Reduced Kidney Function as a Risk Factor for Incident Heart

Failure: The Atherosclerosis Risk in Communities (ARIC)

Study
Anna Kottgen,* Stuart D. Russell, Laura R. Loehr, Ciprian M. Crainiceanu,
Wayne D. Rosamond, Patricia P. Chang, Lloyd E. Chambless,**and Josef Coresh*
Departments of *Epidemiology and Biostatistics, Johns Hopkins Bloomberg School of Public Health, Welch Center for
Prevention, Epidemiology and Clinical Research, Johns Hopkins Medical Institutions, and Department of Medicine,
Johns Hopkins University School of Medicine, Baltimore, Maryland; and Departments of Epidemiology, Medicine, and
**Biostatistics, University of North Carolina, Chapel Hill, North Carolina

Reduced kidney function is a risk factor for cardiovascular morbidity and mortality, and both heart failure (HF) and kidney
failure incidences are increasing. This study therefore sought to determine the effect of decreased kidney function on HF
incidence in a population-based study of middle-aged adults. From 1987 through 2002, 14,857 participants of the Atheroscle-
rosis Risk in Communities (ARIC) study who were free of prevalent HF at baseline were followed for incident HF
hospitalization or death (International Classification of Diseases, Ninth Revision/10th Revision 428/I50). Estimated GFR
(eGFR) was calculated using the abbreviated Modification of Diet in Renal Disease (MDRD) Study equation, and kidney
function was categorized as normal (eGFR >90 ml/min per 1.73 m2; n 7143), mildly reduced (eGFR 60 to 89 ml/min per 1.73
m2; n 7311), and moderately/severely reduced (eGFR <60 ml/min per 1.73 m2; n 403). Cox proportional hazards models
were used to control for demographic and cardiovascular risk factors; analyses were stratified by the presence of coronary
heart disease at baseline. During a mean follow-up of 13.2 yr, 1193 participants developed HF. The incidence of HF was
three-fold higher for individuals with eGFR <60 ml/min per 1.73 m2 compared to the reference group with eGFR >90 ml/min
per 1.73 m2 (18 versus 6 per 1000 person-years). The overall adjusted relative hazard of developing HF was 1.94 (1.49 to 2.53)
for individuals with eGFR <60 ml/min per 1.73 m2 compared to the reference group and was significantly increased for
individuals with and without prevalent coronary heart disease at baseline. A substantially greater decline in kidney function
occurred in individuals concomitant with HF hospitalization/death compared to those who did not develop HF. In summary,
middle-aged adults with moderately/severely reduced kidney function are at high risk for developing HF.
J Am Soc Nephrol 18: 13071315, 2007. doi: 10.1681/ASN.2006101159

R
educed kidney function has been established as a risk
factor for cardiovascular disease (CVD) in several re-
cent studies, both in populations at high risk for CVD
and in the general population (17). Moderately reduced kid-
ney function is very common, affecting an estimated 8.3 million
US adults (8). Specifically, reduced kidney function has been
proposed as a risk factor for deterioration of prevalent heart
failure (HF) as well as a risk factor for incident HF (9 15).
However, most previous studies were restricted to subgroups
such as elderly individuals (9 11,13), predominantly white in-
dividuals (16), or individuals with preexisting coronary heart
disease (CHD) (12). These individuals might be at increased
risk for incident HF as a result of advanced age or comorbidi-
ties. Therefore, we sought to determine the role of impaired
kidney function as a risk factor for incident HF in a large,
Received October 26, 2006. Accepted January 31, 2007.
Published online ahead of print. Publication date available at www.jasn.org.
Address correspondence to: Dr. Josef Coresh, 2024 E. Monument Street, Suite
2-600, Baltimore, MD 21287. Phone: 410-955-0495; Fax: 410-955-0476; E-mail:
coresh@jhu.edu
population-based, biracial study of middle-aged US adults, the
Atherosclerosis Risk in Communities (ARIC) Study. We hy-
pothesized that individuals with reduced kidney function are
at increased risk for incident HF and sought to estimate both
the absolute risk and the adjusted relative risk.
Chronic kidney disease (CKD) and HF often occur together
(1,14,17,18), but relatively few studies have data on the decline
in kidney function in relation to incident HF. A recent study of
individuals with left ventricular systolic dysfunction reported
significantly higher mortality for those with a more rapid com-
pared to those with a slower decline in kidney function (19).
Using data from multiple ARIC study visits, we also investi-
gated the changes in kidney function in the years before and
after the first HF hospitalization.
Finally, previous studies did not account for the impact of
measurement error and biologic variability in serum creatinine
on the association between reduced kidney function and inci-
dent HF. Therefore, it is useful to use models that take into
account variability in estimated kidney function that is assessed
using a creatinine-based estimating equation, a procedure that
is feasible in a large population-based study but subject to

Copyright 2007 by the American Society of Nephrology ISSN: 1046-6673/1804-1307

1308 Journal of the American Society of Nephrology
sizable measurement error as a result of both biologic and
measurement variation (20).
Materials and Methods
Study Population and Design
The ARIC Study is a community-based prospective cohort to inves-
tigate the etiology of atherosclerosis (21). From 1987 to 1989, 15,792
male and female volunteers aged 45 to 64 were recruited by probability
sampling from four US communities (Forsyth County, NC; suburban
Minneapolis, MN; Washington County, MD; and Jackson, MS). The
baseline visit (visit 1) included at-home interviews, laboratory measure-
ments, and clinic examinations. The study participants returned for
additional visits in 1990 to 1992 (visit 2), 1993 to 1995 (visit 3), and 1996
to 1998 (visit 4). Details of the enrollment process and study procedures
are fully described elsewhere (21).
Baseline Variables and Data Collection
The study participants provided information on demographic, socio-
economic, health behavior, risk factor control, and medical history
variables to a trained interviewer at the baseline visit. Resting, seated
systolic (SBP) and diastolic blood pressure (DBP) measurements were
recorded by certified technicians using a random-zero sphygmoma-
nometer, and the average of the second and third readings was used.
Body mass index was calculated as measured weight (kg)/height (m2).
Medication use was self-reported and verified by the inspection of
medication bottles. For laboratory measurements, fasting blood sam-
ples were drawn, centrifuged, frozen, and shipped to ARIC study
laboratories for analysis (22). Measurement of serum glucose, albumin,
and plasma LDL and HDL followed standard ARIC protocols (23,24).
Ultrasound examinations to assess the presence of carotid atheroscle-
rosis have been described previously (25). Serum creatinine was mea-
sured using a modified kinetic Jaffe reaction from serum samples at
visits 1 and 2 and plasma samples at visit 4 (24). Kidney function was
then assessed by calculation of the estimated GFR (eGFR) using the
abbreviated Modification of Diet in Renal Disease (MDRD) Study equation
(26): eGFR (ml/min per 1.73 m2 ) 186.3*(serum creatinine in
mg/dl1.154)*(age0.203)*(0.742 if female)*(1.21 if black). For obtaining
more valid estimates from this formula, creatinine values were standard-
ized across ARIC study visits, and creatinine values from all ARIC visits
were calibrated using regression to the Cleveland Clinic laboratory, where
the MDRD equation was developed, by subtraction of 0.24 mg/dl from
visits 1 and 2 and addition of 0.18 mg/dl to visit 4 (2,27).
Assessment of HF
Study participants with evidence of prevalent HF (n 752) at base-
line were excluded from analyses. Prevalent HF was defined as the
reported current intake of HF medication at visit 1 (n 83) or evidence
of manifest HF as defined by the Gothenburg criteria stage 3 (n 669),
which require the presence of specific cardiac and pulmonary symp-
toms as well as medical treatment of HF (28,29).
Incident HF was defined as the first HF hospitalization or HF coded
as the underlying cause of death and was identified through review of
local hospital discharge lists that showed HF in any position and
county death certificates. Hospitalizations were coded as heart failure
(428) using the International Classification of Diseases Code, Ninth Revision
(ICD-9), and deaths were coded as HF (428 and I50) using the ICD-9
and ICD-10. All cohort hospitalizations and deaths that occurred before
January 1, 2003, were included.
Statistical Analyses
Analyses were limited to 14,857 participants, excluding those with
prevalent HF (n 752) and those who were missing information that
J Am Soc Nephrol 18: 13071315, 2007
was needed to calculate eGFR (race, n 47; baseline serum creatinine,
n 136). Current smoking (yes, no), completed years of education
(12, 12 to 16, 16), and evidence of atherosclerosis of the common
carotid arteries on ultrasound examination (shadowing or plaque on
either side, none) were categorized. Use of antihypertensive medication
at baseline included the use of , , and calcium channel blockers,
angiotensin-converting enzyme inhibitors, diuretics, or a combination
thereof. SBP (mmHg) and age (years) were analyzed continuously.
Prevalent diabetes was defined as a fasting serum glucose level 126
mg/dl, nonfasting glucose level 200 mg/dl, report of a physician
diagnosis of diabetes, or current use of diabetes medication. Prevalent
CHD was based on evidence of previous myocardial infarction by
electrocardiogram at visit 1, history of physician-diagnosed myocardial
infarction, and previous coronary reperfusion procedure (bypass, an-
gioplasty). Anemia was defined as present for hemoglobin values 13.5
g/dl in men and 12 g/dl in women. Kidney function was categorized
using national guidelines (30) as normal (eGFR 90 ml/min
per 1.73 m2), mildly decreased (eGFR 60 to 89 ml/min per 1.73 m2), and
moderately or severely (eGFR 60 ml/min per 1.73 m2). The last category
includes individuals with moderately (eGFR 30 to 59 ml/min per 1.73 m2)
and severely (eGFR 30 ml/min per 1.73 m2) decreased kidney function,
because only 25 individuals had severely decreased kidney function at
baseline. eGFR as a measure of kidney function was also modeled
continuously and was explored for deviations from linearity using
linear spline models. To account for measurement error in eGFR, these
analyses were repeated using the simulation extrapolation (SIMEX)
methodology (31).
Analyses were conducted using Stata version 9.2 software (32). De-
mographic and health characteristics by eGFR and HF status were
compared using 2 tests, t tests, and ANOVA as applicable. HF inci-
dence rates and 95% confidence intervals (CI) were calculated using
time-to-event methods. The proportion of individuals who remained
free of incident HF at any time during follow-up was calculated using
the Kaplan-Meier method. For all survival analyses, the follow-up time
was defined as the period from entry into the study (visit 1) to the first
HF hospitalization, death from HF, or up to the time an individual left
the study. Individuals who were free of HF by January 1, 2003, were
subject to administrative censoring. Relative hazards (RH) of incident
HF in those with a reduced compared with those with a normal level of
kidney function were calculated using Cox proportional hazards mod-
els. Extended models that subsequently adjusted for covariates as well
as for interactions between covariates were compared with nested
models using likelihood ratio tests. The proportionality assumption of
all Cox models was assessed by inspection of the complementary
log(log[survival function]) curves. Changes in eGFR between study
visits were plotted using an Epanechnikov kernel density estimator for
smoothing (33), and the proportion of individuals with a decline of
eGFR 25% (34) between study visits was assessed by HF status.
Results
Of the 14,857 participants who were available for analysis,
7143 (48.1%) had an eGFR of 90 ml/min per 1.73 m2 at visit 1,
7311 (49.2%) had an eGFR of 60 to 89 ml/min per 1.73 m2, and
403 (2.7%) had an eGFR of 60 ml/min per 1.73 m2. Partici-
pants were followed for an average (SD) of 13.2 (3.0) years. Of
the total study population, 45.6% were male, 25.9% were black,
and the mean age (SD) was 54.1 (5.8) years at baseline. During
the course of follow-up, 2079 (14.0%) study participants died.
J Am Soc Nephrol 18: 13071315, 2007 Kidney Function and Incident Heart Failure 1309

Baseline Characteristics
The mean baseline eGFR (SD) of the total study population
was 93.2 (20.7) ml/min per 1.73 m2. The baseline characteristics
of participants in each of the three eGFR categories are pre-
sented in Table 1. In general, the prevalence of cardiovascular
risk factors except for male gender and smoking was signifi-
cantly higher with reduced kidney function.
Baseline characteristics of participants by HF status are com-
pared in Table 2. Because the presence of CHD at baseline was
significantly related to both main exposure and outcome, anal-
yses were conducted to stratify the group of individuals who
developed HF by the presence of CHD at baseline. A total of 180
(15.5%) individuals who developed HF had evidence of CHD at
baseline, and these individuals differed significantly with respect
to several cardiovascular risk factors from those who developed
HF without prevalent CHD at baseline (Table 2).
Incidence of HF
During the course of follow-up, 1193 study participants de-
veloped HF. Of these, 1187 (99.5%) were identified through a
hospitalization. The overall incidence rate of HF was 6.1 per
1000 person-years. Crude and age-adjusted incidence rates of
HF are presented in Table 3; the reduction of age-adjusted HF
incidence rates compared to crude HF incidence rates in the
categories of reduced kidney function reflects the higher pro-
portion of older individuals with reduced kidney function.
Crude and age-adjusted incidence rates of HF were signifi-
cantly higher in individuals with baseline eGFR 60 ml/min
per 1.73 m2 as compared with those with eGFR 90 ml/min
per 1.73 m2. Again, analyses were stratified by presence of CHD
at baseline, yielding crude HF incidence rates in those with
prevalent CHD at baseline that were up to fives times as high
as in those without prevalent CHD (Table 3).
Survival Analyses
Figure 1 shows the cumulative incidence of HF during the
course of follow-up by category of eGFR and presence (Figure
1a) or absence (Figure 1b) of CHD at baseline. A higher pro-
portion of individuals with eGFR 60 ml/min per 1.73 m2
developed incident HF as compared to individuals in either of
the other two categories of renal function. The difference be-
tween the cumulative incidence curves was highly significant
(logrank test P 0.001 for no CHD, 0.01 for CHD at baseline).
After 13 yr of follow-up, 19.0% of individuals with eGFR 60
ml/min per 1.73 m2 and no baseline CHD experienced an HF
hospitalization/HF death, compared to 51.1% with baseline
CHD (Figure 1, a and b).
Analyses using Cox proportional hazards models yielded
crude RH of incident HF of 1.05 (95% CI 0.93 to 1.18) for
individuals with eGFR of 60 to 89 ml/min per 1.73 m2 and 3.24
(95% CI 2.56 to 4.09) for those with eGFR 60 ml/min per 1.73
m2 compared to those with eGFR 90 ml/min per 1.73 m2.
Models that subsequently adjusted for baseline demographic
factors and traditional and nontraditional CVD risk factors
attenuated the RH: the fully adjusted model (adjusted for age,

Table 1. Baseline characteristics of study population (n 14,857) by eGFR category (ml/min per 1.73 m2)a
eGFR 90 eGFR 60 to 89 eGFR 60
Characteristic P
(n 7143; 48.1%) (n 7311; 49.2%) (n 403; 2.7%)

Age (yr) 53.5 (5.8) 54.5 (5.6) 56.6 (5.8) 0.0001

Male gender 44.0 47.8 35.7 0.0001
Black race 37.7 14.5 23.8 0.0001
Educational levelb
11 yr completed 25.9 20.1 27.8
12 to 16 yr completed 39.7 42.2 40.9
17 yr completed 34.4 37.8 31.3 0.0001
BMI (kg/m2)b 27.6 (5.6) 27.4 (4.8) 28.5 (5.1) 0.0001
Current smokersb 30.7 21.7 20.1 0.0001
Diabetesb 12.5 8.8 23.6 0.0001
Antihypertensive medicationb 25.6 27.3 55.6 0.0001
SBPb 121.9 (18.9) 119.9 (18.3) 126.2 (22.6) 0.0001
Prevalent CHDb 3.5 4.7 7.3 0.0001
LDL cholesterol (mg/dl)b 135.5 (39.2) 139.4 (38.9) 145.7 (44.0) 0.0001
HDL cholesterol (mg/dl)b 53.1 (17.4) 50.5 (16.7) 49.3 (17.3) 0.0001
Serum creatinine (mg/dl) 0.98 (0.14) 1.18 (0.15) 1.94 (2.0) 0.0001
Anemiab 14.2 8.1 21.3 0.0001
Carotid atherosclerosisb 7.8 7.9 12.0 0.02
Serum albumin (g/dl) 3.85 (0.27) 3.89 (0.26) 3.82 (0.33) 0.0001
a
Data are percentages for dichotomous variables and mean (SD) for continuous variables. BMI, body mass index; CHD,
coronary heart disease; eGFR, estimated GFR; SBP, systolic blood pressure.
b
Missing values (number missing): Level of education, 24; BMI, 11; current smokers, 15; diabetes, 28; SBP, 3;
antihypertensive medication, 8; prevalent CHD, 302; LDL, 303; HDL, 101; anemia, 107; carotid atherosclerosis, 453.
1310 Journal of the American Society of Nephrology J Am Soc Nephrol 18: 13071315, 2007

Table 2. Baseline characteristics of individuals who developed HF versus those who did not develop HFa
HF No HF
Characteristic
No CHD CHD All All P
(n 980)b (n 180)b (n 1193) (n 13,664)

Age (yr) 56.9 (5.4) 57.2 (5.1) 56.9 (5.4) 53.8 (5.7) 0.0001
Male gender 50.3 79.4d 55.2 44.8 0.0001
Black race 35.2 20.6d 32.6 25.3 0.0001
Education levelc
11 yr completed 39.8 41.7 40.1 21.6
12 to 16 yr completed 36.2 36.1 36.2 41.4
17 yr completed 24.0 22.2 23.7 37.0 0.0001
BMI (kg/m2)c 29.6 (6.3) 29.6 (5.5) 29.6 (6.2) 27.4 (5.1) 0.0001
Current smokersc 38.1 32.8 37.3 25.0 0.0001
Diabetesc 30.9 32.4 31.0 9.2 0.0001
Antihypertensive medicationc 44.9 71.1d 48.8 25.4 0.0001
SBPc 130.7 (22.3) 126.1 (22.3)d 129.9 (22.3) 120.3 (18.2) 0.0001
LDL cholesterol (mg/dl)c 141.9 (40.7) 146.2 (40.2) 142.4 (40.6) 137.3 (39.1) 0.0001
HDL cholesterol (mg/dl)c 47.2 (16.1) 40.2 (12.6)d 46.1 (15.9) 52.2 (17.1) 0.0001
Anemiac 13.9 15.2 13.8 11.2 0.01
Carotid atherosclerosisc 15.4 22.0d 16.2 7.3 0.0001
Serum albumin (g/dl) 3.8 (0.3) 3.8 (0.3) 3.8 (0.3) 3.9 (0.3) 0.0001
Serum creatinine (mg/dl) 1.22 (1.0) 1.25 (0.45) 1.23 (0.96) 1.1 (0.3) 0.0001
eGFR (ml/min per 1.73 m2)
90 46.2 37.8 44.9 48.3
60 to 90 46.9 55.5 48.3 49.3
60 6.9 6.7 6.8 2.4 0.0001
a
Data are percentages for dichotomous variables and mean (SD) for continuous variables. HF, heart failure.
b
Data are missing for prevalent CHD on 302 individuals; therefore, numbers do not add up.
c
Missing values (number missing): Level of education, 24; BMI, 11; current smokers, 15; diabetes, 28; SBP, 3;
antihypertensive medication, 8; prevalent CHD, 302; LDL, 303; HDL, 101; anemia, 107; carotid atherosclerosis, 453.
d
Individuals with and without CHD and CHD were statistically significantly different (P 0.05).

gender, race, level of education, SBP, use of antihypertensive
medication, smoking, diabetes, body mass index, CHD at base-
line, LDL and HDL cholesterol, serum albumin, anemia, and
carotid atherosclerosis) yielded RH of 1.10 (95% CI 0.97 to 1.26,
eGFR 60 to 89 ml/min per 1.73 m2) and 1.94 (95% CI 1.49 to
2.53, eGFR 60 ml/min per 1.73 m2; Table 4), and fit the data
significantly better than a model that adjusted for demograph-
ics and traditional CVD risk factors only (likelihood ratio test
P 0.001). The significantly increased RH of incident HF in
individuals with eGFR 60 ml/min per 1.73 m2 compared to
those with eGFR 90 ml/min per 1.73 m2 was observed in
individuals with and without CHD at baseline (Table 4). More-
over, the significant findings comparing individuals with eGFR
60 ml/min per 1.73 m2 to those with eGFR 90 ml/min per
1.73 m2 and using the fully adjusted model held when individ-
uals who were free of prevalent CHD and remained free of
incident CHD up to and at the incidence of HF were evaluated
(RH 2.12; 95% CI 1.48 to 3.02; n 13,146). Finally, the associa-
tion of reduced kidney function with HF incidence remained
similar when SBP and the intake of antihypertensive medica-
tion were incorporated as time-varying covariates into the fully
adjusted Cox model (eGFR 60 to 89 ml/min per 1.73 m2: RH
1.11 [95% CI 0.97 to 1.26]; eGFR 60 ml/min per 1.73 m2: RH
2.06 [95% CI 1.59 to 2.69] compared with eGFR 60 ml/min per
1.73 m2 for all individuals, n 13,711).
eGFR was also modeled continuously using a piece-wise
linear spline model with and without accounting for measure-
ment error in eGFR (SD of eGFR measurement error 8.8; reli-
ability coefficient 0.82). Models that incorporated as many as
five spline terms were explored. Figure 2 shows this model as
well as a model that retains the only statistically significant
knot at 90 ml/min per 1.73 m2. Below an eGFR of 90 ml/min
per 1.73 m2, the RH of incident HF was 1.21 (95% CI 1.14 to
1.29) per 10 ml/min per 1.73 m2 lower eGFR, whereas it was
0.94 (95% CI 0.91 to 0.99) per 10 ml lower eGFR for eGFR values
90 ml/min per 1.73 m2 (P interaction 0.008). The increase
in RH of HF was not significantly different in individuals with
eGFR 60 ml/min per 1.73 m2 compared with those with eGFR
of 60 to 89 ml/min per 1.73 m2 (P interaction 0.145). Account-
ing for measurement error in eGFR led to a 33% higher RH of
incident HF below an eGFR of 90 ml/min per 1.73 m2: The RH
of HF was 1.28 (95% CI 1.19 to 1.37) per 10 ml/min per 1.73 m2
lower eGFR.
A significant interaction of race and eGFR was present: Using
the fully adjusted model and individuals with eGFR 90 ml/min
per 1.73 m2 as the reference category, the RH of HF in individuals
J Am Soc Nephrol 18: 13071315, 2007 Kidney Function and Incident Heart Failure 1311

Table 3. Incidence rates of HF in the ARIC cohort by baseline eGFRa

HF
Parameter
No CHD CHD Allb

Baseline eGFR 90 ml/min per 1.73 m2

no. of events 452 68 536
person-years 90,071 2738 94,810
crude IR 5.0 24.8 5.7
95% CI 4.6 to 5.5 19.3 to 31.5 5.2 to 6.2
age-adjusted IR 5.0 24.6 5.5
95% CI 4.5 to 5.4 19.1 to 31.5 5.1 to 6.0
Baseline eGFR 60 to 89 ml/min per 1.73 m2
no. of events 460 100 576
person-years 91,831 3573 97,352
crude IR 5.0 28.0 5.9
95% CI 4.6 to 5.5 22.8 to 34.0 5.4 to 6.4
age-adjusted IR 3.8 25.3 4.6
95% CI 3.4 to 4.3 19.9 to 31.1 4.1 to 5.1
Baseline eGFR 60 ml/min per 1.73 m2
no. of events 68 12 81
person-years 4296 208 4577
crude IR 15.8 57.7 17.7
95% CI 12.3 to 20.1 29.8 to 100.7 14.1 to 22.0
age-adjusted IR 11.4 36.8 12.7
95% CI 8.2 to 15.8 12.6 to 107.8 9.3 to 17.1
a
Crude rates of incident HF per 1000 person-years with 95% confidence interval (CI), and adjusted to the mean age (54.1 yr)
for all groups. ARIC, Atherosclerosis Risk in Communities; IR, incidence rate.
b
Includes 302 individuals with missing information on CHD.

Figure 1. Cumulative incidence of heart failure (HF) stratified by category of estimated GFR (eGFR) and presence of coronary heart
disease (CHD) at baseline in 14,824 Atherosclerosis Risk in Communities (ARIC) Study participants. Cumulative incidence curves
for CHD-negative individuals (a) in the eGFR categories of 90 and 60 to 89 ml/min per 1.73 m2 are virtually indistinguishable.

with eGFR of 60 to 89 ml/min per 1.73 m2 was significantly higher
in black compared to white individuals (1.57 [95% CI 1.25 to 1.98]
versus 0.94 [95% CI 0.81 to 1.10]). In individuals with eGFR 60
ml/min per 1.73 m2, the RH of HF was higher, although not
statistically significant, in black compared to white individuals
(2.40 [95% CI 1.57 to 3.67] versus 1.69 [95% CI 1.21 to 2.36]). There
was no significant interaction between reduced eGFR and diabe-
tes status at baseline or between reduced eGFR and hypertension
status at baseline (hypertension defined as SBP 140 mmHg, DBP
90 mmHg, or the current intake of antihypertensive medication).
Significant interactions between age in 5-yr intervals and both
diabetes and CHD did not alter the association of reduced kidney
function and incident HF.
To address the temporality of a decline in kidney function
1312 Journal of the American Society of Nephrology J Am Soc Nephrol 18: 13071315, 2007

Table 4. Adjusted relative hazard (95% CI) of HF by category of eGFRa

HF (No. of Events/At Risk)
Baseline eGFR Category (ml/min per 1.73 m2)
No CHD (894/13,146) CHD (166/557) Allb (1060/13,703)

90 1.0 (Reference) 1.0 (Reference) 1.0 (Reference)

60 to 89 1.10 (0.95 to 1.27) 1.20 (0.85 to 1.69) 1.10 (0.97 to 1.26)
60 1.83c (1.37 to 2.45) 2.58c (1.34 to 4.98) 1.94c (1.49 to 2.53)
a
Fully adjusted model included the following baseline variables (covariates in italics contributed significantly to the overall
model): Age, race, gender, level of education, prevalent CHD in the overall model, SBP, antihypertensive medication, diabetes, smoking,
BMI, LDL and HDL cholesterol, serum albumin, anemia, and carotid atherosclerosis.
b
A total of 302 individuals in the All group were missing information on prevalent CHD at baseline.
c
P 0.05.

Figure 3. Percentage change in eGFR between study visits 2 and

Figure 2. Relative hazard (RH) of incident HF across the range
of continuous GFR values before and after accounting for mea-
surement error in eGFR. eGFR is modeled using linear spline
models with one knot (at 90 ml/min per 1.73 m2; solid lines)
and five knots (at 60, 75, 90, 105, and 120 ml/min per 1.73 m2;
dashed lines). Estimates are shown before (black) and after
(gray) accounting for measurement error. Tick marks along the
x axis indicate eGFR values for individual participants.
4 in individuals who developed HF between visits 2 and 4
(black) versus those who did not develop HF by visit 4 (gray).
Percentages in boxes refer to the proportion of individuals with
an eGFR decline of 25% (25.7% of those who developed HF
and 6.6% of those who did not develop HF between visits 2 and
4). Vertical lines represent the mean percentage change in eGFR
between study visits 2 and 4 in those who developed (black)
and did not develop (gray) HF.

and the first HF hospitalization/HF death, we compared the Discussion

percentage change in eGFR between the study visits of those
who developed the study outcome between visits 2 and 4 (on
average 3 and 9 yr after baseline) and of those who did not
develop HF by visit 4. A 25% eGFR decline between visits 1
and 2 was present to a similar extent in those who experienced
the first HF hospitalization/HF death between visits 2 and 4
and those who did not (7.8% versus 8.6%). However, a 25%
eGFR decline in the time interval between visits 2 and 4 was
present to a much greater degree in individuals who experi-
enced the first HF hospitalization/HF death during that time
interval compared to those who did not (25.7% versus 6.6%; P
0.001; Figure 3). Therefore, a high rate of decline in kidney
function several years before HF hospitalization is not a very
strong risk factor for HF incidence, whereas the first HF hos-
pitalization is strongly associated with having a substantial
decrement in kidney function detected at a visit that on average
was approximately 2 yr after the hospitalization.
In this analysis of the ARIC Study as a prospective, commu-
nity-based, biracial sample of middle-aged adults, reduced kid-
ney function was an independent risk factor for incident HF
hospitalization or HF death. This finding is in agreement with
most previously published studies (9,10,12,13,15). The two
studies that found no significant association between reduced
eGFR and incident HF after adjustment for cardiovascular risk
factors used serum creatinine instead of a GFR estimating equa-
tion as a measure of kidney function (16,35).
National guidelines define normal kidney function as eGFR
90 ml/min per 1.73 m2. Our analyses of eGFR as a continuous
variable show that the higher risk for HF starts below this
threshold, and the risk increases further as eGFR is 60 ml/min
per 1.73 m2. At baseline, 3.0% of our study population had mod-
erately or severely reduced kidney function (eGFR 60 ml/min
per 1.73 m2) and would therefore be classified as having CKD
stage 3 or higher following national guidelines (30). This propor-
tion is in agreement with numbers reported for the general US
J Am Soc Nephrol 18: 13071315, 2007
population as represented in the Third and Fourth National
Health and Nutrition Examination Surveys (NHANES III/IV)
(8,36). After exclusion of 4.8% of the total study population be-
cause of prevalent HF, these 3% were reduced by 10%, confirming
the observation from previous studies that CKD is overrepre-
sented in individuals with prevalent HF (17,18,37). That almost
half of our study population had an eGFR of 60 to 89 ml/min per
1.73 m2 at baseline corresponding to mildly reduced kidney func-
tion is in agreement with estimates from the population-based
data of NHANES III for this age range (8). Therefore, HF risk
relates to estimates of kidney function in a large proportion of the
population with the caveat that GFR estimates in this mildly
decreased range are substantially less reliable and are strongly
influenced by the source population (20). In addition, our obser-
vation that 8.0% of our study population experienced incident HF
hospitalization/HF death during the course of follow-up corre-
sponds well to data from other large US prospective studies of
community-dwelling individuals of comparable age (38,39).
The study populations baseline distribution of characteris-
tics by outcome (HF) is consistent with the most common cause
of HF. Considering that CHD is the underlying cause for up to
70% of HF cases in the United States (38 40), one would expect
to observe a statistically significantly higher proportion of in-
dividuals who develop HF with risk factors for CHD. Because
CHD is also associated with CKD (2,4,5,7,41), we stratified our
analyses by the presence of this important potential confounder
at baseline. Still, the fully adjusted RH of incident HF in indi-
viduals without CHD at baseline and eGFR 60 ml/min per
1.73 m2 compared to those with eGFR 90 ml/min per 1.73 m2
was substantial at 1.83 (95% CI 1.37 to 2.45; Table 4). The fully
adjusted model included adjustment for two other important
potential confounders, SBP and diabetes, as well as for baseline
anemia and serum albumin. Because BP elevations, anemia,
and lower serum albumin are known consequences of reduced
kidney function, adjusting for them to some extent overadjusts
the true association between reduced kidney function and in-
cident HF. However, evidence exists that the effect of anemia
and low serum albumin as a result of albuminuria, inflamma-
tion, and malnutrition are associated with increased cardiovas-
cular risk beyond simply being measures of reduced kidney
function (42 45).
Many patients with HF have impaired kidney function
(14,19). Our study supports this observation, because the pro-
portion of individuals with eGFR 60 ml/min per 1.73 m2 was
approximately three times greater in individuals who were
excluded because of prevalent HF than in individuals who
were classified as free of HF at baseline (7.7% versus 2.7%). This
observation can result from two causal pathways: Reduced
eGFR leading to HF, or HF leading to reduced eGFR. Our
observation that reduced eGFR is an independent risk factor for
incident HF hospitalization/HF death is consistent with the
hypothesis that reduced kidney function contributes to mech-
anisms that lead to incident HF. Volume overload and renin-
angiotensin-aldosterone systemmediated increase in BP and
cardiac remodeling (46) are known consequences of CKD,
which likely contribute to this direction of the observed asso-
ciation between reduced eGFR and incident HF. CKD-related
Kidney Function and Incident Heart Failure 1313
abnormalities in nitric oxide balance, coagulation/fibrinolysis,
homocysteine, inflammation, and lipids may contribute as well.
Moreover, excess comorbidities, lesser use of beneficial thera-
pies, and excess toxicities from conventional therapies are im-
portant to consider as reasons for elevated cardiovascular risk
in CKD in addition to the vascular pathobiology of CKD (42).
Prevalent HF that leads to worsening kidney function (17,18)
can plausibly be explained by decreased renal perfusion as a
result of reduced cardiac output as well as neurohumoral
mechanisms. The observation that reduced renal function oc-
curred concomitantly with our study outcome could also be
due either to subclinical HF or to a common underlying cause,
such as CVD first manifesting itself in the kidneys.
The inability to account for the effect of subclinical HF on
worsening kidney function over the entire length of follow-up
is a limitation of our findings. Of the exclusions for prevalent
HF, 89% were based on the Gothenburg criteria, which are
reported to have high specificity (96%; SD 0.8) but only mod-
erate sensitivity (41%; SD 4.2) for diagnosis of HF in the com-
munity (47). We therefore might have failed to exclude some
individuals with prevalent HF. However, the number of mid-
dle-aged individuals who had HF and were missed by our
exclusion criteria is likely to be small given the low prevalence
of HF in this middle-aged, community-based population. Ad-
ditional sensitivity analyses that were conducted to address this
issue showed that exclusion of events that occurred in the first
3 and the first 6 yr of follow-up did not alter our results.
Other limitations of our study are related to the lack of chart
abstraction of HF events by an adjudication committee and to
the estimation of GFR using the MDRD Study equation. We
therefore conducted sensitivity analyses to address the possi-
bility that acute renal failure was misclassified as HF, excluding
from our analyses 126 individuals with ICD-9 codes 584.00 to
584.99 for acute renal failure; our significant results remained
unchanged. Whereas estimates that incorporate serum creati-
nine are valid at lower eGFR levels, they are less precise at
higher eGFR levels (20,48). Also, the indirect calibration of the
ARIC serum creatinine values to the Cleveland Clinic labora-
tory where the MDRD equation was developed may have
resulted in some inaccuracy of the eGFR. However, this would
have been true at all creatinine levels, producing a relatively
small effect on the association of eGFR with HF. The existence
of only one measurement of serum creatinine per study visit
leaves potential for misclassification of eGFR category. As a
result, because of regression dilution bias, the RH that we
observed in the category of lowest eGFR may have underesti-
mated the association observed without error, and estimates
that we present in Figure 3 correcting for measurement error
may be more appropriate. Moreover, we cannot exclude the
possibility that individuals have changed exposure category
after the last measurement of serum creatinine at visit 4 but
before their developing HF. Finally, although we adjusted for a
large number of traditional and nontraditional CVD risk fac-
tors, we cannot rule out residual confounding by unmeasured
factors.
Our study has several advantages, including a large, pro-
spective, community-based setting that represents an age
1314 Journal of the American Society of Nephrology
group that is at risk for development of kidney dysfunction as
well as HF. Furthermore, the ARIC study population is biracial,
and RH of HF that were derived from our study should there-
fore be more generalizable to the middle-aged US population.
In addition, the repeated visits of ARIC study participants
allowed for investigation of the temporality of change in eGFR
in relation to our study outcome. Moreover, we could account
not only for prevalent CHD at baseline but also for incident
CHD up to the HF event in our analyses. Finally, we were able
to account for the effect of measurement error in eGFR.
Conclusion
Moderately and severely reduced kidney function as defined
by national guidelines is an independent risk factor for incident
HF hospitalization or HF death in a population-based sample of
middle-aged individuals. eGFR as an important risk marker
should be calculated not only by nephrologists but also by
cardiologists and could then help in risk stratification and
management of patients who are at risk for HF.
Acknowledgments
The ARIC study is carried out as a collaborative study supported by
National Heart, Lung, and Blood Institute contracts N01-HC-55015,
N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-
HC-55021, and N01-HC-55022.
We thank the staff and participants of the ARIC study for their
important contributions, and Kathryn Carson for assistance with data
preparation.
Disclosures
None.
References
1. Mann JF, Gerstein HC, Pogue J, Bosch J, Yusuf S: Renal
insufficiency as a predictor of cardiovascular outcomes
and the impact of ramipril: The HOPE randomized trial.
Ann Intern Med 134: 629 636, 2001
2. Manjunath G, Tighiouart H, Ibrahim H, MacLeod B, Salem DN,
Griffith JL, Coresh J, Levey AS, Sarnak MJ: Level of kidney
function as a risk factor for atherosclerotic cardiovascular out-
comes in the community. J Am Coll Cardiol 41: 4755, 2003
3. Fried LF, Shlipak MG, Crump C, Bleyer AJ, Gottdiener JS,
Kronmal RA, Kuller LH, Newman AB: Renal insufficiency
as a predictor of cardiovascular outcomes and mortality in
elderly individuals. J Am Coll Cardiol 41: 1364 1372, 2003
4. Sarnak MJ, Levey AS, Schoolwerth AC, Coresh J, Culleton B,
Hamm LL, McCullough PA, Kasiske BL, Kelepouris E, Klag
MJ, Parfrey P, Pfeffer M, Raij L, Spinosa DJ, Wilson PW;
American Heart Association Councils on Kidney in Cardio-
vascular Disease, High Blood Pressure Research, Clinical Car-
diology, and Epidemiology and Prevention: Kidney disease
as a risk factor for development of cardiovascular disease: A
statement from the American Heart Association Councils on
Kidney in Cardiovascular Disease, High Blood Pressure Re-
search, Clinical Cardiology, and Epidemiology and Preven-
tion. Circulation 108: 2154 2169, 2003
5. Shlipak MG, Sarnak MJ, Katz R, Fried LF, Seliger SL,
Newman AB, Siscovick DS, Stehman-Breen C: Cystatin C
J Am Soc Nephrol 18: 13071315, 2007
and the risk of death and cardiovascular events among
elderly persons. N Engl J Med 352: 2049 2060, 2005
6. Rahman M, Pressel S, Davis BR, Nwachuku C, Wright JT Jr,
Whelton PK, Barzilay J, Batuman V, Eckfeldt JH, Farber MA,
Franklin S, Henriquez M, Kopyt N, Louis GT, Saklayen M,
Stanford C, Walworth C, Ward H, Wiegmann T; ALLHAT
Collaborative Research Group: Cardiovascular outcomes in
high-risk hypertensive patients stratified by baseline glomer-
ular filtration rate. Ann Intern Med 144: 172180, 2006
7. Astor BC, Coresh J, Heiss G, Pettitt D, Sarnak MJ: Kidney
function and anemia as risk factors for coronary heart
disease and mortality: The Atherosclerosis Risk in Com-
munities (ARIC) Study. Am Heart J 151: 492500, 2006
8. Coresh J, Astor BC, Greene T, Eknoyan G, Levey AS: Preva-
lence of chronic kidney disease and decreased kidney func-
tion in the adult US population: Third National Health and
Nutrition Examination Survey. Am J Kidney Dis 41: 112, 2003
9. Gottdiener JS, Arnold AM, Aurigemma GP, Polak JF, Tracy
RP, Kitzman DW, Gardin JM, Rutledge JE, Boineau RC: Pre-
dictors of congestive heart failure in the elderly: The Cardio-
vascular Health Study. J Am Coll Cardiol 35: 1628 1637, 2000
10. Sarnak MJ, Katz R, Stehman-Breen CO, Fried LF, Jenny NS,
Psaty BM, Newman AB, Siscovick D, Shlipak MG; Cardio-
vascular Health Study: Cystatin C concentration as a risk
factor for heart failure in older adults. Ann Intern Med 142:
497505, 2005
11. Bibbins-Domingo K, Chertow GM, Fried LF, Odden MC,
Newman AB, Kritchevsky SB, Harris TB, Satterfield S,
Cummings SR, Shlipak MG: Renal function and heart fail-
ure risk in older black and white individuals: The Health,
Aging, and Body Composition study. Arch Intern Med 166:
1396 1402, 2006
12. Bibbins-Domingo K, Lin F, Vittinghoff E, Barrett-Connor E,
Hulley SB, Grady D, Shlipak MG: Predictors of heart fail-
ure among women with coronary disease. Circulation 110:
1424 1430, 2004
13. Chae CU, Albert CM, Glynn RJ, Guralnik JM, Curhan GC: Mild
renal insufficiency and risk of congestive heart failure in men
and women 70 years of age. Am J Cardiol 92: 682686, 2003
14. Dries DL, Exner DV, Domanski MJ, Greenberg B, Stevenson
LW: The prognostic implications of renal insufficiency in
asymptomatic and symptomatic patients with left ventricular
systolic dysfunction. J Am Coll Cardiol 35: 681689, 2000
15. Smith GL, Lichtman JH, Bracken MB, Shlipak MG, Phillips
CO, DiCapua P, Krumholz HM: Renal impairment and
outcomes in heart failure: Systematic review and meta-
analysis. J Am Coll Cardiol 47: 19871996, 2006
16. Culleton BF, Larson MG, Wilson PW, Evans JC, Parfrey PS,
Levy D: Cardiovascular disease and mortality in a com-
munity-based cohort with mild renal insufficiency. Kidney
Int 56: 2214 2219, 1999
17. Forman DE, Butler J, Wang Y, Abraham WT, OConnor
CM, Gottlieb SS, Loh E, Massie BM, Rich MW, Stevenson
LW, Young JB, Krumholz HM: Incidence, predictors at
admission, and impact of worsening renal function among
patients hospitalized with heart failure. J Am Coll Cardiol
43: 61 67, 2004
18. Gottlieb SS, Abraham W, Butler J, Forman DE, Loh E,
Massie BM, OConnor CM, Rich MW, Stevenson LW,
Young J, Krumholz HM: The prognostic importance of
different definitions of worsening renal function in conges-
tive heart failure. J Card Fail 8: 136 141, 2002
J Am Soc Nephrol 18: 13071315, 2007
19. Khan NA, Ma I, Thompson CR, Humphries K, Salem DN,
Sarnak MJ, Levin A: Kidney function and mortality among
patients with left ventricular systolic dysfunction. J Am Soc
Nephrol 17: 244 253, 2006
20. Stevens LA, Coresh J, Greene T, Levey AS: Assessing kid-
ney function: Measured and estimated glomerular filtra-
tion rate. N Engl J Med 354: 24732483, 2006
21. The Atherosclerosis Risk in Communities (ARIC) Study:
Design and objectives. The ARIC investigators. Am J Epi-
demiol 129: 687702, 1989
22. National Heart, Lung, and Blood Institute Atherosclerosis Risk
in Communities (ARIC) Study. Operations Manual 7: Blood
Collecting and Processing, Bethesda, National Heart, Lung,
and Blood Institute, 1988
23. National Heart, Lung, and Blood Institute Atherosclerosis Risk
in Communities (ARIC) Study. Operations Manual 8: Lipid and
Lipoprotein Determinations, Bethesda, National Heart, Lung,
and Blood Institute, 1987
24. National Heart, Lung, and Blood Institute Atherosclerosis Risk
in Communities (ARIC) Study. Operations Manual 10: Clinical
Chemistry Determinations, Bethesda, National Heart, Lung,
and Blood Institute, 1987
25. National Heart, Lung, and Blood Institute Atherosclerosis Risk
in Communities (ARIC) Study. Operations Manual 6a,b: Ultra-
sound Assessment, Scanning, Reading, Bethesda, National
Heart, Lung, and Blood Institute, 1987, 1989
26. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D:
A more accurate method to estimate glomerular filtration
rate from serum creatinine: A new prediction equation.
Modification of Diet in Renal Disease Study Group. Ann
Intern Med 130: 461 470, 1999
27. Coresh J, Astor BC, McQuillan G, Kusek J, Greene T, Van
Lente F, Levey AS: Calibration and random variation of the
serum creatinine assay as critical elements of using equa-
tions to estimate glomerular filtration rate. Am J Kidney Dis
39: 920 929, 2002
28. Eriksson H, Caidahl K, Larsson B, Ohlson LO, Welin L,
Wilhelmsen L, Svardsudd K: Cardiac and pulmonary
causes of dyspnoea: Validation of a scoring test for clinical-
epidemiological useThe Study of Men Born in 1913. Eur
Heart J 8: 10071014, 1987
29. Wilhelmsen L, Eriksson H, Svardsudd K, Caidahl K: Im-
proving the detection and diagnosis of congestive heart
failure. Eur Heart J 10[Suppl C]: 1318, 1989
30. National Kidney Foundation: K/DOQI clinical practice
guidelines for chronic kidney disease: Evaluation, classifi-
cation, and stratification. Am J Kidney Dis 39: S1S266, 2002
31. Carroll RJ, Ruppert D, Stefanski LA, Crainiceanu C: Mea-
surement Error in Nonlinear Models: A Modern Perspective,
2nd Ed., New York, Chapman & Hall CRC, 2006
32. StataCorp: Stata Statistical Software: Release 9.0 [computer
program], College Station, TX, StataCorp., 2005
33. Epanechnikov VA: Nonparametric estimation of a multi-
variate probability density. Theory of Probability and Its
Applications 14: 153158, 1969
34. Muntner P, Coresh J, Smith JC, Eckfeldt J, Klag MJ: Plasma
lipids and risk of developing renal dysfunction: The Ath-
Access to UpToDate on-line is available for additional clinical information
at http://www.jasn.org/
Kidney Function and Incident Heart Failure 1315
erosclerosis Risk in Communities Study. Kidney Int 58:
293301, 2000
35. Garg AX, Clark WF, Haynes RB, House AA: Moderate
renal insufficiency and the risk of cardiovascular mortality:
Results from the NHANES I. Kidney Int 61: 1486 1494, 2002
36. Coresh J, Byrd-Holt D, Astor BC, Briggs JP, Eggers PW,
Lacher DA, Hostetter TH: Chronic kidney disease aware-
ness, prevalence, and trends among US adults, 1999 to
2000. J Am Soc Nephrol 16: 180 188, 2005
37. Hillege HL, Nitsch D, Pfeffer MA, Swedberg K, McMurray
JJ, Yusuf S, Granger CB, Michelson EL, Ostergren J, Cornel
JH, de Zeeuw D, Pocock S, van Veldhuisen DJ; Candesar-
tan in Heart Failure Assessment of Reduction in Mortality
and Morbidity (CHARM) Investigators: Renal function as a
predictor of outcome in a broad spectrum of patients with
heart failure. Circulation 113: 671 678, 2006
38. He J, Ogden LG, Bazzano LA, Vupputuri S, Loria C, Whel-
ton PK: Risk factors for congestive heart failure in US men
and women: NHANES I epidemiologic follow-up study.
Arch Intern Med 161: 996 1002, 2001
39. Barker WH, Mullooly JP, Getchell W: Changing incidence and
survival for heart failure in a well-defined older population,
19701974 and 19901994. Circulation 113: 799805, 2006
40. Gheorghiade M, Bonow RO: Chronic heart failure in the
United States: A manifestation of coronary artery disease.
Circulation 97: 282289, 1998
41. Coresh J, Astor B, Sarnak MJ: Evidence for increased car-
diovascular disease risk in patients with chronic kidney
disease. Curr Opin Nephrol Hypertens 13: 73 81, 2004
42. McCullough PA: Why is chronic kidney disease the
spoiler for cardiovascular outcomes? J Am Coll Cardiol
41: 725728, 2003
43. Go AS, Yang J, Ackerson LM, Lepper K, Robbins S, Massie
BM, Shlipak MG: Hemoglobin level, chronic kidney dis-
ease, and the risks of death and hospitalization in adults
with chronic heart failure: The Anemia in Chronic Heart
Failure: Outcomes and Resource Utilization (ANCHOR)
Study. Circulation 113: 27132723, 2006
44. Hillege HL, Fidler V, Diercks GF, van Gilst WH, de Zeeuw D,
van Veldhuisen DJ, Gans RO, Janssen WM, Grobbee DE, de
Jong PE; Prevention of Renal and Vascular End Stage Disease
(PREVEND) Study Group: Urinary albumin excretion predicts
cardiovascular and noncardiovascular mortality in general pop-
ulation. Circulation. 106: 17771782, 2002
45. de Zeeuw D, Parving HH, Henning RH: Microalbuminuria
as an early marker for cardiovascular disease. J Am Soc
Nephrol 17: 2100 2105, 2006
46. Sun Y: The renin-angiotensin-aldosterone system and vas-
cular remodeling. Congest Heart Fail 8: 1116, 2002
47. Fonseca C, Oliveira AG, Mota T, Matias F, Morais H, Costa
C, Ceia F; EPICA Investigators: Evaluation of the perfor-
mance and concordance of clinical questionnaires for the
diagnosis of heart failure in primary care. Eur J Heart Fail 6:
813 820, 821 822, 2004
48. Coresh J, Stevens L: Kidney function estimating equations:
Where do we stand? Curr Opin Nephrol Hypertens 15: 276
284, 2006