The effects of aerobic exercise on eGFR, blood pressure and VO2peak in patients with chronic kidney disease stages 3-4: A systematic review and meta-analysis

Background CKD is associated with several comorbidities, cardiovascular disease being the most significant. Aerobic training has a beneficial effect on cardiovascular health in healthy and some well-defined non-healthy populations. However, the effect of aerobic training on glomerular filtration rate in patients with CKD stages 3–4 is unclear. Objective To review the effects of aerobic exercise training on kidney and cardiovascular function in patients with chronic kidney disease (CKD) stages 3–4. Methods A random-effects meta-analysis was performed to analyse published randomized controlled trials through February 2018 on the effect of aerobic training on estimated glomerular filtration rate, blood pressure and exercise tolerance in patients with CKD stages 3–4. Web of Science, PubMed and Embase databases were searched for eligible studies. Results 11 randomized controlled trials were selected including 362 participants in total. Favourable effects were observed on estimated glomerular filtration rate (+2.16 ml/min per 1.73m2; [0.18; 4.13]) and exercise tolerance (+2.39 ml/kg/min; [0.99; 3.79]) following an on average 35-week aerobic training program when compared to standard care. No difference in change in blood pressure was found. Conclusions There is a small beneficial effect of aerobic training on estimated glomerular filtration rate and exercise tolerance, but not on blood pressure, in patients with CKD stages 3–4. However, data are limited and pooled findings were rated as of low to moderate quality.


Objective
To review the effects of aerobic exercise training on kidney and cardiovascular function in patients with chronic kidney disease (CKD) stages 3-4.

Methods
A random-effects meta-analysis was performed to analyse published randomized controlled trials through February 2018 on the effect of aerobic training on estimated glomerular filtration rate, blood pressure and exercise tolerance in patients with CKD stages 3-4. Web of Science, PubMed and Embase databases were searched for eligible studies.

Results
11 randomized controlled trials were selected including 362 participants in total. Favourable effects were observed on estimated glomerular filtration rate (+2.16 ml/min per 1.73m 2 ; [0.18; 4.13]) and exercise tolerance (+2. 39  Introduction subjects, (2) mean baseline eGFR of the population between 59 and 15 ml/min per 1.73m 2 (CKD stage 3 and 4), (3) aerobic exercise includes the intervention method, (4) intervention encompassed at least 3 months of training with at least 2 sessions per week and (5) studies had to be written in Dutch, English, German or French. Non-randomized controlled and observational trials, reviews, meta-analyses, short communication articles, letters to the editor and study protocols were excluded. No publication date or other restrictions were imposed. Studies were searched describing the effect of aerobic exercise compared to standard care on eGFR, blood pressure or VO 2 peak in patients with CKD stages 3-4. Three electronic databases were consulted and searched for eligible papers until February 2018 (i.e. Web of Science, PubMed and Embase) with a supplementary search on clinicaltrials.gov to identify potential missing trials and to retrieve protocols. Reference lists of included studies were screened for additional studies. MeSH-terms or keywords used in the search strategy (S1 Table)

Data extraction and quality assessment
Qualitative (Table 1) and quantitative data were extracted from the included articles. Corresponding authors were addressed to provide missing or tailored data based on our eligibility criteria (e.g. excluding patients with CKD stage 2). Data for between-groups analyses was extracted from the results of patient and control group analyses. Body mass index (BMI) was post hoc included as secondary outcome. Methodological quality of studies was evaluated according the checklist for assessing risk of bias provided by the Cochrane Collaboration [14]. Quality of evidence and strength of recommendations were assessed by the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) method [15].

Statistical analysis for meta-analysis
The 'GRADEpro Guideline Development Tool' and 'Review Manager 5.3.5' from the Cochrane Collaboration were used for quality assessment of evidence and to calculate effect estimates for combinations of single effects from included studies and to subsequently perform subgroup analyses respectively. Treatment effects were calculated as mean differences in outcomes. If eGFR was calculated using another equation than the CKD-EPI or Modified Diet in Renal Disease formula (MDRD), the CKD-EPI formula was applied on mean and standard deviation values of the population [16]. The Dersimonian-Laird inverse variance weighted random effects method was used to pool study findings and heterogeneity was calculated (I 2 , <25% = no heterogeneity and !75% = high heterogeneity) [17]. Sensitivity analysis was planned and subgroup analyses were performed on changes in eGFR to explore sources of heterogeneity and subgroup variables included: intervention type (i.e. aerobic exercise vs. resistance and aerobic exercise) and duration, age and BMI. Age and BMI were categorised based on cut-off points: 60 years old and 32 kg/m 2 respectively. Subgroup breakdown for intervention duration was based on duration more or less than 6 months. Results were considered significant at 95% study confidence interval. Forest plots were used to present findings of the meta-analysis.

Search results
A total of 1038 studies were identified through our database search. After removing duplicates and two screening phases (title/abstract and full text), 11 studies were withheld based on the in-and exclusion criteria (Fig 1).

Study characteristics
Depending on the study and following the ' Kidney [1,20,23,24,28]. Attrition and the other sources of bias performed better compared to the aforementioned bias criteria with a low risk in more than 75% of the studies (Fig 2). Quality of evidence for the eGFR and BMI analyses were rated as of moderate strength deriving from detection bias ( Table 2). Findings of blood pressure and VO 2 peak were valued of poorer quality due to high heterogeneity (i.e. I 2 !75%) and blood pressure outcome was rated of high risk for publication bias based on asymmetrical funnel plots (S1- S5 Fig).
Blood pressure. Ten studies described the effect of aerobic exercise on blood pressure in patients with CKD stages 3-4. For systolic and diastolic blood pressure, respectively, eight and seven studies were included in the meta-analysis covering a total of 133 and 119 patients receiving aerobic exercise compared to 136 and 118 control subjects [1, 19- Aerobic exercise training of a 32-week duration on average showed no effect on blood pressure in the between-groups analysis (Fig 3). However, mean decreases in systolic blood pressure of 5.20 mmHg [1.00; 9.4] (I 2 = 15%) in patients assigned to aerobic exercise (Fig 4)  Exercise tolerance. Ten studies examined exercise tolerance, as quantified by VO 2 peak in CKD stages 3-4. These studies compared 161 participants receiving aerobic exercise with 164 control subjects [1,[19][20][21][22][23][24][25][26]28]. Similarly, over an intervention period of on average 32 weeks, meta-analysis of the VO 2 peak showed a within-group increase of 1.70 ml/kg/min [0.65; 2.74] (I 2 = 6%) amongst patients following aerobic exercise training (Fig 4). When compared to standard care, aerobic training improved VO 2 peak by 2.39 ml/kg/min [0.99; 3.79] (I 2 = 90%) (Fig 3). Subgroup analyses. Analysis performed on changes in eGFR outcome with studies categorised mean age <60 or !60 years old showed a difference disfavouring the elderly (p = 0.004; I 2 = 88%) (S8 Fig). Also a subgroup difference with mean BMI 32 kg/m 2 as cut-off point was noted (p<0.001; I 2 = 92.2%); in favour of subjects scoring <32 kg/m 2 (S9 Fig). Leehey et al was excluded in the latter analysis due to a nutritional intervention in the control group [21]. No subgroup differences were found when studies were divided based on intervention duration and type (S10 and S11 Figs).

Discussion
The purpose of this meta-analysis was to evaluate the effects of aerobic exercise training in patients with CKD stages 3-4 on outcomes of cardiovascular and kidney function when compared to standard care. We found that aerobic exercise improved eGFR (2.16 ml/min per 1.73m 2 ), VO 2 peak (2.39 ml/kg/min) and decreased BMI (-0.73 kg/m 2 ) when compared to standard care. In line with the most recent Cochrane review [11], aerobic exercise may thus be beneficial for the CV risk profile. We did not find differences in blood pressure following an aerobic training program compared to standard care. Hence, our study supports European guidelines on physical exercise are also applicable in patients with CKD stages 3-4 [9].
Even though only two studies used exercise tolerance as primary outcome, it is not really surprising to find an increase in VO 2 peak after an aerobic exercise intervention in patients with CKD stages 3-4, as this effect is well established in other settings [31, 32]. We considered an increase of 2.38 ml/min per kg (i.e. +10.9%) in VO 2 peak as clinically relevant based on a minimal clinically important difference of 2 ml/min per kg or 6% reported in studies including patients with other chronic diseases [33][34][35]. Also the accompanying improvements in global and physical aspects of Quality of Life (four included studies assessing Quality of Life found clinical relevant changes) indicated the improvement in VO 2 peak being of clinical relevance [21, 25-27]. Aerobic training has been shown to effect several central and peripheral CV adaptations such as improved cardiac output, decreased peripheral vascular resistance, higher blood volume, expanded capillary volume and increased peripheral O 2 -extraction, improving maximal oxygen uptake [36], with shear stress as key factor in vascular adaptations [37]. It is well documented that plasma volume expands up to 25% after 10-14 exercise sessions through increased plasma albumin levels and sodium retention in sedentary healthy subjects. A sustained 8-10% higher plasma compared to erythrocyte volume has also been noted after 30 days of training in sedentary people [38].
Sympathetic vasomotor over-activity is already present in the early stages of CKD and might play a role in the further deterioration of kidney function [39,40]. Aerobic exercise restores surrogate parameters of sympathetic nervous system over-activity in patients with   (1) glomerular filtration rate (ml/min per 1.73m 2 ), (2) systolic blood pressure (mmHg), (3) diastolic blood pressure (mmHg) and (4) VO 2 peak (ml/kg/min).
the efferent renal arteriole in CKD, suggesting a downregulation of the renin-angiotensin system [43]. Indeed, Cornelissen and Fagard reported a 20% lower plasma renin activity due to aerobic exercise training in healthy sedentary normo-and hypertensive subjects [44]. Endothelial dysfunction is already present in early stages of CKD. The efferent renal arteriole demonstrates a constriction in response to glomerular endothelial injury [45] and may eventually contribute to further deterioration in eGFR and progression to renal microvascular disease [46]. A late-onset decrease in oxidative stress and increase in anti-oxidative factors due to exercise training could prevent progressive fibrosis in the kidneys [47,48]. Exercise-induced blood volume expansion likely contributes to increased renal blood flow and shear stress in the kidney at rest and during exercise [49][50][51], the latter stimulating improvements in endothelial function [52]. However, Van Craenenbroeck et al [26] and Headley et al [27] reported no effect on flow-mediated dilation in patients with CKD stages 3-4 after a 12 and 16 weeks aerobic training regimen respectively. As a possible explanation, the authors advance the relatively short intervention period, which might not be enough to reverse established endothelial dysfunction. The results of this meta-analysis do not confirm a beneficial effect of aerobic exercise training on eGFR as a proxy for kidney function (Fig 4). However, they support the hypothesis that exercise might be beneficial in slowing down the progression CKD (Fig 3). Hence, as a preventive strategy, aerobic exercise training might be useful in patients with CKD stages 3-4.
Blood pressure is recognized as a critical remediable aspect in the maintenance of kidney function [37]. Hypertensive subjects experience a more substantial decrease in blood pressure by exercise training than normotensive subjects; underlying mechanisms to this training effect in (non-)healthy hypertensive subjects are improved vascular function (vascular remodelling and endothelial function), decreased sympathetic vasomotor over-activity and downregulation of the renin-angiotensin system [44]. However, the current meta-analysis did not show significant alterations in blood pressure by aerobic exercise training [53]. Our results are in line with the Cochrane review which reported no change in blood pressure following CV exercise. The Cochrane review, however, found a decrease in systolic and diastolic blood pressure in studies describing high intensity aerobic training (>60% VO 2 peak) and in studies combining aerobic and resistance training [11]. This meta-analysis included one study with such intensities [20] and four using combined exercise training [1,19,21,24], and sensitivity analysis excluding these citations did not alter results. Due to a generally well-controlled blood pressure (<140/ 90 mmHg) at baseline [54], a great effect-size in blood pressure was not expected. Other possible explanations for the lack of significant change in blood pressure could be the use of and change in antihypertensives masking training effects (i.e. only four studies described unchanged medical therapy [22,23,25,26]) and analysis being of low power since blood pressure was examined as secondary outcome in included studies. An unexpected decrease in systolic blood pressure by standard care was observed.
A moderate and high degree of heterogeneity (I 2 = 50% and I 2 = 90%) was found in the meta-analysis of eGFR and VO 2 peak respectively. Part of the heterogeneity in primary outcome might be due to age as elderly experience a smaller training effect size compared to younger subjects [55,56]. The three studies covering the oldest participants noted a lower effect size on eGFR [1,20,21]. Analyses excluding these studies resulted in an improved eGFR of 3.43 ml/min per 1.73m 2 [1.62; 5.23] with a negligible observed heterogeneity (I 2 <25%). Also exclusion of the three studies with the highest mean BMI values [1,19,20] resulted in a negligible heterogeneity and subgroup analyses showed favouring results for patients with better body weight control. A detrimental correlation between BMI and oxidative stress in patients with CKD stages 3-4 [57] could result in additional impairments of endothelial function ultimately obliterating exercise training effects on eGFR. It must be noted that two of three studies with mean BMI !32 kg/m 2 are among those covering the oldest participants [1,20]. Heterogeneity can partially if not almost fully be explained by between-study differences in age and BMI. Even though training modalities are determinants of treatment effects sizes for CV outcomes in patients with CKD [52], nor analyses on training type and duration nor differences in supervision, frequency or intensity of training could explain the high heterogeneity. However, exclusion of the four studies using both aerobic and resistance exercise training also resulted in a negligible heterogeneity (I 2 = 19%). Sensitivity analysis excluding studies which also included patients with CKD stage 2 showed no alterations in effect size and heterogeneity (i.e. 2.16 ml/ min per 1.73m 2 ; I 2 = 50% in the current analysis vs. 2.98 ml/min per 1.73m 2 ; I 2 = 56% in the sensitivity analysis for eGFR, p = 0.08 for subgroup analysis) [19][20][21].
It should be noted that muscle gain following physical exercise also affects eGFR measurements as it increases serum creatinine levels, thereby decreasing eGFR based on equations using creatinine. Hence, an exercise-induced increase in muscle mass may underestimate beneficial effects on kidney function as absolute improvements in eGFR may have been more pronounced than observed.
Quality and strength of evidence regarding VO 2 peak and blood pressure are low to very low, mainly due to high risk of bias and heterogeneity. However, exclusion of articles scoring high risk of bias on at least two different criteria did not alter the overall effect. The pooled findings of change in eGFR following aerobic exercise were rated as of moderate strength.
Even though this article could have different conclusions than the Cochrane review, in particular because only RCTs including patients with CKD stages 3-4 were selected, similar findings were retrieved for blood pressure and VO 2 peak [11].

Limitations
This review has some contextual limitations and reasonable risk of bias. Selective reporting bias could result from retrospective registration of our protocol. Sedentary patients are less likely to engage in exercise intervention trials, for this reason, selection bias can be a pitfall in our meta-analysis, and results do thus not necessarily apply to the overall population with CKD stages 3-4. No studies were included reporting measured GFR and two estimation equations for eGFR were used with a rather low correlation to measured methods (i.e. r = 0.41 for the CKD-EPI and r = 0.27 for the MDRD) and to each other, inducing heterogeneity and additional bias [58]. Hydration status and muscle mass are undisputable confounders in the effects of aerobic exercise training on eGFR and should be taken into account in future studies. Outcome analyses were at risk of insufficient statistical power of individual studies since not more than 3 included studies described a similar primary outcome. Heterogeneity could be overestimated in our meta-analysis including only a small number of studies as we used the DerSimonian-Laird method. However, this is the standard approach applied by Review Manager 5.3.5 [59].

Conclusions and guidelines for further research
This meta-analysis supports a beneficial effect of aerobic exercise on glomerular filtration rate and exercise tolerance in patients with CKD stages 3-4 when compared to standard care. Given the heterogeneous data, the lack of studies on hard and patient relevant endpoints, and at best low to moderate methodological quality of studies, this meta-analysis should act as an impetus for larger and more conclusive interventional research of high methodological quality (e.g. by examining measured GFR as primary outcome and with hydration status as covariate). Consensus regarding exercise prescription for this vulnerable population can only be reached when the question about which frequency, intensity, type and time of exercise is of most benefit, is answered.