Effects of exercise on kidney and physical function in patients with non-dialysis chronic kidney disease: a systematic review and meta-analysis

Patients with non-dialysis chronic kidney disease (CKD) are at greater risk of early mortality and decreased physical function with an advance in the stage of CKD. However, the effect of exercise in these patients is unclear. This meta-analysis aimed to determine the effects of physical exercise training on the risk of mortality, kidney and physical functions, and adverse events in patients with non-dialysis CKD. The meta-analysis conformed to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement and the Cochrane Handbook recommendations. On 16 August 2019, the PubMed, CINAHL, Cochrane Library databases, and Embase were electronically searched, with no restrictions for date/time, language, document type, or publication status, for eligible randomized controlled trials (RCTs) investigating the effects of exercise on mortality and kidney and physical function in patients with non-dialysis CKD. Eighteen trials (28 records), including 848 patients, were analyzed. The effects of exercise on all-cause mortality and estimated glomerular filtration rate were not significantly different from that of usual care. Exercise training improved peak/maximum oxygen consumption compared to usual care. Regular exercise improves physical and walking capacity for patients with non-dialysis CKD. Effect on leg muscle strength was unclear.


Results
In total, 3784 records were identified after the removal of duplicates, and 45 records remained after the screening of titles and abstracts. Further, 17 records were excluded based on the full eligibility criteria. In total, 18 trials (28 records) 8,15, including 848 patients with non-dialysis CKD who had met the eligibility criteria of this review were included in the analysis (Fig. 1). The characteristics of the trials included in this review are described in Table 1. Overall, the records demonstrated a broad range of follow-up duration (median follow-up = 20.5 weeks; range 8−72 weeks) and CKD stage of the trial population (CKD stages 3-4, 9 trials; stages 2-4, 4 trials; stage 3, 1 trial; stages 3-5, 1 trial; stages 1-3, 1 trial; not reported, 2 trials). Categorizations of the types of exercise training in the trials were as follows: center-based exercise = 9 trials, home-based exercise = 4 trials, combined both center and home-based exercise = 6 trials, aerobic exercise = 8 trials, resistance exercise = 2 trials, and combined both aerobic and resistance training = 8 trials.
Quality assessment. The results of the risk of bias assessment in all the trials are summarized in Table 2.
All participants were classified into the exercise training and the usual care groups. Blinding of participants was not possible due to the nature of exercise training. Three trials 17,32,36 showed high risks of bias related to the   Kidney function (eGFR, Scr). Nine trials including 459 participants reported eGFR as outcomes. eGFRs were mostly evaluated using the CKD-Epidemiology Collaboration (EPI) creatinine equation 18,20,28 or the Modification of Diet in Renal Disease (MDRD) formula 8,16,17,19,30,36 . The results showed that the effect of exercise training on eGFR was not significant compared to usual care (mean difference (MD), − 0.34; 95% CI − 1.91 to 1.22; participants = 459, trials = 9; I 2 = 0%) (Fig. 3). Subgroup analysis performed for exercise type, eGFR (< 30 or ≥ 30 mL/ min/1.73 m 2 ), and BMI (< 30 or ≥ 30 kg/m 2 ) showed no evidence of differences between the groups (test for subgroup difference: P = 0.88, 0.19, and 0.58, respectively). Subgroup analysis was not performed for other variables because an I 2 > 50% was not obtained. Only 5 trials including 231 participants reported serum creatinine (Scr) (μmol/L) as outcomes. There was no evidence of effects of physical exercise interventions on Scr improvement compared to usual care (MD, 1.48; 95% CI − 7.50 to 1.31; participants = 231, trials = 10; I 2 = 0%) (see Supplementary Fig. S1 online).   (Fig. 4), although high heterogeneity was detected (P = 0.005, I 2 = 60%). Subgroup analysis performed for each exercise type, eGFR (< 30 or ≥ 30 mL/min/1.73 m 2 ), length of exercise intervention, and the    45-13.42; participants = 323, trials = 7; I 2 = 37%); additionally, heterogeneity was lower in subgroup analysis than in pre-subgroup analysis (see Supplementary Table S1 online). In the meta-regression analyses of 9 trials (excluding one trial 39 due inadequate BMI data), a significant association was observed between the MD of peak/max VO 2 and BMI [slope:-0.555 BMI (95% CI − 0.925 to 0.186), P = 0.009] (see Supplementary Fig. S5 online). The length of exercise intervention and eGFR were not significantly associated with the MD of peak/max VO 2 (P = 0.301 and 0.713, respectively).
Adverse events. Among the included records, 12 trials reported of adverse events related to exercise training; eleven trials showed no adverse event occurred during exercise training and tests. However, only 1 trial 21 reported 11 adverse events possibly related to the study (6 cases of hypotension, 1 case of knee pain, 1 rapid atrial fibrillation case, 1 case of Achilles tendon pain, 1 case of joint pain, and 1 case of chest pain).  Fig. S4 online).

Effects of interventions
Sensitivity analysis and publication bias. Sensitivity analysis was performed on the exclusion of studies for high risks of bias in the overall results, and there were no changes evident compared to the overall results. Funnel plots showed that trials evaluating eGFR were symmetrically distributed; on the contrary, the distribution of the plot in trials using peak/max VO 2 was slightly asymmetrical (see Supplementary Figs. S6, S7 online). In Egger's test, no significant publication bias was observed in trials using eGFR and peak/max VO 2 (P = 0.955 and 0.261, respectively).

Discussion
To the best of our knowledge, this is the first meta-analysis assessing the effects of physical exercise training on the risk of mortality, physical and kidney function, and adverse events exclusively in patients with non-dialysis CKD. The main findings of this review revealed that the effect of exercise training on all-cause mortality and kidney function could not be established in patients with non-dialysis CKD, while exercise training improved physical and walking capacity. A previously reported systematic review which had conducted meta-analysis showed that exercise training significantly improved eGFR compared with usual care in patients with non-dialysis CKD 43 . However, the review had included non-RCT studies 44,45 , and some study participants were included more than twice in the meta-analysis; this may have resulted in selection bias and the overestimation of the effect of exercise on kidney function. Some studies showed that exercise training improved vascular function, attenuated the increase in sympathetic nervous system activity, and reduced blood pressure in patients with non-dialysis CKD 17,18,28,33,35,38,44 . This supports the hypothesis that exercise training could delay the decline in kidney function. However, in our review, the effect of exercise training with moderate intensity on the rate of kidney function decline was found to be inconclusive. The duration of exercise intervention in the included trials may have been insufficient to demonstrate improvement in mortality rates and kidney function. As for non-RCT studies, a retrospective longitudinal cohort study 46 reported that the completion of renal rehabilitation consisting of aerobic and resistance training for a 12-week period was associated with longer event-free survival during the follow-up period (median 34 months). Also, an observational study 47 showed that substitution of sedentary activity with light activity, but not with exercise training, was associated with a lower hazard of death in the CKD group. Similarly, a previous study showed that muscle mass and physical activity affected SCr rather than cystatin C; thus, use of cystatin may be an adequate alternative to assess renal function 48 . Two trials 19,28 included in our review measured cystatin; however, a meta-analysis could not be performed due to the limited number of trials. A future study assessing the effect of longer duration exercise training on mortality and kidney function based on cystatin levels is required. An increase in adverse events in the exercise training group compared with the usual care group could not be determined. Of all studies in this review, only one reported adverse event related to exercise Scientific Reports | (2020) 10:18195 | https://doi.org/10.1038/s41598-020-75405-x www.nature.com/scientificreports/ training 21 . Therefore, a meta-analysis of adverse events could not be conducted; there may have been low rates of adverse events in exercise training overall. Regular exercise training improved physical and walking capacity in patients with non-dialysis CKD; this was consistent with the results of a previous systematic review of the effect of exercise training in patients with CKD, including dialysis patients, kidney transplant patients, and heart failure patients 13,[49][50][51] . Common symptoms of these chronic diseases (CKD including non-dialysis, dialysis, and kidney transplant patients, DM, and heart failure) were loss of muscle strength, lack of physical activity, and reduced physical capacity. A previous study found that physical capacity (e.g., peak/max VO 2 ) was related to the mortality of patients with CKD 52 , suggesting that exercise training increased physical capacity and benefited patients with CKD. The results showed that lower BMI at baseline predicted greater improvements in peak/max VO 2 , while other factors were not significantly associated with the effect of exercise training.
In a previous study that included patients with heart failure, BMI was not associated with an improvement in physical capacity 53 , which is in contrast with our study result. However, explaining the association between BMI and improvement in physical capacity was difficult because only univariate meta-regression analysis was performed in this study due to the small trial sample size. This limits a concurrent consideration of the influence of other factors. Furthermore, adherence to exercise training was reported in only 67% of trials. Adherence rate for exercise training ranged from 63% to 96.9%, possibly affecting its influence on peak/max VO 2 . Further studies should assess the relationships between the effect of peak/max VO 2 and other factors.
Leg muscle strength is an important marker of physical function that predicts mortality in patients with CKD receiving dialysis 54 . However, the effect of resistance exercise training on leg muscle strength was not significant because of the small number of trials involving patients with non-dialysis CKD. Previous reviews showed that progressive resistance training significantly improved standardized muscular strength in patients with CKD on dialysis 49 . Further research is required to determine whether resistance training improves leg muscle strength in patients with non-dialysis CKD.
The generalizability of this review was limited by age and cause of kidney disease. CKD is more common in people aged 65 or more years 1,55 , and diabetes and high blood pressure have been considered as causes of kidney disease 6,56,57 . However, the approximate mean age of participants in the included trials ranged from 50 to 65 years. The number of older adults may increase in the future; therefore, further studies should assess the effect of exercise training on elderly patients with CKD.
There are some limitations to this review. First, complete data were not obtained because there were missing data in some trials, despite efforts in reaching out to the authors. For this reason, there is a possibility of presence of predicted direction of bias. Secondly, some trials have a high risk of bias, especially those related to the predicted direction of bias, because of missing outcome data and information bias due to the absence of blinding. More high-quality RCTs are needed to clarify the effects of exercise training. Trials included in our review were mostly studies with short durations of intervention and follow-up periods were less than 1 year. Thus, the duration of exercise intervention may have been insufficient to show an improvement in the mortality rates and a significant association with exercise training in the meta-regression analysis. Our systematic review did not include non-RCTs, because RCTs are more likely to provide unbiased information about the differential effects of alternative health interventions (clearly defined exercise training or usual care) than non-RCTs. Therefore, inclusion of non-RCTs of good quality with longer follow-up periods could potentially alter the results. Finally, adherence to exercise training was not reported in 33% of trials, and this may have biased the effect of exercise on kidney and physical functions.

Conclusion
Regular aerobic and/or resistance training improves physical and walking capacity for patients with non-dialysis CKD. The effect on mortality, kidney function, and leg muscle strength is inconclusive. Furthermore, few adverse events related to exercise training were reported, suggesting that regular exercise training with moderate intensity for 8 weeks to 1.5 years may be safe for patients with non-dialysis CKD. Future studies and multi-center RCTs with larger sample sizes and cohorts of elderly people are needed to focus on the effect of resistance training in non-dialysis CKD.

Methods
Protocol and registration. The protocol was registered on UMIN Clinical Trial Registry (UMIN ID000039799). The meta-analysis was performed following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement 58,59 . A systematic review was conducted in agreement with the recommendations stated in the Cochrane Handbook 60 .
Eligibility criteria. Types of study. We included all RCTs, cluster-RCTs, and cross-over trials that investigated the effects of physical exercise interventions on physical function, kidney disease, and mortality of patients with non-dialysis CKD. RCTs without appropriate control groups, including those lacking usual care treatment arms, were excluded. Similarly, quasi-RCTs were excluded because their allocation of participants to treatments is not randomized.
Participants. Adult participants older than 18 years of age who were diagnosed with CKD were excluded. CKD was defined according to the Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease 2 . CKD was defined based on the following criteria: (1) abnormal kidney structure or function and (2) presence of CKD for more than 3 months with impaired health status. Abnormal kidney function was defined as decreased GFR (< 60 mL/min/1.73 m 2 ) or detection of one or more abnormalities for markers of kidney damage,

Types of interventions.
Hospital-based or home-based exercise interventions were included if supervised by health professionals or self-training. Similarly, different types of exercise, such as resistance training, aerobic exercise, or both, were included. The interventions were compared to control interventions, such as usual care or no-exercise care, consisting of medical care. Studies with exercise interventions clearly defined for frequency (at least once a week), intensity (using percentage of peak workload/oxygen uptakes, anaerobic threshold, or Borg scales), or duration of exercise (more than one month) were included. Abnormal types of exercise were equally included.
Outcome measures. The primary outcomes were as follows: (1)  Screening the studies. Two authors (KN, TS) independently screened all titles and abstracts for all potential studies against the inclusion criteria. Full reports were obtained for all titles that appeared to meet the inclusion criteria and for those wherein any uncertainty was observed. Subsequently, the two authors screened the full-text reports to determine whether these articles met the inclusion criteria. Reasons for the exclusion of ineligible studies were identified. In case of disagreements, a third reviewer (SY) provided comments and made a final decision. The entire screening process was recorded, and the study selection process is described in the PRISMA flow chart (Fig. 1).

Data extraction.
The two reviewers conducted the data extraction from eligible articles according to the recommendations stated in the PRISMA statement 58,59 . Disagreements were resolved by the third reviewer. Study characteristics and clinical outcome measures were extracted. The extracted data included general information (authors, year of publication, location), participant characteristics (sample size, inclusion/exclusion criteria, randomization process and allocation, mean age, gender, and percentage of patients with diabetes), interventions (the type of intervention, intensity, duration, and frequency), outcome measures (all-cause mortality, kidney function, including eGFR and Scr, physical function markers, including peak VO 2 , muscular leg strength, time of TUG, and 6-min walk distance). The corresponding authors of the included publications were contacted for missing data and further information if considered necessary.
Risk of bias. The risk of bias was assessed using the Revised Cochrane risk-of-bias tool for randomized trials (RoB 2) criteria recommended by the Cochrane Handbook of Systematic Reviews of Interventions 60 . The domains for risk of bias are (1) randomization process, (2) deviations from intended interventions, (3) missing outcome data, (4) measurement of outcome data and (5) selection of the reported results. The risk of bias was categorized as low, some concerns, or high. After the judgment of all domains, the overall risk of bias was assessed by the judgment of all domains as low, some concerns, or high risk of bias. The two reviewers independently conducted the risk assessment, and the third reviewer resolved disagreements.
Data synthesis strategy. Statistical analysis was performed using the Review Manager Software (RevMan V.5.3) to combine and calculate the effect size for each outcome according to the recommendations set by the statistical guidelines described in the Cochrane Handbook of Systematic Reviews of Interventions 60 . Meta-analyses of the data were performed if eligible studies were sufficiently clinical and statistically homogeneous. Clinical heterogeneity was assessed by considering the between-study variability for specific factors such as age or type of exercise interventions. Statistical heterogeneity was tested using the Chi-square test and the I 2 statistic. In the event there was substantial heterogeneity between studies (I 2 > 50% or P < 0.1), the study design and characteristics of the studies were examined. The possible causes of heterogeneity were explored by conducting sub-group, meta-regression, or sensitivity analyses. Random effect models were applied when appropriate. A meta-analysis was conducted if data were appropriate. Dichotomous data (mortality) were described using risk ratios (RR) with a 95% confidence interval (CI). Continuous outcomes were analyzed using weighted mean differences (WMD) (with 95% CI) or SMD (95% CI) if different measurement scales were used.
Subgroup, meta-regression, or sensitivity analyses. Sub-group analysis was performed on all primary outcomes as follows: exercise types (center-based or home-based exercise, or a combination of center-and home-based exercise), basal eGFR (< 30 mL/min/1.73 m 2 or ≥ 30 mL/min/1.73 m 2 ), and basal BMI (< 30 kg/m 2 or ≥ 30 kg/m 2 ). For center-based exercises, patients participated in exercise training sessions under the real-time Scientific Reports | (2020) 10:18195 | https://doi.org/10.1038/s41598-020-75405-x www.nature.com/scientificreports/ supervision of professionals either in a hospital or training center. For home-based exercises, patients performed exercise training at home or in a community setting without real-time supervision of professionals. Furthermore, sub-group and meta-regression analyses were performed to explore the causes of heterogeneity among primary outcomes if an I 2 > 50% was obtained. Meta-regression analysis was performed using the Stata 14 software (www.stata .com). The length of exercise intervention (< 24 weeks, 24-48 weeks, or ≥ 48 weeks) and percentage of patients with DM complications (≥ 50% or < 50%) were used as subgroup factors. In the univariate meta-regression model, eGFR, BMI, or length of intervention were used as independent factors.
Sensitivity analysis was performed to explore the sources of heterogeneity, such as the exclusion of studies with a high risk of bias and the evaluation of meaningful changes in the effect size.
Assessment of publication bias. The potential for publication bias was assessed using funnel plots and Egger`s test if more than ten studies were available.

Data availability
We confirm that the data supporting the findings of this review are available within the article and its supplementary materials.