Blood Pressure and Decline in Kidney Function

Findings from the Systolic Hypertension in the Elderly Program (SHEP)

Abstract

ABSTRACT. The association between BP and decline in kidney function in older persons and the BP component most responsible for kidney disease are unknown. This study investigated the relationship between baseline BP and an incident decline in kidney function among 2181 men and women enrolled in the placebo arm of the Systolic Hypertension in the Elderly Program (SHEP). A decline in kidney function was defined as an increase in serum creatinine equal to or greater than 0.4 mg/dl over 5 yr of follow-up. The incidence and relative risk of a decline in kidney function increased at higher levels of BP for all BP components, independent of age, gender, ethnicity, smoking, diabetes, and history of cardiovascular disease. Systolic BP imparted the highest risk of decline in kidney function. The adjusted relative risk (95% confidence interval) associated with the highest compared with the lowest quartile of BP was 2.44 (1.67 to 3.56) for systolic; 1.29 (0.87 to 1.91) for diastolic; 1.80 (1.21 to 2.66) for pulse; and 2.03 (1.39 to 2.94) for mean arterial pressure. The risk associated with systolic BP remained strong in models containing other BP components, while diastolic, pulse, and mean arterial pressure had no significant association with a decline in kidney function in models containing systolic BP. Therefore, systolic BP is a strong, independent predictor of a decline in kidney function among older persons with isolated systolic hypertension.

Kidney disease is a major public health problem in the United States. The number of persons with treated end-stage renal disease (ESRD) has increased substantially during the past 20 yr (¹). This growth has occurred primarily in individuals older than 65 yr of age (¹). Moreover, ESRD represents a small portion of the kidney disease burden. For every person with ESRD, 29 people have an elevated serum creatinine (²). Kidney dysfunction is especially common among older persons, affecting 10 to 15% of persons older than 70 yr of age (³).

Efforts to reduce the burden of kidney disease require an understanding of kidney disease risk factors. Elevated BP is a strong independent risk factor for developing ESRD (^4,5). However, the association between BP and milder forms of kidney dysfunction is not clear. In addition, the component of BP most responsible for kidney disease is unknown. A growing body of work suggests that, among older persons, systolic BP, and possibly pulse pressure, are more strongly correlated with coronary heart disease, congestive heart failure, and mortality than diastolic BP or mean arterial pressure (^6–13). The relative association of each BP component with kidney dysfunction has not been examined.

To determine the risk of a decline in kidney function associated with BP among older men and women, we prospectively studied 2181 people enrolled in the placebo arm of the Systolic Hypertension in the Elderly Program (SHEP). To determine which BP component carries the greatest risk of an incident decline in kidney function, we compared the effects of four different BP measures: systolic pressure, diastolic pressure, pulse pressure, and mean arterial pressure.

Materials and Methods

SHEP was a randomized, sixteen-center, hypertension treatment trial designed to study the effect of stepped care beginning with the thiazide diuretic chlorthalidone on stroke incidence among older persons with isolated systolic hypertension (¹⁴). Between March 1984 and January 1988, 4736 men and women over the age of 65 yr were randomized. Those undergoing screening were eligible for recruitment when the average of four seated BP measurements, two at each of two baseline visits, was between 160 and 219 mmHg for systolic BP and less than 90 mmHg for diastolic BP. The exclusion criteria included the presence of a major illness, such as a recent myocardial infarction or stroke, cancer, alcoholic liver disease, insulin-treated diabetes mellitus, depression, and kidney failure. Details concerning the screening and recruitment procedures in SHEP have been published elsewhere (¹⁴). For this analysis, persons with a serum creatinine greater than 2.0 mg/dl or with serum creatinine measured after randomization were excluded, leaving 4425 participants. To characterize the natural history of the relationship between BP and decline in kidney function, only the 2181 participants randomized to placebo were included in this analysis.

Measurement

Certified technicians using a random-zero sphygmomanometer assessed each participant’s BP. For the current analysis, the baseline systolic BP and diastolic BP were defined as the average of the four baseline BP. Pulse pressure was defined as the systolic BP minus the diastolic BP. Mean arterial pressure was calculated as ([diastolic BP × 2] + systolic BP) ÷ 3.

Outcome

The principal outcome examined in this analysis was incidence of a decline in kidney function, defined as an increase in follow-up creatinine compared with baseline ≥0.4 mg/dl. Baseline creatinine was measured at the second baseline visit before randomization. Participants who had previously taken antihypertensive medication were observed off their medication for greater than 2 wk before phlebotomy. Serum creatinine was measured yearly thereafter.

A cutpoint of 0.4 mg/dl change in serum creatinine was chosen as the outcome on the basis of an assessment of the year-to-year variability of serum creatinine among SHEP participants. The SD of change in creatinine between yearly visits was approximately 0.2 mg/dl. Therefore, we defined an incident rise in serum creatinine as a change equal to twice this amount, 0.4 mg/dl. To make certain that any observed associations were insensitive to outcome categorization, separate analyses were performed using other outcomes: mean change in serum creatinine, calculated GFR, and increases in serum creatinine of ≥0.5, ≥0.6, and ≥0.7 mg/dl during follow-up (i.e., more stringent definitions of kidney dysfunction).

Statistical Analyses

Baseline characteristics of the study population were examined overall and by systolic BP quartiles using Kruskal-Wallis tests for continuous variables and χ² tests for categorical variables. The unadjusted incidence of a decline in kidney function per 1000 person-yr of observation was calculated by quartile of BP components using Poisson regression (¹⁵). For this calculation, the numerator represents the number of events occurring over the duration of the study; the denominator represents the total number of years of observation. For example, a person who develops kidney dysfunction in year 2 contributes an event to the numerator and 2 yr to the denominator. A person who develops kidney dysfunction in year 5, however, contributes an event to the numerator and 5 yr to the denominator. Therefore, an early event is weighted more heavily than a later event. A second advantage is that people who were lost to follow-up only contribute the time before withdrawal to the denominator. For example, a person lost to follow-up after 3 yr contributes only 3 yr to the denominator.

The relative risk of a decline in kidney function associated with quartiles of BP components was estimated using proportional hazards regression with and without adjustment for age, gender, ethnicity, current smoking status, history of diabetes, and cardiovascular disease (¹⁶). Inclusion of body mass index and HDL did not substantially alter risk estimates, so these models are not presented.

The relative predictive power of BP components was also evaluated in two ways. First, because the component measures of BP have different distributions, we calculated the risk associated with a 1 SD difference at baseline in each component. Second, we constructed a series of six models, including every combination of two BP components. To further examine the collinearity between systolic BP and pulse pressure, we then constructed two models using both components after the method used by Franklin et al. (⁸) In the first model, systolic BP was grouped into quartiles and the effect of pulse pressure was examined within each quartile. In the second model, pulse pressure was grouped into quartiles and the effect of systolic BP was examined within each quartile. Models were compared using likelihood ratio χ².

Antihypertensive medication can affect serum creatinine level. Therefore, we performed an analysis stratified by use of antihypertensive medication before recruitment. In addition, we performed a separate analysis stratified by use of open-label antihypertensive medications during follow-up.

Because kidney disease can increase BP, we also performed an analysis stratified by baseline creatinine to minimize the possibility of reverse causation contributing to the relationship between BP and a subsequent decline in kidney function. Finally, the relative risk of a decline in kidney function was calculated according to quartiles of systolic BP across strata defined by ethnicity and diabetes status at baseline.

Results

The baseline characteristics of the 2181 SHEP participants are shown by quartile of systolic BP in Table 1. Participants with higher systolic BP tended to be older, were more often women, and had higher diastolic BP, pulse pressure, and mean arterial pressure. Their mean heart rate, however, was lower. Other characteristics did not differ substantially among the systolic BP quartiles.

Table 1: . Baseline characteristics, overall, and by systolic BP quartile among 2181 SHEP participants

Over the 5 yr of follow-up, 226 persons experienced an increase in serum creatinine ≥0.4 mg/dl. The unadjusted incidence and adjusted relative risk of a decline in kidney function increased at higher levels of BP for all BP components, but this trend was not statistically significant for diastolic pressure (Table 2 and Figure 1). The risk associated with the highest quartile of BP compared with the lowest was greatest for systolic BP and lowest for diastolic BP (Figure 1). Mean arterial pressure and pulse pressure conveyed intermediate risk. Relative risks were unchanged in unadjusted analyses and in models stratified by gender (data not shown).

Figure 1. : Adjusted relative risk of a decline in kidney function according to quartiles of BP components among 2181 SHEP participants.

Table 2: . Incidence per 1000 person-yr (95% confidence interval) of a decline in kidney function according to quartiles of BP components among 2181 SHEP participants

To compare the relative predictive power of the BP components, we calculated the risk of a decline in kidney function associated with a 1 SD difference in each component at baseline, adjusted for other variables (Table 3). Consistent with the prior analysis, systolic BP conveyed the greatest risk and diastolic BP the lowest. Mean arterial pressure and pulse pressure conveyed intermediate risks.

Because systolic BP, diastolic BP, pulse pressure, and mean arterial pressure are correlated, we constructed joint models consisting of two BP measures. The strong and significant relationship between systolic BP and a decline in kidney function was relatively unchanged when systolic pressure was considered with the other components (relative risks per 1 SD higher systolic BP, 1.35 to 1.42; all P values < 0.001; LR χ² = 46.0, 8 degrees of freedom [df]). In contrast, the risks associated with diastolic, pulse, and mean arterial pressure were reduced and not statistically significant when considered simultaneously with systolic BP (relative risks per 1 SD higher diastolic, pulse, and mean arterial pressure, 1.03, 0.96, and 1.04, respectively; all P values = 0.67; LR χ² = 46.0, 8 df). To further examine the independent effect of systolic BP, we constructed joint models consisting of systolic BP and pulse pressure (Table 4). In the first model, systolic BP was strongly associated with a decline in kidney function, whereas pulse pressure was not. The absence of effect was consistent across strata defined by quartiles of systolic BP. In the second model, pulse pressure was not associated with a decline in kidney function. However, within strata defined by quartiles of pulse pressure, systolic BP was strongly and consistently associated with a decline in kidney function.

Table 3: . Adjusted relative risk (95% confidence interval) of a decline in kidney function associated with a 1 SD higher BP^a

Table 4: . Adjusted relative risk (95% confidence interval) of a decline in kidney function associated with a 1 SD higher BP^a

Among the models containing a single BP component, the model with systolic BP alone was the most predictive, as evidenced by the high likelihood ratio χ² (LR χ² = 45.8, 7 df, Table 3). The models containing two BP components were not more predictive than the model containing systolic BP alone (χ² = 0.2, 1 df, P = 0.90 for difference between the model containing systolic BP alone and the models containing two BP components).

The association between systolic BP and a decline in kidney function remained strong and graded when more stringent definitions of a decline in kidney function were considered (Table 5). The risk estimates associated with systolic BP were similar when analyses were repeated using mean creatinine change and calculated GFR (data not shown). The use of open-label antihypertensive medication before recruitment or during follow-up did not substantially affect the relationship between baseline systolic BP and an incident decline in kidney function (data not shown).

Table 5: . Relative risk (95% confidence interval) of a decline in kidney function according to quartiles of systolic BP among 2181 SHEP participants^a

Table 6 presents the stratified analyses. The risk of a decline in kidney function associated with systolic BP was strong and graded in all groups and similar in persons with a baseline creatinine below 1.0 mg/dl and those above that cutpoint. The association between systolic BP and a decline in kidney function tended to be greater in persons with diabetes than those without and in African Americans compared with European Americans.

Table 6: . Stratified relative risk (95% confidence interval) of a decline in kidney function according to quartiles of systolic BP

Discussion

Our study extends knowledge of the link between BP and kidney disease in several ways. Systolic BP, pulse pressure, and mean arterial BP are predictors of decline in kidney function among older persons with isolated systolic hypertension. Adjusting for other factors minimally affected these relationships. Of these measures, systolic BP has the greatest ability to predict a decline in kidney function. In addition, this association persisted in a group with very low likelihood of baseline kidney dysfunction and in high-risk groups, i.e., diabetic patients and African Americans. The risk associated with systolic BP was somewhat greater in these high-risk groups, but there was no statistically significant interaction.

The strong, graded relationship between BP and ESRD is well described (^4,5). However, the association between BP and milder forms of kidney disease is less clear. Two longitudinal cohort studies have examined the relationship between baseline BP and decline in kidney function in normotensive population samples. In the Bogalusa Heart Study, baseline BP correlated with an increase in creatinine among young normotensive African-American men but not other race-gender groups (¹⁷). In a population sample of middle-aged European Americans, Perneger et al. (¹⁸) demonstrated an association between BP and incident hypercreatinemia in both men and women. Similar studies have not been performed in a population sample of older persons. Among hypertensive persons, several studies have demonstrated a decline in kidney function among persons undergoing BP treatment (^19,20). In two BP treatment trials, baseline BP was associated with decline in kidney function. In the Hypertension Detection and Follow-up Program, higher baseline diastolic BP was associated with decline in kidney function among middle-aged hypertensive men and women (^21,22). In the Multiple Risk Factor Intervention Trial, higher baseline systolic and diastolic BP were associated with decline in kidney function among African-American and European-American men (¹⁹). Our results extend these findings to older men and women with isolated systolic hypertension. This association is seen among persons with relatively low levels of creatinine and tended to be more pronounced in persons at higher risk of renal disease, i.e., those with diabetes and African Americans.

The BP component most responsible for kidney disease is unknown. BP varies over the cardiac cycle and is not well characterized by a single measure. Systolic BP is the maximum achieved BP and diastolic BP, the lowest achieved pressure. Mean arterial pressure represents the average pressure, and pulse pressure the variation in BP over a cardiac cycle. As people age, all components increase until midlife, when diastolic pressure begins to fall while systolic and pulse pressure continue to rise (²³). With regard to organ damage, components may differ in importance and may damage different organs. Among persons older than 60 yr of age, systolic BP and pulse pressure convey the greatest risk of atherosclerosis and coronary heart disease (^{6–8,10–13,24}). The impact of pulse pressure on cardiac events may be due to the increased cardiac load associated with a high systolic BP and the decreased myocardial perfusion associated with a low diastolic BP (²⁵). However, mean arterial pressure is more closely associated with incident stroke than is pulse pressure (^7,25–28). Our results suggest that the risk of a decline in kidney function is more related to systolic BP than pulse pressure, similar to what has been seen for stroke.

This work has several limitations. The sample was limited to older persons with high systolic BP and low diastolic BP. Therefore, we were unable to examine the relationship between BP and kidney dysfunction across the full range of age or BP. In fact, given the absence of a comparison group with normal systolic BP, we likely underestimated the effect of systolic BP on kidney function. In addition, the effect of diastolic BP may be different than that observed given the absence of a group with diastolic BP greater than 90 mmHg. However, these eligibility criteria make this an especially good cohort for examining the effect of pulse pressure. Another limitation is that serum creatinine is an insensitive measure of kidney function, resulting in misclassification of kidney function. Such misclassification would likely create a conservative bias, however, that would lead to underestimating the relative risk. Finally, people at higher risk of events, including decline in kidney function may be more likely to be missing follow-up data (²⁹). This loss to follow-up of high-risk people would likely lead to a conservative bias and subsequent underestimation of risk.

In conclusion, our analysis has shown that systolic BP is a strong independent risk factor for a decline in kidney function among older persons with isolated systolic hypertension. This finding is especially important given that most cases of uncontrolled hypertension in the United States are due to systolic hypertension among older adults (³⁰). Therefore, prevention and treatment of systolic hypertension may help stem the growing epidemic of kidney dysfunction among older persons.

Supported by grants T32 DK07732, T32HL07024, R29DK48362, K24DK02856, and AG01760 from the National Institutes of Health and a grant from Bristol-Myers Squibb.