Atherosclerosis and Lipoproteins |
From the Department of Clinical Biochemistry (B.A-L., A.T-H.), Herlev University Hospital, and the Department of Clinical Biochemistry (B.G.N.), Glostrup University Hospital, University of Copenhagen, Denmark.
Correspondence to Birgit Agerholm-Larsen, Department of Clinical Biochemistry, Herlev University Hospital, Herlev Ringvej 75, DK-2730 Herlev, Denmark. E-mail balarsen{at}image.dk
| Abstract |
|---|
|
|
|---|
Key Words: ACE gene insertion deletion polymorphism ACE activity blood pressure ischemic cardiovascular disease
| Introduction |
|---|
|
|
|---|
We therefore performed meta-analyses of studies on the association between the ACE I/D gene polymorphism and plasma ACE activity; blood pressure; and risk of myocardial infarction, ischemic heart disease, and ischemic cerebrovascular disease on small and large studies separately. We also tested for evidence of sample-size bias.11
| Methods |
|---|
|
|
|---|
Exclusion Criteria
Studies were excluded if participants were nonwhite, if ACE
activity and blood pressure were not measured in healthy control
subjects or information on SD or SEM was not available, or if
associations were not reported for each of the genotypes
(II, ID, DD). Studies performed
exclusively in patients with familial
hypercholesterolemia, diabetes mellitus, or
hypertension were also excluded. The exclusion of studies conducted in
diabetics and hypertensive patients may flatten the association of the
D allele with the end points analyzed. Studies
were selected by 1 author (B.A.L.) and verified by another (B.G.N.).
The exclusion criteria restricted the number of studies to 8 on plasma
ACE activity (Table 1
), 19 on blood
pressure (Table 2
), 19 on myocardial
infarction (Table 3
), 18 on
ischemic heart disease (Table 3
), and 6 on
ischemic cerebrovascular disease (Table 4
), in total 46 studies including 32 715
subjects.
|
|
|
|
Data Extraction
When data were reported separately for different geographical
locations in the same study, this division of data was sustained in the
meta-analyses. For 2 studies on blood pressure, data were
entered separately for the 2 sexes because combined data were not
available.12 13 When >1 control group was reported,
control subjects for meta-analyses were chosen with the
following priority: (1) subjects from a general population sample, (2)
subjects without any acknowledged disease, (3) subjects without
ischemic cardiovascular disease, and (4)
subjects without the end point studied.
Absolute levels of plasma ACE activity ranged considerably from study
to study. We therefore calculated the percent change in ID
and DD subjects relative to II subjects. In most
studies, blood pressure was measured with the subject at rest in the
supine or sitting position, and myocardial infarction was defined as
presence of
2 of the following: characteristic chest pain, elevation
of cardiac enzymes, and electrocardiographic results consistent
with a myocardial infarction. Ischemic heart disease was a
myocardial infarction, angina pectoris, percutaneous
transluminal coronary angioplasty, coronary artery
bypass graft surgery, severe stenosis (>50% to 75%) on
coronary angiography, and/or myocardial ischemia on an
exercise test. Ischemic cerebrovascular disease was sudden
onset of focal neurological symptoms and/or evidence of severe
atherosclerosis in carotid or cerebral arteries and/or
evidence of a brain infarct. Data were extracted by 1 author (B.A.L.)
and verified by another (B.G.N.).
Statistical Methods
A standard procedure was followed for each
meta-analysis: (1) Studies were ordered by increasing
statistical weight in the meta-analyses (fixed-effect model),
and either pooled-effect size or odds ratio (Mantel-Haenszel approach)
was calculated for all studies combined14 (a value of
P<0.05 on z test was considered significant)
(Figures 1 to 5![]()
![]()
![]()
![]()
);
for myocardial infarction, a subgroup meta-analysis of subjects
considered at low risk,1 ie, body mass index and
apolipoprotein B below the median, was also performed. (2) The
meta-analyses were tested for evidence of
heterogeneity between studies (a value of
P<0.05 on
2 test was considered
significant). (3) Funnel plots were visually examined for evidence
of sample-size bias (funnel plots similar to those shown in Figures 1 through 5![]()
![]()
![]()
![]()
), and a statistical test
for sample-size bias (funnel plot asymmetry) was
performed11 (a value of P<0.1 was
considered significant). (4) Meta-analyses on the smallest
and largest studies were performed separately. The "smallest
studies" constituted the cumulation of studies by statistical weight
in the meta-analysis from the studies with the least weight
upward until close to 50% of the total weight. The "largest
studies" were the remaining studies, those that contributed the upper
50% of the total statistical weight in the meta-analysis. (5)
A standardized z statistic was used to assess whether
sufficient evidence existed for agreement or disagreement between large
and smaller studies in each meta-analysis (a value of
P<0.05 on z test was considered
significant).9 (6) Random-effects model
meta-analyses were performed for all studies combined, as well
as for small and large studies separately. It is debatable whether the
fixed-effects model or the random-effects model is the model of choice
in studies like ours6 7 8 9 10 ; we therefore chose to use both
models.
|
|
|
|
|
Aggregation of data on plasma ACE activity is complicated by the fact
that different methods and scales have been used in the studies of
interest. By transforming the data sets onto a natural logarithmic
scale, we circumvented these differences between studies, making the
data available for statistical tests. Data are shown as percentages
based on untransformed data, but all statistical analyses were
carried out on transformed data. However, logarithmically transformed
data cannot be used without modification before entering the
meta-analysis, and the following equations were therefore used
before the meta-analysis: (1)
µ
ln(X)-1/2ln[(SD2/X2)+1]
to calculate the approximated mean value of plasma ACE activity, and
(2)

ln[(SD2/X2)+1]1/2
to calculate the equivalent approximated variance (these equations are
derived from the expressions of means and variance in the logarithmic
normal distribution).
To test the hypothesis that plasma ACE activity is associated with ACE genotype in a dose-dependent pattern (raw data indicate II subjects to have the lowest, ID subjects intermediate, and DD subjects to have the highest ACE activity), we tested whether ID subjects have plasma ACE activity precisely intermediate between the II and DD subjects. A value of P<0.05 suggests lack of a stepwise effect of genotype on plasma ACE activity, that is, the plasma ACE activity in ID subjects is not exactly midway between activity levels in DD and II subjects (there is partial dominance or recessivity depending on the sign of the departure). We used the following equation to test the hypothesis: z=[(2xeffect size(ID versus II)) +(-1xeffect size(DD versus II))]÷[SD(mean)x([1/n(DD)x(-1)2]+(1/n(ID)x22)+{1/n(II)x[-2-(-1)]2})1/2]. Effect size(ID versus II) and effect size(DD versus II) are the estimated effect sizes from the meta-analysis of ID versus II and DD versus II, respectively. SD(mean) is the mean of the 2 SDs for effect size(ID versus II) and effect size(DD versus II), respectively. n(DD), n(ID), and n(II) are the number of subjects included in the meta-analysis for each of the 3 genotypes DD, ID, and II, respectively.
| Results |
|---|
|
|
|---|
Odds ratios for the comparison of DD subjects versus
ID and II subjects combined were, for myocardial
infarction, 1.21 (95% CI, 1.11 to 1.32; P<0.001) (Figure 3
); for ischemic heart disease, 1.16 (1.08 to 1.25;
P<0.001) (Figure 4
); and for ischemic
cerebrovascular disease, 1.18 (1.01 to 1.37; P=0.03) (Figure 5
). The equivalent odds ratio for myocardial infarction when
only low-risk subjects were considered was 1.34 (95% CI, 1.02 to 1.75;
P<0.05).
Heterogeneity
There was evidence of heterogeneity between
studies in the meta-analyses for plasma ACE activity
(ID versus II, P<0.01 and
DD versus II, P<0.001) and for risk
of myocardial infarction (DD versus
ID+II, P<0.001) and ischemic
heart disease (DD versus ID+II,
P<0.01).
Sample-Size Bias
Visual examination of studies included in the
meta-analyses indicated potential sample-size bias for plasma
ACE activity (Figure 1
) and risk of myocardial infarction
(Figure 3
), ischemic heart disease (Figure 4
),
and ischemic cerebrovascular disease (Figure 5
). Tests
for sample-size bias based on funnel plot asymmetry for plasma ACE
activity and risk of myocardial infarction showed positive intercepts
of the regression line deviating from the origin (P=0.07 and
P=0.04, respectively), indicating that small studies in
these meta-analyses showed larger effects than large
studies.11
Small and Large Studies
Compared with II subjects, plasma ACE activity was
increased in ID subjects by +40% (+32% to +48%;
P<0.001) in small studies and by +21% (+15% to +27%;
P<0.001) in large studies (small versus large,
P<0.001) and increased in DD subjects by +71%
(+62% to +80%; P<0.001) in small studies and by +48%
(+41% to +55%; P<0.001) in large studies (small versus
large, P<0.001) (Figure 1
). The data were compatible
with a stepwise effect of genotype on plasma ACE activity only
in large studies (z test: P=0.74; null
hypothesis: stepwise relationship between the 3 genotypes).
Neither in small nor in large studies was systolic or
diastolic blood pressure affected by the ACE
gene polymorphism (Figure 2
, data not shown).
Risk of myocardial infarction was increased in small studies [odds
ratio, 1.47 (1.30 to 1.66; P<0.001)] but not in large
studies [odds ratio, 0.99 (0.88 to 1.12; P=0.91)] (Figure 3
) (small versus large, P=0.001); the equivalent odds
ratios when only low-risk subjects were considered was 2.30 (1.53 to
3.45; P<0.001) for small studies and 0.87 (0.60 to 1.27;
P=0.48) for large studies (small versus large,
P<0.001). Likewise, risk of ischemic heart disease
was increased in small studies [odds ratio, 1.29 (1.15 to 1.43;
P<0.001)], but not in large studies [odds ratio, 1.07
(0.97 to 1.17; P=0.19)] (Figure 4
) (small versus
large, P=0.01). Risk of ischemic cerebrovascular
disease was increased neither in small studies [odds ratio, 1.21 (0.97
to 1.51; P=0.46)] nor in the largest study [odds ratio,
1.15 (0.93 to 1.42; P=0.21)] (Figure 5
) (small
versus large, P=0.74).
Fixed-Effects Versus Random-Effects Model
Effect sizes of ACE genotype on plasma ACE
activity and blood pressure in random-effects models were similar to
those shown in Figures 1
and 2
for fixed-effects models.
Likewise, risk of myocardial infarction, ischemic heart
disease, and ischemic cerebrovascular disease (Figures 3 to 5![]()
![]()
) remained similar when random-effects models rather
than fixed-effects models were used, all studies combined as well as on
small and large studies separately, except for risk of ischemic
cerebrovascular disease for all studies combined, which for the
random-effects model estimated the mean risk across studies to be
statistically nonsignificant [1.21 (0.98 to 1.50);
P=0.08].
| Discussion |
|---|
|
|
|---|
The present meta-analyses differ from the previous ones15 16 17 by inclusion of recently published articles, eg, 2 of the largest studies to date4 5 ; by exclusion of abstract-derived data, studies on nonwhite populations, and studies based entirely on patients with familial hypercholesterolemia, hypertension, or diabetes mellitus and studies that did not report data on each of the 3 genotypes (II, ID, DD) separately; and most importantly, by performing sensitivity analyses on small and large studies separately.
Taken together, the total evidence suggesting a role for the ACE gene polymorphism in ischemic cardiovascular disease is weak: (1) as mentioned above, the association may not exist, and at most the risk is increased 20% for the DD genotype (all studies combined); (2) there is no dose-response effect from II to ID to DD on risk of disease, as for plasma ACE activity; (3) findings between studies are not consistent; and finally, (4) although increased plasma ACE activity was found to predict myocardial infarction in a cross-sectional study,21 the data rather seem to suggest that the modest effect on serum ACE levels in patients is most likely the consequence and not the cause of myocardial infarction. Nevertheless, we cannot exclude the possibility that potential cardiovascular effects of the ACE gene polymorphism are restricted to certain types of patients, ie, those with or without established cardiovascular disease, diabetes, or hypercholesterolemia.
The odds ratio for myocardial infarction for DD versus ID and II was higher in individuals with body mass index and apolipoprotein B below the median than in all individuals combined. In the Copenhagen City Heart Study, the odds ratio among low-risk individuals, defined as nonsmoking, normotensive, nondiabetic subjects (cases, n=81 women and men; controls, n=3266 women and men), was 1.35 (95% CI, 0.85 to 2.16; P=0.22).
In support of the present findings, meta-analyses of small clinical intervention trials showing substantial and highly significant treatment benefits have on several occasions later been contradicted by large clinical intervention trials.6 7 8 9 10 In situations like these, the results of the smaller studies most likely represent genuine observations. However, several circumstances may account for the fact that mainly small studies with positive rather than negative results come to the attention of researchers performing meta-analyses.11
In the present meta-analysis, large studies include 1 to 5 studies, and the comparison between large and small studies is therefore sensitive to the results of these few large studies, including our own,4 5 representing observations in a limited number of contexts. However, the individual results of most of both small and large studies are in agreement with the results of our studies, with the exception of results for myocardial infarction in small studies, in which there is also evidence of sample-size or publication bias. The study by Lindpaintner et al,3 together with our own studies4 5 and a few others, is responsible for the negative results observed in the large studies. This study, being prospective, has the advantage of a longitudinal analysis and therefore of avoiding possible selection by mortality; but on the other hand, the studied population (US male physicians) may not be representative of a "standard sample." Indeed, physicians may be more aware of the control of risk factors and the assumption that medical therapy, in particular aspirin or antihypertensive drugs, reduces the incidence of ischemic heart and cerebrovascular disease.
Conversely, meta-analyses of small clinical intervention trials have on many occasions accurately predicted the results of later large clinical intervention trials.6 7 8 9 10 It therefore seems important also to consider the possibility that the ACE gene polymorphism in fact truly predicts risk of ischemic cardiovascular disease in small studies but not in large studies. It could be that cases were at higher risk of disease or were less contaminated with individuals without the correct diagnosis in small studies than in large studies and that the association therefore was seen only in small studies. The fact that 2 of the largest studies included cases with symptoms of ischemic cardiovascular disease as well as verified severe atherosclerosis speaks against this possibility.4 5 Likewise, control subjects might be better selected in small than in large studies; however, this is an unlikely explanation, because 1 of the large studies was a nested case-control study within a prospective study,3 and in 2 of the other large studies, the controls were individuals from a large general population sample.4 5 It is also possible that the ACE gene polymorphism operates only in certain contexts, those present mainly in small studies, but not in US male physicians3 or Danes.4 5 Finally, asymmetry in effect size or odds ratio relative to sample size could occur by chance alone; however, this is hard to imagine in a total of 4 of 5 end points studied in the present set of meta-analyses.
Our own studies4 5 account for 33% of the total
population in the analyses and most of the subjects included in
the large studies. The analysis of ischemic
cerebrovascular disease and blood pressure is practically limited to
the comparison of our previous studies versus the remaining studies.
The present study, therefore, repeats in great part our previous
observations obtained in the very homogeneous population of
Denmark, results that may reflect a racial difference compared with
other studies and not only the difference between large and small
studies. Because this meta-analysis refers to a genetic factor,
this concern seems relevant. The prognostic value of a genetic marker
may be different in different populations, and a negative result
obtained in a selected population is valuable information that concerns
this specific group of people. Because the availability of large
studies of different populations is still limited, the present
comparison may have the effect of generalizing a specific observation.
However, we recently determined the effect of ACE
genotype on serum ACE activity in 900 women and men from the
Copenhagen City Heart Study58 : 50 subjects with each
of the genotypes DD, ID, and
II in each of the 2 sexes were drawn randomly within each of
the age groups 40 to 49, 50 to 59, and 60 to 69 years. In both women
and men and in all age groups, ACE genotype
explained 30% to 40% of the total variation in serum ACE activity,
and this effect was codominant, with DD subjects having the
highest, ID subjects intermediate, and II
subjects the lowest serum ACE activity (P<0.001 for all
pairwise comparisons). This rules out the possibility that the lack of
association between ACE genotype and risk of
ischemic cardiovascular disease in the
Copenhagen City Heart Study is a result of the ACE gene not
being operative in Danes. Taken together, the genotype
frequencies and the effect on serum ACE activity in the Copenhagen City
Heart Study are similar to those found in most other white populations
(Figure 1
).58 Thus, it is reasonable that the
prognostic value of this specific genetic marker should be the same in
Danes as in other white populations, unless one believes that
ACE genotype does not exert its effect through serum
ACE activity.
A plausible mechanism for the lack of effect of the ACE gene
polymorphism on cardiovascular disease could
therefore be that subjects with the DD genotype,
despite having increased levels of plasma ACE, do not necessarily
produce increased amounts of angiotensin II. This
hypothesis is supported by the facts that (1) the ACE gene
polymorphism did not affect blood pressure, despite large effects
on ACE activity, and (2)
40% of angiotensin I may be
converted to angiotensin II by pathways other than
ACE.59 Nevertheless, because ACE inhibition is well
known to reduce blood pressure, other potential ACE
genotypes that reduce plasma ACE activity to levels seen in
subjects treated with ACE inhibitors may influence
cardiovascular disease, because plasma ACE activity in
this situation may become rate limiting in the production of
angiotensin II.
A limitation of the present results is that information on use of ACE inhibitors was either absent or limited in the studies included in the present analyses and thereby impossible to correct for in the meta-analyses. ACE inhibitors influence plasma ACE levels, blood pressure, and risk of ischemic cardiovascular disease. Accordingly, if the use or effect of ACE inhibitors differs according to ACE genotype, this could have influenced our results. Another limitation is that most studies on ischemic cardiovascular disease, with the exception of the US Physicians Health Study, included only nonfatal cases.3 If the DD genotype was in fact associated with the incidence of ischemic cardiovascular disease, the results may reflect an underestimation of the importance of the DD genotype. In fact, a study of deaths from definite or possible myocardial infarction in Belfast, Ireland, found the D allele frequency to be increased (P<0.02) among the cases of fatal myocardial infarction.60 A final limitation of the present meta-analyses is that data were taken directly from published articles and not from the original data sets provided by the various authors.
In conclusion, the present meta-analyses cast doubt on the value of the ACE DD genotype as a marker for increased risk of ischemic cardiovascular disease. The DD and ID genotypes seem to be associated with a 50% and a 20% increase in plasma ACE activity, respectively.58 New studies on the ACE gene polymorphism should study the mechanism by which the ACE gene polymorphism may affect ACE activity and focus on diseases known or suspected to be influenced by elevated levels of plasma ACE activity61 62 63 64 rather than on ischemic cardiovascular disease.
| Acknowledgments |
|---|
Received February 8, 1999; accepted July 1, 1999.
| References |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
J. B. van der Net, J. van Etten, M. Yazdanpanah, G. M. Dallinga-Thie, J. J.P. Kastelein, J. C. Defesche, R. P. Koopmans, E. W. Steyerberg, and E. J.G. Sijbrands Gene-load score of the renin-angiotensin-aldosterone system is associated with coronary heart disease in familial hypercholesterolaemia Eur. Heart J., June 1, 2008; 29(11): 1370 - 1376. [Abstract] [Full Text] [PDF] |
||||
![]() |
E. Zintzaras, G. Raman, G. Kitsios, and J. Lau Angiotensin-Converting Enzyme Insertion/Deletion Gene Polymorphic Variant as a Marker of Coronary Artery Disease: A Meta-analysis Arch Intern Med, May 26, 2008; 168(10): 1077 - 1089. [Abstract] [Full Text] [PDF] |
||||
![]() |
F. Bonnet, S. Patel, M. Laville, B. Balkau, A. Favuzzi, L. D. Monti, N. Lalic, M. Walker, and on behalf of the European Group for the Study of I Influence of the ACE Gene Insertion/Deletion Polymorphism on Insulin Sensitivity and Impaired Glucose Tolerance in Healthy Subjects Diabetes Care, April 1, 2008; 31(4): 789 - 794. [Abstract] [Full Text] [PDF] |
||||
![]() |
D. K. Arnett and for the Writing Group Summary of the American Heart Association's Scientific Statement on the Relevance of Genetics and Genomics for Prevention and Treatment of Cardiovascular Disease Arterioscler. Thromb. Vasc. Biol., August 1, 2007; 27(8): 1682 - 1686. [Full Text] [PDF] |
||||
![]() |
A. Muthumala, H. Montgomery, J. Palmen, J. A. Cooper, and S. E. Humphries Angiotensin-Converting Enzyme Genotype Interacts With Systolic Blood Pressure to Determine Coronary Heart Disease Risk in Healthy Middle-Aged Men Hypertension, August 1, 2007; 50(2): 348 - 353. [Abstract] [Full Text] [PDF] |
||||
![]() |
K. D. Marciante, J. C. Bis, M. J. Rieder, A. P. Reiner, T. Lumley, S. A. Monks, C. Kooperberg, C. Carlson, S. R. Heckbert, and B. M. Psaty Renin-Angiotensin System Haplotypes and the Risk of Myocardial Infarction and Stroke in Pharmacologically Treated Hypertensive Patients Am. J. Epidemiol., July 1, 2007; 166(1): 19 - 27. [Abstract] [Full Text] [PDF] |
||||
![]() |
D. K. Arnett, A. E. Baird, R. A. Barkley, C. T. Basson, E. Boerwinkle, S. K. Ganesh, D. M. Herrington, Y. Hong, C. Jaquish, D. A. McDermott, et al. Relevance of Genetics and Genomics for Prevention and Treatment of Cardiovascular Disease: A Scientific Statement From the American Heart Association Council on Epidemiology and Prevention, the Stroke Council, and the Functional Genomics and Translational Biology Interdisciplinary Working Group Circulation, June 5, 2007; 115(22): 2878 - 2901. [Abstract] [Full Text] [PDF] |
||||
![]() |
E. E. Ntzani, E. C. Rizos, and J. P. A. Ioannidis Genetic Effects versus Bias for Candidate Polymorphisms in Myocardial Infarction: Case Study and Overview of Large-Scale Evidence Am. J. Epidemiol., May 1, 2007; 165(9): 973 - 984. [Abstract] [Full Text] [PDF] |
||||
![]() |
C. Kooperberg, J. C. Bis, K. D. Marciante, S. R. Heckbert, T. Lumley, and B. M. Psaty Logic Regression for Analysis of the Association between Genetic Variation in the Renin-Angio |