Chronic kidney disease (CKD) is a significant global public health concern, characterized by a progressive loss of kidney function that impairs patients' quality of life, functional status, and increases overall morbidity and mortality. Individuals with CKD, including those not requiring kidney replacement therapy (non–dialysis-dependent CKD, NDD-CKD), often exhibit physical inactivity and reduced exercise capacity.

For patients with NDD-CKD, regular physical activity confers numerous benefits. For instance, aerobic exercise has been shown to improve peak oxygen uptake (VO₂ peak),¹ preserve endothelial function in conduit arteries, and enhance microvascular reactivity. Resistance training, even in patients adhering to low-protein diets, has been shown to increase muscle mass and strength.¹ These factors are critical for mitigating the prevalent issue of sarcopenia in CKD. Beyond its effects on physical parameters, exercise favorably modifies cardiovascular risk factors and psychosocial well-being in patients.

In 2014 and 2015, two key systematic reviews by Heiwe et al² and Barcellos et al³ evaluated exercise therapy in CKD patients, including those on maintenance dialysis. Although these reviews included studies on NDD-CKD patients, many of the included trials focused on dialysis patients. While the number of randomized controlled trials (RCTs) on exercise therapy for NDD-CKD patients has since been increasing, the evidence base for this population remains limited compared to that for dialysis patients⁴.

The aim of this study is to clarify the evidence regarding the effectiveness and the safety of exercise therapy in patients with NDD-CKD. We will achieve this by identifying high-quality studies (only RCTs) from the existing literature and synthesizing the currently available evidence.

Search strategy

To ensure transparency, accountability, and to prevent research duplication, this systematic review and meta-analysis protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration no. CRD42024536731). The methodology for this review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines.⁵ PubMed and Ichushi-Web, a database of Japanese medical literature, were searched. The search strategy was based on terms related to the population and intervention, and was restricted by publication date (supplementary materials 1). A previous systematic review for the Renal Rehabilitation Guidelines⁶ used an identical search strategy and included literature published from January 1, 1970, to April 5, 2017 (Fig. 1). To avoid duplication of this prior work, we limited our search to literature published from March 6, 2017, to February 29, 2024. In addition, we included only RCTs from the reference list of the previous guideline for analysis. The search was updated immediately before the final analysis to incorporate the most current evidence. Only studies published in English were included. Four authors (S.N., A.H., T.H., and S.I.) were divided into two pairs to conduct the literature screening. Each pair independently performed study selection. After the search, all documents were divided between the pairs for a primary screening of titles and abstracts. Documents eligible for secondary screening were then re-divided, and the full text of each document was reviewed by one of the two pairs. Discrepancies were resolved by consensus with two senior authors (T.U. and S.T.), one of whom (S.T.) was the corresponding author. Cohort studies and publications without available full text were excluded.

Eligibility Criteria

Eligible studies were RCTs involving adult patients with CKD stage 1 or greater who were not undergoing dialysis. No restrictions were applied based on gender, race, nationality, or exercise modality. Studies on patients undergoing hemodialysis or peritoneal dialysis, kidney transplant recipients, or adolescents (age < 18 years) were excluded. In this study, we restricted comparators of the included RCTs to those assigned to a sedentary control group or a control group receiving no exercise training.

Data extraction

Data were extracted using a standardized data extraction form. The following information was extracted from each study: first author's surname, publication year, study characteristics (country, design, and sample size), participant characteristics (age, sex, and CKD stage), intervention characteristics (modality, delivery, intensity, duration, and frequency), and outcomes.

Study outcomes

The primary outcomes of this review were all-cause mortality, kidney function, and hospitalization. For kidney function, we used the following surrogate markers: the estimated glomerular filtration rate using serum creatinine (eGFRcr), the urine protein-creatinine ratio (U-PCR), and the urine albumin-creatinine ratio (U-ACR). Secondary outcomes included exercise capacity, ambulatory function, muscle strength, HR-QoL, body mass index (BMI), lipid profiles, hemoglobin A1c (HbA1c), and falls associated with exercises. Effect measures were reported as either ratio measures (e.g., risk ratio, odds ratio) or difference measures (e.g., mean difference, risk difference).

Risk of bias assessment

Four reviewers independently assessed the risk of bias and methodological quality of the included studies. We used the Cochrane Risk-of-Bias Tool (RoB 2)⁷ for risk of bias assessment, the robvis⁸ for making the visual summary.

Data synthesis and analysis

A systematic review and meta-analysis were conducted. For outcomes amenable to meta-analysis, the analysis was performed using Review Manager 5.4. Meta-analyses were conducted using random-effects models, with effect measures calculated from means, standard deviations (SDs), and sample sizes. If two or more studies reported the same variable, a meta-analysis was performed for that variable.

Mean difference (MD) with 95% confidence intervals (CIs) was used for meta-analysis when studies reported outcomes using the same measurement scale. For variables with different measurement scales, standardized mean difference (SMD) with 95% CI was calculated. Statistical heterogeneity was assessed using the I² statistic (inconsistency test), with values greater than 40% considered to indicate substantial heterogeneity.

Analysis of subgroups or subsets

Subgroup analyses were conducted to identify potential explanations for heterogeneity and to provide valuable suggestions for tailoring interventions in clinical practice if differences in outcomes were observed. Obesity was specifically considered because its presence may affect the effectiveness and safety of exercise training. The Japan Society for the Study of Obesity has adopted the BMI of 25 or higher as the standard for obesity classification.⁹ Consequently, this study employed 25 as the cutoff value of BMI for evaluation.

The initial search identified 9131 articles (Fig. 1). After a primary screening of titles and abstracts, 27 RCTs, all from PubMed, were selected for full-text review. During the secondary screening, studies were excluded if they were unsuitable for analysis (e.g., not focusing on NDD-CKD patients or potentially unreliable reporting). In addition, we conducted full-text review on the 20 references of the previous guideline. Finally, a total of 18 RCTs were included in the meta-analyses (Table 1), and additional data were extracted from one post hoc analysis.²¹ This cohort comprised nine RCTs included in the previous guideline^10–18 and nine newly identified RCTs.^{19, 20, 22–28} Of these, 16 studies that reported on the eight primary outcomes were included in the quantitative meta-analysis.^{11–22, 24, 25, 27, 28}

Among the fifteen studies included in the meta-analysis, only two were from Japan,^{20, 25} while the other thirteen were international reports. Participants’ BMI at baseline is greater than 25 in most trials (12 of 15).^{11–18, 21, 22, 24, 28} The duration of the exercise interventions across all studies ranged from 8 to 52 weeks.

Figure 1. Flow diagram of systematic literature search on exercise therapy in patients with NDD-CKD

[ We would like to place Table 1 here, right under Fig. 1. ]

Primary outcomes – mortality, kidney function, hospitalization

No RCTs investigated the effects of exercise intervention on all-cause mortality or hospitalization.

A meta-analysis was performed on eleven RCTs^{10–19, 27} that assessed kidney function as an outcome. This analysis focused on eGFRcr and proteinuria. Ideally, the assessment of kidney function before and after an exercise intervention should be based on eGFR using cystatin C (eGFRcys), as exercise-induced increases in muscle mass can elevate serum creatinine levels, potentially lowering eGFRcr. However, only one RCT¹² reported pre- and post-intervention eGFRcys. Therefore, we conducted a meta-analysis on eGFRcr. While none of the individual results from each study were significant, most (eight out of eleven) studies reported favorable results for eGFRcr in the intervention groups (Fig. 2.1). The pooled analysis suggested a smaller decline of eGFRcr (MD, 0.96 mL/min/1.73 m² ; 95% CI, -1.25 to 3.17). In subgroup analyses, a meta-analysis was conducted, specifically focusing on RCTs of which participants’ mean BMI at baseline was lower than 25. Two RCTs^{15, 18} were included in this subgroup analysis, and were consistent with the results of the previous meta-analysis (Fig. 2.1.2). To assess the degree of proteinuria, we conducted meta-analyses using the SMD. There was no significant difference of both U-PCR and U-ACR (Fig. 2.2).

Figure 2. Forest Plot: Pooled Effect of Exercise on eGFRcr, U-PCR, and U-ACR in NDD-CKD patients

Secondary outcome - exercise capacity, ambulatory function, muscle strength

A meta-analysis of nine RCTs^{10–14, 19, 21, 22, 27} revealed a significant improvement in exercise capacity with exercise intervention. The mean VO₂ peak in the intervention group was higher than the control group (MD, 2.13 mL/kg/min; 95% CI, 0.60 to 3.66) (Fig. 3.1).

A meta-analysis of six RCTs^{11, 14, 19, 23–25} using the 6-minute walk test (6MWT) was performed to evaluate ambulatory function. The walking distance in the intervention group was longer than control group (MD, 51.16 m; 95% CI 29.75 to 72.57) (Fig. 3.2). While the I² value of the forest plot was 27%, the results of individual studies appeared heterogeneous on visual inspection of the forest plot.

We conducted two meta-analyses to assess muscle strength: one on handgrip strength and one on the sit-to-stand (STS) test. The meta-analysis of three RCTs^{15, 18, 25} that measured handgrip strength did not show a significant difference (Fig. 3.3). Similarly, the meta-analysis of three RCTs^{11, 17, 24} that measured the STS test did not show a significant difference in the intervention group, although the slight improvement was found in the intervention group (SMD, 0.86; 95% CI -0.41 to 2.13) (Fig. 3.4). In this meta-analysis, the I² value was 93%, indicating significant heterogeneity, which was also evident from the forest plot.

Figure 3. Forest Plot: Pooled Effect of Exercise on VO₂ peak, 6MWT, handgrip strength, and sit-to-stand test

Legend of Fig. 3: ^a Confidence interval calculated by Wald-type method, ^b Tau² calculated by Restricted Maximum-Likelihood method

Secondary outcome - QoL

The meta-analyses about HR-QoL included data from RCTs that used the Kidney Disease Quality of Life (KDQOL) instrument, the 12-Item Short Form Health Survey (SF-12), the 36-Item Short Form Health Survey (SF-36), or the RAND scoring (RAND-12 or RAND-36). The meta-analysis of three RCTs evaluating the KDQOL instrument^{17, 19, 25} revealed significant improvements of two subscales: Effects of Kidney Disease and Symptom/Problem List (Fig. 4.2, 4.3). The meta-analyses of six RCTs^{14, 15, 19, 24, 25, 27} that used the SF-12, SF-36, or RAND-12 demonstrated significant improvement in the Physical Component Summary (PCS) (Fig. 4.4). Furthermore, the meta-analyses of five RCTs^{11, 15, 17, 18, 25} that used the SF-36 or RAND-36 demonstrated significant improvement in the four categories; General health, Role limitations due to physical health, Vitality, and Bodily pain (Supplementary materials, Figure S1).

Figure 4. Forest Plot: Pooled Effect of Exercise on HR-QoL Across Multiple Subscales in NDD-CKD patients

Legend of Fig. 4: ^a Confidence interval calculated by Wald-type method, ^b Tau² calculated by Restricted Maximum-Likelihood method, ^c Values of Thompson et al for PCS and MCS are physical health composite and mental health composite, respectively.

Secondary outcome - metabolic parameters

The meta-analysis of BMI included eight RCTs,^{12–14, 16, 17, 21, 25, 27} revealing a modest reduction in BMI in the intervention group (Fig. 5.1). This meta-analysis had an I² value of 65%, indicating substantial statistical heterogeneity.

Separate meta-analyses were performed for total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) (n = 7 each),^{12, 14, 16, 17, 21, 22, 27} and triglycerides (TG) (n = 4).^{14, 16, 21, 27} A meta-analysis of TC showed the significant positive MD value, which means there was a smaller reduction in TC level in the intervention group than the control group (Fig. 5.2). Meta-analyses of LDL-C, HDL-C, and TG did not show significant differences (Fig. 5.3–5.5). Although the mean baseline BMI of participants of all included RCTs was greater than 25, the meta-analyses of TC, LDL-C, and TG showed substantial heterogeneity, with I² values greater than 40%.

The meta-analysis of HbA1c included three RCTs, showing no significant reduction in the intervention group (Fig. 5.6) and yielded an I² value of 30%. In the trials where participants’ baseline BMI was greater than 25, HbA1c tended to decrease in the intervention group, though this was not significant.

Figure 5. Forest Plot: Pooled Effect of Exercise on BMI, lipid profiles, and HbA1c in NDD-CKD patients

Legend of Fig. 5: ^a Confidence interval calculated by Wald-type method, ^b Tau² calculated by Restricted Maximum-Likelihood method

Secondary outcome – falls associated with exercises and other adverse events

Adverse events, such as musculoskeletal injuries, were reported in some of the included studies during the research period. However, there was no significant increase in the incidence of these events in the intervention groups compared to the control groups.

Heterogeneity

In accordance with our a priori protocol, heterogeneity was defined as an I² value greater than 40% in the forest plots. Heterogeneity was observed for several outcomes, including STS test, HR-QoL (“Burden of kidney diseases”, “Physical functioning”, and “Role limitations due to physical health”), BMI, and lipid profiles (TC, LDL-C, and TG).

Study Quality Assessment

Risk of bias was assessed for all included studies (Supplementary materials, Figure S2). Due to the nature of the exercise intervention, blinding of participants was not feasible. Consequently, all studies were rated as having "Some concerns" or "High risk" for overall risk of bias. Furthermore, since the QoL of the subjects was assessed using self-reported questionnaires, all studies measuring HR-QOL were rated as "High risk" for the fourth domain (Bias in measurement of the outcome).

Publication Bias

Funnel plots were created for the included literature, and the presence of publication bias was evaluated. Although a limited number of studies were included in the analysis, the funnel plot appeared symmetrical, and no obvious publication bias was detected (Supplementary materials, Figure S3).

Our systematic review and meta-analysis confirmed that exercise therapy is a safe and effective intervention for patients with NDD-CKD. The key findings of this study were significant improvements in exercise capacity and ambulatory function. We also observed a modest improvement of muscle strength and eGFRcr. These benefits on physical function, exercise capacity, and muscle strength are consistent with the recent literature.^{6, 29, 30}

Exercise is known to mitigate several CKD-related complications, such as physical inactivity, reduced exercise capacity, and sarcopenia, by improving cardiovascular function and muscle mass. The observed improvements in HR-QoL suggest that these physical benefits may translate into a better overall quality of life for patients. Despite these positive findings, the presence of heterogeneity in our meta-analyses warrants a cautious interpretation of some of the results.

For exercise capacity and ambulatory function, although the I² value was low (33% and 27%, respectively), the visual assessment of the forest plot suggested underlying heterogeneity. Differences in patient characteristics may have contributed to this observed heterogeneity. For instance, the mean baseline age and BMI of participants of the Tang et al. study,¹⁹ which showed an exercise-favoring result, was 45.1 years and 23.6, respectively. In contrast, the mean baseline age and BMI of participants of the Leehey et al. study,¹⁴ which showed a neutral result, was substantially higher at 66.0 years and 36.8, respectively.

Similarly, significant heterogeneity was observed for STS test. The study by Barcellos et al²² may have contributed to this, as it reported no improvement in muscle strength. The mean age of the intervention group in this study was 65 years, and the study population included a higher proportion of women (male-to-female ratio of 1:2), which are potential sources of heterogeneity.

For BMI as an outcome, the forest plot showed conflicting results among the included studies. Two RCTs (Baria et al¹² and Greenwood et al¹³) reported a significant reduction in the intervention group, while other RCTs found no significant change. This heterogeneity can be attributed, in part, to specific studies where the mean baseline BMI values of the intervention groups (Leehey et al¹⁴: 36.2, Headley et al¹⁶: 32.7, and Miele et al²¹: 34.9) were notably higher than those in the other trials.

For lipid profiles, we were unable to identify any specific baseline characteristics that would account for the substantial heterogeneity. In addition, the results about lipid profiles should be interpreted with caution, as it was likely influenced by confounding dietary factors. For example, in the trial by Miele et al,²¹ the control group demonstrated a more significant reduction in post-intervention lipid intake than the intervention group. Although the meta-analysis revealed that the reduction in TC was significantly greater in the control group, the influence of confounding factors, such as these changes in dietary content, must be considered. Consequently, the observed changes in lipid profiles cannot be solely attributed to the exercise intervention, and further research controlling for dietary variables is warranted.

Our meta-analysis of kidney function, as measured by eGFRcr, did not yield a significant difference. This finding may be attributable to the limited number of available studies and the substantial heterogeneity of the data (e.g., variations in exercise modalities, patient populations, and intervention durations). Furthermore, reliance on eGFRcr may have underestimated the true effect of exercise, as exercise-induced increases in muscle mass can elevate serum creatinine levels, potentially masking a positive effect on kidney function. Future studies should use more reliable markers, such as eGFRcys, to accurately assess the impact of exercise intervention on kidney function of NDD-CKD patients.

Our study has several strengths, primarily due to the inclusion of only RCTs. This rigorous approach minimizes the risk of confounding and selection bias, thereby providing a higher level of evidence compared to observational studies. Furthermore, we used a predefined protocol registered with PROSPERO to ensure transparency and prevent research duplication. Finally, our team of independent reviewers assessed the risk of bias and methodological quality using validated tools. Any discrepancies identified during the review process were resolved through thorough discussion among the reviewers. This was facilitated by close communication and regular interaction among our team members, ensuring a high degree of consensus and rigor in our findings.

However, our study also has several limitations. The number of studies included in the meta-analysis was relatively small, and most of the included trials focused on patients with obesity, which limits the generalizability of our findings to non-obese CKD patients. The duration of most interventions was also relatively short, ranging from 8 to 52 weeks. Consequently, we were unable to evaluate the efficacy of exercise intervention from a longer-term perspective, such as over five or ten years. Furthermore, while the efficacy of specific exercise methods, such as blood flow restriction (BFR) training and high-intensity interval training (HIIT), has been reported in patients with CKD,³¹ our study did not evaluate these modalities individually. Consequently, the optimal type of exercise to prescribe for CKD patients remains undetermined. Finally, we could not account for all potential confounding factors, such as dietary variables, which may have influenced the outcomes.

Exercise therapy is a safe and effective strategy for improving exercise capacity, ambulatory function, and QoL in patients with NDD-CKD. Our findings support the incorporation of exercise therapy into routine clinical care. However, the current analysis does not provide sufficient evidence to demonstrate that exercise therapy attenuates loss of kidney function in patients with NDD-CKD. Thus, further research is needed to assess the long-term impact on hard clinical endpoints, which would provide definitive evidence to overcome existing barriers to implementation.