If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Background: Chronic kidney disease (CKD) is a risk factor for cardiovascular disease (CVD). Concurrently, CVD may promote CKD, resulting in a vicious cycle. We evaluated this hypothesis by exploring whether CKD and CVD have an additive or synergistic effect on future cardiovascular and mortality outcomes. Methods: Patients were pooled from 4 community-based studies: Atherosclerosis Risk in Communities, Framingham Heart, Framingham Offspring, and Cardiovascular Health Study. CKD is defined by an estimated glomerular filtration rate less than 60 mL/min/1.73 m2 (<1 mL/s/1.73 m2). Baseline CVD included myocardial infarction, angina, stroke, transient ischemic attack, claudication, heart failure, and coronary revascularization. The primary outcome is a composite of cardiac events, stroke, and death. Secondary outcomes included individual components. Multivariable analyses using Cox regression examined differences in study outcomes. The interaction of CKD and CVD was tested. Results: The study population included 26,147 individuals. During 10 years, 4% (n = 2,927) of individuals with no CKD or CVD developed the primary outcome, 33% (n = 518) with only CKD, 37% (n = 1,260) with only CVD, and 66% (n = 459) with both. Both CKD (hazard ratio [HR], 1.26; 95% confidence interval [CI], 1.16 to 1.35; P < 0.0001) and CVD (HR, 1.83; 95% CI, 1.72 to 1.95; P < 0.0001) were independent risk factors for the primary outcome. The interaction term CKD × CVD was not statistically significant (HR, 0.98; 95% CI, 0.85 to 1.13; P = 0.74). Similar results were obtained for secondary outcomes. Conclusion: CKD and CVD are both strong independent risk factors for adverse cardiovascular and mortality outcomes in the general population. Although individuals with both risk factors are at extremely high risk, there does not appear to be a synergistic effect of CKD and CVD on outcomes.
Many prior studies evaluating the relationship between CKD and CVD examined either incident CVD or recurrent CVD. Studies focusing on recurrent CVD events consistently showed an independent association for CKD,
Just as CKD is a risk factor for CVD, CVD also may cause CKD and promote progression of kidney disease; mechanisms include heart failure promoting kidney function decline and atherosclerosis promoting progression of renovascular disease.
Subsequently, as kidney function declines, a number of cardiovascular risk factors may be exacerbated, including volume expansion, derangements in calcium-phosphate metabolism, dyslipidemia, and hypertension.
This may further worsen CVD. Several have termed these relationships between the kidney and heart the “cardiorenal syndrome” and suggested that the interplay may promote a vicious cycle in which each promotes progression of the other.
Therefore, we hypothesized that CKD may confer a greater magnitude of risk for cardiovascular events and all-cause mortality in individuals with prevalent CVD. To evaluate this hypothesis, we examined the importance of CKD, CVD, and their interaction on future CVD and mortality outcomes in a large pooled database from the general population.
This post hoc analysis of pooled subject-level data from 4 community-based longitudinal limited-access data sets tests the hypothesis that CKD is a more marked risk factor for adverse outcomes in individuals with a history of prior CVD.
Data from the following studies were used: Atherosclerosis Risk in Communities (ARIC) Study, the Cardiovascular Health Study (CHS), the Framingham Heart Study (FHS), and the Framingham Offspring Study (Offspring). ARIC enrolled 15,792 participants aged 45 to 64 years between 1987 and 1989, whereas CHS enrolled 5,201 subjects 65 years and older between 1989 and 1990. CHS recruited an additional 687 African Americans in 1992 and 1993. FHS began in 1948 with residents of Framingham, MA, and Offspring recruited children and spouses of children of FHS participants in 1971. Serum creatinine levels were assessed at baseline in ARIC and CHS, at the 15th biennial examination (1977 to 1979; n = 2,632) in FHS, and at the second examination (1979 to 1983; n = 3,863) in Offspring; these examinations are considered the baseline period for our analyses. Details of recruitment for all studies are described elsewhere.
Pooling these cohort studies provides a large sample that allows us to examine subgroup associations in subjects with moderate and severe CKD and enhances generalizability by drawing from multiple communities with a broad age range.
Ascertainment of Level of Kidney Function
Calibration of creatinine level for use in glomerular filtration rate (GFR)–estimating equations was accomplished by indirectly calibrating mean individual study creatinine values to mean Third National Health and Nutrition Examination Survey values for a given age, race, and sex. Including an additional calibration factor for proper use in the 4-variable Modification of Diet in Renal Disease Study equation, this resulted in serum creatinine level adjustments of −0.24 mg/dL (−21 μmol/L) in ARIC, −0.11 mg/dL (−10 μmol/L) in CHS, −0.04 mg/dL (−4 μmol/L) in the CHS African-American cohort, −0.22 mg/dL (−20 μmol/L) in FHS, and −0.32 mg/dL (−28 μmol/L) in Offspring.
Subjects are defined as having CKD when baseline GFR is less than 60 mL/min/1.73 m2 (<1 mL/s/1.73 m2). Subjects with estimated GFR less than 15 mL/min/1.73 m2 (<0.25 mL/s/1.73 m2) were excluded from the study to avoid confounding by imminent kidney failure and treatment by dialysis or kidney transplantation.
Baseline characteristics included demographics (age, sex, race, education), lifestyle (smoking, alcohol), past medical history (baseline CVD, diabetes mellitus, hypertension), medication use (antihypertensive and diabetes medications), physical examination findings (body mass index, systolic and diastolic blood pressure), left ventricular hypertrophy (LVH) by electrocardiogram, and laboratory variables (total cholesterol, high-density lipoprotein [HDL] cholesterol, GFR estimated from serum creatinine, serum glucose, hematocrit).
Race is defined as white or African American, with Framingham cohorts assumed to be entirely white.
Level of education was dichotomized by high school graduation status. Cigarette smoking and alcohol use were dichotomized as current users and nonusers. Diabetes is defined by the use of insulin or oral hypoglycemic medications or fasting glucose level of 126 mg/dL or greater (≥7.0 mmol/L). Hypertension is defined by systolic blood pressure of 140 mm Hg or greater, diastolic blood pressure of 90 mm Hg or greater, or antihypertensive medication use at baseline. Body mass index was calculated by using the formula: weight (kg)/height2 (m), and LVH is defined by electrocardiographic criteria.
Baseline CVD included a history of myocardial infarction (MI), angina, stroke, transient ischemic attack, and intermittent claudication as defined by consensus committees for the respective studies. In addition, baseline CVD included a history of congestive heart failure in CHS, FHS, and Offspring (not ascertained in ARIC) and a history of angioplasty or coronary bypass procedures in ARIC and CHS (not available in the Framingham cohorts). Methods used for collection of baseline data and CVD events by each of these studies are described elsewhere.
From the initial pooled cohort of 28,175 individuals, we excluded data for 575 subjects who had missing information on age, race, sex, or creatinine level or were of nonwhite/non–African American race; 36 subjects with GFR less than 15 mL/min/1.73 m2 (<0.25 mL/s/1.73 m2); 93 subjects who did not provide permission to release data; 3 subjects without follow-up data; and 556 subjects with missing data for the presence of baseline CVD. Of the remaining 26,912 individuals, 967 (3.6%) were missing other data points, such as systolic blood pressure or total cholesterol level, yielding a final sample of 25,945 (Fig 1).
Follow-Up Time and Outcomes
We censored all subjects at 10 years for follow-up times to be similar in the pooled cohort. The primary study outcome is a composite end point of MI and fatal coronary heart disease (CHD), stroke, and all-cause mortality. MI is defined by the occurrence of both clinically recognized and silent infarctions. Silent MI was ascertained either at scheduled follow-up visits or by interim electrocardiogram performed for other clinical indications. Secondary outcomes included the individual analysis of MI/fatal CHD, stroke, and all-cause mortality.
Analysis and Statistical Methods
Chi-square tests and analysis of variance were used to compare baseline data between subjects with and without CKD. Kaplan-Meier survival analysis was used to estimate nonparametric survival distribution among study participants by CKD status.
Cox proportional hazards regression was used to examine differences in principal study outcomes between the respective comparison groups while adjusting for covariates. This initial additive model has no interaction term and assumes that for all values of the exposure variable (CKD), the effect on the outcome is constant regardless of the value of a second exposure variable (CVD). We then tested the interaction term CKD × CVD. In a model with a significant interaction term, the effect of 1 exposure (eg, CKD) on the outcome varies according to the value of a second exposure (eg, CVD); this interaction model is referred to as “nonadditive” or “multiplicative.” To evaluate whether specific risk factors would impact on the presence or absence of an interaction between CVD and CKD, we first built a parsimonious model with terms only for CVD, CKD, and their interaction. We then fitted this model by using forward selection and evaluated the significance of the interaction term after the addition of each variable.
Candidate covariates included traditional CVD risk factors described in the Framingham population; namely, age, sex, smoking status, total cholesterol level, HDL cholesterol level, diabetes mellitus, and history of hypertension, as well as race, LVH, alcohol use, body mass index, anemia, and education status. Variables were retained for P less than 0.05 in the multivariable model. A term for original study was included in all models. Nonlinear relationships between covariates and outcomes were tested by including squared terms in models and retaining them when significant (P < 0.05). A priori, we tested for an additional interaction term between race and CKD because we had previously noted this relationship.
Evaluation of the secondary outcomes of MI/fatal CHD, as well as stroke and all-cause mortality, was performed by using covariates that were significant in the fully adjusted model for the composite outcome.
We performed a sensitivity analysis after imputing data for 967 individuals missing single data points. Single imputation of these missing variables was based on age, sex, race, and prior CVD-stratified means. Furthermore, because heart failure may influence adverse outcomes independent of atherosclerosis, an additional sensitivity analysis was performed after reclassifying subjects as not having CVD if their only CVD history included heart failure. To further explore whether relationships among CKD, CVD, and outcomes were modified by other risk factors, we tested interactions between both CKD and other covariates and also CVD and other covariates. When a significant interaction was noted (P < 0.05 for the interaction term) in composite outcome models, the population was dichotomized based on the interaction (eg, subgroups of individuals with and without the particular risk factor that interacted with CKD or CVD) and the interaction term CKD × CVD was tested in both subgroups.
Data were analyzed using SAS, version 9.1 (SAS Institute, Cary, NC).
Baseline characteristics of the 26,912 subjects eligible for the study, stratified by prior CVD and CKD status, are listed in Table 1. Notably, 17.7% of individuals with a baseline history of CVD versus 7.4% without CVD (P < 0.001) had CKD at entry, and 31.3% of individuals with baseline CKD versus 14.4% of individuals without CKD had a history of CVD (P < 0.001).
Table 1Baseline Characteristics of the Cohort Stratified by CKD and CVD Status
CKD and CVD (n = 759)
CKD and No CVD (n = 1,664)
No CKD With CVD (n = 3,519)
No CKD or CVD (n = 20,970)
72.4 ± 9.0
68.1 ± 11.4
62.2 ± 10.4
56.2 ± 11.2
High school graduate (%)
Medical history (%)
Medication use (%)
Body mass index (kg/m2)
27.1 ± 4.7
26.9 ± 4.5
27.8 ± 5.2
27.3 ± 4.9
Systolic blood pressure (mm Hg)
137.7 ± 24.5
134.7 ± 22.1
128.7 ± 21.9
124.6 ± 19.6
Diastolic blood pressure (mm Hg)
71.6 ± 12.5
73.6 ± 11.5
72.3 ± 11.7
74.3 ± 11.1
Serum creatinine (mg/dL)
1.4 ± 0.4
1.3 ± 0.3
0.9 ± 0.2
0.8 ± 0.2
Estimated GFR (mL/min/1.73 m2)
49.6 ± 9.4
51.7 ± 7.6
90.2 ± 20.7
93.9 ± 21.4
Total cholesterol (mg/dL)
217.5 ± 46.0
223.8 ± 45.6
217.7 ± 43.5
217.7 ± 42.5
HDL cholesterol (mg/dL)
46.6 ± 14.8
52.1 ± 16.0
48.1 ± 15.8
52.3 ± 16.3
42.1 ± 4.9
42.2 ± 4.5
42.7 ± 4.3
42.4 ± 4.1
Original cohort (%)
NOTE. Values expressed as percent or mean ± SD. The total number of patients exceeds those in primary analyses because of missing data. P < 0.001 for comparisons among groups for all variables. To convert serum creatinine in mg/dL to μmol/L, multiply by 88.4; GFR in mL/min/1.73 m2 to mL/s/1.73 m2, multiply by 0.01667; HDL and total cholesterol in mg/dL to mmol/L, multiply by 0.02586.
Event rates are listed in Table 2. Individuals with both prior CVD and CKD had the highest event rates (116 events and 54.8 events/1,000 person-years for the composite and cardiac outcomes, respectively; Fig 2). In a parsimonious model including only terms for CKD, CVD, and their interaction, the interaction term was borderline significant and carried a negative coefficient, indicating a less-than-additive effect of CKD and CVD. The interaction term lost statistical significance when the first variable, age, was added to the model and remained nonsignificant when additional variables were added (data not shown).
Table 2Adverse Event Rates Stratified by Baseline CVD and CKD Status
CKD and CVD
CKD and No CVD
No CKD With CVD
No CKD or CVD
Rate/1,000 person y
Rate/1,000 person y
Rate/1,000 person y
NOTE. Values expressed as number (percent) unless noted otherwise. Number of events includes all individuals and does not take into account censoring for missing data.
In full multivariable models, both baseline CVD and CKD were independent risk factors for all study outcomes (Table 3; Fig 3). Although individuals with both CVD and CKD were at greatest risk, the interaction term CVD × CKD did not approach statistical significance for any study outcome. The hazard ratio (HR) for the interaction term in a model using the composite outcome was 0.98 (95% confidence interval [CI], 0.85 to 1.13; P = 0.74), HR for cardiac outcome was 1.05 (95% CI, 0.84 to 1.31; P = 0.68), HR for stroke outcome was 0.97 (95% CI, 0.73 to 1.30; P = 0.85), and HR for mortality outcome was 0.95 (95% CI, 0.81 to 1.12; P = 0.53).
Table 3HRs for Study Outcomes in Multivariable Analyses Before the Addition of Interaction Terms to the Models
An additional term for HDL squared was included in all models to account for nonlinearity of this variable.
NOTE. Values expressed as HR (95% CI). HR for age reflects risk associated with a 10-year increase; systolic blood pressure, a 10-mm Hg increase; HDL cholesterol level, a 10-mg/dL (0.26-mmol/L) increase; body mass index, a 2-unit increase; and total cholesterol level, a 20-mg/dL (0.52-mmol/L) increase. Terms for study of origin also were included in multivariable models.
The composite outcome includes cardiac events (MI and fatal CHD), fatal and nonfatal stroke, and all-cause mortality.
† An additional term for HDL squared was included in all models to account for nonlinearity of this variable.
Results based on imputed models did not significantly differ from nonimputed models. Additionally, the interaction between CVD and CKD remained nonsignificant in models that reclassified subjects as not having CVD if their only CVD history included heart failure. Because we previously found a significant interaction between African-American race and CKD, we tested this interaction term in multivariable models.
This did not result in a significant change in P for the interaction between CVD and CKD.
Finally, in fully adjusted main-effects models, anemia was associated with a statistically significant increase in composite and mortality events, but not cardiac or stroke events (Table 3). The interaction term between CKD and anemia was statistically significant for all outcomes (HR for composite outcome, 1.40; 95% CI, 1.13 to 1.72; P = 0.001; HR for cardiac outcome, 1.11; 95% CI, 1.00 to 1.98; P = 0.05; HR for stroke outcome, 1.70; 95% CI, 1.08 to 2.65; P = 0.02; and HR for mortality outcome, 1.36; 95% CI, 1.08 to 1.70; P = 0.01), consistent with increased risk for adverse outcomes in individuals with both anemia and CKD. In subgroup analyses of individuals with and without anemia, the interaction term CKD × CVD remained nonsignificant for all outcomes. Other interaction terms did not achieve statistical significance in main-effects models.
In the current study, we confirm that individuals with CKD and prior CVD are at high risk for adverse outcomes; however, we did not appreciate a synergistic effect of these risk factors on adverse events. Results of this study add to the current state of knowledge by helping define the risk for adverse cardiac and mortality events associated with CKD. Although the presence of either CKD or CVD independently predicted all study outcomes, the magnitude of risk for cardiac events, stroke, and mortality in individuals with both CKD and CVD was particularly marked.
We previously noted synergistic relationships in individuals with diabetes between anemia and CKD and in individuals with CKD between anemia and LVH. Findings in these studies suggested that CKD and CVD risk factors may have complex overlapping interrelationships.
Kidney disease and CVD have many shared risk factors, most notably diabetes and hypertension. Additionally, upregulation of the renin-angiotensin-aldosterone system and the sympathetic nervous system in patients with CVD also may contribute to systemic vascular disease, including kidney disease.
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 Research, Clinical Cardiology, and Epidemiology and Prevention.
we suspected there would be a synergistic relationship such that the presence of CKD enhanced the risk for adverse outcomes attributable to CVD.
However, in this study, the interrelationship between CKD and prior CVD was only additive and not synergistic. This finding is surprising because: (1) more consistent results had been seen implicating CKD as an independent risk factor for recurrent CVD events
Importantly, most recurrent CVD studies tracked patients immediately after MI, potentially resulting in a greater recurrent CVD event rate than in a purely prevalent CVD population, thereby influencing the statistical relationship between CVD and CKD. This analysis shows the difficulties of interpreting HRs based on different models, let alone different studies. Interestingly, in our previous work analyzing incident and recurrent CVD in 2 distinct analyses, we found that the hazard associated with CKD was 1.19 for the composite outcome and 1.09 for the cardiac outcome for incident events,
These were based on models with identical covariates; only coefficients for these covariates differed, with certain covariates assuming different importance depending on prior history of CVD. In the current study, because comorbid conditions are identically adjusted, we are able to directly assess the importance of prevalent CVD in relationship to CKD.
The absence of a statistically significant interaction between CKD and CVD could imply that more of the suspected circular risk is accounted for by risk factors included in our models (blood pressure, LVH) and less by risk factors not included in our analysis (inflammation, calcium-phosphorus metabolism, malnutrition). However, even when we investigated this hypothesis by adding individual terms to a parsimonious model that only included terms for CKD, CVD and their interaction, the interaction term lost statistical significance when the first variable, age, was added to the model. In addition, when we added other significant interactions to the model, namely, the interaction between CKD and anemia, the interaction between CKD and CVD remained nonsignificant. We therefore hypothesize that much of the adverse risk associated with CKD reflects a greater burden and duration of CVD, thereby incorporating some of the adverse effects of CVD into the importance of CKD; in essence, the presence of CKD is identifying individuals with worse CVD. Alternatively, the presence of CKD may impact on progression of subclinical CVD. By the time clinically apparent CVD has resulted, the additional risk previously attributable to CKD is incorporated into the statistical term for CVD, making any interaction statistically insignificant. Accordingly, there may be many individuals in our study with subclinical CVD that has not yet manifest for whom the interaction may be important; this “misclassification” could bias our results to the null. Finally, it is possible that only a small portion of the CKD population has enhancement of CVD risk caused by CKD and that our study is underpowered to identify individuals in whom CKD promotes additional CVD risk. However, it is notable that all point estimates for the HR associated with the interaction term are close to 1.0.
A major strength of our study is that it assesses the importance of kidney function in a generalizable adult US population, with rigid ascertainment of cardiovascular events. Additionally, we use GFR-estimating equations, rather than serum creatinine values alone, with calibration to the Cleveland Clinic standard for use of the Modification of Diet in Renal Disease equation. Furthermore, with nearly 27,000 subjects, many of whom had prevalent CVD, more than 4,400 composite and 2,350 cardiac events, and a relatively parsimonious model, we are confident that we would detect a clinically significant interaction between CVD and CKD.
Our study also has several limitations. First, serum creatinine level used to define CKD was assessed at only a single visit. However, given that patients were free of acute illnesses at the time of enrollment, creatinine level likely is a reflection of stable kidney function. Second, there were borderline differences in outcomes based on the study of origin, with the Offspring population at significantly lower risk than the others. This factor most likely is caused by inherent differences between the younger Offspring study and the 3 other cohorts. However, we forced terms for study of origin into all our analyses to account for these and any other differences associated with study. Third, it is possible that there is bias from unmeasured variables. For example, if individuals with CKD are more likely to be treated for comorbidities, an interaction between CKD and CVD may be obscured. While this is possible, we would assert that individuals with CKD generally comprised an undertreated population.
Fourth, we are including individuals with and without CVD in the same model, thereby making the statistical assumption that risk factors have the same importance in all individuals. Finally, all except 70 individuals with CKD in this cohort had stage 3 disease with a mean GFR of 51 mL/min/1.73 m2 (0.85 mL/s/1.73 m2), and it is possible that enhancement of risk does not occur until more severe kidney disease is present. However, based on Third National Health and Nutrition Examination Survey data, more than 90% of individuals with diminished kidney function in the United States were defined as stage 3,
making findings in this report relevant to the care of the vast majority of individuals with CKD.
Overall, our study shows that both CKD and prior CVD are independent risk factors for adverse CVD and mortality outcomes in the general population, and individuals with both CKD and CVD are at very high risk, but there is no synergistic effect of these risk factors on outcomes. This suggests that much of the shared risk associated with CVD and CKD may be accounted for by characteristics intrinsic to CVD and CKD and is less dependent on the interaction between these 2 disease states.
ARIC, CHS, Framingham Heart, and Offspring studies are conducted and supported by the National Heart, Lung and Blood Institute in collaboration with the individual study investigators. This report was not prepared in collaboration with the study investigators and does not necessarily reflect the opinions or views of the study investigators or the National Heart, Lung and Blood Institute.
Prevalence of chronic kidney disease and decreased kidney function in the adult US population.
Originally published online as doi:10.1053/j.ajkd.2006.05.021 on July 13, 2006.
Support: Grant support from R21 DK068310, K23 DK71636, T32 DK007777, and Amgen Inc, Thousand Oaks, CA. Study sponsors were not involved in data analysis or interpretation of findings. Potential conflicts of interest: None.