This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.47 Relevant studies were identified by searching the PubMed, Scopus, Web of Science and EBSCO databases. The search terms were the following: “cardiovascular disease*” OR “chronic disease*” OR “heart disease*” OR hypertension OR “peripheral arterial disease” OR diabetes mellitus OR “coronary artery disease” OR “heart failure” OR myocardial infarction OR “vascular disease*” in combination with “endothelial progenitor cells” OR “progenitor cells” OR “colony-forming units” OR “stem cells” OR “circulating angiogenic cells” in combination with “exercise training” OR exercise OR “cardiac rehabilitation” OR “physical exercise” OR “physical activity” OR “exercise therapy” OR “rehabilitation exercise” and avoiding the terms infant* OR child* OR rat* OR mice OR mouse OR animal* OR cancer OR neoplasm* OR neoplasia. All the keywords were MeSH terms, with the exception of “acute myocardial infarction”, “progenitor cells” and “circulating angiogenic cells”. Considering that EPCs were first isolated by Asahara et al. in 1997,48 we limited the search to studies published between January 1997 and January 31, 2018. In addition, we manually searched the reference lists of landmark studies and review articles on this topic to identify any relevant studies that electronic searches might have missed.

Studies were eligible if they (i) were RCTs; (ii) included adults (≥18 years old) with established CVD; (iii) compared an experimental group receiving a structured program of exercise training, either exclusively or as a component of a cardiac rehabilitation program; (iv) included a control group receiving only standard medical care (that did not perform exercise training); (v) included an exercise training program lasting at least one week; (vi) assessed circulating EPC levels before and after the intervention; and (vii) were written in English. We excluded literature review papers, letters to the editor, abstracts published in conference proceedings, studies that assessed the acute effects of a single exercise session, studies involving healthy subjects or patients with other clinical conditions (e.g. kidney disease), and animal model studies.

Initially a single author screened the titles and discarded duplicates. Subsequently, two authors independently determined whether studies fulfilled the criteria for inclusion in this review through examination of screened titles and abstracts. The same two authors then analyzed the full articles and used a standardized form to determine eligibility for inclusion in the review. Any disagreements regarding the inclusion of studies were resolved by discussion with a third author. The process of identification and selection of the studies is displayed in Figure 1.

Figure 1.

Flow diagram of the literature search.

(0.44MB).

Data were extracted for the main outcome (circulating EPC levels) as well as for the study design, population, details of the intervention (frequency, mode, duration, intensity and characteristics of the exercise) and control groups, antibodies used to identify EPCs, units of EPC measurement, functionality of EPCs in vitro (proliferative capacity, migratory capacity, number of colony forming unit endothelial cells), time of blood sample collection for EPC analysis, and exercise capacity. The methodological quality of the studies was determined using the Physiotherapy Evidence Database (PEDro) scale. This scale comprises 10 items related to internal validity and interpretability; a higher score indicates better methodological quality.49 Based on the PEDro score, the quality of the included RCTs were described as following: 6-10 for high-quality studies; 4-5 for fair-quality studies; and ≤3 for poor-quality studies.49 Two authors independently extracted the above-mentioned data using a customized form and evaluated methodological quality, while a third author resolved disagreements.

The data were analyzed using STATA/SE software, version 14 (StataCorp, College Station, TX, USA). The standardized mean difference between the pre-post of the intervention group versus the control group was calculated for EPC levels using Cohen's d as the effect size statistic. The significance of the pooled effect size was estimated based on 95% confidence intervals (CI). Data not provided in the main text or tables were obtained by contacting the authors or (when there was no response) extracted from figures. One study50 included two intervention groups (a younger group ≤55 years old and an older group ≥65 years old); the data from these groups were analyzed as independent samples.

The heterogeneity of results across studies was quantified by I2 statistics as follows: low risk of heterogeneity (0-40%); moderate risk of heterogeneity (30-60%); substantial heterogeneity (50-90%); or considerable heterogeneity (75-100%). Significance levels were expressed as p-values.51

Subgroup analyses were performed based on the clinical condition: HF or arterial disease (peripheral and coronary).

Random-effects meta-regression was used to assess whether results differed according to the mean age of participants and the length of the intervention, since these could be considered sources of heterogeneity.

Publication bias was identified by visual inspection of funnel plots and the method proposed by Egger.52

The initial search identified 651 titles. After exclusion of 227 duplicates, the titles and abstracts of the remaining 424 articles were analyzed, following which 332 articles were excluded, resulting in 92 studies qualifying for full-paper assessment. Of the 92 potentially eligible articles, only four25,50,53,54 met the inclusion criteria (Figure 1). After screening the reference lists of relevant articles on this topic, two additional papers, not retrieved from the database search, were found.55,56 A total of six papers were therefore included in the systematic review and meta-analysis.

These six studies included 236 participants, 119 in exercise training and 117 in control groups (Table 1). The sample size ranged from 21 to 64 participants and overall the studies involved more men (n=189, almost 80%) than women (n=47). Two studies54,56 included coronary artery disease (CAD) patients (n=85), three studies50,53,55 involved patients with HF with reduced ejection fraction (n=111), and one study included patients with peripheral arterial disease (PAD) (n=40) with mild to severe claudication.25 All studies identified and quantified circulating EPCs by flow cytometry using a combination of specific antibodies: three studies used the specific antibodies CD34+/KDR+,50,54,56 two CD45dim/CD34+/KDR+,53,55 one CD34+/CD133+/KDR+25 and one CD133+/KDR+50. Three studies quantified EPCs as the percentage of mononuclear cells,25,53,55 two as the absolute number per ml of peripheral blood,25,50 one as the percentage of absolute cell counts divided by lymphocyte counts,56 and one as cells per 100 000 laser detection events.54 Blood collection for EPC determination was performed 24-48 hours,54 72 hours,25 and up to a week53 after the last exercise session. Some studies did not provide the time of blood collection.50,55,56 Only one study reported the time of day of blood collection, between 8:00 and 9:00 am.53

Regarding the characteristics of exercise training, all studies used aerobic exercise, either exclusively25,50,53–55 or combined with resistance exercise.56 The length of exercise interventions ranged from four50 to 24 weeks,25 with a frequency of two,25 three,54,56 or five50,53,55 sessions per week. The duration of the exercise sessions ranged from 30 to 60 min. Exercise intensity was determined using claudication symptoms,25 or heart rate at the ventilatory anaerobic threshold,53,55 at 80% of maximum heart rate (HRmax)56 or at 70% of maximum oxygen uptake.50 Only one study involved resistance exercise determined by intensity assessed through perceived exertion and HRmax, at a similar intensity to aerobic exercise (80% of HRmax).56 One study did not report either exercise session duration or exercise intensity.54

When exercise training was compared to control, EPC levels were higher in the exercise training groups (ES=0.57; 95% CI: 0.01, 1.12), with considerable heterogeneity (I2=75.6%; p<0.001) (Figure 2).

Figure 2.

Forest plots of effect size for exercise training interventions in patients with cardiovascular disease according to clinical condition. CI: confidence interval; EPCs: endothelial progenitor cells; ES: effect size.

(0.2MB).

Subgroup analyses were performed based on the clinical condition, i.e. patients with HF and patients with arterial disease (CAD or PAD). Studies enrolling patients with CAD and PAD were grouped together, because only one study in patients with PAD was included in this review. In patients with HF, a significant increase in circulating EPC levels was seen in the exercise training group (ES=0.52; 95% CI: 0.15-0.90), with low heterogeneity (I2=0.0%; p=0.648) (Figure 2).

In patients with arterial disease, there was no significant increase in EPC levels in the exercise training groups (ES=0.67; 95% CI: -0.70-2.04), with considerable heterogeneity (I2=91.2%; p<0.001) (Figure 2). The two studies54,56 that enrolled patients with CAD showed no changes in circulating EPC numbers after the exercise training program, while the study25 that assessed the effects of aerobic exercise in the mobilization of EPCs in patients with PAD and intermittent claudication showed a significant increase in circulating EPC levels.

The random-effects meta-regression model showed that age (Figure 3) and length of the intervention (Figure 4) are not related to heterogeneity across studies (p=0.821 for age, p=0.113 for length of the intervention).

Figure 3.

Random effects meta-regression for age.

(0.06MB).

Figure 4.

Random effects meta-regression for length of intervention.

(0.06MB).

The pooled effect size estimate was not significantly altered in magnitude or direction when individual study data were removed from the analysis one at a time.

The studies had a mean score of 6 on the PEDro scale, ranging from 5 to 7. Two studies scored 5, which indicates fair methodological quality; the remaining studies were classified as high methodological quality RCTs (Table 1).

There was no significant publication bias for exercise training intervention versus control, as demonstrated by the asymmetry of the funnel plot (Figure 5) and Egger's test (p=0.324).

Figure 5.

Assessment of potential publication bias by Egger's test.

(0.07MB).

Some limitations of this review should be acknowledged. A major limitation is the considerable heterogeneity among studies. Differences in the methodological assessment of EPCs may have contributed to this level of heterogeneity. EPCs are a relatively scarce cell population, which makes their quantification by flow cytometry a challenging task.66 The best EPC phenotype to determine EPCs in a clinical setting appears to be CD34+/KDR+/CD45dim.66–68 The results of different studies may also to some extent be influenced by the lack of consistency in isolation strategies, use of lineage markers to determine EPCs, and units used to report EPCs.22 In this regard, the use of cells per ml as units to quantify EPCs50 is a major limitation, due to the hemodilution that can be seen in patients with HF. Also, several studies did not clearly state the time of day of blood collection and the time from the last exercise session, as EPCs appear to display diurnal variation69 and the timing of blood collection post-exercise could reflect the effect of the last exercise session.23,70 The differences between studies in disease severity, medication use and cardiovascular risk factors may also have contributed to the high heterogeneity. Data from some studies were extracted from figures, which is another potential source of error. The lack of a clear description of the characteristics (e.g., intensity, mode, duration, frequency) of the exercise training programs was also a limitation.

This analysis has several implications for further research. In most studies, the sample was composed mainly (or exclusively) of male patients; therefore, future studies should include mixed-gender samples and sub-analysis by gender, when possible, to ascertain the possible existence of between-gender differences, since hormones could influence EPC levels.71 It would be useful to assess endothelial function and correlate any changes with changes in EPC levels; the only study included in this review that assessed the relation between endothelial function and EPCs found no correlation. Although the effect of exercise on both endothelial function and EPC levels shares common mechanisms (decreasing inflammatory cytokine concentrations and increasing endothelial nitric oxide synthase activity and endothelium-dependent nitric oxide production),12,29,72 the relation between the two variables still lacks clarification. Future studies should use standard gating strategies and assessment protocols for the quantification of EPCs.