Kidney International, Vol. 58 (2000), pp.
293–301
Plasma lipids and risk of developing renal dysfunction:
The Atherosclerosis Risk in Communities Study
PAUL MUNTNER, JOSEF CORESH, J. CLINTON SMITH, JOHN ECKFELDT, and MICHAEL J. KLAG
Departments of Epidemiology, Biostatistics, Medicine, Health Policy and Management, The Johns Hopkins University Schools
of Hygiene and Public Health and Medicine, Baltimore, Maryland; Department of Pediatrics, University of Mississippi School
of Medicine, Jackson, Mississippi; and Department of Laboratory Medicine and Pathology, University of Minnesota,
Minneapolis, Minnesota, USA
Plasma lipids and risk of developing renal dysfunction: The
Atherosclerosis Risk in Communities Study.
Background. Animal and in vitro data suggest that dyslipi-
demia plays an important role in the initiation and progression
of chronic renal disease, but few prospective studies have been
conducted in humans.
Methods. We studied the relationship of plasma lipids to a
rise in serum creatinine of 0.4 mg/dL or greater in 12,728
Atherosclerosis Risk in Communities (ARIC) participants with
baseline serum creatinine that was less than 2.0 mg/dL in men
and less than 1.8 mg/dL in women.
Results. During a mean follow-up of 2.9 years, 191 persons
had a rise in creatinine of 0.4 mg/dL or greater, yielding an
incidence rate of 5.1 per 1000 person years. Individuals with
higher triglycerides and lower high-density lipoprotein (HDL)
and HDL-2 cholesterol at baseline were at increased risk for
a rise in creatinine after adjustment for race, gender, baseline
age, diabetes, serum creatinine, systolic blood pressure, and
antihypertensive medication use (all P trends ⱕ0.02). The ad-
justed relative risk for the highest versus lowest quartile of
triglycerides was 1.65 (95% CI, 1.1, 2.5, P ⫽ 0.01) and for HDL
was 0.47 (95% CI, 0.3, 0.8, P ⫽ 0.003). These associations were
significant in participants with normal creatinine (defined as
⬍1.4 mg/dL for men and ⬍1.2 mg/dL for women), with diabe-
tes, and without diabetes. The effect of high triglycerides was
independent of plasma glucose, but was weaker and less consis-
tent after further adjustment for fasting insulin in nondiabetics.
Conclusions. High triglycerides and low HDL cholesterol,
but not low-density lipoprotein cholesterol, predict an in-
creased risk of renal dysfunction. The treatment of these lipid
abnormalities may decrease the incidence of early renal dis-
ease.
Epidemiological studies have shown that hypertension
and dyslipidemia are both risk factors for the develop-
ment of cardiovascular disease [1–5]. Hypertension has
Key words: chronic renal disease, triglycerides, cholesterol, glomerular
injury, progression of renal disease.
Received for publication April 29, 1999
and in revised form September 15, 1999
Accepted for publication February 1, 2000
 2000 by the International Society of Nephrology
293
also been demonstrated to have adverse effects on the
kidney and is strongly associated with initiation [6] of
renal disease and accelerated loss of kidney function
[7–9]. More recently, it has been suggested that dyslipide-
mia may contribute to accelerated development of renal
insufficiency [10–14]. Although there is substantial ani-
mal evidence supporting a causal relationship between
lipid levels and the development of glomerular injury
leading to glomerulosclerosis [15, 16], human data are
sparse [17–19]. To date, most prospective studies of lipids
and renal disease have focused on the progression of
established renal disease rather than renal disease initia-
tion [20–23]. There is only one study, to our knowledge,
that has focused on lipids predicting renal disease initia-
tion [10]. Therefore, little is known about the prospective
relationship between lipids and declines in renal function
among persons with normal renal function. Such infor-
mation is necessary to identify modifiable risk factors
that predict development and progression of renal dis-
ease.
We sought to determine the association of plasma
lipids with loss of renal function and the clinical onset
of mild renal insufficiency. To achieve this goal, we pro-
spectively assessed 12,728 participants of the Atheroscle-
rosis Risk in Communities (ARIC) study during three
years of follow-up.
METHODS
The sample population for this study was drawn from
the ARIC Study cohort, which was comprised of popula-
tion-based samples from four U.S. communities: Forsyth
County, North Carolina; Jackson, Mississippi; suburban
Minneapolis, Minnesota; and Washington County, Mary-
land. The complete design of ARIC has been published
previously [24]. In brief, 15,792 men and women ages 45
to 64 were recruited from 1987 through 1989 through
the use of probability sampling techniques. Follow-up
visits occurred in three-year intervals, at which time par-
294 Muntner et al: Lipids and renal disease
ticipants underwent standardized examinations in field
center clinics. These follow-up visits included measure-
ments of demographic, lifestyle, and physiological char-
acteristics. This report is based on data from the baseline
examination and the first follow-up examination of the
cohort (1987 through 1989 and 1990 through 1993, re-
spectively). The overall ARIC cohort follow-up rate for
visit 2 exceeded 93%.
Exclusion criteria
Persons (N ⫽ 53) with severe hypercreatinemia (Cr ⱖ
2.0 mg/dL for males, Cr ⱖ 1.8 mg/dL for females) at
baseline were excluded because of the high probability of
substantial renal dysfunction [25]. Participants on lipid-
lowering medications at baseline (N ⫽ 563) were also
excluded because of concerns that their baseline lipid
measurements may not reflect their true risk profile [26].
By necessity, persons with missing data for lipids or creat-
inine at baseline (N ⫽ 623), missing covariates (sex, race,
diabetes status, antihypertensive medication use, N ⫽
48), or missing creatinine data at follow-up (N ⫽ 1300)
were excluded. Of the 1300 participants with missing
creatinine at follow-up, 307 had died, while 38 were miss-
ing serum measurements, and 955 (6% of visit 1 partici-
pants) did not attend ARIC visit 2. ARIC participants
who did not fast for eight hours prior to their serum
draw were excluded (N ⫽ 413) from all analyses. In
addition, we limited our analysis to white and African
American participants because of small numbers of par-
ticipants from other races (N ⫽ 48). These exclusion
criteria left us with a population of 12,728 for our primary
analyses.
Measurements
Serum and ethylenediaminetetraacetic acid (EDTA)
plasma from all participants were drawn at each visit.
Participants were asked to fast at least 12 hours before
the blood draw. Blood was drawn from the antecubital
vein while the participants were seated, and blood chem-
istries were performed at the ARIC central chemistry
laboratory at the University of Minnesota and plasma
lipids at the ARIC central lipid laboratory in Houston,
Texas, USA. Plasma cholesterol, plasma triglycerides,
and high-density lipoprotein (HDL) cholesterol were de-
termined using enzymatic methods and low-density lipo-
protein (LDL) cholesterol was calculated using the Frie-
dewald equation [27]. Apolipoproteins A-1 and B were
measured using radioimmunoassay and Lp(a) using a
double-antibody enzyme-linked immunosorbent assay
(ELISA) technique [27] for apo(a) detection. ARIC
technicians, trained and certified in the use of a random-
zero sphygmomanometer, took three blood pressure
measurements following a standardized procedure; an
average of the second and third measurements was used
to estimate blood pressure. Gender and age information
was abstracted from questionnaires. Baseline informa-
tion on the use of antihypertensive and lipid-lowering
medications was determined through interview and ex-
amination of pill bottles brought to the study interview.
The presence of hypertension was defined as a systolic
blood pressure ⱖ140 mm Hg, a diastolic blood pressure
ⱖ90 mm Hg, or the use of antihypertensive medications.
Diabetes mellitus was defined as a fasting glucose of
ⱖ126 mg/dL, nonfasting glucose of ⱖ200 mg/dL, or his-
tory/treatment of diabetes. Serum glucose was assayed
by the hexokinase method. A central laboratory mea-
sured fasting insulin with a nonspecific radioimmuno-
assay.
Definition of outcome
At both the baseline and follow-up visits, serum creati-
nine was measured using a modified kinetic Jaffe method
[27]. Assessment of the methodological and day-to-day
variability within ARIC participants revealed that 0.18
mg/dL (methodogical variability, SD ⫽ 0.0494, and
within-person variability, SD ⫽ 0.043) was the minimal
change in creatinine at which 95% confidence existed
that a true change had occurred [28]. We defined an
incident rise in serum creatinine as a change greater than
twice this amount (ⱖ0.4 mg/dL) during the approxi-
mately three-year follow-up period. A change in esti-
mated creatinine clearance during follow-up was used
as an additional outcome measure of change in renal
function. Creatinine clearance was estimated using the
Cockroft-Gault formula [29] adjusted for body size by
dividing by body surface area [30] and multiplying by
1.73. The outcome for this analysis was defined as indi-
viduals whose calculated creatinine clearance declined
by 25% or more.
Statistical methods
Plasma lipids assessed at baseline were the primary
exposures of interest. These included total cholesterol;
HDL cholesterol and its subfractions, HDL-2, and HDL-3
cholesterol; LDL cholesterol; apolipoprotein A-1; apoli-
poprotein-B; Lp(a); and triglycerides. Because of their
skewed distribution, triglycerides were log transformed.
Other variables were chosen as covariates, based on a
priori considerations, because they were thought to in-
fluence change in serum creatinine. These included the
following clinical and demographic characteristics: gen-
der, race, age, diabetes mellitus, systolic blood pressure,
and baseline creatinine. Since some antihypertensive
medications are associated with dyslipidemia [31, 32], we
also adjusted for the type of antihypertensive medication
being taken. Antihypertensive medications were divided
into four categories: diuretics, ACE inhibitors, thiazides,
and any other anti-hypertensive medication.
Length of follow-up between the ARIC baseline visit
and the first follow-up visit (ARIC visit 2) was deter-
 Muntner et al: Lipids and renal disease 295

mined for each participant, and the rise in serum creati- Table 1. Baseline characteristics of the study population and cases
who subsequently developed a rise in creatinine ⱖ0.4 mg/dL
nine incidence rate was calculated by quartile of lipids.
Trends in crude incidence rates across lipid quartiles Characteristic ARIC population Cases
were determined using Poisson regression [33]. Poisson N 12,728 191
regression was also used to determine the adjusted (for Male gender 45% 51%
Black race 23% 47%
the a priori chosen clinical and demographic characteris- Diabetes mellitus 10% 29%
tics) relative risks and 95% confidence intervals of a rise On antihypertensive medications 23% 43%
in serum creatinine with the lowest quartile of lipid level Hypertensive 32% 62%
as the comparison group. The statistical significance of Mean (SD) Mean (SD)
the trends in risk of a rise in serum creatinine across Age years 54 (6) 56 (6)
lipid quartiles was determined for the adjusted models Creatinine mg/dL 1.09 (0.19) 1.15 (0.29)
SBP mm Hg 121 (18) 133 (26)
using the Wald test and quartile number as a continuous DBP mm Hg 74 (11) 78 (14)
variable. Total cholesterol mg/dL 215 (42) 222 (52)
Analyses with lipid levels as continuous variables were Triglycerides mg/dL 128 (79) 160 (121)
Loge triglycerides log mg/dL 4.7 (0.5) 4.9 (0.6)
used to confirm the associations detected using quartile HDL cholesterol mg/dL 53 (18) 50 (15)
analyses. Relative risks in the continuous analyses are HDL-2 cholesterol mg/dL 15 (9) 13 (8)
quoted for differences that approximate the median dif- HDL-3 cholesterol mg/dL 39 (11) 37 (11)
LDL cholesterol mg/dL 135 (39) 135 (39)
ference between the first and fourth quartiles (threefold Lp(a) 98 (105) 133 (134)
higher triglyceride levels and 40 mg/dL higher HDL cho- Loge Lp(a) 4.0 (1.2) 4.3 (1.2)
lesterol). Analyses with lipids as continuous variables Apolipoprotein A-I mg/dL 133 (31) 129 (31)
Apolipoprotein B mg/dL 93 (29) 97 (34)
were conducted for the overall population and stratified
by creatinine at baseline: normal creatinine (Cr ⱕ 1.3 The study population was limited to persons without severe hypercreatinemia
(serum creatinine ⱕ2.0 mg/dL for males; ⱕ1.8 mg/dL for females) and not on
mg/dL for males, Cr ⱕ 1.1 mg/dL for females, N ⫽ lipid-lowering medications.
11,323) and hypercreatinemia (1.4 ⱕ Cr ⱕ 1.9 mg/dL for Abbreviations are: SBP, systolic blood pressure; DBP, diastolic blood pres-
sure; ARIC, Atherosclerosis Risk in Communities study.
males, 1.2 ⱕ Cr ⱕ 1.7 mg/dL for females, N ⫽ 1818).
Analyses were also conducted stratified by race and dia-
betes and hypertension status at baseline.
Because higher triglyceride levels and lower HDL cho- cholesterol, and apolipoprotein A-1 demonstrated lower
lesterol levels are markers of the insulin-resistance syn- incidence rates (all P ⱕ 0.05). A greater than twofold
drome [34–36], additional analyses were performed using difference in incidence rates across lipid quartiles oc-
continuous lipid levels and adjusting for baseline glucose curred with HDL cholesterol. The incidence of a creati-
and insulin levels. Insulin levels in diabetics are consid- nine rise was not associated with total cholesterol, LDL
ered not reliable; therefore, the effect of lipid levels was cholesterol, or apolipoprotein B levels (P ⫽ 0.31, 0.66,
assessed applying insulin levels in nondiabetics and a and 0.33, respectively).
dummy variable for insulin levels for ARIC participants As shown in Table 1, individuals who later developed
with diabetes mellitus. Finally, Poisson regression mod- a rise in creatinine were substantially more likely to be
els were fit using a decline in creatinine clearance, ap- black, have diabetes, and use antihypertensive medica-
proximated by the Cockroft-Gault formula, 25% or more tions at baseline. In addition, cases had a higher mean
as the outcome. All analyses were also conducted with- age, baseline serum creatinine, and blood pressure. The
out adjustment for baseline serum creatinine and yielded bottom row of each lipid listing in Table 2 shows the
similar results. relative risk of a rise in creatinine after adjustment for
all of these, and other, characteristics. Quartiles of tri-
glycerides, HDL cholesterol, and HDL-2 cholesterol
RESULTS maintained strong dose response associations with an
Baseline characteristics of the ARIC participants are incident rise in creatinine (P ⱕ 0.05) after adjusting for
displayed in Table 1. The mean length of follow-up be- race, gender, age, and baseline systolic blood pressure,
tween the ARIC baseline visit and the first follow-up visit type of antihypertensive medication use, diabetes melli-
was 2.9 years. During follow-up, 1.7% of the participants tus status, and creatinine. Additionally, the relative risk
(N ⫽ 191) experienced a creatinine rise of ⱖ0.4 mg/dL comparing the highest with the lowest quartile of HDL
(characteristics are displayed in Table 1), yielding an and HDL-2 cholesterol, and triglycerides were all sig-
overall incidence rate of 5.1 per 1000 person years of nificantly different from 1.0. The risk of a rise in creati-
follow-up. In contrast, only 0.2% had a fall of ⱕ0.4 mg/dL. nine in the highest quartile of triglycerides was 1.65 times
Higher quartiles of triglycerides and Lp(a) were asso- that for the lowest quartile (P ⫽ 0.01). Persons in the
ciated with a higher incidence of creatinine rise (Table 2). highest quartiles of HDL cholesterol and HDL-2 choles-
In contrast, higher quartiles of HDL, HDL-2, and HDL-3 terol were 0.47 and 0.57, respectively, times as likely to
296 Muntner et al: Lipids and renal disease

Table 2. Incidence, adjusteda relative risks and 95% confidence intervals of a rise in creatinine ⱖ0.4 mg/dL from baseline to the 3-year
follow-up by lipid quartiles at baseline in 13,141 participants
Quartile
Lipid 1 2 3 4 P trend
Total cholesterol mg/dL
Range ⱕ186 187–212 213–239 ⱖ240
N events 45 47 45 54
Rate per 1000 patient-years 4.8 4.9 4.9 5.9 0.31
Adjusted relative riska 1.0 1.02 (0.7,1.5) 0.96 (0.7,1.5) 1.06 (0.7,1.6) 0.85
Triglycerides
Range ⱕ78 79–109 110–155 ⱖ156
N events 38 36 50 67
Rate per 1000 patient-years 4.0 3.9 5.4 7.2 0.0009
Adjusted relative riska 1.0 0.99 (0.6,1.6) 1.31 (0.9,2.0) 1.65 (1.1,2.5) 0.008
HDL cholesterol mg/dL
Range ⱕ41 42–51 52–63 ⱖ64
N events 62 48 55 26
Rate per 1000 patient-years 6.8 5.1 5.8 2.8 0.0009
Adjusted relative riska 1.0 0.73 (0.5,1.1) 0.86 (0.6,1.3) 0.47 (0.3,0.8) 0.01
HDL-2 cholesterol mg/dL
Range ⱕ9 10–13 14–19 ⱖ20
N events 72 41 49 29
Rate per 1000 patient-years 6.6 4.4 5.6 3.5 0.01
Adjusted relative riska 1.0 0.65 (0.4,1.0) 0.84 (0.6,1.2) 0.57 (0.4,0.9) 0.05
HDL-3 cholesterol mg/dL
Range ⱕ31 32–38 39–46 ⱖ47
N events 56 53 51 31
Rate per 1000 patient-years 6.3 5.2 5.5 3.5 0.02
Adjusted relative riska 1.0 0.89 (0.6,1.3) 0.99 (0.7,1.5) 0.67 (0.4,1.1) 0.17
LDL cholesterol mg/dL
Range ⱕ109 110–132 133–159 ⱖ160
N events 48 44 50 49
Rate per 1000 patient-years 5.0 4.9 5.2 5.4 0.66
Adjusted relative riska 1.0 0.97 (0.6,1.5) 0.97 (0.7,1.4) 0.90 (0.6,1.3) 0.62
Lp(a)
Range ⱕ3.14 3.15–4.09 4.10–4.97 ⱖ4.98
N events 40 42 44 65
Rate per 1000 patient-years 4.3 4.4 4.7 7.0 0.01
Adjusted relative riska 1.0 0.96 (0.6,1.5) 0.83 (0.5,1.3) 1.10 (0.7,1.7) 0.70
Apolipoprotein A mg/dL
Range ⱕ111 112–130 131–151 ⱖ152
N events 62 45 47 37
Rate per 1000 patient-years 6.6 4.8 5.1 4.1 0.03
Adjusted relative riska 1.0 0.73 (0.5,1.1) 0.79 (0.5,1.2) 0.66 (0.4,1.0) 0.08
Apolipoprotein B mg/dL
Range ⱕ73 74–89 90–110 ⱖ111
N events 46 41 55 49
Rate per 1000 patient-years 4.8 4.5 5.7 5.5 0.33
Adjusted relative riska 1.0 0.87 (0.6,1.3) 1.03 (0.7,1.5) 0.88 (0.6,1.3) 0.73
Adjusted relative risks compare quartile of plasma lipids to the lowest quartile (quartile 1) of lipid level.
a
Adjusted for age, race, gender, and baseline systolic blood pressure, diabetes mellitus, antihypertensive medication use and serum creatinine

experience a rise in creatinine compared with those in
the lowest quartile (P ⬍ 0.02 for each).
Levels of plasma triglycerides and HDL cholesterol
were inversely associated with each other (correlation
of ⫺0.5). To discern whether the effects of these lipids
were independent, a Poisson regression model with
plasma triglycerides and HDL cholesterol as well as clini-
cal and demographic variables was fit. To enhance the
power of the analysis, lipids were modeled as continuous
variables with triglycerides log transformed because of
their skewed distribution. In a model containing HDL
cholesterol, log triglycerides, and demographic and clini-
cal characteristics, log triglycerides remained predictive
of an incident rise in creatinine, while HDL cholesterol
did not. The risk of a rise in creatinine was 1.45 (95%
CI, 1.0, 2.1, P ⫽ 0.04) times greater for threefold higher
(1.09 loge units) triglycerides; HDL cholesterol was not
associated with loss of renal function after adjustment
for triglycerides (RR ⫽ 0.72 for an 40 mg/dL increase
in HDL cholesterol; 95% CI, 0.5, 1.1, P ⫽ 0.15). Because
of the high correlation between HDL and HDL-2 choles-
terol (r ⫽ 0.81), HDL-2 was not included in this model.
Using HDL-2 cholesterol rather than HDL cholesterol
yielded similar results.
 Muntner et al: Lipids and renal disease
To determine the consistency of the relationship be-
tween increased triglyceride levels and a rise in creati-
nine, analyses were conducted stratified by race, creati-
nine level, diabetes mellitus, and hypertension status at
baseline (Fig. 1). For each threefold higher triglycerides,
the adjusted relative risk of a rise in creatinine was 2.39
(P ⫽ 0.001) among African Americans and 1.31 (P ⫽
0.20) among whites, a difference in effect of triglycerides
that was not statistically significant (P interaction ⫽
0.08). The adjusted relative risk among persons with
normal creatinine at baseline was 1.68 (P ⫽ 0.005).
Among persons with borderline or definite hypercreati-
nemia, the relative risk was 1.35 (P ⫽ 0.35). This disparity
was consistent with random variation (P interaction ⫽
0.78). The adjusted relative risk of a rise in creatinine
among nondiabetics at baseline was 1.48 (P ⫽ 0.04) for
each threefold higher triglycerides and 2.44 (P ⫽ 0.007)
among diabetics (P interaction ⫽ 0.19). Additionally, for
each threefold higher triglycerides, the adjusted relative
risk of a rise in creatinine was 1.65 (P ⫽ 0.05) and 1.57
(P ⫽ 0.03) for normotensives and hypertensives, respec-
tively. Body mass index was not associated with declining
renal function, and further adjustment for it yielded no
meaningful changes in the association between lipids and
renal dysfunction.
Adjustment for serum glucose as a continuous variable
297
Fig. 1. Adjusted relative risk of an incident
rise in creatinine between ARIC visit 1 and
visit 2 for a threefold higher baseline plasma
triglyceride level (1.10 loge units) overall and
in selected subgroups. Relative risks were ad-
justed for sex, race, and baseline age, diabetes
mellitus status, systolic blood pressure, type
of hypertension medication use, and creati-
nine in a Poisson regression model.
in addition to the a priori chosen clinical and demo-
graphic characteristics only slightly attenuated the asso-
ciations between triglycerides and HDL and a rise in
creatinine (model 3 compared with model 2; Table 3).
Further adjustment for insulin levels in nondiabetics sub-
stantially attenuated the associations between HDL cho-
lesterol, triglycerides, and a rise in creatinine (model 4;
Table 3). Among nondiabetics, adjustment for insulin
negated the associations between both triglycerides and
HDL cholesterol and a rise in creatinine (both P ⬎ 0.20).
Next, the data were analyzed using change in creati-
nine clearance as the outcome, defined as a ⱖ25% de-
crease in creatinine clearance. There were 407 incident
cases during follow-up. Over 93% of cases defined by a
rise in creatinine met this alternate outcome definition,
but only 48% of persons having a ⱖ25% decrease in
calculated creatinine clearance had a creatinine increase
ⱖ0.4 mg/dL. After adjusting for age, sex, race, diabetes,
systolic blood pressure, type of antihypertensive medica-
tion, and baseline creatinine clearance, the relative risk
of developing a ⱖ25% decrease in creatinine clearance
was 1.54 (95% CI 1.2, 1.9, P ⫽ 0.001) for each threefold
higher triglycerides. Redefining the outcome as a larger
decline (ⱖ30%, ⱖ40%) in creatinine clearance resulted
in progressively larger relative risks. For each threefold
higher triglycerides, the relative risks of developing a
298 Muntner et al: Lipids and renal disease

Table 3. Relative risks and 95% confidence intervals for a rise in creatinine for threefold higher plasma triglycerides and 30 mg/dL
higher HDL cholesterol
Overall No Diabetes Mellitus Diabetes Mellitus
Lipid Model RR (95% CI) P value RR (95% CI) P value RR (95% CI) P value
Plasma 1. Crude 1.77 (1.3,2.4) 0.0001 1.44 (1.0,2.0) 0.04 1.77 (1.0,3.2) 0.05
triglycerides 2. Adjusteda 1.64 (1.2,2.2) 0.002 1.48 (1.0,2.1) 0.04 2.44 (1.3,4.7) 0.007
3. Glucoseb 1.49 (1.1,2.1) 0.02 1.42 (1.0,2.1) 0.07 1.88 (1.0,3.7) 0.07
4. Fullc 1.37 (1.0,1.9) 0.07 1.25 (0.8,1.9) 0.28 — —
HDL 1. Crude 0.53 (0.4,0.8) 0.0008 0.66 (0.4,1.0) 0.03 0.38 (0.1,1.1) 0.06
cholesterol 2. Adjusteda 0.58 (0.4,0.9) 0.01 0.62 (0.4,1.0) 0.03 0.31 (0.1,1.0) 0.05
3. Glucoseb 0.63 (0.4,0.9) 0.03 0.65 (0.4,1.0) 0.05 0.39 (0.1,1.2) 0.11
4. Fullc 0.70 (0.5,1.1) 0.10 0.75 (0.5,1.2) 0.22 — —
a
Adjusted for sex, race, and baseline age, antihypertensive medications, serum creatinine, and systolic blood pressure
b
Adjusts for baseline glucose in addition to the factors in model 2
c
Adjusts for baseline glucose, insulin in nondiabetics, and the factors in model 2

large decline in creatinine clearance (ⱖ30%, ⱖ40%)
were 1.68 (P ⫽ 0.004) and 4.67 (P ⫽ 0.001), both of
which were statistically significant. Adjustment for fast-
ing insulin and glucose had only a minimal effect on the
magnitude of these associations, and triglycerides, but
not HDL cholesterol, were a statistically significant pre-
dictor of a decline in creatinine clearance. After adjust-
ment for insulin, glucose, and the other covariates, a
threefold higher amount of triglycerides was associated
with a relative risk of 1.51 (95% CI, 1.2 to 2.0; P ⫽ 0.003)
for a 25% decline in creatinine clearance.
As mentioned previously, ARIC participants on lipid-
lowering medications were excluded from these analyses.
When the 563 participants on lipid-lowering medications
were included, the results remained virtually identical in
both magnitude and significance. Re-analyzing the data
excluding persons who were placed on lipid-lowering
medications during follow-up (N ⫽ 605, 22 cases of renal
dysfunction) did not substantially change the results of
any of our analyses.
DISCUSSION
This study is the first to our knowledge to assess pro-
spectively the association between a large number of
lipids and the subsequent decline in renal function in
the general population. In this study, HDL, HDL-2 cho-
lesterol, and triglycerides were all significant predictors
of a rise in serum creatinine, a measure of renal function.
These associations persisted after adjustment for demo-
graphic and clinical information, including sex, race, age,
systolic blood pressure, diabetes mellitus status, and type
of antihypertensive medication use. Among the lipids
we investigated, triglycerides had the strongest and most
statistically significant association with a future decline
in renal function. After adjustment for clinical and demo-
graphic factors, the highest quartile of log triglycerides
imparted a 65% higher risk of developing a rise in creati-
nine ⱖ0.4 mg/dL within three years compared with the
lowest quartile of triglycerides. Even after adjustment
for other lipid levels, demographic, and clinical charac-
teristics, higher triglycerides at baseline were associated
with a significantly higher risk of creatinine rise during
follow-up. When insulin levels among nondiabetics were
considered in the analysis, however, the effect of triglyc-
erides was reduced and no longer significant. In contrast,
when a 25% or greater decline in creatinine clearance
was used as the outcome, higher plasma triglycerides
predicted a loss of renal function independent of insulin
levels. These results raise the possibility that the insulin-
resistance syndrome may underlie or mediate the ob-
served associations between lipids and a loss of renal
function. Although BMI is frequently associated with
both renal disease and the insulin-resistance syndrome,
further adjustment for BMI yielded no changes in the
association between lipids and renal dysfunction.
Possible mechanisms whereby dyslipidemia causes re-
nal injury have been reviewed [8, 37, 38]. While the
pathophysiology by which hypercholesterolemia may
cause damage to the kidney is uncertain, experimental
data support the hypothesis that glomerulosclerosis and
atherosclerosis share common mechanisms [38]. Similar-
ities in anatomical origin and functional properties of
glomerular mesangial cells to those of vascular smooth
muscle cells provide clues to analogous cellular re-
sponses common to both atherosclerosis and glomerulo-
sclerosis [37, 38]. Many of the studies reviewed have
shown that glomerular lipid deposits and the appearance
of foam cells accompany renal injury in models of hyper-
lipidemia. Experimental data also suggest that dyslipide-
mia influences renal injury. In several animal models,
feeding high cholesterol diets were associated with the
initiation of renal dysfunction [38], and pharmacologic
lowering of lipids with either clofibrate or lovastatin re-
sulted in a reduction of renal injury [39–41].
Several prospective studies in humans have shown an
association between dyslipidemia and renal disease pro-
gression among patients with proven renal disease
[18–21]. In the Modification of Diet in Renal Disease
 Muntner et al: Lipids and renal disease
(MDRD) study, low HDL cholesterol independently
predicted the rate of renal decline among 840 partici-
pants, but triglycerides were not examined [20]. In
smaller studies, total cholesterol [21, 22] and LDL choles-
terol [22], or apo(B) [22, 23] predicted renal disease
progression. A detailed lipoprotein analysis in 44 pa-
tients [18] showed a strong correlation of the rate of
glomerular filtration rate decline with the plasma concen-
tration of triglyceride-rich apoB containing lipoproteins
but no significant association with cholesterol-rich apoB-
containing lipoproteins. Only one prospective study has
examined dyslipidemia and the risk of renal function
decline among persons without known renal disease. In
a subset of 2700 males from the Helsinki Heart Study
(HHS), a randomized clinical trial originally designed to
investigate the ability of gemfibrozil to prevent coronary
heart disease, low HDL predicted a rise in creatinine,
while no association was seen with triglycerides [10].
The association between HDL cholesterol and a rise in
creatinine was present in both the gemfibrozil treatment
and placebo groups of the HHS, while no association
between triglyceride levels and a rise in creatinine was
present in either group. The difference between our
study and the HHS in the relative importance of HDL
and triglycerides may be due to restriction of the Helsinki
study to persons with elevated non-HDL cholesterol. In
addition, the analysis in the HHS focused on the mean
progression rate in all individuals, not the development
of hypercreatinemia as in our study. The mean progres-
sion rate was not related to age, a well-demonstrated
risk factor for loss of renal function, and surprisingly was
more strongly related to HDL cholesterol than blood
pressure. Results from the present study provide addi-
tional evidence that an excess of triglyceride-rich lipo-
proteins and associated low HDL cholesterol plays a
larger role in the decline of renal function than total
cholesterol or LDL cholesterol. Our results extend the
HHS findings to a different population with a broader
range of lipid measurements.
Although serum creatinine is quite commonly used as
a measure of renal function, it has been criticized for
being unreliable [42, 43]. Some people with low serum
creatinine may suffer from renal dysfunction, while some
people with high serum creatinine do not always have
renal impairment [44]. Nevertheless, we feel it appro-
priate to use the change in serum creatinine between
baseline and follow-up as a measure of renal impairment
for three reasons. First, most large prospective cohort
studies investigating the risk of renal disease progression
use serum creatinine as a measure of renal function
[9, 10] since direct measurement of the glomerular filtra-
tion rate is substantially more expensive and time con-
suming. Second, there is a strong correlation between
declines in the glomerular filtration rate assessed using
125
I-iothalamate and changes in serum creatinine when
299
the length of follow-up is sufficiently long. In the pilot
phase of the MDRD study [45], this correlation was 0.70
for 15 months of follow-up. The correlation in this study
with three years of follow-up is likely to be even greater.
Third, our results remained very similar when several
different definitions of substantial renal functional de-
cline were applied. While examination of proteinuria
would have been helpful as an additional confirmation
of renal dysfunction, urine was not collected in the ARIC
study.
We use the term early renal disease progression to
reflect the fact that the majority of the individuals are
making the transition from a “normal” serum creatinine
to an “abnormal” serum creatinine. Although the results
of this study hold for patients with a normal serum creati-
nine at baseline, we cannot determine that these patients
had no renal pathology at baseline. Thus, we cannot
conclude that lipids affect renal disease initiation rather
than early progression.
A possible concern with our analysis is the phenome-
non of regression to the mean. Because of variability
or measurement error, participants with a low serum
creatinine measurement at baseline may have a higher
measurement at follow-up because their baseline mea-
surement was lower than their long-term average (or
“true”) value [46]. These participants’ serum creatinine
may have merely regressed back to its long-term mean
level. We identified 32 ARIC participants whose creati-
nine was 0.2 mg/dL or greater below the mean at baseline
and whose creatinine measurement at the follow-up visit
was 0.4 mg/dL higher. Removing these 32 cases and
repeating our analyses yielded nearly identical results.
Alternatively, defining cases as individuals who had a
normal serum creatinine at baseline and a rise in creati-
nine of ⱖ0.4 mg/dL to at least 0.2 mg/dL greater than
the upper limit of normal yielded similar results as well.
Even though the consistent prospective association
between lipids and a rise in creatinine was statistically
significant and independent of other risk factors, the
potential for reverse causality is of concern. Lipoprotein
abnormalities are very common in end-stage renal dis-
ease patients [47–49]. The most common abnormalities
are elevated triglycerides and low HDL cholesterol. In
this analysis of ARIC data, lipids were measured only
three years prior to the follow-up determination of creat-
inine at visit 2. Dyslipidemia may have been present
because of subclinical renal disease at baseline. It is possi-
ble that like hypertension, dyslipidemia accelerates renal
disease progression and renal disease increases dyslipi-
demia. An appropriate approach to reduce the effect of
subclinical disease is to perform a study with longer fol-
low-up. Serum creatinine measurements will soon be
available from the nine-year follow-up of the ARIC co-
hort. These data will provide an excellent opportunity
300 Muntner et al: Lipids and renal disease
to more fully determine the effect that dyslipidemia has
on the incidence of declining renal function.
In conclusion, this study implicates hypertriglyceride-
mia in the initiation of mild renal insufficiency. Specifi-
cally, our study shows that persons with low HDL choles-
terol and hypertriglyceridemia at baseline have a higher
risk of having a loss of renal function. While several
clinical trials have shown that the use of lipid-lowering
medications and low cholesterol diets are effective in
raising HDL cholesterol and lowering triglycerides in
persons with renal insufficiency [50–53], it is not known
whether lipid-lowering therapy reduces renal disease
progression. Additional studies, of the long-term effects
of dyslipidemia on mild renal insufficiency, the benefits
of lipid-lowering therapy in preventing renal insuffi-
ciency, and the role of the insulin resistance syndrome
on the development of early renal dysfunction need to
be performed.
ACKNOWLEDGMENTS
The ARIC study is carried out as a collaborative study supported
by Contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-
55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022 from the
National Heart, Lung, and Blood Institute. Dr. Muntner was supported
by National Institutes of Health training grant T32HL07024-23. Dr.
Coresh was supported by grants from the National Institute of Diabetes
Digestive and Kidney Disease (DK48362) and the National Center
for Research Resources (GCRC grant RR00722). We wish to thank
the staff at the ARIC field centers and the coordinating center. In
particular, the following persons are acknowledged: Phyllis Johnson,
Marilyn Knowles, and Melisa LaVergne from the University of North
Carolina at Chapel Hill; Amy Haire, Delilah Posey, and Leslie Angel-
Potter from the University of North Carolina at Winston Salem; Mary-
Louise Lauffer, Suzanne Pillsbury, and Anne Safrit from Wake Forest
University, Winston Salem; Cora L.K. Peoples, Cecile Snell, and Betty
S. Warren from University of Mississippi Medical Center at Jackson;
Molly Harrington, Darlene Heath, and Eli Justiniano from University
of Minnesota, Minneapolis Center; Sunny Harrell, Patricia Hawbeaker,
and Joan Nelling from The Johns Hopkins University, Baltimore; Susan
Mitterling, Ashley Ewing, and R. Christy Moore from the University
of Texas Medical School at Houston; Doris J. Harper, Charles E.
Rhodes, and Julita Samoro from Methodist Hospital Atherosclerosis
Clinical Laboratory, Houston; and Debbie Rubin Williams, Patsy
Tacker, and Lily Wang from the ARIC Coordinating Center at the
University of North Carolina at Chapel Hill.
Reprint requests to Josef Coresh, M.D., Ph.D., Welch Center for
Prevention, Epidemiology, and Clinical Research, The Johns Hopkins
Medical Institutions, 2024 East Monument Street, Suite 2-600, Baltimore,
Maryland 21205-2223, USA.
E-mail: coresh@jhu.edu
REFERENCES
1. Neaton J, Wentworth D: Serum cholesterol, blood pressure, ciga-
rette smoking, and death from coronary heart disease: Overall
findings and differences by age for 316,099 white men: Multiple
Risk Factor Intervention Trial Research Group. Arch Intern Med
152:55–64, 1992
2. Castelli WP: Epidemiology of coronary heart disease: The Fra-
mingham study. Am J Med 76:4–12, 1984
3. Stamler J, Dyer AR, Shekelle RB, Neaton J, Stamler R: Rela-
tionship of baseline major risk factors to coronary and all-cause
mortality, and to longevity: Findings from long-term follow-up of
Chicago cohorts. Cardiology 82:191–222, 1993
4. MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton
J, Abbott R, Godwin J, Dyer A, Stamler J: Blood pressure,
stroke, and coronary heart disease. I. Prolonged differences in
blood pressure: Prospective observational studies corrected for
regression dilution bias. Lancet 335:765–774, 1990
5. Stamler J, Stamler R, Neaton JD: Blood pressure, systolic and
diastolic and cardiovascular risks: US population data. Arch Intern
Med 153:598–615, 1993
6. Perneger T, Nieto F, Whelton P, Klag MJ, Comstock GW,
Szklo M: A prospective study of blood pressure and serum creati-
nine: Results from the “CLUE” study and ARIC study. JAMA
269:488–493, 1993
7. Ma KW, Greene EL, Raij L: Cardiovascular risk factors in chronic
renal failure and hemodialysis populations. Am J Kidney Dis
19:505–513, 1992
8. Diamond J, Karnovsky M: Focal and segmental glomerulosclero-
sis: Analogies to atherosclerosis. Kidney Int 33:917–924, 1988
9. Klag M, Whelton P, Randall B, Randall BL, Neaton JD,
Brancati FL, Ford CE, Shulman NB, Stamler J: Blood pressure
and end-stage renal disease in men. N Engl J Med 334:13–18, 1996
10. Manttari M, Tiula E, Alikoski T, Manninen V: Effects of hyper-
tension and dylipidemia on the decline in renal function. Hyperten-
sion 26:670–676, 1995
11. Kamanna VS, Roh DD, Kirschenbaum MA: Atherogenic lipopro-
teins: Mediators of glomerular injury. Am J Nephrology 13:1–5,
1993
12. Wheeler DC, Bernard DB: Lipid abnormalities in the nephrotic
syndrome: Causes, consequences, and treatment. Am J Kidney Dis
23:331–346, 1994
13. Appel G: Lipid abnormalities in renal disease. Kidney Int 39:169–
183, 1991
14. Cramp DG: Plasma lipid alterations in patients with chronic renal
disease. CRC Crit Rev Clin Lab Sci 17:77–101, 1982
15. Kasiske B, O’Donnell M, Garvis W: Pharmacologic treatment
of hyperlipidemia reduces glomerular injury in rat 5/6 nephrectomy
model of chronic renal failure. Circ Res 62:367–375, 1988
16. Kasiske B, O’Donnell M, Schmitz P, Kim Y, Keane WF: Renal
injury in diet induced hypercholesterolemia in rats. Kidney Int
37:880–891, 1990
17. Sammuelsson O, Attman P, Knight-Gibson C, Larsson R, Mulec
H, Wedel H, Weiss L, Alaupovic P: Plasma levels of lipoprotein
(a) do not predict progression of human chronic renal failure.
Nephrol Dial Transplant 11:2237–2243, 1996
18. Sammuelsson O, Attman P, Knight-Gibson C, Larsson R, Mulec
H, Weiss L, Alaupovic P: Complex apolipoprotein B-containing
lipoprotein particles are associated with a higher rate of progression
of human chronic renal insufficiency. J Am Soc Nephrol 9:1482–
1488, 1998
19. Diamond J: Effects of dietary interventions on glomerular patho-
physiology. Am J Physiol 258:F1–F8, 1990
20. Hunsicker LG, Adler S, Gaggiula A, England BK, Greene T,
Kusek JW, Rogers NL, Teschan PE: Predictors of the progression
of renal disease in the Modification of Diet in Renal Disease Study.
Kidney Int 51:1908–1919, 1997
21. Washio M, Okuda S, Ikeda M, Hirakata H, Nanishi F, Onoyama
K, Yoshimura T, Fujishima M: Hypercholesterolemia and the
progression of the renal dysfunction in chronic renal failure pa-
tients. J Epidemiol 6:172–177, 1996
22. Samuelsson O, Mulec H, Knight-Gibson C, Attman PO, Kron
B, Larsson R, Weiss L, Wedel H, Alaupovic P: Lipoprotein
abnormalities are associated with increased rate of progression of
human chronic renal insufficiency. Nephrol Dial Transplant 12:
1908–1915, 1997
23. Samuelsson O, Aurell M, Knight-Gibson C, Alaupovic P, Att-
man PO: Apolipoprotein-B containing lipoproteins and the pro-
gression of renal insufficiency. Nephron 63:279–285, 1993
24. The ARIC Investigators: The Atherosclerosis Risk in Communi-
ties (ARIC) Study: Design and objectives. Am J Epidemiol
129:687–702, 1989
25. Perrone R, Madias N, Levey A: Serum creatinine as an index of
renal function: New insights into old concepts. Clin Chem 38:1933–
1953, 1992
26. Benfante R, Hwang LJ, Masaki K, Curb JD: To what extent do
 Muntner et al: Lipids and renal disease
cardiovascular risk factor values measured in elderly men represent
their midlife values measured 25 years earlier? A preliminary re-
port and commentary from the Honolulu Heart Program. Am J
Epidemiol 140:206–216, 1994
27. National Heart Lung and Blood Institute Atherosclerosis
Risk in Communities (ARIC) Study: Operations Manual No. 7:
Blood Collection and Processing. Bethesda, National Heart, Lung
and Blood Institute, 1987
28. Eckfeldt JH, Chambless LE, Shen YL: Short-term within person
variability in clinical chemistry results: Experience from the Ath-
erosclerosis Risk in Communities Study. Arch Pathol Lab Med
118:496–500, 1994
29. Cockroft DW, Gault MH: Prediction of creatinine clearance from
serum creatinine. Nephron 16:31–41, 1976
30. du Bois D, du Bois EF: A formula to estimate the approximate
surface area if height and weight be known. Arch Intern Med
17:863–871, 1916
31. Kasiske BL, Ma JZ, Kalil RSN, Louis TA: Effects of antihyper-
tensive therapy on serum lipids. Ann Intern Med 122:133–141, 1995
32. Weidmann P, Uehlinger DE, Gerber A: Antihypertensive treat-
ment and serum lipoproteins. J Hypertens 3:297–306, 1985
33. STATA: Version 5.0. College Station, TX
34. Dengel DR, Goldberg AP, Mayuga RS, Kairis GM, Weir MR:
Insulin resistance, elevated glomerular filtration fraction, and renal
injury. Hypertension 28:127–132, 1996
35. Reaven GM: The kidney: An unwilling accomplice in syndrome
X. Am J Kidney Dis 30:928–931, 1997
36. Dzurik R, Spustova V, Janekova K: The prevalence of insulin
resistance in kidney disease patients before the development of
renal failure. Nephron 69:281–285, 1995
37. Klahr S, Schreiner G, Ichikawa I: The progression of renal
disease. N Engl J Med 318:1657–1670, 1988
38. Diamond JR, Karnovsky MJ: Focal and segmental glomeruloscle-
rosis: Analogies to atherosclerosis. Kidney Int 33:917–924, 1988
39. Kasiske B, O’Donnell M, Cowardin W, Keane WF: Lipids and
the kidney. Hypertension 15:443–450, 1990
40. Kasiske B, O’Donnell M, Garvis W, Keane WF: Pharmacologic
treatment of hyperlipidemia reduces glomerular injury in rat 5/6
301
nephrectomy model of chronic renal failure. Circ Res 62:367–374,
1988
41. Kasiske B, O’Donnell M, Cleary M, Keane WF: Treatment of
hyperlipidemia reduces glomerular injury in obese Zucker rats.
Kidney Int 33:667–672, 1988
42. Shemesh O, Golbertz H, Kriss J, Myers BD: Limitations of
creatinine as a filtration marker in glomerulopathic patients. Kid-
ney Int 28:830–838, 1985
43. Walser M, Drew HH, LaFrance ND: Reciprocal creatinine slopes
often give erroneous estimates of progression of chronic renal
failure. Kidney Int 36(Suppl 27):S81–S85, 1989
44. Walser M, Drew HH, LaFrance ND: Creatinine measurements
often yielded false estimates of progression in chronic renal failure.
Kidney Int 34:412–418, 1988
45. Modification of Diet in Renal Disease (MDRD) Study Group:
Assessing the progression of renal disease in clinical studies: Effect
of duration of follow-up and regression to the mean. J Am Soc
Nephrol 1:1087–1094, 1991
46. Yudkin PL, Stratton IM: How to deal with regression to the
mean in intervention studies. Lancet 347:241–243, 1996
47. Cameron J: The nephrotic syndrome and its complications. Am J
Kidney Dis 10:157–171, 1987
48. Cutler R: Lipid disorders in renal disease: Prevalence, pathogene-
sis and diagnosis. Dial Transplant 17:533–536, 1988
49. Chan M, Varghese Z, Moorhead J: Lipid abnormailities in ure-
mia, dialysis, and transplantation. Kidney Int 19:625–637, 1981
50. Goldberg A, Applebaum-Bowden D, Bierman EL, Hazzard
WR, Haas LB, Sherrard DJ, Brunzell JD, Huttunen JK, Ehn-
holm C, Nikkila EA: Increase in lipoprotein lipase during clofi-
brate treatment of hypertriglyceridemia in patients on hemodialy-
sis. N Engl J Med 301:1073–1076, 1979
51. Sherrard D, Goldberg A, Hass L, Brunzell JD: Chronic clofi-
brate therapy in maintenance hemodialysis patients. Nephron
25:219–221, 1980
52. Sanfelippo M, Swenson R, Reaven G: Response of plasma triglyc-
erides to dietary changes in patients on hemodialysis. Kidney Int
14:180–186, 1978
53. Cattran D, Steiner G, Fenton S, Ampil M: Dialysis hyperlipid-
emia: Response to dietary manipulation. Clin Nephrol 13:177–182,
1980