Physical Activity and Rapid Decline in Kidney Function Among Older Adults FREE

Chronic kidney disease is one of the fastest growing health conditions in older people. Approximately 30% of individuals older than 70 years have chronic kidney disease, defined by an estimated glomerular filtration rate (eGFR) lower than 60 mL/min/1.73 m², which is less than half of normal for a young, healthy adult.¹ Kidney dysfunction is a major risk factor for cardiovascular events and mortality across multiple populations. In older individuals, a lower eGFR, as estimated by serumcystatin C levels, is linearly related to the risk of cardiovascular events, premature death, and a decline in functional status.^2- 4 The age-associated decline in kidney function is highly variable^5- 6; identifying modifiable factors that could preserve kidney function later in life could have a substantial public health impact.

In the general population, greater physical activity is associated with lower risks of coronary heart disease, stroke, and cardiovascular death.^7- 9 Physical activity confers diverse metabolic benefits that may moderate the long-term risks of glomerulosclerosis and progressive kidney dysfunction. Exercise stimulates glucose uptake by skeletal muscle, thereby reducing insulin secretion and promoting lipolysis.¹⁰ Exercise also contributes to a fall in systemic blood pressure and a reduction in body mass.^11- 14 In contrast, a sedentary lifestyle predisposes to adiposity, which promotes inflammation, insulin resistance, and hypertension.^15- 16 These adverse processes may directly injure the kidney.^17- 18

We hypothesized that greater physical activity would be associated with a lower incidence of kidney function decline among older adults. To test this hypothesis we evaluated physical activity level among 4011 participants in the Cardiovascular Health Study (CHS),^2- 4 a community-based study of ambulatory adults 65 years or older. Because exercise may influence serum creatinine levels via changes in muscle mass, we estimated longitudinal changes in kidney function using serial measurements of cystatin C levels, which is less dependent on muscle mass.¹⁹

The CHS^2- 4 is a community-based prospective cohort study of cardiovascular disease among people 65 years or older. The design and recruitment criteria have been previously described.^20- 21 Briefly, 5201 men and women 65 years or older who were ambulatory and not institutionalized were randomly selected and enrolled from Medicare eligibility lists in 4 US communities in 1989 through 1990; an additional 687 African American participants were recruited and enrolled in 1992 through 1993. Subjects were excluded from CHS if they required a wheelchair in the home, were institutionalized, required a proxy to give consent, were receiving hospice care, were planning to move out of the area within 3 years, or were undergoing irradiation or chemotherapy for cancer. Each center's institutional review committee approved the study, and all participants gave informed consent. The baseline evaluation included a standardized physical examination, diagnostic testing, laboratory evaluation, and questionnaires regarding health status, medical history, and cardiovascular and lifestyle risk factors.

Beginning with the 5888 CHS participants from the baseline examination (1989-1990 for the main CHS cohort and 1992-1993 for the African American cohort), we excluded 1452 participants who did not complete at least 2 cystatin C level measurements that were necessary to calculate a slope. To focus on adults who had the capacity to exercise, we further excluded 280 participants who were unable to complete 1 or more basic household chores (walking around at home, getting out of bed, dressing, bathing, and/or using the toilet) and 9 participants who were unable to complete the timed 15-foot walk. Finally, we excluded 63 participants for missing data on key physical activity variables or diabetes status and 73 participants with a baseline eGFR higher than 120 mL/min/1.73 m². We chose to exclude participants with a very high eGFR because serologic markers poorly correlate with kidney function within this eGFR range²² and because we detected considerable regression to the mean for these outlying eGFR values (49 of 73 showed rapid progression of kidney disease). Following these exclusions, 4011 participants were included in the analyses.

Questionnaires were administered at baseline to estimate each participant's self-reported walking pace, exercise intensity, number of blocks walked weekly, and leisure-time activity levels. During the baseline examination, participants were asked whether they had engaged in any of the following 15 leisure-time activities in the prior 2 weeks: swimming, hiking, aerobics, tennis, jogging, racquetball, walking, gardening, mowing, raking, golfing, bicycling, dancing, calisthenics, and riding an exercise cycle. The intensity of each activity has been established and validated by the Minnesota Heart Survey.²³ Participant responses regarding each type of activity, frequency, and duration were used to calculate leisure-time physical activity, expressed in kilocalories per week.

We summed leisure-time activity (ordinal score of 1-5 for quintiles) and pace of walking (ordinal score of 1-3 for pace <2 mph, 2-3 mph, or >3 mph) into a single physical activity score variable according to a previous CHS analysis of lifestyle factors and diabetes²⁴ to represent the joint association of these variables with kidney function decline. We tested the interaction of leisure-time activity and pace of walking and found that the interactions were additive, not multiplicative (likelihood ratio test, P = .42). Kilocalories expended per week and walking pace were weakly correlated (r = 0.16). We also examined leisure-time activity and walking pace individually and further evaluated exercise intensity and the number of blocks walked per week.

The study outcome was a rapid decline in kidney function, defined previously in CHS by the loss of more than 3.0 mL/min/1.73 m² per year in eGFR.^25- 26 An annual eGFR loss of 3.0 mL/min/1.73 m² corresponds to the 25% of the CHS cohort with the most rapid decline in eGFR and represents a magnitude of change that is more than 3 times greater than the rate previously described in studies of aging. A change of this magnitude is beyond the range of noise in measurement.⁶

We used serum cystatin C levels to estimate GFR because serum creatinine levels, the traditional serologic marker of kidney function, depend on muscle mass, which declines with older age and may be influenced by exercise.¹⁹ Cystatin C levels were measured from frozen serum samples stored at −70°C using a particle-enhanced immunonephelometric assay (N Latex Cystatin C; Dade Behring [now Siemens Healthcare Diagnostics Inc], Deerfield, Illinois) with a nephelometer (BNII, Siemens Healthcare Diagnostics Inc). The assay is stable through several freeze-thaw cycles.²⁷ We calculated eGFR at the CHS baseline (1989-1990), year 3 (1992), and year 7 (1996-1997) examinations using the following equation:

eGFR_{cystatin C} = 76.7 × (cystatin C)^−1.18

Derived in a recent pooling study of 3418 adults who underwent simultaneous cystatin C measurements and gold standard radionuclide measurements of GFR, this equation explains approximately 82% of the variation in directly measured GFR.²⁸

Participants completed standardized interviews and an extensive examination. Medical records were reviewed and standardized criteria applied to verify the presence of self-reported cardiovascular diseases.²⁹ Medications were ascertained using the inventory method in which participants brought prescription and nonprescription medication bottles to each study examination.²⁰ Diabetes was defined as the presence of a fasting glucose level lower than 140.5 mg/dL or current treatment with insulin or oral hypoglycemic agents. (To convert glucose to millimoles per liter, multiply by 0.0555.) Systolic and diastolic blood pressures were calculated from the mean of 2 consecutive readings taken while the patient was seated. Carotid ultrasonography was performed to measure the maximum stenosis of the internal and common carotid arteries.³⁰ A water-sealed, Collins Survey II spirometer (WE Collins, Braintree, Massachusetts) was used to measure forced expiratory volume in 1 second, according to American Thoracic Society criteria.³¹ Phlebotomy was performed under fasting conditions, and the blood was analyzed at the 4 field centers for levels of hemoglobin, high-density lipoprotein cholesterol, triglycerides, serum albumin, creatinine, and fibrinogen.³² C-reactive protein was measured in the entire cohort using stored plasma and a validated in-house high-sensitivity enzyme-linked immunosorbent assay.³³

We tabulated baseline participant characteristics according to physical activity category. We calculated at-risk time for each participant as time from the baseline examination until the final cystatin C measurement, and we calculated the slope of kidney function change for each individual using linear regression. We created the binary outcome variable of rapid kidney function decline based on an a priori defined slope cut point of −3.0 mL/min/1.73 m² per year. We used Poisson regression with robust variance estimation to estimate the association of physical activity covariates with rapid kidney function decline after adjusting for potential confounding variables.³⁴ We chose covariates as potential confounding factors based on plausibility that they could confound the association of physical activity level with kidney function decline, and we investigated groups of potential confounding factors by constructing nested multivariate models.

We conducted sensitivity analyses to evaluate whether associations of physical activity with change in kidney function change were robust after excluding participants who had prevalent cardiovascular disease and those with poor or fair self-reported health status. We used the likelihood ratio test to evaluate whether associations of physical activity and rapid kidney function decline differed according to baseline kidney function, sex, race, diabetes, and body mass index (BMI). We evaluated the continuous slope of eGFR in a secondary analysis. All P values were 2 tailed (α = .05). All analyses were performed using STATA software, release 10.1 (StataCorp LP, College Station, Texas).

The 1452 participants who were excluded owing to lack of follow-up kidney function measurements were older than participants who completed follow-up measurements (75.3 years vs 72.0 years, respectively), had a lower combined physical activity score (4.4 vs 5.3), a lower likelihood of being white (76.2% vs 88.5%), a lower likelihood of being female (51% vs 59%), and a lower baseline eGFR (69 mL/min/1.73 m² vs 78 mL/min/1.73 m²). Of the 1452 participants excluded, 1212 died prior to the scheduled follow-up kidney measurements (83.5%). The remainder of analyses pertain to the 4011 included study participants.

Demographics, comorbid diseases, and laboratory characteristics differed between participants in the highest vs lowest physical activity groups (Table 1). The highest physical activity group was characterized by a greater proportion of men and white subjects, a higher education level, a lower prevalence of cardiovascular diseases, better lung function, and leaner body mass. Baseline eGFR was modestly higher among participants in the highest physical activity group.

Quintiles of leisure-time physical activity for the entire CHS cohort were defined by cut points of 105.0, 480.0, 1012.5, and greater than 2088.0 kcal/wk. Interpretations of these values in terms of activity types and durations are summarized in Table 2.

There were 1663 participants who completed 2 cystatin C measurements and 2348 participants who completed 3 measurements in a median follow-up time of 7 years. The mean and median annual declines in eGFR_{cystatin C} were 1.73 and 1.55 mL/min/1.73 m², respectively (interquartile range, 0.33-2.96 mL/min/1.73 m² per year). There were 958 participants with a rapid decline in kidney function (23.9%), defined as greater than 3.0 mL/min/1.73 m² per year loss in eGFR_{cystatin C} (4.1 events per 100 person-years).

The age-, race-, and sex-adjusted rate of rapid kidney function decline decreased in graded fashion with greater physical activity scores (Figure), ranging from 15.8 rapid decline events per 100 person-years among participants in the highest physical activity group (physical activity score of 8) to 30.2 rapid decline events per 100 person-years among participants in the lowest physical activity group (physical activity score of 2).

After adjustment for demographics, prevalent cardiovascular disease, medication use, smoking, alcohol use, BMI, blood pressure, and laboratory measurements, greater physical activity scores were associated with statistically lower risks of rapid kidney function decline (Table 3). Further adjustment for subclinical disease measurements (ankle arm index, lung function, and common carotid intima-media thickness), impaired fasting glucose levels, and self-reported health status did not materially alter these estimates. After full adjustment, the 2 highest physical activity scores combined^7- 8 were associated with an estimated 28% lower adjusted risk of rapid kidney function decline (95% confidence interval [CI], 21%-41% lower) compared with the 2 lowest physical activity scores combined.^2- 3 Other physical activity and function measures, including total kilocalories of leisure-time physical activity, walking pace, and exercise intensity were also associated with a statistically lower risk of rapid kidney function decline, but number of blocks walked per week was not.

To evaluate whether observed associations of physical activity with kidney function decline might reflect poor health status among individuals with the lowest physical activity scores, we repeated our analyses removing subjects with prevalent cardiovascular disease and those with fair or poor self-reported health status (Table 4). The size of the association between physical activity and rapid kidney function decline was similar in all restricted subgroups and in the subsets of participants with a baseline eGFR of lower than 60, 60 to 90, and 90 to 119 mL/min/1.73 m² (Table 5) (P = .46 for interaction). No statistical interaction of physical activity and rapid kidney function decline was observed for sex (P = .87), race (P = .48), BMI (P = .22), diabetes (P = .97), or prevalent cardiovascular disease (P = .71).

Comparing participants in the 2 lowest (scores of 2-3) vs the 2 highest (scores of 7-8) physical activity groups, we found a mean difference in annual eGFR decline of −0.31 mL/min/1.73 m² per year (95% CI, −0.55 to −0.06) after full adjustment. For leisure-time physical activity groups (quintiles of kilocalories per week), comparing participants in the lowest quintile vs those in the highest quintile, we found that the difference was −0.39 mL/min/1.73 m² per year (95% CI, −0.65 to −0.13 mL/min/1.73 m² per year).

We observed an association of greater physical activity levels with a lower risk of rapid decline in kidney function among a general population of older adults. Associations were consistent across different types of self-reported physical activity and function, increased in magnitude with the intensity and amount of physical activity, and persisted after adjustment for well-measured clinical and subclinical disease characteristics. Kilocalories of leisure-time physical activity and exercise intensity were the 2 physical activity and function characteristics that were most strongly associated with rapid kidney function decline, whereas the number of blocks walked per week and walking pace were less strongly associated. Physical activity was associated with a statistically significant but small difference in the mean decline in eGFR assessed continuously. If these observed associations are causal, then exercise could represent a viable means to prevent progressive kidney disease in this vulnerable population.

To our knowledge, these are the first data to demonstrate an association of physical activity with the long-term change in kidney function among older adults. Kronborg et al³⁵ evaluated sex-specific risk factors for the change in kidney function in a nondiabetic Norwegian population. In age-adjusted analyses, lesser physical activity was associated with a greater increase in the serum creatinine level over time among women. However, these associations did not persist after adjustment in either sex. Previous studies of physical activity are hampered by the use of serum creatinine levels to estimate kidney function. Since exercise may increase muscle mass or limit the decline in muscle mass that occurs with inactivity, benefits of exercise on kidney function may be obscured by a concomitant rise in the serum creatinine level.

A small nonrandomized study of the effect of regular aquatic exercise in patients with moderate chronic renal failure assigned 17 adults with chronic renal failure to low-intensity aerobic exercise in a pool during a period of 12 weeks, twice a week, with sessions lasting for 30 minutes.³⁶ Nine matched controls remained sedentary. The participants in the exercise group showed significant decreases in mean (SD) cystatin C levels: from 1.7 (0.2) mg/L at baseline to 1.4 (0.1) mg/L at 12 weeks; while no such change was noted in the control group: from 1.7 (0.3) mg/L at baseline to 2.0 (0.5) mg/L at 12 weeks. Also, the estimated mean (SD) creatinine clearance was enhanced in the exercise group: increased from 62.9 (5.9) mL/min/1.73 m² at baseline to 67.1 (7.0) mL/min/1.73 m² at 12 weeks; while it remained relatively constant in the control group: 69.8 (12.3) mL/min/1.73 m² at baseline vs 66.3 (13.2) mL/min/1.73 m² at 12 weeks. (To convert creatinine clearance to milliliters per second per square meter, multiply by 0.0167.)

Exercise has both short-term and long-term beneficial effects on metabolism in nondiabetic subjects. In controlled trials, moderate physical activity improves fasting and postprandial glucose-insulin homeostasis, induces and maintains weight loss, raises high-density lipoprotein cholesterol levels, lowers low-density lipoprotein cholesterol and triglyceride levels, lowers blood pressure, and probably lowers inflammation and improves endothelial function.^13,37- 43 These metabolic benefits may affect the risk of kidney disease incidence and progression. Among more than 10 000 nondiabetic participants in the Atherosclerosis Risk in Communities Study,⁴⁴ components of the metabolic syndrome, specifically insulin resistance, were associated with a greater incidence of chronic kidney disease. In a nondiabetic Norwegian population,³⁵ the nonfasting insulin to glucose ratio was associated with a greater decline in kidney function among both men and women.

The most important limitation of this observational study is the potential for confounding because many healthy characteristics are linked with a greater desire and capacity to exercise. Indeed, in this study population, lower physical activity was associated with a number of health factors, including smoking, higher BMI, and a higher proportion of clinical and subclinical cardiovascular disease. The CHS provides a unique opportunity to evaluate physical activity and kidney function decline in older adults because both exposure and outcome were assessed using validated methods and because cardiovascular risk factors and health status characteristics were carefully measured, increasing the ability to adjust for confounding. It is important to note that some of the adjustment covariates, namely systolic blood pressure, BMI, and cholesterol and C-reactive protein levels could also be mediators of the effect of physical activity on kidney function decline, thereby potentially attenuating the size of the observed associations.

Survivorship bias represents a second potential limitation. Analyses were limited to participants who survived for at least 3 years to undergo a second measurement of kidney function. Participants who did not return for a second study visit had lower baseline physical activity levels than those who were included in the study. If excluded individuals also had greater declines in kidney function, then associations of physical activity with rapid kidney function decline could be underestimated.

Participant questionnaires were used to define physical activity variables. Although physical activity has been associated with several clinical outcomes in the CHS,^45- 46 some measurement error related to the ascertainment of physical activity characteristics is expected. Given the prospective nature of the data collection, it is probable that such error was random with respect to the decline in kidney function; such measurement error would be expected to dilute the study findings.

In conclusion, we present prospective data demonstrating an association of greater physical activity with a lower risk of rapid kidney function decline in a general population of older adults. Associations were independent of measured comorbidity, were consistent across different types of physical activity characteristics, strengthened with greater physical activity levels, and are supported by biologic evidence demonstrating effects of exercise on metabolic pathways that directly affect kidney function. These findings suggest a causal relationship of exercise with a lower risk of kidney disease progression in older people; however, this observational study cannot prove a cause-effect relationship. These findings motivate further studies to evaluate whether exercise represents a viable method for protecting against age-related decline in kidney function.

Correspondence: Cassianne Robinson-Cohen, MS, Department of Epidemiology, University of Washington, Box 357236, Seattle, WA 98195 (cassyrc@u.washington.edu).

Accepted for Publication: September 4, 2009.

Author Contributions: Ms Robinson-Cohen had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Robinson-Cohen, de Boer, Sarnak, Shlipak, Siscovick, and Kestenbaum. Acquisition of data: Robinson-Cohen, Katz, Shlipak, and Siscovick. Analysis and interpretation of data: Robinson-Cohen, Katz, Mozaffarian, Dalrymple, de Boer, Shlipak, Siscovick, and Kestenbaum. Drafting of the manuscript: Robinson-Cohen. Critical revision of the manuscript for important intellectual content: Robinson-Cohen, Katz, Mozaffarian, Dalrymple, de Boer, Sarnak, Shlipak, Siscovick, and Kestenbaum. Statistical analysis: Robinson-Cohen, Katz, Dalrymple, de Boer, and Kestenbaum. Obtained funding: Sarnak, Shlipak, and Siscovick. Study supervision: Robinson-Cohen, Shlipak, and Siscovick.

Financial Disclosure: None reported.

Funding/Support: This research was supported by National Institutes of Health Research Project Grant R01 AG 027002. The CHS was supported by contracts N01-HC-35129, N01-HC-45133, N01-HC-75150, N01-HC-85079 through N01-HC-85086, N01-HC-15103, N01-HC-55222, and U01 HL080295 from the National Heart, Lung, and Blood Institute; by the National Institute of Neurological Disorders and Stroke; and by grant R01AG027002 from the National Institutes on Aging.