Chronic Adaptations to Eccentric Cycling Training: A Systematic Review and Meta-Analysis

This study aimed to investigate the effects of eccentric cycling (ECCCYC) training on performance, physiological, and morphological parameters in comparison to concentric cycling (CONCYC) training. Searches were conducted using PubMed, Embase, and ScienceDirect. Studies comparing the effect of ECCCYC and CONCYC training regimens on performance, physiological, and/or morphological parameters were included. Bayesian multilevel meta-analysis models were used to estimate the population’s mean difference between chronic responses from ECCCYC and CONCYC training protocols. Group levels and meta-regression were used to evaluate the specific effects of subjects and study characteristics. Fourteen studies were included in this review. The meta-analyses showed that ECCCYC training was more effective in increasing knee extensor strength, vastus lateralis fiber cross-sectional area, and six-minute walking distance compared to CONCYC. Moreover, ECCCYC was as effective as CONCYC in decreasing body fat percentage. CONCYC was more effective in increasing V˙O2max and peak power output attained during concentric incremental tests. However, group-level analyses revealed that ECCCYC was more effective than CONCYC in improving V˙O2max in patients with cardiopulmonary diseases. ECCCYC is a viable modality for exercise interventions aiming to improve parameters of muscle strength, hypertrophy, functional capacity, aerobic power, and body composition, with more advantages than CONCYC training in improving neuromuscular variables.


Introduction
Over the last decades, sports scientists and physiologists have advanced our knowledge of the mechanisms and applications of eccentric (i.e., lengthening) muscle contractions. Given its distinctive characteristics, eccentric contractions may induce different morphological, neuromuscular, and metabolic adaptations compared to isometric and concentric muscle actions [1,2]. Evidence from isokinetic exercise indicates that eccentric contractions are more effective in increasing muscle strength than concentric contractions [3]. Additionally, there is robust evidence supporting the utilization of eccentric exercise for the treatment of tendinopathies [4][5][6] and injury prevention and rehabilitation [7][8][9][10][11]. Recently, there has been a growing interest in the utilization of eccentric exercises in the treatment of individuals with poor exercise tolerance, such as elderly individuals and patients with cardiopulmonary diseases [12,13].
In fact, the low-energy cost associated with eccentric muscle work makes eccentric exercises a feasible alternative to counteract sarcopenia and declines in the functional capacity of older individuals, especially those with clinical conditions, since eccentric exercises may provide sufficient stimulus to trigger neuromuscular adaptations without 2 of 24 imposing severe cardiovascular and/or respiratory burden [13][14][15][16]. In this context, eccentric exercise via motorized cycle ergometers, named eccentric cycling (ECC CYC ), comprises a safe way to exercise eccentrically, avoiding impact on lower limb joints and falls, as well as allowing appropriate quantification of negative work absorbed during exercise bouts. ECC CYC can be categorized as a moderate-load eccentric exercise [12] consisting of submaximal eccentric exercise performed continuously for long periods (i.e., 10-30 min). In practice, pedaling eccentrically consists in resisting (i.e., attempting to brake) the motordriven backward movement of the pedals. Thus, during ECC CYC sessions, one must produce a large number of eccentric contractions with the locomotor muscles, mainly the knee and hip extensor muscles [17], in a coordinated manner, trying to maintain the level of resistive force prescribed, usually displayed on a screen connected to the cycle ergometer.
Interestingly, in addition to the known potent stimulus of eccentric exercises for muscle strength development, chronic ECC CYC interventions can confer improvements in aerobic power, exercise tolerance, and body composition parameters [14,[18][19][20][21]. Hence, ECC CYC has the potential to improve distinctive performance and physiological/morphological parameters by inducing both energetic/metabolic and tensional adaptive stimuli within the same session, characterizing a time-effective modality [22]. However, some studies only indicate neuromuscular and morphological muscle adaptations following ECC CYC interventions [23][24][25][26], and there is evidence suggesting that eccentric endurance exercises may not have an impact on aerobic metabolism [27].
These disagreements in the chronic responses to ECC CYC may be related to the lack of information regarding the eccentric exercise intensity continuum, which makes the prescription of ECC CYC load based on a specific level of homeostasis disturbance related to a targeted adaptation difficult. Additionally, the fitness level of the individuals submitted to ECC CYC interventions could explain, in part, the controversial results of ECC CYC adaptations since the moderate intensity-like behavior (for details, please see Barreto et al. [22]) of the physiological responses to ECC CYC may be insufficient to promote aerobic/metabolic adaptations in healthy active and highly trained individuals. Therefore, the present systematic review aimed to determine the effectiveness of ECC CYC interventions in improving different performance, physiological, and morphological parameters, as well as investigating the impact of the intervention and population characteristics on these training outcomes. We choose to compare ECC CYC with concentric cycling (CON CYC ) interventions since CON CYC protocols are widely used for exercise treatment in clinical populations [28,29] and to improve aerobic and body composition parameters in the general population [30].

Materials and Methods
The present review is part of a larger systematic review project that aims to investigate the acute and chronic physiological responses to ECC CYC compared to CON CYC . The original protocol was prospectively registered within the Open Science Framework (https://osf.io/sa6g3, accessed on 1 December 2022). The search strategy and selection of studies included both transversal and longitudinal studies that investigated the responses to ECC CYC bouts and training protocols, respectively. The present review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [31].

Study Eligibility
The PICO framework (Population, Intervention, Comparator, and Outcome) was used to establish the eligibility criteria. The population included male and female human subjects without restrictions on age, health condition, or level of physical fitness. The exercise intervention must have comprised ECC CYC training protocols lasting at least one week. The comparator consisted of CON CYC training of the same duration. Investigations that reported outcomes related to neuromuscular function, aerobic power, functional capacity, and body composition before and after ECC CYC and CON CYC training protocols were considered for inclusion. Only original investigations published in peer-reviewed outlets and written in English were considered eligible. Studies that investigated the responses to single-leg ECC CYC training or ECC CYC performed with the upper limbs were excluded.

Information Sources
The searches were conducted until February 2021 and updated in November 2021 using the online electronic databases PubMed, ScienceDirect, and Embase. Studies meeting the inclusion criteria identified in the reference lists of the included articles were also included in the present review. Missing data and/or information from the selected studies were requested via electronic mailing to the authors.

Search Strategy
The search strategy was designed to track all possible studies on the topic of "eccentric cycling". Five target studies on this topic were used to formulate the search strategy [24,[32][33][34][35]. The title and abstract of the records were examined for the identification of possible search terms using the word frequency analysis tool of the PubMed database.
The search strategy was verified by identifying the recognized relevant studies in the results of preliminary searches and by identifying new relevant studies obtained through changing search terms. Hence, we adopted the following search strategy through titles and abstracts of indexed documents: ("eccentric" OR "eccentrically" OR "negative work") AND ("cycling" OR "bicycle" OR "pedaling" OR "pedalling" OR "ergometer" OR "ergometry"). No temporal clipping was established.

Study Selection
Two authors (R.V.B. and L.C.R.L.) carried out the selection of studies independently, using a freely available software-Rayyan QCRI (https://www.rayyan.ai/, accessed on 1 December 2022) [36]. Following the exclusion of duplicates, the titles and abstracts were screened, and irrelevant records were removed. Subsequently, the full-text articles were reviewed and assessed for eligibility.

Data Extraction
The data of each study included in this review were extracted separately by two authors (R.V.B. and L.C.R.L) into an Excel spreadsheet (Microsoft, Redmond, WA, USA). The same authors then compared their spreadsheets and addressed the inconsistencies through discussion. When necessary, data were extracted from figures using the freely available software Web Plot Digitizer (https://automeris.io/WebPlotDigitizer, accessed on 1 December 2022).

Data Items
Information about the publication (i.e., author, year, journal, and digital object identifier-DOI), study design (e.g., randomized or quasi-randomized), population (i.e., sample size, age, height, body mass, maximal oxygen uptake ( . VO 2max ), and health condition), intervention (i.e., intensity, duration, and type of the sessions, number of sessions per week, pedal cadence, and duration of the ECC CYC training protocol), comparator (i.e., intensity, duration, and type of the sessions, number of sessions per week, pedal cadence, and duration of the CON CYC training protocol), and outcomes (i.e., mean and standard deviation (SD) of the performance, physiological, and/or morphological parameter assessed before and after CON CYC and ECC CYC training period and the inferential statistics parameters) was extracted from all studies included in this review.
The terms . VO 2max and PPO were used to represent measures of maximal and peak oxygen consumption and maximal and peak power output, respectively, obtained from an incremental CON CYC test performed until exhaustion or a symptom-limited incremental CON CYC test.

Study Quality Assessment
The methodological quality of the selected reports was rated using the Physiotherapy Evidence-Based Database (PEDro) scale [37]. The PEDro scale comprises 11 criteria related to the external (item 1) and internal (items 2-9) validity of the study, as well as the statistical procedures (items [10][11]. Except for the first item, which was not utilized to generate the PEDro score, the report receives one point for each satisfying item. Therefore, the higher the study score (with 10 being the highest), the higher the study's quality [37]. Two authors (R.V.B. and L.C.R.L.) assessed the quality of the included studies independently, and discrepancies were handled by a third author (B.S.D.).

Effect Sizes Calculation
Three different effect sizes were calculated for each variable: (1) the mean percent difference between pre-and post-ECC CYC training, (2) the mean percent difference between pre-and post-CON CYC training, and (3) the net effect between training modalities (i.e., the mean difference between pre-to-post ECC CYC and pre-to-post CON CYC ). Thus, a negative net effect means that the change induced by ECC CYC training in the analyzed variable was % greater than the change induced by CON CYC training and vice versa.
The precision of each effect size was determined by the standard error (SE) of the pre-to-post change for CON CYC and ECC CYC conditions. The SE of the change for each condition (i.e., CON CYC and ECC CYC ) was calculated by dividing the SD of the pre-to-post change by the square root of the sample size. The SE of the net difference between CON CYC and ECC CYC was calculated as the sum of the pre-to-post SE of each condition via their variances as follows: Note that SE of net difference was derived to present to the reader the "observed" effect sizes in the studies; the net mean difference between the conditions was determined during the statistical analysis, as detailed below. Within the meta-analytic model, the weight of each effect size was set by the inverse of the squared SE (1/SE 2 ). Thus, effects derived from studies with less variability in responses and/or a greater number of participants exerted greater weight on meta-analyzed effect size. The SD of the pre-to-post change was determined using the exact p-values via t-statistics, confidence intervals, F-values [38], and raw data, or was determined from data extracted from figures. When it was not possible, the SD of the change was inputted by the mean correlation coefficient (r) derived from pre-to-post scores of the studies in which the inferential statistics deemed its determination possible [38]. A moderate correlation coefficient (r = 0.50) was adopted for the imputation of SD of the change when none of the alternatives described above was possible.
Pre-to-post effect sizes and their respective SE were converted to percentage units by dividing by the mean of CON CYC and ECC CYC conditions, respectively, and multiplying it by 100.

Statistical Analysis
The meta-analyses were conducted within a Bayesian framework using multilevel models. Analyses were performed in the statistical software R (v4.0; R Core Team [2020], Vienna, Austria) in its graphical interface RStudio (v1.2.5; Boston, MA, USA). The brms package [39] was used for analyses, which allowed the adjustment of multilevel Bayesian models using Stan (i.e., a platform for statistical modeling and high-performance statistical computation) [40].
We derived the effects for each arm to verify the pre-to-post change for both CON CYC and ECC CYC interventions and evaluate the contrast effect between the two conditions. Hence, two analysis models were carried out for each outcome. The first model was composed of a linear meta-regression analysis, where the response variable was the effect sizes and the covariate (i.e., model fixed effect) was the ECC CYC and CON CYC conditions coded numerically as 0 and 1, respectively. Ergo, the intercept of the meta-regression provided the population (fixed effect) meta-analyzed average effect of the ECC CYC condition, and the slope of the regression provided the average net difference between ECC CYC and CON CYC conditions. The second model included only the pre-to-post effect sizes of CON CYC to derivate the population's meta-analyzed average effect of this condition. Random effects in the first model included the intercept and slope of each study for each time point in the case of repeated measures (i.e., each pair of CON CYC and ECC CYC effect sizes for each time point), the intercept and slope of each health condition of the participants, and the duration of the interventions (i.e., specific group-levels). For the second model, random effects were set for each effect size, the health condition of the participants, and the duration of the intervention identities (intercepts).
When the outcome provided more than ten included studies, a meta-regression with more than the two conditions as covariates was carried out [38]. Fixed effects (i.e., covariates) included the conditions (ECC CYC and CON CYC ), the duration of the intervention (linear as days), the interaction conditions × duration of the intervention, and participants' . VO 2max (linear as mL/kg/min). Random effects included the intercept, slope of condition, slope of intervention duration, and slope interaction conditions × duration of the intervention.
Weakly-informative Student's t prior distributions (df = 3, µ = 0, and σ = 10) were used for fixed-effects models, and half Student's t-distributions (df = 3 and σ = 10) were used for between-group-level variance effects (i.e., τ values). Model fitting was performed using Markov Chain Monte Carlo (MCMC) methods, more specifically, the No-U-Turn sampler (NUTS) implemented in Stan. For each model, four chains were run in parallel with 4000 iterations and a warm-up of 1000 iterations. The convergence of the models was verified with Gelman-Rubin diagnostics (R) [41].
To deal with repeated measures in the meta-analyses of studies with more than one effect for the same participant, variance-covariance matrices were calculated. When the information provided in the studies was insufficient to determine the correlation between the dependent effect sizes to perform the matrix calculation, a moderate coefficient of correlation (r = 0.50) was assumed between the effects derived from the same participants. Furthermore, sensitivity analyzes were also performed using the values of r = 0.30 and 0.70 in the calculation of matrices (see Supplementary Material Figure S1, which presents the results of sensitivity analyses). For meta-regressions in which the covariate . VO 2max was missing, we input the data during model fitting using a multivariate model as described elsewhere [42].
Heterogeneity between the effects and group levels was presented as SD (tau-τ). All posterior data generated by MCMC were reported as medians with two-tailed 95% credible intervals (CrI). Furthermore, considering the complete posterior distributions, the probability (in %) of the effect being greater than zero (p > 0) was presented; that is, the area of the posterior distribution located above zero. The area of the posterior distribution of pre-to-post training effect sizes located above zero indicates the probability of the training modality (i.e., CON CYC or ECC CYC ), inducing an increase in the variable from pre-training measures. The area of the posterior distribution of net effects between training modalities located above zero indicates the probability of the CON CYC inducing a greater increase in the variable and vice versa.

Study Selection
The literature search yielded a total of 992 results. A total of 628 titles and abstracts were screened after duplicate removal, and 105 full-text articles were read and assessed for eligibility. Four additional articles were retrieved from reference lists. A total of 14 articles were included in this review ( Figure 1).
The literature search yielded a total of 992 results. A total of 628 titles and abstracts were screened after duplicate removal, and 105 full-text articles were read and assessed for eligibility. Four additional articles were retrieved from reference lists. A total of 14 articles were included in this review ( Figure 1).  Table 1 summarizes the main characteristics of the included studies. Most of the studies (57%) included in this systematic review recruited patients with cardiopulmonary  Table 1 summarizes the main characteristics of the included studies. Most of the studies (57%) included in this systematic review recruited patients with cardiopulmonary diseases. Specifically, three studies recruited coronary artery disease patients [19,34,43], two studies enrolled chronic heart failure patients [18,44], and three studies reported data from chronic obstructive pulmonary disease patients [14,35,45]. Among the remaining studies, two have investigated the effects of ECC CYC and CON CYC training in healthy individuals [24,46], two recruited sedentary participants [23,25], one evaluated amateur cyclists [26], and one evaluated obese adolescents [20]. LaStayo et al. [46] Healthy individuals 6 weeks 2-5 sessions 10-30 min 100-300W during the first 3 wk, and then, PO was adjusted to match VO 2 between modes/50-100W during the first 3 wk and then, PO was adjusted to match VO 2 between modes Knee extensors IPT (n = 9)
The intensity of ECC CYC and CON CYC training sessions of the studies included in this review was prescribed based on individuals' rate of perceived exertion (RPE), heart rate (HR) corresponding to the ventilatory threshold, percentage of PPO, . VO 2max , or HR max attained during the CON CYC incremental test and percentage of age-predicted HR max . In 14% of the included studies, the intensity of the sessions was defined so that the participants of the two groups (i.e., ECC CYC and CON CYC ) exercised at the same PO [23,25]; in 22% of the studies, the training intensity was chosen to match VO 2 during the ECC CYC and CON CYC sessions [20,34,46]; and 36% of the studies set the intensity of both training groups to elicit the same relative HR [14,19,24,35,43]. Only two studies (14%) matched ECC CYC and CON CYC training intensity with the same RPE [26,45], and two studies (14%) prescribed training intensities based on safe RPE and HR values for the investigated population without establishing equalization criteria between groups [18,44].

Quality Assessment
The scores on the PEDro scale ranged from 3 to 8 points (mode = 6 points; mean = 6 points) ( Table 2). Most studies (79%) included in this review presented "good" (i.e., 6-8 points) methodological quality, while two studies (14%) presented "fair" (i.e., 4-5 points) methodological quality and one study (7%) was rated with "poor" methodological quality [47]. Most studies (86%) included in this review were randomized controlled clinical trials, except for two studies [25,45], which adopted a quasi-randomized design, where participants were allocated into groups according to their forced expiratory air volume in the first second (FEV 1 ) and age [45], and maximal isometric voluntary force of the knee extensor muscles [25]. None of the studies scored on criteria 5 and 6 of the scale (i.e., criteria related to blinding of subjects and therapists who administered the intervention, respectively). Blinding of assessors was performed in five studies, which scored at criterion 7 [14,19,20,35,44]. Eleven of the included studies assessed knee extensors IPT (n participants = 212) ( Figure 2). The estimated average effect of the difference between pre-and post-CON CYC training IPT values showed that CON CYC was effective in improving IPT (µ = 5.83% [3.01%, 8.79%]; p > 0 = 100%) (Figure 2A). The estimated average effect of the difference between pre-and post-ECC CYC training also indicated that ECC CYC was effective in improving IPT (µ = 13.02% [9.19%, 17.17%]; p > 0 = 100%) ( Figure 2B). The meta-analyzed net effect showed a more favorable effect of ECC CYC than CON CYC on IPT (µ = −6.82% [−11.16%, −2.82%]) ( Figure 2C). The posterior density of the average net effect indicated no probability (p > 0 = 0%) of the CON CYC inducing greater improvements in IPT compared to ECC CYC training. The heterogeneity between CON CYC training effects was lower (τ = 4.58% [3.15%, 6.78%]; Figure 2D) than the heterogeneity between ECC CYC training effects (τ = 7.55% [5.47%, 10.8%]; Figure 2E) and heterogeneity between net effects (τ = 5.9% [3.52%, 9.21%]; Figure 2F).     Figure 3 presents the conditional effects, the modifying effects, and the residual heterogeneity derived from the meta-regression. The effects on knee extensors IPT were modified mainly by the duration of the intervention. The meta-regression revealed that the difference between ECC CYC and CON CYC training effects on IPT becomes greater as the intervention duration increases. The aerobic fitness expressed as the relative . VO 2max (i.e., mL/kg/min) showed no important modification to the changes in IPT following the two types of cycling training. Small to moderate heterogeneity was observed in the meta-regression-derived effects.   . VO 2maxmaximal oxygen uptake. * Reference condition adjusted to CON CYC , with 7 days of intervention, and subjects . VO 2max of 16 mL/kg/min.
It is well established that the high levels of force produced during eccentric contractions constitute an optimal stimulus for muscle strength development [3,48]. Moreover, evidence suggests that strength gains induced by eccentric exercises are very specific to the type of contraction as well as the movement velocity produced during training sessions [3]. In the meta-analysis published by Roig et al. [3], they compared the effects of eccentric and concentric training on muscle strength, and it was found that eccentric training was more effective in improving total strength (i.e., an average of isometric, concentric, and eccentric maximal force) and eccentric strength, but not concentric and isometric strength. Our results corroborate with previous literature reporting strength adaptations following eccentric and concentric training and showing greater effectiveness of ECC CYC compared to CON CYC in increasing muscle strength. It is important to emphasize that, in most of the studies included in the meta-analyses of strength variables, the workload used during ECC CYC was greater than that used during CON CYC training sessions, which may explain the superior strength gains observed for ECC CYC compared to CON CYC [1,3]. Furthermore, we found greater effectiveness of ECC CYC in improving muscle strength from all modes of contraction (i.e., isometric, concentric, and eccentric), which contradicts previous results suggesting poor transferability of strength gains induced by eccentric exercises to different types of contraction [3,49,50]. Nevertheless, the posterior distribution of average net effects on IPT, ICPT, and IEPT indicated probabilities of 82% to 100% of ECC CYC inducing greater improvements in muscle strength than CON CYC training. Moreover, there was evidence that the superiority of ECC CYC in increasing IPT compared to CON CYC training becomes greater as intervention duration increases. Therefore, the existing data indicate that the prescription of ECC CYC aiming to develop locomotor muscle strength is more advantageous in comparison to CON CYC training.
Previous systematic reviews indicated that eccentric exercise might generate greater muscle hypertrophy than concentric exercise due to the greater mechanical tension imposed on muscle fibers during eccentric actions [3,51]. Accordingly, the present results showed that ECC CYC was more effective than CON CYC training in increasing f-CSA. It is important to note that the estimated average effect showed a small difference (µ = −3%) between changes induced by ECC CYC and CON CYC and similar probabilities for each modality to induce greater changes than the other (i.e., 44% of probability of the CON CYC being more effective and 56% of probability of the ECC CYC being more effective). A possible explanation for the discrepancy observed between the results of the present study (which indicate a small difference between muscle hypertrophy induced by ECC CYC vs. CON CYC training), and the current literature (which suggests the superiority of eccentric contractions in producing muscle hypertrophy) could be the distinctive characteristics of muscle remodeling to eccentric and concentric muscle overloading [52]. It has been shown that changes in muscle size are associated with changes in f-CSA following concentric exercises, but muscle growth may occur without significant changes in f-CSA following eccentric exercises. Instead, muscle hypertrophy following eccentric exercises may occur through increased sarcomere length and/or the addition of sarcomeres in series [21,[52][53][54]. Unfortunately, the only measurement of muscle hypertrophy that was possible to meta-analyze was f-CSA due to the lack of use of more reliable methods to evaluate muscle size (e.g., magnetic resonance image, computerized tomography, or ultrasonography) in investigations assessing muscle hypertrophy following ECC CYC and CON CYC . Hence, more studies are needed to establish the effectiveness of ECC CYC in inducing muscle hypertrophy in comparison to CON CYC .
Unlike most eccentric exercise modalities, ECC CYC is performed at submaximal intensities, with large muscles working continuously for long periods (~10 to 30 min) [12]. Hence, in addition to the potent adaptive stimulus provided by lengthening contractions to the neuromuscular system [1], it has been suggested that ECC CYC may also induce aerobic adaptations [16,22,55]. Our study provides novel information showing that the difference between ECC CYC and CON CYC training effects (i.e., net effect) on . VO 2max is small (µ = 2%), with CON CYC being slightly more effective than ECC CYC . The meta-analysis of pre-to-post training effects showed 7% [95% CrI 0.5%, 12%] and 4% [95% CrI −7%, 15%] average increases in . VO 2max following CON CYC and ECC CYC training, respectively, which suggests that ECC CYC can also positively affect aerobic fitness. Indeed, we found that . VO 2max enhancement is quite likely (80% of probability) following ECC CYC training. Interestingly, group-level analysis of the population revealed greater probabilities of patients with cardiopulmonary disease (p < 0 = 69%) and obese adolescents (p < 0 = 58%) to benefit from greater increases in . VO 2max following ECC CYC compared to CON CYC training, while the estimated posterior densities for healthy individuals, amateur cyclists and sedentary participants showed smaller probabilities of greater increases in . VO 2max following ECC CYC compared to CON CYC training (p < 0 = 40%, 38%, and 10%, respectively). It is important to note that ECC CYC and CON CYC training protocols adopted in studies using sedentary participants were performed at the same PO [23,25], whereas the remaining sub-groups of group-level analysis of the population used ECC CYC and CON CYC training protocols performed at the same relative HR [19,24], at the same VO 2 [20,43], or similar RPE [26,44,45]. Thus, as previously discussed in the literature [22], ECC CYC sessions performed at the same PO as CON CYC sessions may produce an insufficient stimulus to improve the aerobic capacity of some populations since ECC CYC would be performed at a lower metabolic demand in such a condition.
The only meta-analyzed variable in this study for which CON CYC training induced considerably greater increases compared to ECC CYC was PPO. The results showed that the changes in PPO induced by CON CYC were 7% [95% CrI 2%, 11%] greater than the changes induced by ECC CYC training. Moreover, the posterior distribution of the estimated average effect indicated a 100% of probability of the CON CYC being more effective than ECC CYC in improving PPO. However, the meta-analysis of pre-to-post training effects showed an average increase of 11% [95% CrI 5%, 16%] following ECC CYC training, with a posterior probability of 100% of ECC CYC inducing a positive effect on PPO. Moreover, the meta-regression of conditional effects on PPO revealed an important effect of intervention duration in the difference between training modalities, which decreases as intervention duration increases. In other words, the effectiveness of the modalities in improving PPO tends to be similar as intervention duration increases. This may be associated with long-lasting muscle remodeling and delayed manifestation of gains in concentric muscle power following eccentric training regimens [56][57][58]. Nevertheless, the results obtained in the present study indicate that ECC CYC can increase PPO achieved during a CON CYC incremental test, but it is less effective than CON CYC training in increasing this variable.
One of the most discussed applications of ECC CYC is its implementation into exercise programs for special populations (i.e., frail individuals and patients with chronic diseases) [12,13]. In this context, the improvement in functional capacity is one of the main objectives of the exercise intervention. The distance covered in the six-minute walking test is an important measure of functional capacity and is associated with quality of life and longevity in older individuals and patients with chronic diseases [59]. The obtained data indicate that ECC CYC training induces greater increases of 6MWD compared to CON CYC training. Although the estimated difference between changes induced by ECC CYC and CON CYC on 6MWD was small (µ = 2%), the posterior distribution of the average population effect revealed a higher probability (p < 0 = 78%) of ECC CYC training inducing a greater increase in 6MWD compared to CON CYC . The meta-analyses of pre-to-post training effects showed an average increase in 6MWD of 11% [95% CrI 7%, 15%] versus 8% [95% CrI 5%, 12%] following ECC CYC and CON CYC training, respectively, with both modalities presenting high probabilities of being effective in increasing 6MWD. However, most studies included in the meta-analysis of 6MWD effects reported lower cardiovascular burden [18,44,45] and lower sensation of dyspnea [14,45] during ECC CYC compared to CON CYC training sessions. This can be considered an important advantage of ECC CYC interventions for exercise treatment of clinical populations. Importantly, all studies included in the meta-analysis of 6MWD effects involved patients with cardiopulmonary diseases. Therefore, the present evidence supports the utilization of ECC CYC training for the development of functional capacity in patients with cardiopulmonary diseases.
The reduction in body fat content is an important outcome in the treatment of chronic diseases and obesity [60]. For this purpose, aerobic exercises such as running and CON CYC are widely utilized [30]. The results of this study showed that ECC CYC training may be as effective as CON CYC training in reducing body fat percentage. The pre-to-post training average effects showed that both cycling modes induced~1% reductions in BF%. The studies included in the meta-analysis of BF% effects were conducted with obese adolescents [20] and patients with cardiopulmonary diseases [34,35]. Thus, it is still unknown if ECC CYC is effective in decreasing BF% in healthy, active individuals. It has been suggested that metabolic substrate utilization during eccentric exercise differs from concentric exercise, with increased fat oxidation rate and reduced glucose utilization during eccentric modalities [61]. Conversely, there is evidence showing similar utilization of energetic substrate between ECC CYC and CON CYC sessions when performed at the same VO 2 [62]. The studies included in the analysis of BF% changes involved ECC CYC and CON CYC interventions performed with similar metabolic demands (i.e., VO 2 ), which may be an important factor influencing the magnitude of BF% changes following ECC CYC [22]. To date, it is unclear whether ECC CYC is effective in decreasing BF% of all types of the population, but the present data support its utilization, at least for obese adolescents and patients with cardiopulmonary diseases.

Limitations
Some limitations should be considered when interpreting the current findings. The number of studies included in the meta-analysis of ICPT, IEPT, f-CSA, and BF% effects was small (4, 3, 3, and 3 studies, respectively), which may have affected the accuracy of the estimates of combined effects (i.e., average population effect) [63]. Moreover, the studies included in this review were conducted with different populations and training protocols. Thus, meta-regressions and group-level analyses were conducted to verify the impact of the population and intervention characteristics on effect sizes. However, meta-regression analyses were only possible for two variables (i.e., IPT and PPO), with . VO 2max and intervention duration as covariates, and group-level analyses were conducted considering the healthy condition of the participants and intervention duration as random effects. Hence, it was not possible to verify the impact of other populations and intervention characteristics in the chronic adaptations to ECC CYC compared to CON CYC , such as the sex of the participants or the intensity used during training. Due to scarce evidence on the mechanisms underpinning the chronic adaptations to ECC CYC , it was necessary to interpret and discuss some of our results in light of evidence from other eccentric exercise modalities. Additionally, there is a lack of studies investigating the chronic effects of ECC CYC in trained athletes and healthy, active individuals.

Future Perspectives
Currently, the positive benefits of eccentric exercise modalities in improving strength and muscle mass in different populations are widely recognized [2,3,10,12,51]. The present data extend our knowledge on the possible applications of eccentric muscle work, indicating that ECC CYC training may be used as a time-effective modality to improve distinctive performance, physiological, and morphological parameters. The results of this study also indicate that the effectiveness of ECC CYC training in inducing different adaptations, such as improving neuromuscular function and aerobic power at the same time, may be influenced by the fitness level of the participants and the intensity of the exercise sessions. Due to the unique force production-to-energy demand relationship of ECC CYC , the metabolic disturbances during low-demanding ECC CYC sessions (e.g., ECC CYC sessions performed at the same PO of CON CYC ) may not be sufficient to trigger aerobic adaptations in sedentary individuals [25]. On the other hand, ECC CYC prescribed at a metabolic load similar to that elicited by CON CYC training was effective in improving both neuromuscular function and aerobic power parameters in clinical patients [19,43,45]. Future investigations should address the impact of different training intensities in the adaptations promoted by ECC CYC training in distinct populations to optimize ECC CYC prescription and individualization.

Conclusions
The current evidence indicates that ECC CYC is a feasible modality for exercise interventions aiming to improve parameters of muscle strength, hypertrophy, functional capacity, aerobic power, as well as body composition, with greater effectiveness than CON CYC training in improving strength-related variables. Additionally, ECC CYC may be more advantageous than CON CYC training in improving the aerobic power of patients with cardiopulmonary diseases. Therefore, ECC CYC training can be integrated as a time-effective modality in exercise interventions aiming to improve key physical/physiological param-eters associated with good quality of life and healthy aging, such as leg muscle strength, aerobic power, and whole-body fat content. Furthermore, ECC CYC constitutes a feasible alternative to CON CYC for exercise treatment of patients with cardiopulmonary disease given its tolerability and greater effectiveness.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph20042861/s1, Figure S1: Sensitivity analyses; Figure S2: Forest plot of effect sizes on isokinetic concentric torque; Figure S3: Forest plot of effect sizes on isokinetic eccentric torque; Figure S4: Forest plot of effect sizes on fiber cross-sectional area; Figure