Association of physical activity with risk of chronic kidney disease in China: A population-based cohort study

Highlights

• •Total physical activity was inversely associated with the risk of chronic kidney disease (CKD).
• •Nonoccupational physical activity was inversely associated with the risk of CKD.
• •Both low-intensity and moderate- to vigorous-intensity physical activity were inversely associated with the risk of CKD.
• •Physical activity could be a target for intervention in CKD, no matter what type or intensity.

Abstract

Background

Information on the association between physical activity (PA) and the risk of chronic kidney disease (CKD) is limited. We aimed to explore the associations of total, domain-specific, and intensity-specific PA with CKD and its subtypes in China.

Methods

The study included 475,376 adults from the China Kadoorie Biobank aged 30–79 years during 2004–2008 at baseline. An interviewer-administered questionnaire was used to collect the information about PA, which was quantified as metabolic equivalent of task hours per day (MET-h/day) and categorized into 4 groups based on quartiles. Cox regression was used to analyze the association between PA and CKD risk.

Results

During a median follow-up of 12.1 years, 5415 incident CKD cases were documented, including 1159 incident diabetic kidney disease (DKD) cases and 362 incident hypertensive nephropathy (HTN) cases. Total PA was inversely associated with CKD risk, with an adjusted hazard ratio (HR, 95% confidence interval (95%CI)) of 0.83 (0.75–0.92) for incident CKD in the highest quartile of total PA as compared with participants in the lowest quartile. Similar results were observed for risk of DKD and HTN, and the corresponding HRs (95%CIs) were 0.75 (0.58–0.97) for DKD risk and 0.56 (0.37–0.85) for HTN risk. Increased nonoccupational PA, low-intensity PA, and moderate-to-vigorous-intensity PA were significantly associated with a decreased risk of CKD, with HRs (95%CIs) of 0.80 (0.73–0.88), 0.85 (0.77–0.94), and 0.85 (0.76–0.95) in the highest quartile, respectively.

Conclusion

PA, including nonoccupational PA, low-intensity PA, and moderate-to-vigorous-intensity PA, was inversely associated with the risk of CKD, including DKD, HTN, and other CKD, and such associations were dose dependent.

Graphical abstract

1. Introduction

Chronic kidney disease (CKD) has been a major public health problem worldwide and is closely associated with the risk of a range of adverse outcomes, including cardiovascular events and death. It is predicted that CKD will rise from 16th to 5th place among the leading causes of premature death between 2016 and 2040.¹ In 2016, the global prevalence of CKD hit 13.4%. In recent years, the incidence of CKD has been increasing in developing countries due to the significant increase in non-communicable diseases, particularly diabetes and hypertension.² According to the Chinese Kidney Disease Scientific Report, CKD includes 5 major subtypes, among which diabetic kidney disease (DKD) and hypertensive nephropathy (HTN) are prevalent in urban areas of China.³

Previous studies have suggested that physical activity (PA) is of significance for the prevention of CKD.^4,5 The MJ longitudinal cohort study conducted in Taiwan, China reported that habitual PA was inversely associated with CKD, with a hazard ratio (HR, 95% confidence interval (95%CI)) of 0.91 (0.85–0.96) in the high PA level (≥16.50 metabolic equivalent of task hours per week (MET-h/week)) compared to the very low PA level (<3.75 MET-h/week).⁵ However, prospective studies on the association between PA and CKD risk remain insufficient and inconsistent.6, 7, 8 PA was not associated with CKD in the Framingham Heart Study,⁶ whereas the Tehran Lipid and Glucose Study observed an adverse effect of PA on CKD among males.⁸ In addition, existing studies mainly focus on leisure-time PA.^4,9 Nevertheless, occupational PA accounts for a large proportion of daily PA, and low-intensity PA (LPA) is feasible and sustainable.^10,11 Although the PA paradox suggests that occupational PA may be detrimental,¹² Yamamoto et al.¹³ showed a 12% lower risk of CKD (HR = 0.88, 95%CI: 0.86–0.96) when comparing standing/walking at work to sedentary behavior. Thus, further studies on different domains and intensities of PA are needed, especially in China where evidence is more limited.

Therefore, this study aimed to explore the association between total, domain-specific, and intensity-specific PA and CKD risk based on data from the China Kadoorie Biobank (CKB) study, so as to help take appropriate measures to improve the prevention of CKD.

2. Methods

2.1. Study population

The CKB study is an ongoing population-based prospective cohort of over half a million Chinese adults. Details of the study design and methods of the CKB have been reported previously.^14,15 In brief, the study recruited 512,724 participants aged 30–79 years from 10 geographically diverse regions (5 urban and 5 rural) across China during 2004–2008 at baseline. In addition, about 5% of the surviving participants were randomly selected for resurveys every 4–5 years after completing the baseline survey. The study was approved by the Ethical Review Committee of the Chinese Center for Disease Control and Prevention in Beijing, China and the Oxford Tropical Research Ethics Committee at the University of Oxford, UK. All participants provided written informed consent before taking part in the study.

In the present study, we excluded participants who were previously diagnosed by a physician with CKD (n = 7574), coronary heart disease (n = 15,472), stroke (n = 8884), or cancer (n = 2578) at baseline in order to minimize their reverse causality on PA levels (Some participants have one or more disease.). We also excluded those with implausibly small, large, or conflicting levels of PA (n = 6185), those with missing data for body mass index (BMI; n = 2), and those lost to follow-up shortly after baseline (n = 1). After exclusions, 475,376 participants remained for the final analysis.

2.2. Assessment of PA

The questions on PA and sedentary leisure time were adapted from validated questionnaires in previous studies,^16,17 with additional modifications after a CKB pilot study. An interviewer-administered questionnaire was used during the baseline survey and subsequent resurveys to collect information about the type, frequency, and duration of PA in 4 domains (occupation, commuting, housework, and leisure-time activity) as well as about the number of leisure hours spent on sitting activities per week over the previous 12 months. Detailed information on the questionnaire can be found in previous publications (Supplementary Table 1).¹¹ An MET was assigned to each PA to assess its intensity level according to the updated 2011 Compendium of Physical Activity.¹⁸ Each PA level was calculated by multiplying the MET value for that activity by the hours spent on that activity per day (MET-h/day). Domain-specific PA level was calculated by summing all the MET-h/day spent on occupational or nonoccupational (commuting, housework, and leisure time) PA. Similarly, intensity-specific PA level, which includes LPA (1.5–2.9 METs) and moderate-to-vigorous-intensity PA (MVPA; moderate, ≥2.9–5.9 METs; vigorous, ≥5.9 METs), was calculated in the same way. The total PA level was obtained by summing PA levels in all domains or intensities.

About 1300 participants completed the same questionnaire twice within 1.5 years after the baseline survey. We tested the reproducibility of total PA between the 2-round surveys. The intraclass correlation coefficient was 0.59.¹⁹

2.3. Assessment of covariates

Eligible participants completed an interviewer-administered laptop-based questionnaire on their sociodemographic characteristics (age, sex, education, occupation, annual household income, and marital status), lifestyle factors (tobacco smoking; alcohol consumption; and consumption of red meat, fresh vegetables, and fresh fruits), and personal medical history (hypertension and diabetes) at baseline. Height, body weight, waist circumference, and blood pressure were measured by trained staff using calibrated instruments at baseline. All participants provided a 10 mL non-fasting blood sample for a quick on-site test of random plasma glucose. BMI was calculated as weight (kg) divided by height square (m²). According to the National Health Commission of the People's Republic of China, prevalent hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, self-reported doctor-diagnosed hypertension, or self-reported use of antihypertensive drugs at baseline. Prevalent diabetes was defined as fasting blood glucose ≥ 7.0 mmol/L, random blood glucose ≥ 11.1 mmol/L, or self-reported doctor-diagnosed diabetes.

2.4. Assessment of CKD

Incident CKD cases were obtained through linkage to the local mortality and disease surveillance points system and the national health insurance system and supplemented by annual active confirmation through visiting local communities or directly contacting participants. All events were coded using the International Classification of Disease, 10th Revision by trained staff blinded to the baseline information of participants. In the present study, we identified CKD cases with major subtypes (DKD and HTN) according to China Kidney Disease Network³ and added CKD (International Classification of Disease, 10th Revision: N18) without unspecified subtypes as other CKD²⁰ (Supplementary Table 2). Participants were followed from the date of baseline questionnaire completion to the date of diagnosis of CKD, death, loss to follow-up, or December 31, 2018, whichever came first.

2.5. Statistical analysis

The exposure, which included total, domain-specific, and intensity-specific PA, was categorized into 4 groups based on their quartiles. Means or percentages of baseline characteristics were calculated across total PA categories using linear regressions for continuous variables or logistic regressions for categorical variables and adjusting for age, sex, and study area as appropriate.

Cox proportional hazards models were used to estimate the HRs and 95%CIs for the associations between PA and CKD incidence, with age as the underlying time scale and stratified by age at baseline (in 5-year intervals) and 10 study areas. Multivariate models were adjusted for age at baseline; sex; education (primary school and below, middle school and high school, or college and above); occupation (manual, non-manual, or not working); annual household income (<RMB10,000; RMB10,000–19,999; >RMB19,999); marital status (married or unmarried); smoking status (never or occasional, ex-regular, and among current daily smokers (cigarettes/day)): <15, 15–24, >24); alcohol consumption (never or occasional, ex-regular, weekly, and among daily drinkers (g/day): <15, 15–29, >29–59, >59); consumption frequencies of red meat, fresh vegetables, and fresh fruits (0, 0.5, 2.0, 5.0, 7.0 days/week, continuous); sedentary leisure time (continuous); BMI (continuous); prevalent hypertension and diabetes (presence or absence). For domain-specific (occupational or nonoccupational) and intensity-specific (LPA or MVPA) PA analysis, these domains or intensities were included in the adjusted models. In addition, to examine the linear trend of the association, the median of each category of PA was included in the model as a continuous variable.

A single measurement of PA tends to underestimate the actual association of the usual PA with CKD risk because of within-person variation or measurement error.²¹ Repeat measurement of PA at the second resurvey among approximately 20,000 participants was used to correct for regression dilution bias. The mean usual PA in each baseline category was assigned as the mean PA of the second resurvey in the corresponding category.²²

Subgroup analysis was conducted to examine whether the association differed by age (<60 or ≥60 years), sex (male or female), region (rural or urban), tobacco smoking (current daily, or not current daily), alcohol consumption (current weekly or not current weekly), BMI (<24.0 or ≥24.0 kg/m²), sedentary leisure time (<3 h or ≥3 h), or prevalence of hypertension (no or yes) and diabetes (no or yes) at baseline. The p values for interaction were corrected using the false discovery rate.

Sensitivity analyses excluded participants who had CKD in the first 2 years and 5 years of follow-up as well as those who had respiratory diseases or diabetes at baseline. We additionally adjusted for waist circumference or self-rated health status at baseline. We further regrouped PA categories using quantiles by sex. In addition, we conducted a competing risk analysis defining mortality as the competing risk.

The statistical analyses were conducted using Stata 15.0 (StataCorp., College Station, TX, USA). Statistical significance was set at 2-tailed p_trend < 0.05.

3. Results

3.1. Baseline characteristics of participants by total PA

Among all 475,376 participants, 40.62% were men, 43.12% resided in urban areas, and the age at baseline was 51.47 ± 10.51 years (mean ± SD). The baseline total PA was 21.70 ± 13.75 MET-h/day. The mean values in the lowest (Q1), second (Q2), third (Q3), and highest (Q4) quartiles were 7.09, 14.31, 24.12, and 41.30 MET-h/day, respectively. Participants who were more physically active were more likely to be young, rural residents, agricultural and industrial workers, current daily smokers, and to have lower sedentary leisure time, BMI, and prevalence of hypertension and diabetes (Table 1).

Table 1.

Characteristics of study participants by baseline total physical activity.

Characteristics	Baseline total physical activity (MET-h/day)				ptrend
	Q1	Q2	Q3	Q4
Participant (n)	118,889	118,895	118,812	118,780
Mean age (year)	57.13	52.42	49.15	47.16	<0.001
Women (%)	59.25	65.94	60.43	51.88	<0.001
Urban residents (%)	51.43	49.01	37.74	34.28	<0.001
Middle school and above (%)	49.70	50.98	49.52	46.51	<0.001
Annual household income >RMB19,999 (%)	39.88	46.13	44.28	41.83	0.652
Agricultural and industrial workers (%)	30.88	51.32	69.97	76.04	<0.001
Married (%)	90.32	91.20	91.27	91.24	<0.001
Current daily smokers (%)	28.69	28.79	29.18	29.76	<0.001
Current weekly drinkers (%)	13.70	15.41	15.66	15.40	<0.001
Regular consumption of foods (%)a
Meat	47.75	48.02	45.71	47.46	0.011
Fresh vegetables	98.27	98.75	97.62	98.63	<0.001
Fresh fruits	29.00	30.18	26.34	25.12	<0.001
Sedentary leisure time (h/day)	3.49	3.17	2.71	2.59	<0.001
BMI (kg/m2)	23.83	23.74	23.44	23.36	<0.001
Hypertension (%)	35.09	34.65	32.30	32.21	<0.001
Diabetes (%)	6.52	5.60	4.42	3.83	<0.001

Notes: The quartiles of total physical activity were Q1: 0–11.05, Q2: >11.05–18.23, Q3: >18.23–30.71, Q4: >30.71 MET-h/day, respectively. Values are means or percentages of participants adjusted for age, sex, and region, where appropriate.

Abbreviations: BMI = body mass index; MET-h/day = metabolic equivalent of task hours per day.

^aValues indicated the consumption frequency as “≥4 days/week”.

3.2. Association of total PA with CKD incidence

During a median of 12.1 years (interquartile range = 1.95 years; 5.6 million person-years) of follow-up, we documented 5415 incident CKD cases, including 1159 DKD cases, 362 HTN cases, and 4280 other CKD cases. Upon multivariate adjustment, total PA was inversely associated with the risk of incident CKD in a dose–response manner (Fig. 1 and Supplementary Table 3). Compared with participants in Q1 of total PA, the adjusted HRs (95%CIs) for incident CKD in Q2–Q4 were 0.87 (0.81–0.94), 0.86 (0.79–0.95), and 0.83 (0.75–0.92), respectively (p for trend = 0.005). Similarly, the adjusted HRs (95%CIs) in Q4 were 0.75 (0.58–0.97) for DKD (p for trend = 0.037) and 0.56 (0.37–0.85) for HTN (p for trend = 0.010).

Fig. 1.

Associations of usual total physical activity with risk of CKD. Usual total physical activities in different categories were 13.07, 15.89, 20.13, and 24.75 MET-h/day, respectively. Models were stratified by baseline age groups and study regions and adjusted for age; sex; education; occupation; household income; marital status; tobacco smoking; alcohol consumption; sedentary leisure time; consumption frequency of red meat, fresh vegetables, and fresh fruits; BMI; and prevalent hypertension and diabetes at baseline. The squares represent HRs, and the vertical lines represent 95%CIs. The dashed lines represent a linear trend of the association between usual total physical activity and the risk of CKD. The numbers above the vertical lines are point estimates for HRs, and the numbers below refer to the number of events. 95%CI = 95% confidence interval; BMI = body mass index; CKD = chronic kidney disease; DKD = diabetic kidney disease; HR = hazard ratio; HTN = hypertensive nephropathy; MET-h/day = metabolic equivalent of task hours per day.

The associations of total PA with incident CKD and its subtypes remained stable after exclusions, additional adjustments to covariates, and categorizing the PA levels by sex. The results did not differ substantially after a competing risk analysis defining mortality as the competing risk (Supplementary Table 4).

3.3. Association of domain-specific PA with CKD incidence

For occupational PA, only participants in Q4 had a lower risk of incident HTN (HR = 0.42, 95%CI: 0.21–0.84), whereas increased nonoccupational PA was significantly associated with a decreased risk of CKD, DKD, and HTN incidence, with an HR (95%CI) of 0.80 (0.73–0.88) for CKD, 0.76 (0.63–0.93) for DKD, and 0.69 (0.48–0.99) for HTN in Q4 as compared to Q1 (Fig. 2 and Supplementary Table 5).

Fig. 2.

Associations of domain-specific physical activity with risk of CKD. Usual occupational physical activities in different categories were 2.56, 7.84, 12.03, and 17.11 MET-h/day, and usual nonoccupational physical activity in different categories were 6.22, 8.20, 9.77, and 11.11 MET-h/day, respectively. Models were stratified by baseline age groups and study regions and adjusted for age; sex; education; occupation; household income; marital status; tobacco smoking; alcohol consumption; sedentary leisure time; nonoccupational physical activity (or occupational physical activity); consumption frequency of red meat, fresh vegetables, and fresh fruits; BMI; and prevalent hypertension and diabetes at baseline. The squares represent HRs, and the vertical lines represent 95%CIs. The dashed lines represent a linear trend of the association between usual domain-specific physical activity and the risk of CKD. The numbers above the vertical lines are point estimates for HRs. 95%CI = 95% confidence interval; BMI = body mass index; CKD = chronic kidney disease; DKD = diabetic kidney disease; HR = hazard ratio; HTN = hypertensive nephropathy; MET-h/day = metabolic equivalent of task hours per day.

3.4. Association of intensity-specific PA with CKD incidence

For intensity-specific PA, increased LPA was associated with a decreased risk of CKD and DKD, with an HR (95%CI) of 0.85 (0.77–0.94) for CKD and 0.75 (0.60–0.94) for DKD in Q4. Similarly, the HR (95%CI) in Q4 of MVPA was 0.85 (0.76–0.95) for CKD, 0.77 (0.60–1.00) for DKD, and 0.60 (0.38–0.92) for HTN (Fig. 3 and Supplementary Table 6).

Fig. 3.

Associations of intensity-specific physical activity with risk of CKD. Usual LPA in different categories were 6.44, 8.73, 9.85, and 10.75 MET-h/day, and usual MVPA in different categories were 5.65, 6.90, 9.29, and 16.52 MET-h/day, respectively. Models were stratified by baseline age groups and study regions and adjusted for age; sex; education; occupation; household income; marital status; tobacco smoking; alcohol consumption; sedentary leisure time; MVPA (or LPA); consumption frequency of red meat, fresh vegetables, and fresh fruits; BMI; prevalent hypertension and diabetes at baseline. The squares represent HRs, and the vertical lines represent 95%CIs. The dashed lines represent a linear trend of the association between usual intensity-specific physical activity and the risk of CKD. The numbers above the vertical lines are point estimates for HRs. 95%CI = 95% confidence interval; BMI = body mass index; CKD = chronic kidney disease; DKD = diabetic kidney disease; HR = hazard ratio; HTN = hypertensive nephropathy; LPA = low-intensity physical activity; MET-h/day = metabolic equivalent of task hours per day; MVPA = moderate-to-vigorous-intensity physical activity.

3.5. Subgroup analysis

In the analysis of the association between total PA and CKD, no statistically significant heterogeneity was observed for these strata (all false discovery rate corrected p for interaction > 0.05) (Supplementary Table 7).

4. Discussion

In this large-scale, population-based cohort study in the Chinese population, total PA was inversely associated with CKD incidence and its major subtypes, including DKD and HTN. Such associations were similar for nonoccupational PA, and both LPA and MVPA were also inversely associated with CKD.

In line with a previous meta-analysis showing that people with the highest level of PA had a 16% lower risk of CKD compared to people with the lowest level of PA,²³ our results found a comparable magnitude of risk in a smaller range of total PA (Q4: ≥24.75 MET-h/day vs. Q1: <13.07 MET-h/day). This suggests total PA might be an important modifiable risk factor for CKD. Comparable results were demonstrated with incident end-stage renal disease in a study using similar assessment methods for PA.²⁴ In contrast, some studies have observed either an increased risk or no effect associated with total PA on CKD.^6,8 The inconsistency is due at least partly to different definitions of PA and to the limited statistical power of a relatively small sample size (fewer than 10,000 participants). Thus, more studies with a detailed assessment of total PA are necessary to confirm our findings.

This study explored the association between different domains of PA and CKD risk, including occupational PA, which few studies have looked at to date. Similar to the PA paradox,¹² our findings showed no significant association between occupational PA and CKD, probably due to the low intensity and long duration of occupational PA as well as the fact of the increase in heart rate and blood pressure for 24 h. Fortunately, our study reported no harmful effects of occupational PA on CKD, thus normal levels of occupational PA can be maintained. Our results indicated that nonoccupational PA played a more critical role in CKD risk among participants. Similarly, Kronborg et al.²⁵ showed that high PA during non-working hours among women predicted an increase in estimated glomerular filtration rate. Previous studies focusing on PA during leisure time^4,5,9 and commuting²⁶ also demonstrated inverse associations with CKD. Therefore, we can say that sufficient nonoccupational PA, including leisure-time, household, and commuting physical activities, is beneficial for preventing CKD.

According to recommendations from the World Health Organization,²⁷ the intensity of PA is also critical to health. Jafar et al.²⁸ reported that a 24% lower risk of end-stage kidney disease was associated with MVPA compared to those with no PA, and the association appeared to be dose dependent with higher intensities of PA. Our results confirmed a beneficial association with MVPA and, furthermore, found an association with LPA, which suggests that for those unable to engage in MVPA, LPA is a viable alternative for lowering CKD risk.

Previous studies have demonstrated a significantly lower risk of DKD among people with high levels of PA,29, 30, 31 but these studies were mainly focused on leisure-time PA. Our study went further, finding that total PA, including nonoccupational PA, LPA, and MVPA, was inversely associated with DKD risk. This should motivate individuals, especially those with diabetes, to maintain sufficient and regular PA. As for HTN risk, there is limited evidence. Our study showed PA to be associated with a lower risk of HTN. Chang et al.³² reported a similar result in hypertensive patients and demonstrated that a higher risk of CKD was significantly associated with low (odds ratio = 1.39; 95%CI: 1.01–1.90) and moderate levels of PA (odds ratio = 1.39; 95%CI: 1.04–1.86) as compared to a high level of PA.

There are several biological pathways that could be contributing to the beneficial effects of PA on CKD.³³ First, PA has been shown to regulate the synthesis and degradation of nitric oxide to maintain normal endothelial and renal function.³⁴ Second, PA can reduce insulin resistance, which was associated with a higher incidence of CKD in the Atherosclerosis Risk in Communities Study.^35,36 Third, systematic inflammation can lead to albuminuria and decreased renal function, while exercise may reverse the inflammatory process by reducing inflammatory cytokines such as leptin, tumor necrosis factor-α, interleukin-1, and interleukin-6 and by cultivating an anti-inflammatory environment.³⁷ Diabetes and hypertension may lead to oxidative stress, insulin resistance, and endothelial dysfunction, whereas PA has beneficial effects in terms of these metabolic disorders.²⁴

Our study was a prospective cohort study with a large sample size and detailed information about domain-specific and intensity-specific PA. Additionally, we calculated usual PA using data from the second resurvey to adjust for possible regression dilution bias in our analysis.

However, some limitations should be considered. First, PA was assessed by a self-report questionnaire. Therefore, recall bias may be unavoidable, and self-recall of LPA is generally poorer than that of MVPA. Second, although we excluded participants with previously physician-diagnosed CKD, coronary heart disease, stroke, or cancer at baseline, reverse causality may still exist due to the long latency period of CKD and other underlying disorders. However, the association remained unchanged in the sensitivity analysis following the exclusion of participants whose CKD occurred in the first 2 and 5 years of follow-up and those who had respiratory diseases at baseline. Third, although established and potential risk factors for CKD were adjusted for, the results may still suffer from residual confounding due to unknown or unmeasured factors.

5. Conclusion

Our study demonstrated that higher level of PA was associated with a lower risk of CKD in a dose–response manner. Additionally, nonoccupational PA, LPA, and MVPA were all inversely associated with CKD risk. Therefore, PA may be a target for intervention in CKD. We recommend people increase their levels of PA, no matter what type or intensity, to prevent CKD.

Data sharing

The access policy and procedures are available at www.ckbiobank.org.