Effect of Kidney Function on Relationships between Lifestyle Behaviors and Mortality or Cardiovascular Outcomes: A Pooled Cohort Analysis

Significance Statement

Healthy lifestyle behaviors reduce death and cardiovascular disease among individuals with preserved kidney function, but the benefits of these behaviors among those with reduced kidney function, a group at higher risk for cardiovascular disease, have not been established. In this pooled analysis of three community-based cohort studies that included 27,271 adults, healthy lifestyle behaviors (not smoking, maintaining body mass index in the recommended range, engaging in regular physical activity, consuming a healthy diet, and no more than a moderate intake of alcohol), analyzed individually and in combination, were associated with significantly reduced risk of death and cardiovascular disease events among individuals with or without reduced kidney function. These findings highlight the importance of lifestyle behaviors as potentially modifiable risk factors for people with kidney disease.

Visual Abstract

Abstract

Background

Adherence to healthy behaviors reduces risks of cardiovascular disease and death in the general population. However, among people with kidney disease, a group at higher risk for cardiovascular disease, such benefits have not been established.

Methods

We pooled data from three cohort studies with a total of 27,271 participants. Kidney function was categorized on the basis of eGFR (≥60, 45 to <60, and <45 ml/min per 1.73 m²). We used proportional hazard frailty models to estimate associations between healthy behaviors (not smoking, at recommended body mass index [BMI], physical activity, healthy diet, and moderate to no alcohol intake) and outcomes (all-cause death, major coronary events, ischemic stroke, and heart failure events).

Results

All recommended lifestyle behaviors were significantly associated with lower risks of death, regardless of eGFR. Not smoking (versus current) and any moderate to vigorous physical activity (versus none) was significantly associated with reduced risks of major coronary and heart failure events, regardless of eGFR. Any (versus no) moderate or vigorous physical activity significantly associated with decreased risk of ischemic stroke events only among those with eGFR ≥60. Moderate to no daily alcohol intake (versus excessive) was significantly associated with an increased risk of major coronary events, regardless of eGFR. For heart failure events, a BMI of 18.5 to 30 associated with decreased risk, regardless of eGFR. Across all eGFR categories, the magnitude of risk reduction for death and all cardiovascular outcomes increased with greater numbers of recommended lifestyle behaviors.

Conclusions

Recommended lifestyle behaviors are associated with lower risk of death and cardiovascular disease events among individuals with or without reduced kidney function, supporting lifestyle behaviors as potentially modifiable risk factors for people with kidney disease.

Cardiovascular disease is the leading cause of death in the United States.¹ In epidemiologic studies, lifestyle behaviors that are generally regarded as healthy have been shown to strongly influence the risks of cardiovascular disease, independent of traditional risk factors and medications.^2–15 However, many of these studies excluded groups at high risk for cardiovascular disease events, such as those with kidney disease.

Adults with kidney disease, estimated to be 15% of the US population, have a ten-fold higher risk of cardiovascular disease compared with the general population.^16,17 Analyses of the effect of healthy lifestyle behaviors on preventing cardiovascular disease and death among individuals with kidney disease have not shown consistent results.^18–22 Because the effect of healthy lifestyle behaviors among individuals with kidney disease is unclear, recommendations for lifestyle behaviors in kidney disease clinical practice guidelines are based largely on general population studies.²³

This study aimed to determine the association of lifestyle behaviors, with the risk of all-cause death and cardiovascular events, including a composite of coronary events, ischemic stroke, and heart failure (HF) among adults with and without reduced kidney function. Due to a higher prevalence and differences in pathophysiology of vascular disease among those with kidney disease compared with the general population, we hypothesized that recommended lifestyle behaviors would have less benefit on all outcomes, and particularly on cardiovascular outcomes, among individuals with reduced kidney function, compared with those with preserved kidney function. We used a pooled dataset created from three community-based prospective cohort studies to evaluate the separate and joint associations of lifestyle behaviors with all-cause death and cardiovascular disease events among individuals with different levels of kidney function.

Methods

Data Sources

We obtained public-use datasets through the National Heart, Lung, and Blood Institute (NHLBI) Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC) for three US cohort studies of cardiovascular disease risk factors: the Atherosclerosis Risk in Communities Study (ARIC), the Cardiovascular Health Study (CHS), and the Multi-Ethnic Study of Atherosclerosis (MESA). ARIC enrolled 15,792 participants aged 45–64 years from four communities (Jackson, Mississippi; Forsyth County, North Carolina; Minneapolis, Minnesota; and Washington County, Maryland) from 1987 to 1989. CHS enrolled 5201 participants aged 65 years and older between 1989 and 1990 from four communities (Sacramento, California; Washington County, Maryland; Forsyth County, North Carolina; and Pittsburgh, Pennsylvania), then purposely recruited 687 Black patients between 1992 and 1993 from three (excluding Washington County) of the four communities. MESA enrolled 6814 participants 45–84 years of age from six field centers from 2000 and 2002 (Baltimore, MD; Chicago, IL; Forsyth County, NC; Los Angeles, CA; New York, NY; St Paul, MN). We used follow-up data until 2017 for ARIC and MESA, and 2009 for the CHS that were available through BioLINCC. Participants provided informed consent before cohort enrollment.^24–29 This study was deemed exempt by the University of Pennsylvania Institutional Review Board.

Study Sample

Our sample of 27,271 was generated by including cohort members with a serum creatinine measurement at baseline and excluding 181 individuals with a body mass index (BMI) <18.5 kg/m² because eGFR categorization is challenging in this group because of low muscle mass. We included those with cardiovascular disease at baseline to maximize generalizability. To estimate the GFR, we utilized the CKD Epidemiology Collaboration equation,³⁰ then each participant was grouped into one of three eGFR categories (≥60, 45 to <60, and <45 ml/min per 1.73 m²; hereafter referred to as: ≥60, 45 to <60, and <45, respectively) that corresponded to preserved, mild-to-moderately decreased, or moderate-to-severely decreased kidney function in accordance with the Kidney Disease Improving Global Outcomes classification system.³¹

Baseline Variables

Baseline data collection methods have been described.^{24–27,29,32–39} For our analyses, diabetes was defined by self-report, use of glucose-lowering agents, or fasting blood glucose ≥126 mg/dl. Hypertension was determined by self-report, measured systolic BP (SBP) >140 mm Hg, or use of antihypertensive medications. Cardiovascular disease at baseline was defined as a history of HF, myocardial infarction (MI), coronary revascularization, stroke, transient ischemic attack, or claudication. In all three studies, BP was recorded as an average of two readings after resting in a seated position.

Measures of Lifestyle Behaviors

We evaluated five lifestyle behaviors on the basis of prior literature demonstrating associations with the risk of cardiovascular disease or all-cause death: smoking, BMI (as an indicator of healthy weight maintenance), physical activity, alcohol intake, and dietary habits.^{4,9,12,14,40–42} Smoking status was self-reported. BMI (kg/m²) was calculated from measured height and weight. Weekly moderate-to-vigorous physical activity was estimated on the basis of self-reported frequency and duration of activities^32–35 that corresponded to ≥3 metabolic equivalents.⁴³ Dietary composition was determined from food frequency questionnaires.^36–39 A diet score (0–5) was constructed by assigning one point for each component recommended by the American Heart Association (AHA): fruits and vegetables (≥4.5 cups or servings/d), fish (≥3.5 oz or ≥2 servings/wk), whole grains (≥3 1 oz servings/d), sodium (<1500 mg/d), and sugar-sweetened beverages (≤36 oz/wk or ≤4 glasses/wk).⁴⁴ Self-reported alcohol intake was classified as none or moderate (≤28 g/d, or ≤2 standard alcoholic drinks, for men and ≤14 g/d, or ≤1 standard alcoholic drink, for women) or excessive (>28 g/d for men, >14 g/d for women) on the basis of national guidelines.⁴⁵ Each lifestyle behavior was dichotomized as “recommended” versus “not recommended” on the basis of prior literature, including the AHA Life’s Simple 7 three categories of cardiovascular health (ideal, intermediate, and poor): BMI (18.5 to <30 versus ≥30 kg/m²), weekly moderate to vigorous physical activity (any versus none), diet score (≥2 versus 0–1), smoking status (never/former versus current), and daily alcohol intake (none/moderate versus excessive).^{4,9,12,13,40–42,46,47} The AHA intermediate and ideal categories were combined to define recommended because our primary analyses included those with a history of cardiovascular disease, and by definition, individuals with prevalent cardiovascular disease can, at best, achieve intermediate cardiovascular health. The AHA poor category defined not recommended behaviors. The dichotomization using ideal/intermediate versus poor, compared with ideal versus intermediate/poor AHA categories has been shown to have consistent associations with clinical outcomes.⁷

A lifestyle behavior score ranging from 0 (least healthy) to 5 (healthiest) was generated as the sum of recommended lifestyle behaviors. Because only 0.2%, 2.3%, and 12.9% of participants had lifestyle behavior score of 0, 1, and 2, respectively, we collapsed the lower three categories into a category named “0–2”.

Follow-Up and Outcomes

Each cohort ascertained death and cardiovascular events during follow-up.^26,48–51 All of the studies had medical adjudication committees and standardized diagnoses (which were similar between studies) on the basis of medical records (see Supplemental Methods for further details). Our study outcomes included all-cause death and the initial occurrence after cohort entry of each of the following: a composite of major coronary events (definite or probable MI, definite fatal coronary heart disease, or revascularization procedure including coronary angioplasty or other revascularization procedure), ischemic stroke, and HF events. Participants were followed from baseline to death, loss to follow-up, or end of cohort follow-up, whichever occurred first.

Statistical Analyses

Comparisons of baseline characteristics across eGFR groups were made using chi-squared tests, ANOVA, or Kruskal-Wallis tests, as appropriate. Proportional hazards regression models were used to estimate the relationship of dichotomous lifestyle behaviors with each outcome, with the not recommended behavior serving as the reference group. To account for the study-to-study heterogeneity that may be caused by differences in the cohort studies’ patient populations or methods of data collection, frailty models were used for the time to event analyses.⁵² Frailty models treat the study as a random effect instead of fixed. The frailty term is an unobservable random variable that corresponds to each individual’s underlying modification to the baseline hazard function to account for unobserved heterogeneity.⁵³ For all-cause death, adjusted Cox models were used. In the presence of competing events (i.e., death unrelated to major coronary events, ischemic stroke, and HF events), cause-specific hazards models were generated treating the competing events as censored. Proportional hazards models were also used to estimate the association of the lifestyle behavior score as a categorical variable with each outcome, with a lifestyle behavior category of 0–2 serving as the reference group. The proportional hazards frailty models were adjusted for age, sex, race/ethnicity, education, baseline cardiovascular disease, hypertension, diabetes, BMI, eGFR, SBP, and diastolic BP, and a cohort study was included as a random effect. Proportionality assumptions were visually examined with Schoenfeld residuals. Kaplan-Meier curves with log-rank testing were assessed for equality of survival distributions for all-cause death between lifestyle behavior score and eGFR groups. Fine and Gray’s competing risk regression models and Gray’s tests assessed the equality of cumulative event probabilities for the composite of major coronary events, ischemic stroke, and HF events. Effect modification was tested using interaction terms of lifestyle behaviors and eGFR categories (≥60, 45 to <60, and <45 ml/min per 1.732).

Missingness for covariates ranged from 0% to 0.2%, For lifestyle behaviors, the missingness was 0.2% for smoking, 2.2% for alcohol, 4.4% for diet score, and 2.4% for physical activity. Multiple imputation was performed for any missing lifestyle behavior variables (n=1265) with the following variables: sex, BMI, age, race, education, SBP, diastolic BP, baseline eGFR, diabetes, hypertension, history of cardiovascular disease, medications (ACE inhibitor, statins, aspirin, betablockers), and total cholesterol.⁵⁴ Final models are on the basis of the imputed results. In sensitivity analyses, we modeled the individual lifestyle behavior regression analyses for diet and BMI as continuous variables, assessed the relationships of potassium and sodium intake with outcomes of interest, and repeated the lifestyle behavior score regression analyses excluding those with cardiovascular disease at baseline. Analyses were performed using STATA, version 16 (Stata Corp, College Station, TX). All P values were two tailed (α=0.05).

Results

Baseline characteristics of the study population are shown in Table 1. There were 16,347 participants with eGFR ≥60, 8257 participants with eGFR 45 to <60, and 2667 participants with eGFR <45. Participants with eGFR <45 were older and had less education, higher SBP, and more diabetes, cardiovascular disease, and hypertension. Healthy diet scores were more frequent among those with an eGFR 45 to <60. Moderate to vigorous physical activity, BMI ≥30 kg/m², current smoking, and excessive alcohol use were more common among those with eGFR ≥60.

Table 1.

Baseline demographics, clinical characteristics, and lifestyle behaviors by eGFR category (n=27,271)

Baseline Characteristics	All (n=27,271)	eGFR ≥60 (n=16,347)	eGFR 45 to <60 (n=8257)	eGFR <45 (n=2667)	P Valuea
Cohort study, n (%)					<0.001
Multiethnic study of atherosclerosis	6740 (24.7)	5861 (35.9)	726 (8.8)	153 (5.7)
Cardiovascular health study	14,833 (54.4)	7806 (47.8)	5817 (70.5)	1210 (45.4)
Atherosclerosis risk in communities study	5698 (20.9)	2680 (16.4)	1714 (20.8)	1304 (48.9)
Demographics
Age, mean (SD), yr	60.1 (10.2)	58.8 (9.8)	60.3 (9.8)	66.9 (10.8)	<0.001
Male, %	45.4	46.1	43.7	46.4	0.001
Race, %					<0.001
White	67.8	59.6	80.1	80.1
Black	23.8	28	17.5	17.6
Hispanic	5.5	8.2	1.4	1.2
Other	3.0	4.3	1.1	1.1
Education, %					<0.001
Less than high school	23.3	22.6	22.6	30.4
High school/some college	44.4	44.2	44.7	44.0
4 yr degree or more	32.3	33.2	32.7	25.7
Clinical characteristics
Systolic BP, mean (SD), mm Hg	137 (20)	135 (20)	140 (19)	142 (22)	<0.001
Diastolic BP, mean (SD), mm Hg	84 (16)	82 (15)	87 (15)	81 (17)	<0.001
eGFR, mean (SD), ml/min per 1.73 m2	66.1 (18.3)	77.2 (14.1)	53.6 (4.3)	36.7 (8.0)	<0.001
Total cholesterol, mean (SD), mg/dL	208.9 (41)	205.5 (40)	213.9 (40.9)	214.6 (44)	<0.001
Diabetes, %	13.9	13.9	11.5	20.8	<0.001
Cardiovascular diseaseb, %	14.4	10.6	17.0	30.0	<0.001
Hypertension, %	44.7	41.3	44.1	67.6	<0.001
HF, %	3.1	2.2	4.0	5.9	<0.001
Ace-inhibitor use, %	6.5	6.7	5.2	8.8	<0.001
Betablocker use, %	10.5	9.7	10.6	15.0	<0.001
Statin use, %	4.5	5.6	2.8	3.0	<0.001
Aspirin use, %	37.6	33.8	43.7	42.4	<0.001
BMI, mean (SD), kg/m2	27.7	27.8 (5.3)	27.5 (4.8)	27.6 (18.8)	0.01
Sodium intake, mg/d, median (IQR)	1641 (1175–2381)	1654 (1185–2390)	1565 (1129–2179)	1923 (1312–2965)	<0.001
Potassium intake, mg/d, median (IQR)	2321 (1720–3047)	2353 (1735–3085)	2326 (1741–3033)	2103 (1562–2829)	<0.001
Lifestyle behaviors, %
BMI					<0.001
≥30 kg/m2	27.2	28.1	25.6	26.2
18.5 to <30 kg/m2	72.8	71.9	74.5	73.8
Moderate-vigorous physical activity					<0.001
0 min/wk	48.5	46.7	51.2	51.3
≥1 min/wk	51.5	53.3	48.8	48.8
Diet score (range 0–5)c					<0.001
0–1	39.6	41.7	35.8	38.6
≥2	60.4	58.3	64.2	61.4
Smoking status					<0.001
Current	19.9	21.7	17.9	14.4
Never/former	80.1	78.3	82.1	85.6
Daily alcohol intake					<0.001
Excessived	8.6	9.2	8.1	6.1
None/moderate	91.4	90.8	91.9	93.9
Lifestyle Behavior Scoree					<0.001
0–2	15.3	16.3	14.1	13.2
3	29.6	30.3	28.5	28.9
4	35.8	34.9	37.0	37.6
5	19.3	18.6	20.4	20.4

IQR, interquartile range.

^aX² test of difference across eGFR categories.

^bAny cardiovascular disease: history of HF, MI, coronary revascularization, stroke, transient ischemic attack, or claudication.

^cDefined as sum of each AHA-recommended dietary component: fruits and vegetables (≥4.5 cups or servings/d), fish (≥3.5 oz or ≥2 servings/wk), whole grains (≥3 1 oz servings/d), sodium (<1500 mg/d), and sugar-sweetened beverages (≤36 oz/wk or ≤4 glasses/wk).

^dExcessive: >28 g/d, or >2 standard drinks, for men and >14 g/d, or >1 standard drink, for women.

^eDefined as sum of each healthy lifestyle behavior criterion met: BMI (18.5 to <30 kg/m² [1 point] versus ≥30 [0 points]), physical activity (any weekly moderate to vigorous physical activity [1 point] versus no moderate to vigorous physical activity (0 points)], diet score [≥2 (1 point) versus 0–1 (0 points)], smoking status (never/former [1 point] versus current [0 points], and daily alcohol intake (no to moderate [1 point] versus excessive [0 points]).

Individual Lifestyle Behaviors and Outcomes

As shown in Table 2, during a median follow-up of 16.7 years (interquartile range, 8.6–22.7), there were 5238 composite major coronary events (1.3 in 100 person-years), 1985 ischemic strokes (0.5 in 100 person-years), 4473 HF events (1.2 in 100 person-years), and 9776 deaths from all-causes (2.3 in 100 person-years).

Table 2.

Outcome events and event rate (per yr) overall and by eGFR category

Outcome Events	Overall	eGFR ≥60	eGFR 45 to <60	eGFR <45
All-cause death
Total all-cause deaths, n	9756	4784	3264	1708
Death rate (per 100 person-yr), median (IQR)	2.36 (2.31–2.41)	2.02 (1.97–2.08)	2.29 (2.21–2.37)	4.99 (4.76–5.23)
Noncoronary heart disease deaths, n	7962	3993	2657	1312
Noncoronary heart disease death rate (per 100 person-yr), median (IQR)	1.93 (1.88–1.97)	1.69 (1.64–1.74)	1.86 (1.79–1.93)	3.83 (3.63–4.05)
Composite of major coronary eventsa
Total events, n	5238	2584	1827	827
Event rate (per 100 person-yr), median (IQR)	1.34 (1.30–1.37)	1.15 (1.10–1.19)	1.36 (1.30–1.43)	2.52 (2.36–2.70)
Incident MI events	2802	1381	994	427
Incident MI rate (per 100 person-yr), median (IQR)	0.69 (0.67–0.72)	0.60 (0.57–0.63)	0.72 (0.67–0.76)	1.26 (1.14–1.38)
Fatal coronary heart disease events, n	1794	791	607	396
Fatal coronary heart disease event rate (per 100 person-yr), median (IQR)	0.43 (0.41–0.45)	0.33 (0.31–0.36)	0.43 (0.39–0.46)	1.16 (1.05–1.28)
Revascularization procedures, n	2582	1377	940	265
Procedure event rate (per 100 person-yr), median (IQR)	0.64 (0.62–0.67)	0.60 (0.57–0.63)	0.69 (0.64–0.73)	0.79 (0.69–0.88)
Ischemic stroke
Total ischemic strokes, n	1985	986	682	317
Ischemic stroke event rate (per 100 person-yr), median (IQR)	0.48 (0.46–0.50)	0.42 (0.39–0.45)	0.48 (0.45–0.52)	0.92 (0.82–1.03)
Heart failure
Total heart failure events, n	4473	2168	1520	785
Heart failure event rate (per 100 person-yr), median (IQR)	1.16 (1.12–1.19)	0.97 (0.93–1.01)	1.15 (1.09–1.21)	2.49 (2.32–2.67)

^aDefinite or probable myocardial infarction, definite fatal coronary heart disease, and revascularization procedure including coronary angioplasty or other revascularization procedure.

All of the recommended lifestyle behaviors were associated with reduced all-cause death, regardless of eGFR (Figure 1A, Table 3). Former/never (versus current) smoking and any (versus no) moderate to vigorous weekly physical activity were associated with a protective effect of major coronary events (adjusted hazard ratio [aHR] 0.62; 95% confidence interval [95% CI], 0.57 to 0.66 and aHR, 0.85; 95% CI, 0.80 to 0.90, respectively) and HF events (aHR, 0.48; 95% CI, 0.45 to 0.52 and aHR, 0.84; 95% CI, 0.79 to 0.90, respectively) regardless of eGFR (Figure 1B, Table 3). None/moderate (versus excessive) daily alcohol intake was associated with an increased risk of major coronary events (aHR, 1.29; 95% CI, 1.16 to 1.44). Any (versus no) moderate to vigorous physical activity was associated with decreased risk of ischemic stroke events only among those with eGFR ≥60 (aHR, 0.80; 95% CI, 0.70 to 0.92; P value for interaction by eGFR group = 0.001) (Figure 1C, Table 3). For HF events, a BMI of 18.5 to <30 (versus ≥30) kg/m² was associated with decreased risk events regardless of eGFR (aHR, 0.87; 95% CI, 0.78 to 0.97) (Figure 1D, Table 3).

Figure 1.

Risk of adverse outcomes according to individual lifestyle behaviors. (A) All-cause death, (B) composite of major coronary events, (C) ischemic stroke, and (D) HF events. Plots show HR and error bars indicate 95% CI.

Table 3.

Association of individual lifestyle behaviors with all-cause death, major coronary events, ischemic stroke, and HF

Individual Lifestyle Behaviors	All eGFR Categories		P Value for Interaction by eGFR Category
	HR (95% CI)	P value
All-cause death
Smoking status			0.12
Current	Ref
Former/Never	0.46 (0.43 to 0.48)	<0.0001
BMI			0.76
≥30 kg/m2	Ref
18.5–<30 kg/m2	0.84 (0.78 to 0.91)	<0.0001
Diet score			0.68
0–1	Ref
≥2	0.93 (0.89 to 0.97)	0.001
Physical activity			0.42
0 min/wk	Ref
≥1 min/wk	0.80 (0.76 to 0.84)	<0.0001
Daily alcohol intake			0.06
Excessivea	Ref
None/moderate	0.87 (0.81 to 0.93)	<0.0001
Major coronary events
Smoking statusb			0.18
Current	Ref
Former/Never	0.62 (0.57 to 0.66)	<0.0001
BMIc			0.38
≥30 kg/m2	Ref
18.5–<30 kg/m2	0.98 (0.89 to 1.08)	0.74
Diet scoreb			0.74
0–1	Ref
≥2	1.01 (0.95 to 1.07)	0.86
Physical activityd			0.22
0 min/wk	Ref
≥1 min/wk	0.85 (0.80 to 0.90)	<0.0001
Daily alcohol intakeb			0.97
Excessivea	Ref
None/moderate	1.29 (1.16 to 1.44)	<0.0001
Ischemic stroke
Smoking status			0.04
Current	Ref
Former/never	0.59 (0.53 to 0.66)	<0.0001
BMI			0.27
≥30 kg/m2	Ref
18.5 to <30 kg/m2	0.93 (0.79 to 1.09)	0.37
Diet score			0.36
0–1	Ref
≥2	0.92 (0.83 to 1.01)	0.08
Physical activity			0.001
0 min/wk	Ref
≥1 min/wk	0.91 (0.82 to 1.00)	0.06
Daily alcohol intake			0.47
Excessivea	Ref
None/moderate	0.95 (0.81 to 1.12)	0.57
Heart failure
Smoking status			0.99
Current	Ref
Former/never	0.48 (0.45 to 0.52)	<0.001
BMI			0.54
≥30 kg/m2	Ref
18.5 to <30 kg/m2	0.87 (0.78 to 0.97)	0.006
Diet score			0.60
0–1	Ref
≥2	0.99 (0.93 to 1.05)	0.77
Physical activity			0.30
0 min/wk	Ref
≥1 min/wk	0.84 (0.79 to 0.90)	<0.001
Daily alcohol intake			0.07
Excessivea	Ref
None/moderate	1.00 (0.89 to 1.11)	0.11

Models adjusted for age, race, sex, education, baseline eGFR, hypertension, BMI, SBP, diastolic BP, history of diabetes, history of cardiovascular disease, aspirin use, ACE-inhibitor use, betablocker use, statin use, total cholesterol, eGFR category, and original cohort as a random effect.

^aExcessive: >28 g/d, or >2 standard drinks, for men and >14 g/d, or >1 standard drink, for women.

^bAdjustment did not include history of cardiovascular disease or hypertension due to numerical infeasibility.

^cAdjustment did not include history of cardiovascular disease or total cholesterol due to numerical infeasibility.

^dAdjustment did not include history of cardiovascular disease due to numerical infeasibility.

In sensitivity analyses, an increase in BMI was associated with increased risk of major coronary and HF events regardless of eGFR category. An increase in diet score was associated with decreased risk of all-cause death and major coronary events regardless of kidney function (Supplemental Table 1). An increase in dietary sodium intake was associated with increased risk of all-cause death and increased potassium intake was associated with a decreased risk of stroke, but increased risk of HF events across eGFR groups (Supplemental Table 1).

Lifestyle Behavior Score and Outcomes

The frequencies of lifestyle behavior score are reported in Table 1. For each eGFR group, those with higher lifestyle behavior scores had better survival (Supplemental Figure 1; log-rank test, P<0.001) and lower incidence of cardiovascular disease events (Supplemental Figures 2–4; Gray’s tests P<0.001). Lifestyle behavior score categories of 3, 4, and 5 (versus 0–2) were each associated with dose-related lower risks of all-cause death, major coronary events, ischemic strokes, and HF events, regardless of kidney function (Figure 2, A–D, Table 4).

Figure 2.

Risk of adverse outcomes according to combined lifestyle behavior score. (A) All-cause death, (B) composite of major coronary events, (C) ischemic stroke, and (D) HF events. Plots show HR and error bars indicate 95% CI.

Table 4.

Association of composite lifestyle behavior score with all-cause death, major coronary events, ischemic stroke, and HF

Lifestyle Behavior Score	All eGFR Categories		P Value for Interaction by eGFR Category
	HR (95% CI)	P value
All-cause death
0–2	Ref		0.13
3	0.72 (0.67 to 0.76)	<0.0001
4	0.54 (0.51 to 0.58)	<0.0001
5	0.48 (0.45 to 0.52)	<0.0001
Major coronary eventsa
0–2	Ref		0.82
3	0.87 (0.80 to 0.95)	0.002
4	0.78 (0.72 to 0.85)	<0.001
5	0.73 (0.65 to 0.81)	<0.001
Ischemic stroke
0–2	Ref		0.16
3	0.80 (0.70 to 0.92)	0.002
4	0.73 (0.63 to 0.84)	<0.0001
5	0.62 (0.52 to 0.74)	<0.0001
Heart failure
0–2	Ref		0.41
3	0.73 (0.67 to 0.80)	<0.0001
4	0.63 (0.57 to 0.69)	<0.0001
5	0.56 (0.50 to 0.63)	<0.0001

Models adjusted for age, race, sex, education, baseline eGFR, hypertension, BMI, SBP, diastolic BP, history of diabetes, history of cardiovascular disease, aspirin use, ACE-inhibitor use, betablocker use, statin use, total cholesterol, eGFR category, and original cohort as a random effect.

^aAdjustment did not include history of cardiovascular disease and hypertension due to numerical infeasibility.

When eGFR was analyzed as a continuous variable (per ml/min per 1.73 m²), interactions were evident between smoking and all-cause death (β coefficient −0.005; 95% CI, −0.008 to −0.003) and major cardiovascular events (β coefficient −0.005; 95% CI, −0.009 to −0.001), and between daily alcohol intake and all-cause death (β coefficient −0.008; 95% CI, −0.011 to −0.004) and HF events (β coefficient −0.008; 95% CI, −0.001 to −0.002) (Supplemental Table 2). The magnitude of protection associated with not smoking (compared with current smoking) or associated with no or moderate alcohol intake (compared with excessive intake) was greater among individuals with a higher baseline eGFR value. Results of sensitivity analyses excluding participants with baseline cardiovascular disease history were qualitatively similar to results of the primary analyses (Supplemental Table 3).

Discussion

In this study of individuals with a broad range of kidney function, we found that recommended lifestyle behaviors, analyzed individually and in combination, were associated with reduced risk of death and cardiovascular disease events. For all-cause death, each of the five recommended lifestyle behaviors was protective, and there was a strong graded lower risk as the number of recommended lifestyle behaviors increased. With the exception of nonsmoking and physical activity, relationships of individual recommended lifestyle behaviors were less evident for cardiovascular disease outcomes than for death, regardless of kidney function. When considered in combination, adherence to a greater number of recommended behaviors was clearly associated with reduced risk of death and cardiovascular disease events, regardless of kidney function.

Although most individuals with kidney disease die from cardiovascular disease,^16,55 evidence is limited regarding optimal management and prevention of cardiovascular disease in this population. Kidney disease is often an exclusion criterion for cardiovascular outcomes trials, and recommendations for lifestyle behaviors included in kidney disease clinical practice guidelines are based largely on studies in the general population. The relevance of these recommendations to people with kidney disease who have both higher risk and different underlying pathophysiology of cardiovascular disease has not been established.^16,17,56

We expected that healthy behaviors would have less of an effect on the risk of death and cardiovascular events among those with reduced, compared with preserved, kidney function, because of the higher prevalence, and greater severity of vascular disease among those with reduced kidney function.^57,58 However, our study indicates the relative risk reductions associated with recommended behaviors are similar, regardless of the level of kidney function. Because the individuals with worse kidney function have a higher risk of death and cardiovascular events than those with preserved kidney function, the similar relative reductions in risk with healthy behaviors across eGFR categories translates into greater absolute reductions in outcome events for those with more severe kidney dysfunction than for those with more preserved kidney function.

The finding that recommended lifestyle behaviors are associated with better outcomes across levels of kidney function may reflect favorable effects of healthy behaviors on traditional cardiovascular disease risk factors, such as hypertension, obesity, insulin resistance, and hyperlipidemia, which are common to people with and without kidney disease.⁶ However, healthy behaviors may also counteract cardiovascular risk factors unique to kidney disease, such as albuminuria and mediators of inflammation and oxidative stress.^59,60 Albuminuria, which has been linked to endothelial dysfunction and incident cardiovascular disease, can be reduced with dietary changes, such as moderate salt reduction.⁶¹ Physical activity may be particularly relevant for individuals with kidney disease because it lowers C-reactive protein levels in a dose-response manner,⁶² and potentially reduces frailty because physical activity has been associated with an increase in cardiovascular fitness and muscle strength.⁶³ The effect of a healthy lifestyle on novel cardiovascular disease risk factors in kidney disease deserves further investigation.

We did not observe consistent associations between adherence to the AHA-recommended diet pattern and lower risk across all cardiovascular disease outcomes. In previous studies of cohorts that, for the most part, were without kidney disease, healthier diets were associated with lower risk of mortality and cardiovascular disease events.^7,9 Our lack of finding an association with a recommended diet could be due to our use of a less-sensitive measure of a healthy diet, because we included the AHA intermediate and ideal categories to define a healthy diet (score ≥2), and several of the previous studies used an ideal category of the AHA-recommended diet (score >4). Additionally, for those with decreased kidney function, a diet rich in fruits and vegetables may result in hyperkalemia, and thereby possibly counteract beneficial effects of a “heart-healthy” diet. Additional research on effects of diet is needed among those with and without kidney disease.

Healthy lifestyle behaviors may be most effective when present in combination. Prior studies have demonstrated that, when analyzed at the per-person level, the number of healthy lifestyle behaviors has a strong graded inverse relationship with death,^7,13 MI,⁶⁴ cardiovascular mortality,^7,8,10,13 stroke, and HF.^65,66 However, most of these studies excluded individuals with a history of kidney disease. This study extends the investigation of potential protective effects of lifestyle behaviors to individuals with reduced kidney function, and found a similar graded inverse relationship of three or more recommended lifestyle behaviors with death and cardiovascular events regardless of kidney function.

This study further confirms that lifestyle behaviors should be viewed as major independent risk factors for poor health that require screening and treatment among all individuals, regardless of kidney function. In the past two decades, participation in healthy lifestyle behaviors has decreased,¹³ yet less than half of patients report being counseled about healthy behaviors during their most recent physician office visit.^67,68 Further research is needed to understand how patients and providers prioritize these behaviors.

The results of our study need to be interpreted in the context of several limitations. First, the lifestyle behaviors are susceptible to measurement error due to self-report and possible inadequacies in data harmonization across studies. Second, the analyses incorporated measures of lifestyle behaviors only at baseline, and it is possible that behaviors changed over time. Third, dichotomizing the lifestyle behaviors with graded effects on risk would result in different interpretation of associations, compared with evaluating behaviors as continuous measures or by restricting to more extreme scores. Therefore, our findings likely underestimate the importance of each lifestyle behavior and their combined effects. Additional limitations include insufficient power to detect modest associations among the low kidney function group, and residual confounding and inability to establish causality due to the observational nature of the study.

Despite these limitations, our study has important strengths, including a large sample size, the use of data from three well-established cohorts, and inclusion of a substantial number of individuals with impaired kidney function. Additional strengths include rigorous assessments of lifestyle behaviors and outcomes that should minimize measurement error and misclassification, detailed data on important baseline risk factors to reduce confounding, and a diverse study population to increase generalizability.

In summary, healthy lifestyle behaviors were associated with lower risk of death and most of the cardiovascular disease outcomes, regardless of kidney function. Our findings extend previous demonstrations of protective effects of healthy lifestyle behaviors in the general population to those with kidney disease, a highly prevalent condition associated with markedly increased risks of cardiovascular disease morbidity and mortality. Our study supports the importance of lifestyle as a potentially modifiable risk factor for individuals with and without kidney disease.

Disclosures

A. Anderson reports personal fees from Kyowa Hakko Kirin, outside the submitted work. H. Feldman reports consultancy agreements with Kyowa Hakko Kirin Co, Ltd. (ongoing), National Kidney Foundation, InMed, Inc., and DLA Piper, LLP; reports receiving research funding from Regeneron; honoraria from Rogosin Institute (invited speaker); reports being a scientific advisor or member of the Steering Committee, Chronic Renal Insufficiency Cohort Study, Editor-in-Chief, and National Kidney Foundation (Member of Advisory Board). J. Hsu reports being a scientific advisor or member of AJKD as a Statistics/Epidemiology Editor and PLOS ONE as a Statistical Advisory Board member. L. Dember receives compensation from the National Kidney Foundation for her role as a Deputy Editor of the American Journal of Kidney Diseases; and reports consulting fees from Merck and GlaxoSmithKline.

Funding

This work was supported in part by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant K23 DK118198-01A1 (to S. Schrauben). A. Anderson reports grants from National Institutes of Health/NIDDK, during the conduct of the study.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020040394/-/DCSupplemental.

Supplemental Methods. Ascertainment of outcome events.

Supplemental Figure 1. Kaplan-Meier curves for overall survival by eGFR category and Lifestyle Behavior Scores.

Supplemental Figure 2. Cumulative incidence functions for major coronary disease events across eGFR categories and Lifestyle Behavior Scores.

Supplemental Figure 3. Cumulative incidence functions for ischemic stroke events across eGFR categories and Lifestyle Behavior Scores.

Supplemental Figure 4. Cumulative incidence functions for HF events across eGFR categories and Lifestyle Behavior Scores.

Supplemental Table 1. Association of continuous measures of diet, BMI, sodium intake, and potassium intake with all-cause death, major coronary events, ischemic stroke, and HF by eGFR category.

Supplemental Table 2. Assessment of statistical interactions between lifestyle behaviors and clinical outcomes by continuous eGFR (centered).

Supplemental Table 3. Associations of Lifestyle Behavior Scores and all-cause death, major coronary events, ischemic stroke, and HF events by eGFR category among participants without a history of cardiovascular disease at baseline.