese re-
some have challenged this notion. But which side is cor- searchers are indolent and weak? No, of course not. Be-
rect, and why can’t we come to a definitive conclusion cause the Japanese public is skeptical about the benefits
after all this time and with more and more scientific data of medical therapy? No, they generally accept everything
available? We believe the answer is very simple: for the physicians say; unfortunately, this is also complicated by
side defending this so-called cholesterol theory, the the fact that physicians don’t have enough time to study
amount of money at stake is too much to lose the fight. the cholesterol issue by themselves, leaving them simply
The issue of cholesterol is one of the biggest issues in to accept the information provided by the pharmaceutical
medicine where the law of economy governs. Moreover, industry. Reading through this supplementary issue, it
advocates of the theory take the notion to be a simple, ir- will become clear why Japan can be the starting point for
refutable ‘fact’ and self-explanatory. They may well think the anti-cholesterol theory campaign. The relationship
that those who argue against the cholesterol theory—ac- between all-cause mortality and serum cholesterol levels
tually, the cholesterol ‘hypothesis’—are mere eccentrics. in Japan is a very interesting one: mortality actually goes
We, as those on the side opposing the hypothesis, under- down with higher total or low density lipoprotein (LDL)
stand their argument very well. Indeed, the first author of cholesterol levels, as reported by most Japanese epidemi-
this supplementary issue (TH) had been a very strong be- ological studies of the general population. This relation-
liever and advocate of the cholesterol hypothesis up until ship cannot be observed as easily in other countries, ex-
a couple of years after the Scandinavian Simvastatin Sur- cept in elderly populations where the same relationship
vival Study (4S) reported the benefits of statin therapy in exists worldwide. The mortality from coronary heart dis-
The Lancet in 1994. To be honest with the readers, he used ease in Japan has accounted for around just 7% of all-
to persuade people with high cholesterol levels to take cause mortality for decades; a much lower rate than seen
statins. He even gave a talk or two to general physicians in Western countries. The theory that the lower the cho-
promoting the benefits of statins. Terrible, unforgivable lesterol levels are, the better is completely wrong in the
mistakes given what we came to know and clearly know case of Japan—in fact, the exact opposite is true. Because
now. Japan is unique in terms of cholesterol-related phenom-
In this supplementary issue, we explore the back- ena, it is easy to find flaws in the cholesterol hypothesis.
ground to the cholesterol hypothesis utilizing data ob- Based on data from Japan, we propose a new direction in
tained mainly from Japan—the country where anti-cho- the use of cholesterol medications for global health pro-
lesterol theory campaigns can be conducted more easily motion; namely, recognizing that cholesterol is a negative
than in any other countries. But why is this? Is it because risk factor for all-cause mortality and re-examining our
the Japanese researchers defending the hypothesis receive use of cholesterol medications accordingly. This, we be-
© 2015 S. Karger AG, Basel
0250–6807/15/0668–0001$39.50/0
E-Mail [email protected]
www.karger.com/anm
lieve, marks the starting point of a paradigm shift in not than before, and we hope that we lay out the case for why
only how we understand the role cholesterol plays in a paradigm shift in cholesterol treatment is needed, and
health, but also how we provide cholesterol treatment. sooner rather than later. We would like to stress in closing
The guidelines for cholesterol are thus another area of that we have received no funding in support of writing or
great importance. Indeed, the major portion of this sup- publishing this supplementary issue, and our conflicts of
plementary issue (from Chapter 4 onward) is given over interest statements are given in full at the end.
to our detailed examination and critique of guidelines
published by the Japan Atherosclerosis Society. We ded-
icate a large portion of this work to these guidelines be- A Note on the Units Used in This Issue
cause they are generally held in high regard in Japan, and
the country’s public health administration mechanism We use two unit systems to report blood cholesterol
complies with them without question. Physicians, too, and triglyceride concentrations: mmol/l and mg/dl. This
tend to simply obey the guidelines; their workloads often is because Japanese researchers (and probably American
don’t allow them to explore the issue rigorously enough researchers) are not well accustomed to using the mmol/l
to learn the background truth and they are afraid of liti- system. If the original papers we cite used the mmol/l
gation if they don’t follow the guidelines in daily practice. system, we use that system first followed by mg/dl, and
These chapters clearly describe some of the flaws in the vice versa. The following equations are used in this sup-
guidelines—flaws which are so serious that it becomes plementary issue for most of the part: 1 mmol/l = 38.67
clear that times must change and the guidelines must be mg/dl for cholesterol and 1 mmol/l = 89 mg/dl for tri-
updated. glycerides (we use two significant digits for the coeffi-
Our purpose in writing this supplementary issue is to cient for triglycerides because it depends on fatty acid
help everyone understand the issue of cholesterol better species).
2 Ann Nutr Metab 2015;66(suppl 4):1–116 Hamazaki/Okuyama/Ogushi/Hama
DOI: 10.1159/000381654
Ann Nutr Metab 2015;66(suppl 4):1–116
DOI: 10.1159/000381654
Chapter 1 Cholesterol and Mortality
Summary: All-cause mortality is the most appropriate all-cause mortality, the following paradoxical situation
outcome to use when investigating risk factors for life- might result.
threatening disease. Section 1 discusses all-cause mortal- Suppose that A is a potentially life-threatening disease
ity according to cholesterol levels, as determined by large and that B is a risk factor for A, then people who are at the
epidemiological studies in Japan. Overall, an inverse best end of the risk factor B continuum in terms of all-
trend is found between all-cause mortality and total (or cause mortality are probably the healthiest and last to be
low density lipoprotein [LDL]) cholesterol levels: mortal- treated for risk factor B. This simple principle should be
ity is highest in the lowest cholesterol group without ex- respected across the board, otherwise many of us will be
ception. If limited to elderly people, this trend is univer- treated with unnecessary medicines. It is highly unfortu-
sal. As discussed in Section 2, elderly people with the nate, then, that this is the case for cholesterol in Japan,
highest cholesterol levels have the highest survival rates and probably in all advanced countries. This section dis-
irrespective of where they live in the world. cusses the relationship between cholesterol levels and all-
cause mortality in Japan, and you may find that what you
learned about cholesterol is not actually the case.
(1) Cholesterol and All-Cause Mortality in Japan Fig. 1-1 shows the relationship between all-cause mor-
tality and LDL cholesterol levels in Japan, as determined
All-cause mortality is the most appropriate outcome by the largest epidemiological study—the Ibaraki Prefec-
for both interventional and epidemiological studies to ture Health Study—carried out in Japan in recent years
use when investigating risk factors for life-threatening [2]. Men and women (n = 91,219) aged 40–79 years with
disease. As shown in table 1, the PDQ® Levels of Evi- no history of stroke or coronary heart disease (CHD)
dence for Adult and Pediatric Cancer Treatment Studies were followed for 10.3 years. The hazard ratio (HR) of all-
(National Cancer Institute) classify total mortality (i.e., cause mortality adjusted for age and many potential con-
overall survival from a defined time) as the most impor- founding factors was calculated according to LDL choles-
tant outcome to patients, the most easily defined, and the terol levels and revealed that all-cause mortality was es-
least subject to investigator bias [1]. The table was origi- sentially inversely correlated with LDL cholesterol levels
nally created for cancer treatment studies, but points in both men and women.
A-D can be seen to apply to any life-threatening disease. The first reaction of well-informed advocates of the
If focus is placed on cause-specific mortality instead of cholesterol theory to the findings of this Japanese study
© 2015 S. Karger AG, Basel
0250–6807/15/0668–0001$39.50/0
E-Mail [email protected]
www.karger.com/anm
Table 1. Strength of endpoints
Commonly measured endpoints for adult and pediatric cancer treatment studies are listed below in descending order of strength:
A. Total mortality (or overall survival from a defined time).
This outcome is arguably the most important one to patients and is also the most easily defined and least subject to
investigator bias.
B. Cause-specific mortality (or cause-specific mortality from a defined time).
Although this may be of the most biologic importance in a disease-specific intervention, it is a more subjective endpoint than
total mortality and more subject to investigator bias in its determination. This endpoint may also miss important effects of
therapy that may actually shorten overall survival.
C. Carefully assessed quality of life.
This is an extremely important endpoint to patients. Careful documentation of this endpoint within a strong study design is
therefore sufficient for most physicians to incorporate a treatment into their practices.
D. Indirect surrogates.
i. Event-free survival.
ii. Disease-free survival.
iii. Progression-free survival.
iv. Tumor response rate.
These endpoints may be subject to investigator interpretation. More importantly, they may, but do not automatically, translate
into direct patient benefit such as survival or quality of life. Nevertheless, it is rational in many circumstances to use a treatment
that improves these surrogate endpoints while awaiting a more definitive endpoint to support its use.
This list can be found in PDQ® Levels of Evidence for Adult and Pediatric Cancer Treatment Studies [1].
Men Women
1.0 1.0
y 0.8 0.8
alit
ort
m 0.6 0.6
e
s
u
HR for all-ca 00..24 00..24 CHD deaths
*
0 0
(1) (2 (3) (4) (5) (1) (2) (3) (4) (5)
Serum LDL cholesterol category Serum LDL cholesterol category
(1) <80 (2) 80–99 (3) 100–119 (4) 120–139 (5) 140+ (mg/dl)
(1) <2.06 (2) 2.06–2.57 (3) 2.58–3.09 (4) 3.10–3.61 (5) 3.62+ (mmol/l)
Fig. 1-1. Relationship between serum low density lipoprotein ease (CHD) deaths. The height of the bar for CHD deaths is set
(LDL) cholesterol level and the hazard ratio (HR) for all-cause according to the ratio between the numbers of CHD deaths and
mortality: the Ibaraki Prefecture Health Study [2]. A total of 30,802 all-cause deaths in the respective groups. The width of each col-
men and 60,417 women were followed for a median 10.3 years. umn is proportional to the number of participants in that group.
HRs were adjusted for age and potential confounding factors The vertical lines represent 95% confidence intervals. HRs for all-
(blood pressure categories, anti-hypertensive medication use, dia- cause mortality for each standard deviation increment of LDL cho-
betes mellitus, lipid medication use, body mas index, gamma-glu- lesterol were 0.88 (0.85–0.91) and 0.90 (0.86–0.93) for men and
tamyl transferase, smoking status, alcohol consumption, kidney women, respectively. * Significantly different from reference group
dysfunction, and high density lipoprotein cholesterol and triglyc- (<80 mg/dl) with regard to CHD deaths.
eride categories). Dark gray shading represents coronary heart dis-
4 Ann Nutr Metab 2015;66(suppl 4):1–116 Hamazaki/Okuyama/Ogushi/Hama
DOI: 10.1159/000381654
would be that this kind of phenomenon can be easily ex- found to be a negative risk factor for all-cause mortal-
plained by the presence of participants with an as yet sub- ity (fig. 1-2). Of note, the mortality rates due to cancer
clinical serious disease (e.g., hidden cancer) where some in men and to respiratory disease without cancer (most-
of them who had lower cholesterol levels died during the ly pneumonia) in men and women were lowest in the
study period (reverse causality). To exclude this possibil- highest cholesterol groups.
ity of reverse causality, the authors of the study re-ana- In 2007, Kirihara et al. performed a meta-analysis of
lyzed the data excluding deaths that occurred within the the relationship between total cholesterol levels and all-
first 2 years after baseline measurement and, interesting- cause mortality in Japan [5]. Because Japanese diets have
ly, found that their initial results were not substantially been changing for decades, reports published before 1995
changed [2]. were not included, leaving 5 for analysis. As shown in fig.
Fig. 1-1 also shows the relationship between LDL cho- 1-3, the results indicate that total cholesterol levels ≥240
lesterol levels and CHD mortality by sex. In men, the haz- mg/dl (≥6.22 mmol/l) should not in fact be regarded as a
ard ratio of CHD mortality is significantly higher than lipid disorder. The issue of familial hypercholesterolemia
that of the lowest group. However, in women, no differ- (FH) will be discussed separately below.
ences were observed between any groups. Fig. 1-1 is a One of the most recent, large epidemiological studies
good representation of the situation in Japan with regard in Japan is the Jichi Medical School Cohort Study, a com-
to cholesterol. This relationship between cholesterol and munity-based cohort study conducted in 12 rural areas in
CHD mortality is not linked to genetic differences be- Japan [6]. The study participants were 12,334 healthy
tween Western and Japanese populations. Japanese emi- adults aged 40–69 years who were followed for a mean
grants to Hawaii, where Japanese culture is still preserved 11.9 years. As shown in fig. 1-4, HRs for all-cause mortal-
to a certain degree, had CHD mortality rates intermediate ity were significantly higher in the lowest cholesterol
between those of Japanese men living in Japan and those groups than in the reference groups for both men and
of Japanese American men living in San Francisco, where women. Even the exclusion of deaths within 5 years of
the latter group had CHD mortality similar to the gen- baseline did not change the relationship between low
eral population in San Francisco [3]. cholesterol levels and high mortality. The same case was
It seems, then, that cholesterol is not necessarily a del- apparent when deaths due to liver disease were excluded.
eterious substance after all, and may actually be a marker This finding is incompatible with the notion that associa-
of healthy life and healthy organs. tion between low cholesterol and high mortality might be
Another large-scale epidemiological study conduct- due to the presence of participants with liver disease in
ed in Japan is the Isehara Study, which analyzed data the lowest cholesterol group. We will discuss liver disease
collected from the annual checkups of residents in Is- specifically in later chapters (Chapter 2, Section 4 and
ehara City (population: about 100,000) between 1994 Chapter 5, Section 1).
and 2004 [4]. During the study period, Japanese citizens All-cause mortality in men is U-shaped. In the Jichi
aged ≥40 years were eligible for annual health checkups Medical School Cohort Study, the effects of the presence
provided by local governments in accordance with the of participants with FH tended to be exaggerated in the
Health Care Act for the Elderly (1982; succeeded by highest cholesterol group; in other words, more partici-
new legislation in 2007). The final database containing pants with FH were concentrated in the highest choles-
data on 8,340 men (aged 64±10 years) and 13,591 wom- terol group. This phenomenon tends to appear when the
en (61±12 years) was compiled after applying the fol- participants’ age range is rather low (none aged ≥70 at
lowing exclusion criteria: death within 1 year of base- baseline) and when there are many recruiting areas. For
line, incomplete lipid data, attended single check-up example, if there is only one recruiting area, the more par-
only, and serum triglyceride levels beyond the Friede- ticipants that are recruited, the closer the ratio of partici-
wald equation limits (400 mg/dl or 4.5 mm; 198 men, pants with FH to all participants becomes to the general
126 women). Mean blood LDL cholesterol levels were ratio of FH in Japan (0.2%). However, if there are many
calculated for individuals from all available LDL choles- recruiting areas, selection bias of participants with FH
terol values except their last checkup. Mean follow up may raise the FH ratios to much higher than 0.2% and all-
for all subjects in the final database was 7.1 years (6.7 cause mortality in the highest cholesterol group may in-
years for men, 7.3 years for women). LDL cholesterol crease as a consequence. We will return to this important
levels were divided into 7 groups at 20-mg/dl issue of the proportion of subjects with FH later (Chapter
(0.5-mmol/l) intervals. LDL cholesterol was again 3, Section 3). At any rate, what we can say already is that
Cholesterol and Mortality Ann Nutr Metab 2015;66(suppl 4):1–116 5
DOI: 10.1159/000381654
Men
Other causes
Trauma Ischemic heart disease
*
3,500 * Cerebrovascular Respiratory disease
s Other heart diseases Malignancy
ar 3,000
e
y
n- 2,500 Reference
o
s
per 2,000
0
00 1,500
0,
0
h/1 1,000
at
e 500
D
0
<80 80–99 100–119 120–139 140–159 160–179 (cid:150)(cid:20)(cid:27)(cid:19) (mg/dl)
<2.1 2.1–2.5 2.6–3.0 3.1–3.5 3.6–4.0 4.1–4.6 (cid:150)(cid:23)(cid:17)(cid:26) (mmol/l)
a Blood LDL cholesterol level
Women
2,500
s
ar
e
on-y 2,000 * * *
ers Reference
0 p 1,500
0
0
0,
10 1,000
h/
at
De 500
0
<80 80–99 100–119 120–139 140–159 160–179 (cid:150)(cid:20)(cid:27)(cid:19) (mg/dl)
<2.1 2.1–2.5 2.6–3.0 3.1–3.5 3.6–4.0 4.1–4.6 (cid:150)(cid:23)(cid:17)(cid:26) (mmol/l)
b Blood LDL cholesterol level
Fig. 1-2. Low density lipoprotein (LDL) cholesterol and mortality was 7.1 years. Cox’s proportional hazards regression analysis was
in (a) men and (b) women: the Isehara Study [4]. Over 11 years employed to calculate age-adjusted relative risks in both men and
(1994–2004), 8,340 men (aged 64±10 years) and 13,591 women women. * p = 0.001, Cox’s proportional hazard regression analysis
(61±12 years) were followed in Isehara City, Japan. Deaths during with Bonferroni adjustment. The width of each column is propor-
the first year of follow up were excluded. Mean follow-up period tional to the number of participants in that group.
high cholesterol levels are not a risk factor for all-cause es 2012 (JASG2012) [8], which was published in June
mortality. 2012, depends almost exclusively on NIPPON DATA80
We will finish this section by briefly mentioning the for the risk calculation of CHD death.
findings of another Japanese epidemiological study, To sum up, almost all Japanese epidemiological stud-
NIPPON DATA80 [7]. They are of particular interest be- ies show that high cholesterol levels are a good marker of
cause this is the only epidemiological study performed in longevity. Unfortunately, however, many Japanese doc-
Japan that has ever found high cholesterol levels to be a tors try to reduce patients’ cholesterol levels without due
significant risk factor for all-cause mortality. Describing consideration of these overall findings. The Japan Ath-
the precise picture of this study warrants an entire chapter erosclerosis Society (JAS) first published the Guidelines
(Chapter 5), but for now it is suffice to say that the study for Diagnosis and Treatment of Atherosclerotic Diseases
is of considerable interest because the most important in 1997 and has revised them several times since. How-
part of the Japan Atherosclerosis Society Guidelines for ever, optimal cholesterol or LDL cholesterol levels in
the Prevention of Atherosclerotic Cardiovascular Diseas- terms of all-cause mortality have not been given as yet.
6 Ann Nutr Metab 2015;66(suppl 4):1–116 Hamazaki/Okuyama/Ogushi/Hama
DOI: 10.1159/000381654
Men Women
2.0 2.0
y
alit
ort **
m 1.5 1.5
e *
s
u
a
c
of all- 1.0 ** ** 1.0
sk Ref Ref
e ri 0.5 0.5
v
ati
el
R 0 0
<160 160–199 200–239 (cid:150)240 <160 160–199 200–239 (cid:150)240
(mg/dl)
<4.14 4.14–5.17 5.18–6.21 (cid:150)(cid:25)(cid:17)(cid:21)(cid:21) <4.14 4.14–5.17 5.18–6.21 (cid:150)(cid:25)(cid:17)(cid:21)(cid:21)
(mmol/l)
Serum total cholesterol levels
Fig. 1-3. Cholesterol and all-cause mortality in Japan: meta-analy- each cholesterol group. The width of each column is proportional
sis [5]. This meta-analysis included five reports and excluded re- to the number of participants in that group. Total number of sub-
ports published before 1995, those based on a cohort <5,000 sub- jects: 173,539. * p = 0.02, ** p = 0.0001.
jects, and those with no information on the number of deaths in
2
Men Women
＊＊ *
y
alit
ort
m
e
s
u 1
a
c
all-
or
R f
H
0
<160 160–199 200–239 (cid:150)240 <160 160–199 200–239 (cid:150)240
(mg/dl)
<4.14 4.14–5.16 5.17–6.20 (cid:150)(cid:25)(cid:17)(cid:21)(cid:20) <4.14 4.14–5.16 5.17–6.20 (cid:150)(cid:25)(cid:17)(cid:21)(cid:20)
(mmol/l)
Total cholesterol level
Fig. 1-4. Total cholesterol and all-cause mortality: Jichi Medical lationship between low cholesterol levels and increased mortality
School Cohort Study [6]. More than 12,000 men and women aged did not change even after excluding deaths due to liver disease or
40–69 years in 12 different areas in Japan were followed for a mean deaths within 5 years of baseline. The width of each column is pro-
11.9 years. Cox’s proportional hazards model was employed with portional to the number of participants in that group. * Signifi-
adjustment for age, systolic blood pressure, high density lipopro- cantly different from reference group (160–199 mg/dl, 4.14–5.16
tein cholesterol, smoking, drinking, and body mass index. The re- mmol/l). HR = Hazard ratio.
Cholesterol and Mortality Ann Nutr Metab 2015;66(suppl 4):1–116 7
DOI: 10.1159/000381654
Men Women
1.0
nderol
y aest
alithol
all-cause mortding to total c 0.5
HRs for HD accor ortality
I m
D
H
I
I. <50 mmol/l (194 mg/dl)
II. 5.0–5.9 (195–231) 0
III. 6.0–6.9 (232–270) I II III IV I II III IV
IV. (cid:150)7.0 (271) Cholesterol category
Fig. 1-5. Hazard ratios (HRs) for all-cause mortality and ischemic deaths within the same column. HRs for IHD mortality in choles-
heart disease (IHD) according to total cholesterol level in Norway: terol categories II to IV were not significantly different from those
HUNT 2 Study [9]. A total of 52,087 Norwegians aged 20–74 years in cholesterol category I in men or women. The width of each col-
were followed to calculate cause-specific mortality for 10 years. umn is proportional to the number of participants in that group.
HRs were adjusted for age, smoking, and systolic blood pressure. The HR increments for 1 mmol/l (39 mg/dl) of cholesterol were
The height of the black bar denoting IHD deaths is set according 0.98 (0.93–1.03) for men and 0.94 (0.89–0.99) for women.
to the ratio between the numbers of IHD deaths and all-cause
(2) Elderly People with High Cholesterol Levels Live ings were generalizable, clinical and public health recom-
Longer Irrespective of Where They Live mendations on the ‘dangers’ of cholesterol should be re-
vised; this would be especially true for women, for whom
Before describing the relationship between all-cause moderately elevated cholesterol (by current standards)
mortality and serum total or LDL cholesterol levels in el- might prove to be not just harmless, but even beneficial.
derly people, let’s start by discussing it in the general pop- Turning our focus now to elderly people, for the data
ulation. In well-developed countries, the relationship we currently have available, the situation is perfectly uni-
does not look like that found in Japan, where, as discussed form across the world: the higher the total cholesterol lev-
in Section 1, the higher the cholesterol levels, the lower els, the lower the all-cause mortality rate. Fig. 1-6 shows
the mortality rate. However, according to Petursson et al, mortality in three groups of the oldest residents in the
a phenomenon similar to that seen in Japan exists in the Leiden 85-Plus Study, conducted in Leiden, the
general population of Norway [9]. In their Nord-Trøndelag Netherlands [10]. Total cholesterol concentrations were
Health Study (HUNT 2, 1995–1997), 52,087 Norwegians measured in 724 participants with a median age of 89
aged 20–74 years were followed for cause-specific mortal- years, and mortality risks were calculated over 10 years of
ity over 10 years. As shown in fig. 1-5, the higher the total follow up. Participants with the highest total cholesterol
cholesterol levels, the lower the mortality rate in women. levels (≥6.5 mmol/l) had a lower mortality risk than those
Interestingly, the results for both men and women are al- with the middle range of total cholesterol levels (5.0–6.4
most identical to those of the Jichi Medical School Study mmol/l, middle risk) or those with the lowest range (<5.0
(fig. 1-4). mmol/l). The latter group had the highest risk. Mortality
In their report, Petursson et al. indicated possible er- risks were adjusted for age, sex, and cardiovascular risk
rors in the cardiovascular disease risk algorithms of many factors, and the highest total cholesterol group owed its
clinical guidelines [9]. They concluded that, if their find- longevity to lower mortality from cancer and infection.
8 Ann Nutr Metab 2015;66(suppl 4):1–116 Hamazaki/Okuyama/Ogushi/Hama
DOI: 10.1159/000381654
<5.0 mmol/l (193 mg/dl)
1.0
1.0
Cumulative all-cause mortalityaccording to cholesterol level 0000....2468 5.0–6.4 mmol/l (193–251 m(cid:150)pg(cid:25) /(cid:17)l(cid:24)od(cid:3)gl(cid:80))-r(cid:80)an(cid:82)k(cid:79)(cid:18) (cid:79)=(cid:3)(cid:11) (cid:21)0(cid:24).0(cid:21)0(cid:3)(cid:80)01(cid:74)(cid:18)(cid:71)(cid:79)(cid:12) Relative risk for all-cause mortalityaccording to cholesterol quartile 0.5
0
0
0 2 4 6 8 10
Total cholesterol quartile 1 2 3 4
Years of follow-up
3.85 4.61 5.15 5.99 (mmol/l)
Mean total cholesterol
149 178 199 232 (mg/dl)
Fig. 1-6. All-cause mortality of residents aged ≥85 years according
to total cholesterol level: Leiden 85-Plus Study [10]. A total of 724
residents in Leiden, the Netherlands, were followed for nearly 10 Fig. 1-7. Age-adjusted relative risk for all-cause mortality by serum
years. There were 203, 350, and 171 participants in the low, middle, cholesterol quartile: part of the Honolulu Heart Program [11]. A
and high cholesterol groups, respectively. Mortality risks were ad- total of 3,572 Japanese-American men aged 71–93 years were fol-
justed for age, sex, and cardiovascular risk factors (history of dia- lowed from baseline (1991–1993) to the end of 1996. There were
betes mellitus, myocardial infarction, cerebrovascular accident, 727 deaths (20% of the total) over the study period. Exclusion of
smoking, and hypertension) with Cox’s multivariate proportional the deaths during the first year (106 deaths) did not markedly
hazards model. (Remade with permission from the publisher.) change the results.
As part of the Honolulu Heart Program, serum total cording to baseline plasma LDL tertiles: ≤89, >89 to ≤115,
cholesterol concentrations were measured in 3,572 and >115 mg/dl (≤2.30, >2.30 to ≤2.97, and >2.97 mmol/l).
Japanese-American men aged 71–93 years (between 1991 Patients with the highest baseline LDL cholesterol levels
and 1993) [11]. A total of 727 deaths were registered be- had significantly better outcomes, while those with the
tween baseline and the end of 1996. Compared with the lowest LDL cholesterol levels had the highest mortality.
first (lowest) quartile of total cholesterol, the relative risks The same trend was also observed in those taking statins
for all-cause mortality adjusted for age were significantly (fig. 1-9). Low LDL cholesterol levels predicted less favor-
low in the other quartiles, with the third quartile being saf- able outcomes in patients with heart failure whether they
est, followed by the fourth and second quartiles (fig. 1-7). were taking statins or not.
In another study, the Vorarlberg Health Monitoring In a prospective cohort study with a 6-year follow-up
and Promotion Programme conducted in Austria, 67,413 period conducted in Kuopio, Finland, Tuikkala et al. in-
men and 82,237 women aged 20–95 years underwent var- vestigated the association between total cholesterol levels
ious examinations over a 15-year period (1985–1999), and all-cause mortality in 490 home-dwelling elderly per-
and relations between measured variables and death were sons (aged ≥75 years, men 28%) who did not use lipid-
analyzed [12]. Cox’s proportional hazards models were modifying agents [14]. In a propensity score-adjusted
used to assess the age-adjusted associations between total model using total cholesterol <5 mmol/l (<193 mg/dl, the
cholesterol levels and mortality. In both men and women lowest tertile) as a reference, HRs for all-cause mortality
in the 50–64 and ≥65 age groups, total cholesterol con- became lower with increasing cholesterol tertiles (fig.
centrations were a negative risk factor for all-cause mor- 1-10). The inverse association between serum total cho-
tality (fig. 1-8). lesterol and mortality is often interpreted to be due to
Charach et al. investigated the association between confounding by chronic diseases, but mortality in this
LDL cholesterol levels and clinical outcomes in 297 pa- study was found not to be associated with the following
tients with severe heart failure (mean age 71±11 years, concomitant diseases or health status: history of hyper-
men 73%) in Israel [13]. Mean follow up was 3.7 years tension, current hypertension, heart disease, stroke, ob-
(8 months-11.5 years). The patients were grouped ac- structive pulmonary disease, history of cancer, and de-
Cholesterol and Mortality Ann Nutr Metab 2015;66(suppl 4):1–116 9
DOI: 10.1159/000381654
Men 1.0
<187 mg/dl 4.84 mmol/l High LDL
187–248 mg/dl 4.84–6.41 mmol/l
0.8
>248 mg/dl 6.41 mmol/l
1.5 al Low LDL
v
ality survi 0.6
ort 1.0 ve
m ati
e ul 0.4
us m Middle LDL
a u
c C
All- 0.5
0.2
p = 0.029 p = 0.049
0
a <50 50–64 (cid:150)(cid:25)(cid:24)(cid:3)(cid:11)(cid:92)(cid:72)(cid:68)(cid:85)(cid:86)(cid:12) 0
0 10 20 30 40 50 0 10 20 30 40 50
Women Months
Entire cohort (n = 297) Statin users only
2.5
<184 mg/dl 4.76 mmol/l a (mean age: 71 years) b (n = 166)
184–244 mg/dl 4.76–6.31 mmol/l
2.0 >244 mg/dl 6.31 mol/l
y
alit Fig. 1-9. Cumulative survival rate of patients with severe heart fail-
ort 1.5 ure according to baseline cholesterol level: study in Israel [13]. A
m
e total of 297 patients with severe heart failure were followed for a
us 1.0 mean 3.7 years. (a) Survival rates were compared according to ter-
a
All-c tsielxes, loeff tl ovwen dterincsuitlayr l iepjeocptrioonte ifnra cchtioolnes, tNereowl wYiothrk a Hdjeuasrttm Aesnsto fcoira taigoen,
0.5
functional class, creatinine clearance, diabetes, and hypertension.
(b) Exactly the same trend was observed even if compared between
0
b <50 50–64 (cid:150)(cid:25)(cid:24)(cid:3)(cid:11)(cid:92)(cid:72)(cid:68)(cid:85)(cid:86)(cid:12) patients with ischemic heart disease only (n = 227, p = 0.039).
(Remade with permission from the publisher.)
Fig. 1-8. Fifteen-year follow up of (a) 67,413 men and (b) 82,237
women aged 20–95 years in Vorarlberg, Austria: Vorarlberg
Health Monitoring and Promotion Programme [12]. The width of
a column is proportional to the number of participants. The left 1.0
column in each age group represents the lowest total cholesterol
quartile, the middle column represents the reference group con-
tuaminnin rge pthrees seenctosn tdh ea nhdig thheirsdt cqhuoalretsilteesr ocol mqubairnteilde,. aAnddj uthstee rdi gfhotr caogle- ortality
(Cox’s proportional hazards model). m
e
s
u 0.5
a
c
all-
mentia. The decreasing HR pattern across cholesterol ter- or
tiles did not markedly change between participants with R f
H
any one of the conditions listed above and participants
without any one of them.
In another Finnish study, baseline examinations in- 0
Serum <5 5–5.9 (cid:150)(cid:25) mmol/l
cluding serum cholesterol were performed in 1990 in per- cholesterol <193 193–231 (cid:150)(cid:21)(cid:22)(cid:21) mg/dl
sons selected from the census register in Helsinki (n =
623, aged ≥75 years) who were randomly selected from
the birth cohorts of 1904, 1909, and 1914, and all persons Fig. 1-10. Hazard ratio for all-cause mortality according to total
cholesterol tertile among participants in a home-dwelling elderly
were followed for 17 years [15]. Low cholesterol was
population: study in Kuopio, Finland [14]. A total of 490 home-
found to be associated with poor health and multi-mor-
dwelling residents were followed for 6 years, and cumulative sur-
bidity. Cholesterol <5.0 mmol/l (193 mg/dl) was associ-
vival was calculated in a propensity score adjusted model. HR =
ated with accelerated all-cause mortality and vascular Hazard ratio.
10 Ann Nutr Metab 2015;66(suppl 4):1–116 Hamazaki/Okuyama/Ogushi/Hama
DOI: 10.1159/000381654