Campos H, Moye LA, Glasser SP, Stampfer MJ, Sacks FM. Low-Density Lipoprotein Size, Pravastatin Treatment, and Coronary Events. JAMA. 2001;286(12):1468-1474. doi:10.1001/jama.286.12.1468
Author Affiliations: Department of Nutrition, Harvard School of Public Health (Drs Campos and Sacks); Department of Medicine, Harvard Medical School (Drs Stampfer and Sacks); and Brigham and Women's Hospital (Dr Sacks), Boston, Mass; University of Texas School of Public Health, Houston (Dr Moye); and Division of Epidemiology, University of Minnesota, Minneapolis-St Paul (Dr Glasser).
Context Small low-density lipoprotein (LDL) particle size has been hypothesized
to be a risk factor for coronary heart disease (CHD). Animal models link large
LDL to atherosclerosis. However, the strong association between small LDL
and other risk factors, particularly triglyceride levels, impedes determining
whether LDL size independently predicts CHD in humans.
Objective To examine whether LDL size is an independent predictor of recurrent
coronary events in patients with known CHD, as opposed to a marker for other
Design and Setting Prospective, nested case-control study in the Cholesterol and Recurrent
Events (CARE) trial, a randomized placebo-controlled trial of pravastatin
conducted in 1989-1996.
Participants Survivors of myocardial infarction with typical LDL concentrations (416
cases and 421 controls).
Main Outcome Measure Subsequent myocardial infarction or coronary death during the 5-year
follow-up, analyzed by quintile of LDL particle size and by treatment group.
Results Overall, the mean LDL size was identical in cases and controls (25.6
nm). In patients in the placebo group, large LDL predicted coronary events
in models adjusted only for age (relative risk [RR], 1.79; 95% confidence
interval [CI], 1.01-3.17) and for age and lipid and nonlipid risk factors
(RR, 4.00; 95% CI, 1.81-8.82), comparing those in the highest (mean, 26.6
nm) and lowest (mean, 24.5 nm) quintiles of LDL size. This increased risk
was not present in those taking pravastatin (age-adjusted analysis: RR, 0.98;
95% CI, 0.47-2.04; P = .046 for interaction for a
difference in the effect of LDL size on coronary events between the placebo
and treatment groups; multivariable analysis: RR, 1.33; 95% CI, 0.52-3.38; P = .11 for interaction).
Conclusions Large LDL size was an independent predictor of coronary events in a
typical population with myocardial infarction, but the adverse effect was
not present among patients who were treated with pravastatin. Identifying
patients on the basis of LDL size may not be useful clinically, since effective
treatment for elevated LDL cholesterol concentrations also effectively treats
risk associated with large LDL.
Small-size low-density lipoprotein (LDL) cholesterol has been associated
with coronary disease in several retrospective case-control studies,1- 6
although other studies have not found this association.7- 10
In fact, in 2 studies, larger LDL particles were associated with increased
risk of coronary disease.7,8 Most
prospective studies,11- 13
but not all,14 also found that patients with
coronary disease are more likely than controls to have smaller LDL particles.
However, no study, retrospective or prospective, found that small LDL had
independent predictive value after adjustment for standard lipid risk factors.1- 6,11- 13
One of the difficulties in establishing whether LDL size is an independent
predictor of coronary disease or is merely a marker of other lipid abnormalities
is that the predominance of small LDL is strongly associated with increased
triglyceride and reduced HDL cholesterol concentrations as well as other traits
of the metabolic syndrome.15,16
Predominance of small LDL in patients with coronary disease may simply reflect
these other traits, and so its clinical utility has been questioned.16,17 Indeed, in studies in which triglyceride
concentrations were similar in cases and controls, small LDL was either unrelated9,10,14 or inversely related7,8 with risk of coronary disease. Because
of such strong confounding, the true direct relationship, if any, between
LDL size and coronary disease has been difficult to determine.
The current study uses a prospective, nested case-control design to
examine whether LDL size is an independent predictor of recurrent coronary
events in survivors of myocardial infarction (MI) in the Cholesterol and Recurrent
Events (CARE) trial. This study is unique because it is the largest prospective
study of LDL size and coronary disease to date, and because the lipid levels
are typical of patients with coronary disease.
The CARE trial (conducted in the period 1989-1996) was a randomized,
placebo-controlled trial of pravastatin in 4159 patients who experienced acute
MI 3 to 20 months before enrollment.18 The
eligibility ranges for plasma lipid concentrations were typical of patients
with coronary disease: total cholesterol, less than 240 mg/dL (6.22 mmol/L);
LDL cholesterol, 115 to 174 mg/dL (2.98-4.51 mmol/L); and triglycerides, less
than 350 mg/dL (3.96 mmol/L). The median duration of follow-up was 5 years.
Fasting venous blood was taken from each patient on each of 2 screening visits,
at least 1 week apart, and sent by overnight delivery in cooled containers
to the core laboratory in St Louis, Mo. Plasma was separated in a refrigerated
centrifuge and 1-mL aliquots were placed in polypropylene vials and stored
at –80°C until analysis. Cases were those patients who experienced
a primary end point (coronary death or confirmed MI) during the follow-up
period (n = 486).18 Patients who did not experience
a primary end point were randomly selected and matched to cases by decade
of age (eg, 40-49 years, 50-59 years) and by sex. Sufficient plasma from 2
screening visits was available for LDL size analysis in 416 cases and 421
At the conclusion of the 5-year follow-up period, vials containing frozen
plasma from the 2 screening visits were shipped by overnight delivery to the
laboratory at the Harvard School of Public Health for determination of LDL
size. The LDL peak diameter was determined from whole plasma by lipid-stained,
nondenaturing gel electrophoresis using 2% to 16% polyacrylamide gradient
gels.19,20 Gels were stained for
lipid with Sudan black (Sigma, St Louis, Mo) and scanned with a laser densitometer
(LKB-Ultroscan LX, LKB Instruments Inc, Paramus, NJ). Size of LDL was estimated
from calibration curves that were constructed using latex beads (Duke Scientific
Corp, Palo Alto, Calif) and high molecular-weight standards (Pharmacia AB,
Stockholm, Sweden) as previously described.19
To ensure accuracy in the determination of LDL size, our values were standardized
by sending control samples to Ronald Krauss, MD, at the Donner Laboratory,
Berkeley, Calif. These control samples had LDL diameters smaller than 25.5
nm, between 26.0 and 26.5 nm, and larger than 27.0 nm. The deviation between
our laboratory and the Donner Laboratory was less than 0.25% for all samples.
Subjects with a predominant LDL peak size of 25.5 nm or smaller were classified
as "pattern B," and those with LDL size larger than 25.5 nm were "pattern
A."21 Matched cases and controls were included
together in 1 gel. Laboratory personnel were blinded to case-control status.
The between-run coefficient of variation was 0.95% for the internal control
sample included in every gel, and 0.54% for 20 blinded duplicate samples interspersed
throughout the samples in the study. Plasma triglyceride, LDL cholesterol,
and high-density lipoprotein (HDL) cholesterol concentrations were measured
as previously described.22
Statistical analyses were performed at the University of Texas School
of Public Health, Houston, using SAS version 8.1 (SAS Institute Inc, Cary,
NC). The distribution of LDL size of the controls was used to compute quintiles
and the number of cases and controls in each quintile was then determined.
These quintile categories were also used in subgroup analyses, and therefore
quintiles from controls in subgroup analyses were not evenly distributed.
Multiple logistic regression computed relative risks (RRs) with 95% confidence
intervals (CIs) for case status for the second through fifth quintiles compared
to the first quintile. Tests for linear trend were performed on the RRs across
quintiles, using the median value for each of the quintiles. The univariate
model included age only, while the basic multivariable models used for all
analyses included age, smoking, hypertension, and left ventricular ejection
Other covariates that were associated with LDL size as well as with
coronary events, and thus could be part of a causal pathway between LDL size
and coronary events, were studied in additional models. These covariates were
use of β-adrenergic antagonists, use of diuretics, waist circumference,
history of diabetes, and the standard lipid risk factors LDL cholesterol,
HDL cholesterol, and triglycerides. Waist circumference was the covariate
used to account for obesity since it correlated with LDL size and predicted
recurrent coronary events in this population (4% increase in risk of coronary
events per 2.5 cm of waist circumference, P = .03),22 and because it is now recommended to use in clinical
The protocol specified analyses in the total group regardless of assignment
to placebo or pravastatin, as well as in each treatment group separately.
Tests for interaction between treatment assignment, LDL size, and RR of coronary
events were conducted. The cutoff for statistical significance was P = .05 (2-sided). There were 460 patients in the placebo group and
377 patients in the pravastatin group.
Because of the potential effects of β-blockers on LDL size,4 we repeated the analysis after excluding all those
patients who were taking β-blockers at baseline. In the placebo group,
there were 129 cases and 132 controls not taking β-blockers.
Previous studies carried out in this population22
excluded control patients who had had coronary artery bypass graft (CABG)
surgery, coronary angioplasty, or stroke after randomization during the follow-up.
These subjects were excluded because results from the trial showed that pravastatin
reduced coronary revascularization and stroke,18
and patients with these clinical end points may not be considered event-free
controls. To evaluate whether the inclusion of these subjects in our study
could have affected the results, we also repeated the analysis in the placebo
group including only those controls who had no history of CABG surgery, coronary
angioplasty, or stroke after randomization during the follow-up.
The characteristics of cases and controls are shown in Table 1. The mean LDL size was identical in cases and controls,
25.6 nm. The prevalence of pattern B, denoting a predominance of small LDL,
was 39% in cases and 40% in controls. Patients in the highest quintile of
LDL size were older, less obese, had lower triglyceride and total cholesterol
concentrations, and higher HDL cholesterol concentrations (Table 2). The prevalence of hypertension and diabetes, blood pressure
levels, and medication use was not significantly associated with LDL size.
LDL size was significantly (P<.01) correlated
with triglycerides (r = −0.63), HDL cholesterol (r = 0.50), and LDL
cholesterol (r = 0.13).
Analysis of the total cohort (those randomized to placebo or pravastatin)
showed that LDL size was not a significant predictor of recurrent coronary
events in a multivariable model that included age, smoking, history of hypertension,
and left ventricular ejection fraction (RR, 1.27; 95% CI, 0.82-1.98; P = .28 for the highest quintile compared with the lowest).
When diabetes, waist circumference, and use of β-adrenergic antagonists
and diuretics were added to the model, large LDL was a significant independent
predictor of coronary events (RR, 1.60; 95% CI, 1.00-2.55; P = .049). Further adjustment for plasma triglycerides, LDL cholesterol,
and HDL cholesterol strengthened this association (adjusted RR, 2.10; 95%
CI, 1.20-3.68; P = .01).
The relationship between large LDL size and coronary events was evident
only in the patients in the placebo group (P = .046
for a difference in the effect of LDL size on coronary events between the
placebo and treatment groups). Thus the results are presented separately for
each treatment group. Large LDL was a significant predictor of recurrent coronary
events in the univariate model that only included age as a covariate (Table 3). When smoking, history of hypertension,
left ventricular ejection fraction, use of β-adrenergic antagonists and
diuretics, diabetes, and waist circumference were added to the model, the
risk for recurrent coronary events was increased. Further adjustment for plasma
triglycerides, LDL cholesterol, and HDL cholesterol considerably strengthened
the association between large LDL and the risk of recurrent coronary events
(Table 3 and Figure 1).
Size of LDL was not a predictor of recurrent events in the pravastatin
group. In the age-adjusted model, the RR of a recurrent event for the highest
quintile of LDL size was 0.98 (95% CI, 0.47-2.04; P
= .88 for trend). In a multivariable model adjusted for lipid and nonlipid
risk factors, the RR was only 1.33 (95% CI, 0.52-3.38) (for trend, P = .56; for interaction, P = .11) (Figure 1).
Exclusion of patients who were taking β-blockers did not affect
the findings. In the placebo group, large LDL size was a predictor of recurrent
coronary events in the univariate model that only included age as a covariate
(RR, 2.15; 95% CI, 0.98-4.76; P = .06 for the highest
compared with the lowest quintile). In the multivariable model that included
all the nonlipid risk factors the RR was 2.44 (95% CI, 1.05-5.68; P = .04). Adding triglyceride, LDL cholesterol, and HDL cholesterol
to the model further strengthened the association (RR, 4.22; 95% CI, 1.50-11.92; P = .007).
Similarly, excluding control patients who had had CABG surgery, coronary
angioplasty, or stroke after randomization during the follow-up did not affect
the results. There was a positive trend in the model that included the covariates
age, smoking, history of hypertension, and left ventricular ejection fraction
(RR, 1.77; 95% CI, 0.98-3.22; P = .10 for trend),
which became statistically significant when diabetes, waist circumference,
and the use of β-adrenergic antagonists and diuretics were added to the
model (RR, 2.14; 95% CI, 1.15-3.98; P = .03 for trend).
After adding plasma triglycerides, LDL cholesterol, and HDL cholesterol to
the model, the association between large LDL and the risk of recurrent coronary
events was further increased (RR, 3.66; 95% CI, 1.67-8.02; P = .003 for trend).
We found that large LDL size was a significant predictor of increased
recurrent coronary events in a typical population of survivors of MI. This
association was robust: it was present in both univariate and multivariable
analyses and was independent of other plasma lipid and nonlipid risk factors.
Increased LDL size was a risk factor among patients in the placebo group (4-fold
increase) but not in the pravastatin group. These findings are contrary to
the prevailing view that a predominance of small LDL predicts coronary disease
The association between large LDL and coronary disease has been previously
described. Large LDL was independently associated with increased risk of coronary
disease in Canadian normolipidemic patients.7
Among American Indian communities, large LDL was associated with higher coronary
disease mortality.8 Cross-cultural studies
also suggest that large LDL is associated with clinical coronary disease.
In Finland, where coronary disease incidence is very high, the mean LDL size
is very large (27.1 nm) and the prevalence of pattern B, the small-LDL phenotype,
is very low (15%).14 In contrast, in a healthy
rural Costa Rican population with a low prevalence of coronary disease, LDL
size is smaller (mean, 26.2 nm) and the prevalence of pattern B higher (44%)
than in Finland.24
Because of the focus on small LDL size as a marker of coronary disease,
evidence that large LDL size is atherogenic has not been well recognized.
Large LDL particles, similar in size to those found in humans, predominate
in hypercholesterolemic pigs that are more susceptible to atherosclerosis.25 In nonhuman primates, a diet high in saturated fat
and cholesterol increases LDL size, and the magnitude of this increase is
strongly associated with severity of atherosclerosis.26
In humans as well, diets that are high in saturated fat and cholesterol increase
the plasma concentration of large LDL.27
Large LDL particles are thought to be large because of high cholesterol
ester content. They preferentially bind to isolated arterial proteoglycans,28 delivering more cholesterol per particle to cells
and connective tissue in the arterial wall.26,29
Taken together, evidence from epidemiologic and atherosclerosis studies suggests
that large cholesterol-ester–rich LDLs are atherogenic and predictive
of coronary disease events in humans. Small LDL does have potentially deleterious
properties such as reduced affinity for the LDL receptor,30
longer residence time in plasma,31 increased
susceptibility to oxidation,32 and adverse
effects on the function of vascular cells.15
However, these may not be worse, in vivo, than the harmful properties of large
In our study, large LDL size did not predict coronary events in the
pravastatin group. This may be because pravastatin reduces the concentration
of larger LDL particles, as suggested by a major decrease in average LDL size
(−7 nm, P = .01) and reduced cholesterol ester
content of LDL during pravastatin therapy.33
We investigated reasons why previous studies found that small LDL is
associated with coronary disease. One difference is that the present study
pertained to recurrent events in patients after MI, while previous prospective
studies were in populations that had not experienced a major coronary event
at baseline. However, there has been no difference in how the lipid risk factors
predict initial or recurrent events.18,34,35
In previous studies, small LDL was associated with coronary disease in univariate
analyses that did not include the well-established lipoprotein risk factors.
In prospective studies, small LDL has not remained a significant independent
predictor of coronary disease after adjustment for total cholesterol and triglycerides,12 ratio of total cholesterol to HDL cholesterol,11 or apolipoprotein B,13
all of which were significant predictors despite the inclusion of LDL size
in the model. Similarly, retrospective case-control studies did not find that
small LDL size remained significantly associated with coronary disease in
These findings suggest that confounding by other lipid risk factors accounts
for the association of small LDL with coronary disease.
Comparison of 1 such study, the Physicians' Health Study (PHS),12 with CARE provides insight on the importance of confounding.
In PHS, there was a difference of 53 mg/dL (0.60 mmol/L) in triglyceride concentrations
between MI cases and controls compared with CARE, where this difference was
only 9 mg/dL (0.10 mmol/L). In PHS, cases had significantly smaller LDL compared
with controls, whereas in CARE the mean LDL size was identical in cases and
controls. Thus, confounding between high triglycerides and small LDL was present
in PHS, much more so than in CARE.
In both studies, triglyceride was an independent risk factor. Triglyceride
metabolism is thought to play an important role in determining structure and
composition of LDL,36 with small LDL a byproduct
of hypertriglyceridemia.37 In CARE, the confounding
influence of triglyceride with LDL size was less than in PHS, permitting a
true relationship to emerge between large LDL size and coronary disease. Other
studies failed to detect a significant independent association between large
LDL size and coronary disease, perhaps because of residual confounding by
such highly correlated lipid risk factors. Taken together, these data support
the hypothesis that attenuation of the association between LDL size and other
lipids reveals an independent relationship between large LDL and coronary
One limitation of the study presented herein is that LDL size was not
measured during the follow-up period. Also, while it is unlikely that medications
other than β-blockers and diuretics may have had an effect on LDL size,
the possibility cannot be completely excluded.
In conclusion, large LDL size was an independent predictor of coronary
events in a population typical of patients with cardiovascular disease. This
adverse effect was not present among patients who were treated with pravastatin.
Identifying patients on the basis of LDL size may not be useful clinically,
since effective treatment for elevated LDL cholesterol concentrations also
effectively treats risk associated with large LDL.