Clinical and Subclinical Cardiovascular Disease and Kidney Function Decline in the Elderly

Study Design

The Cardiovascular Heath Study (CHS) is a community-based longitudinal study of adults who were 65 years of age or older at baseline.¹⁵ A main cohort of 5,201 participants was recruited between 1989 and 1990 from four U.S. communities (Sacramento County, California; Forsyth County, North Carolina; Washington County, Maryland; and Allegheny County, Pennsylvania). An additional 687 African-American persons were recruited in 1992 and 1993; the baseline visit for these additional participants was the year 3 visit for the original cohort. Eligible participants were sampled from Medicare eligibility lists in each area. Subjects were excluded if they were institutionalized, required a proxy to give consent, were planning to move out of the area within 3 years after recruitment, required a wheelchair in the home, were receiving hospice care, or were undergoing radiation or chemotherapy for cancer.

Institutional review board approval for the data collection procedures of the Cardiovascular Health Study was obtained at each of the 4 clinical sites and at the Data Coordinating Center (University of Washington). In addition, we obtained institutional review board approval for this study from Tufts-New England Medical Center and University of California, San Francisco.

Measurements of kidney function

Participants were included in this study if they had at least two measurements of kidney function using both creatinine and cystatin C (N = 4,380); 2,396 participants had 3 measures of kidney function; 1,984 participants had 2 measures, and 1,452 had only 1 measure of kidney function. Serum creatinine was assayed by a colorimetric method (Ektachem 700, Eastman Kodak) from frozen sera done near the time of collection. We estimated the GFR with the use of the four-variable version of the Modification of Diet in Renal Disease (MDRD) equation [GFR=186.3×(serum creatinine^−1.154)×(age^−0.203)×1.212 (if black)×0.742 (if female)]; age was updated for each creatinine measure.¹⁶ Creatinine concentrations were indirectly calibrated to the reference laboratory at the Cleveland Clinic, as previously described.¹⁷ Cystatin C was measured from frozen samples that were collected at the 1989–1990 (baseline), 1992–1993 (3rd year of follow-up) and 1996–1997 (7th year of follow-up) visits using a BNII nephelometer (Dade Behring, Inc., Deerfield, Illinois)¹⁸. The year 3 measurements were conducted in 2003, and the baseline and year 7 measures in 2006. All measurements used the same equipment in the same laboratory. We estimated GFR from cystatin C using an equation developed from the pooling of two cohorts with GFR measured from iothalamate (eGFR=76.7×cysC⁻¹.¹⁸).

Change in kidney function was determined by linear regression as the slope per year for each individual. Decline greater than 3 ml/min/1.73m²/year was defined as “rapid kidney decline”, based on its use in other cohorts and its association with adverse outcomes within CHS.^{14, 19–22}

Clinical and Subclinical Cardiovascular Disease

The primary predictor variables for this study were measures of clinical and subclinical cardiovascular disease. Clinical cardiovascular disease measures were categorized as coronary heart disease (myocardial infarction, angina, or revascularization procedure), stroke, or heart failure. A participant was considered to have the specific category of clinical cardiovascular disease if the disease was either present at the baseline visit of CHS or developed prior to the individuals’ final measurement of kidney function. This decision was made because persons with incident CVD events during follow-up were likely to have extensive atherosclerosis at baseline; thus, they were more similar to those participants with prevalent CVD at baseline than those without CVD at baseline or during follow-up. The specific methods for adjudicating prevalent and incident cardiovascular disease have been well–described.^{23, 24} For the diagnosis of prevalent heart failure, self reports were confirmed by components of the physical exam, or if necessary, by a validation protocol that included surveys of treating physicians or review of medical records ²³. For the diagnosis of incident heart failure, a physician’s diagnosis of heart failure was followed by a review of the participant’s medical records.

Analyses of subclinical cardiovascular disease measures as predictors of kidney function decline were limited to participants without any of the above prevalent cardiovascular disease conditions at baseline. For the subclinical measures of carotid intima-medial thickness (IMT), left ventricular (LV) mass, and ankle-arm index (AAI) we categorized into quintiles of each measure and compared each ascending quintile with the lowest. In addition, because no associations were observed across the linear 4 quintiles, we dichotomized at the high quintile for carotid IMT and LV mass; AAI was dichotomized at <0.9. The high quintile of LV mass was gender specific. Systolic dysfunction was determined semi-quantitatively by echocardiography as normal, mildly impaired, or moderately impaired in CHS, corresponding to EF estimates of >55%, 45–55%, and <45%. Less than 100 persons had EF<45%, so we dichotomized EF as <55% for this paper.

The method for AAI was as follows: after five minutes of rest in a supine position, standard blood pressure cuffs were applied to the right arm and both ankles of the participant by a trained technician. Each of the three arteries was identified using palpation or a Doppler stethoscope (8 MHz; Huntleigh Technology, Inc., PLC, Luton, UK), and the blood pressure was measured in rapid succession for each artery using a standard mercury manometer. The ratio of ankle to arm blood pressure was determined separately for each leg; participants were classified as having a low ankle-arm index (AAI) if the ratio was <0.9 in either leg. The AAI was determined in all but 143 of CHS participants. The carotid arteries of participants were evaluated by high-resolution B-mode ultrasonography to determine the maximal intima-media thickness (IMT) of the common carotid and internal carotid arteries²⁵. The maximal IMT was defined for each artery as the mean of the maximal IMT measures of the near and far walls of both the left and right sides. This measure was available in all but 26 CHS participants. Echocardiography was performed using a standardized protocol that has been described in detail elsewhere; baseline echo was available in the original cohort, but not the 687 African-Americans recruited subsequently. For each subject, the baseline echo study was recorded using a Toshiba SSH-160A cardiac ultrasound machine (Toshiba America Medical Systems, Inc., Tustin, CA, USA) onto a super-VHS tape and sent to the Echocardiography Reading Center for review. Left ventricular mass was calculated with M-mode measurements using the Devereux formula and reported in grams.

Covariates for multivariate analyses

Potential covariates were chosen based on their likelihood of confounding the association between cardiovascular measures and kidney function decline. These included demographic characteristics (age, sex, and race); cardiovascular risk factors (hypertension, and diabetes mellitus (defined by use of an oral hypoglycemic agent or insulin or a fasting glucose level ≥ 126 mg/dL), smoking status (ever vs. never), and fasting levels of low-density lipoprotein (LDL) cholesterol and high-density lipoprotein (HDL) cholesterol); inflammatory biomarkers (CRP, IL-6, fibrinogen and hemoglobin concentrations). Models were similar when systolic or diastolic blood pressure measures were used in place of hypertension.

Statistical Analysis

This analysis began with comparisons of participants with and without rapid kidney function decline. Continuous variables were compared using the t-test, and categorical variable by the chi-squared test. Each clinical cardiovascular disease variable was then evaluated separately for its association with kidney function decline, in both linear (continuous outcome) and logistic (dichotomous outcome) regressions. Staged multivariate regression models were implemented as follows: unadjusted, demographic adjusted and multivariate adjusted. The final model included the demographic characteristics and the candidate predictor variables listed above; potential confounders were retained if they had at least a 5% effect on the coefficient for the cardiovascular disease variable after entry into the model.

Baseline kidney function was not included as a covariate in the primary analyses for several reasons^{26, 27}. Firstly, measurement error in baseline kidney function produces biased estimates of the true association between the predictor variable of interest and change in kidney function. Correlations in measurement error over time do not bias the change estimate. Secondly, adjustment for baseline kidney function makes the assumption that regression to the mean within individuals implies regression to the mean between individuals. Finally, by not adjusting for baseline kidney function, we do not duplicate the initial measurement error again within the residual error used in the regression. In sensitivity analyses, we verified that our results would not have been changed had we adjusted for baseline kidney function.

S-Plus (release 8.0, Insightful Inc, Seattle, WA) and SPSS statistical software (release 15.0.1.1, SPSS Inc, Chicago, IL) were used for the analyses. A p-value of <0.05 was considered statistically significant.

Clinical Cardiovascular Disease

In linear regression models with eGFR_creat as the outcome, participants with clinical coronary heart disease experienced a significantly greater loss of kidney function by 0.30 ml/min/1.73m² each year (p=0.004). [Figure 1]. Clinical stroke and heart failure were also associated with faster rates of decline in kidney function 0.4 ml/min/1.73m² per year (p= 0.005) and 0.8 ml/min/1.73m² (p<0.001), respectively. Findings were very similar with eGFR_cysC as the outcome − 0.30 ml/min/1.73m² (p= 0.008) for CHD, 0.30 ml/min/1.73m² (p= 0.05) for stroke and 0.8 ml/min/1.73m² (p< 0.001) for heart failure. For comparison, clinical hypertension (0.30; p< 0.001) and clinical diabetes (0.30; p< 0.001) had similar adjusted associations with kidney function decline as atherosclerotic cardiovascular disease.

Using the dichotomized endpoint of rapid kidney decline, prevalent coronary heart disease was not significantly associated with rapid decline by eGFR_creat in multivariate analysis, but prevalent stroke did have a significant association (Table 2). Conversely, for the outcome of rapid decline based on eGFR_cysC, coronary heart disease was significantly associated with rapid decline, whereas stroke was not. Clinical heart failure was associated with nearly a 2-fold odds of rapid kidney decline after multivariate adjustment using either eGFR_creat or eGFR_cysC.

Subclinical Cardiovascular Disease

After excluding participants with prevalent coronary heart disease, stroke and heart failure, 3,668 participants had repeated measures of kidney function. The baseline eGFR_cysC was 81± 27 ml/min/1.73m² and baseline eGFR_creat was 81± 23 ml/min/1.73m²; mean annual loss of eGFR_creat was −0.3 ± 3.6 ml/min/1.73m² and annual loss of eGFR_cysC was −1.7 ± 2.5 ml/min/1.73m².

In unadjusted linear regression models, AAI< 0.9 was associated with a 0.5 ml/min/1.73m² faster decline in both eGFR_creat [p=0.004] and eGFR_cysC [p=0.02]. However, neither association reached statistical significance after multivariate adjustment (Figure 2). After multivariate analysis, participants in the highest quintile of common carotid IMT had greater loss of eGFR_creat [0.3; p=0.004] and eGFR_cysC [0.2; p=0.03]; the highest quintile of internal carotid IMT was associated with larger declines in eGFR_creat [0.5; p<0.001] than eGFR_cysC in adjusted analyses [0.20; p=0.04] (Figure 2). In contrast, the highest quintile of LV mass was not associated with decline in kidney function modeled by eGFR_cysC or eGFR_creat in either unadjusted or adjusted analyses. LV systolic dysfunction (EF<55%) was not associated with declines in eGFR_creat (0.26; p=0.38) or eGFR_cysC (0.28; p=0.18) in adjusted analyses.

For the outcome of rapid decline by eGFRcreat, low AAI and high common and internal carotid IMT had significant associations after multivariate adjustment [Table 3]. When all subclinical measures were included in the multivariate model, low AAI (1.46; 1.08–1.98) and high internal carotid IMT (1.32; 1.04–1.68) were significantly associated with rapid decline. For rapid decline based on eGFR_cysC, only low AAI had a significant association after multivariate analysis (Table 3); the low quintile of AAI had a somewhat stronger association when compared with the highest quintile (1.67; 1.25 – 2.24). Elevated LV mass and LV systolic dysfunction had no association with rapid kidney decline using either measure of kidney function (Table 3).

Conclusion

In summary, clinical cardiovascular disease, particularly heart failure, was independently associated with declining kidney function in the elderly. Subclinical cardiovascular disease measures had moderate, but significant associations with declining kidney function. These results suggest that atherosclerosis and heart failure have important effects on the progression of kidney disease in older adults. Future study should evaluate whether interventions aimed at cardiovascular prevention can successfully delay the progression of kidney disease in the elderly.