Muscle hypertrophy and ladder‐based resistance training for rodents: A systematic review and meta‐analysis

Abstract This study aimed to review the effects of ladder‐based resistance training (LRT) on muscle hypertrophy and strength in rodents through a systematic review with meta‐analysis. We systematically searched PubMed/Medline, SportDiscuss, Scopus, Google Scholar, Science Direct, and Scielo database on May 18, 2020. Thirty‐four studies were included measuring total (mCSA) or mean muscle fibers cross‐sectional area (fCSA) or maximum load‐carrying capacity (MLCC) or muscle mass (MM). About the main results, LRT provides sufficient mechanical stimulation to increase mCSA and fCSA. Meta‐analysis showed a significant overall effect on the fCSA (SMD 1.89, 95% CI [1.18, 2.61], p < .00001, I 2 = 85%); however, subgroup analysis showed that some muscle types might not be hypertrophied through the LRT. Meta‐analysis showed a significant training effect on the MM (SMD 0.92, 95% CI [0.52, 1.32], p < .00001, I 2 = 72%). Sub‐group analysis revealed that soleus (SMD 1.32, 95% CI [0.11, 2.54], p = .03, I 2 = 86%) and FHL (SMD 1.92, 95% CI [1.00, 2.85], p < .0001, I 2 = 71%) presented significant training effects, despite moderate heterogeneity levels (I 2 = 72%). MLCC increases considerably after a period of LRT, regardless of its duration and the characteristics of the protocols (SMD 12.37, 95% CI [9.36, 15.37], p < .00001, I 2 = 90%). Through these results, we reach the following conclusions: (a) LRT is efficient to induce muscle hypertrophy, although this effect varies between different types of skeletal muscles, and; (b) the ability of rodents to carry load increases regardless of the type and duration of the protocol used.

selection appears to be fundamental to the success of the training protocol, especially concerning its efficiency in stimulating muscle hypertrophy (Kubo, Ikebukuro, & Yata, 2019).
On the other hand, resistance training for rodents has always been a reason for debate (Cholewa et al., 2013;Krause Neto, Silva, Ciena, Anaruma, & Gama, 2016;Strickland & Smith, 2016). Such controversy was generated from the apparent differences found between the equipment used for the training of rodents and humans. However, since the ladder-based resistance training model (LRT) proposed by Hornberger and Farrar (2004), a large amount of evidence has emerged, and greater control of training variables has been investigated and better controlled in experimental resistance training (Krause Neto et al., 2016Luciano et al., 2017;Tibana et al., 2017). Despite this, there is still some doubt as to the efficiency of this training model in stimulating muscle hypertrophy of different muscle types in young and adult rats (Hornberger & Farrar, 2004;Son et al., 2016).
In 2004, Hornberger and Farrar investigated the effects of 8 weeks of LRT on the total muscle cross-sectional area (mCSA), muscle mass, the total, and myofibrillar protein of the soleus, plantaris, flexor hallucis longus (FHL), gastrocnemius, and quadriceps femoris muscles of young rats. After collecting and analyzing the results, the authors demonstrated that only the FHL muscle showed a statistical change in these parameters. This fact raised a hypothesis about a probable inefficiency of this training model in stimulating muscle hypertrophy of more types of muscles, thus limiting the study on other outcomes. However, more recent evidence has shown that some of these outcomes, such as the measurement of muscle mass, might not be effective in showing the real effect of training and thus affect the interpretation of data and the efficiency of the training model. According to Tibana et al. (2017), despite no increased muscle mass, LRT is efficient in inducing a significant increase in the mean muscle fibers cross-sectional area (fCSA) and the capacity of the rodent to carry the load. Yet, other studies corroborate these conclusions, demonstrating that LRT might stimulate hypertrophy of several different types of muscles (Jung et al., 2015;Know, Jang, Cho, Jang, & Lee, 2017;Lim, Gil, Quan, Viet, & Kim, 2018;Luciano et al., 2017;Ribeiro et al., 2017).
Due to the variability of results found in the literature, we find it convenient to carry out a more in-depth investigation of the research associated with the LRT model. Therefore, the present study aimed to review the effects of LRT on muscle hypertrophy in rodents through a systematic review with meta-analysis.

| METHODS
This systematic review followed the PRISMA guideline [preferred reporting items for systematic reviews and meta-analysis] (Liberati et al., 2009). This guideline is widely used in systematic reviews of clinical trials; however, it is possible to adapt this instrument to systematically review the literature covering animal studies (Hooijmans et al., 2014;Hooijmans, de Vries, Rovers, Gooszen, & Ritskes-Hoiting, 2012; Krause Neto, Ciena, Anaruma, de Souza, & Gama, 2015). Therefore, on May 18th, 2020, we searched the PubMed/Medline, SportDiscuss, Scopus, Google Scholar, Science Direct and Scielo databases, using the following Mesh and entry terms and additional key words:((((resistance training OR training, resistance OR strength training OR training, strength OR weight-lifting strengthening program OR strengthening program, weight-lifting OR strengthening programs, weight-lifting OR weight lifting strengthening program OR weight-lifting strengthening programs OR weight lifting exercise program OR exercise program, weight-lifting OR exercise programs, weight-lifting OR weight lifting exercise program OR weight-lifting exercise programs OR weight-bearing strengthening program OR strengthening program, weight-bearing OR strengthening programs, weight-bearing OR weight bearing strengthening program OR weight-bearing strengthening programs OR weight-bearing exercise program OR exercise program, weight-bearing OR exercise programs, weight-bearing OR weight bearing exercise program OR weight-bearing exercise programs OR resistance exercise))) AND (hypertrophy OR hypertrophies OR cell enlargement OR enlargement, cell OR cell size growth OR growth, cell size OR growths, cell size OR size growth, cell OR cell growth in size OR crosssectional area OR muscle strength OR force OR strength OR muscular strength OR torque OR muscular endurance OR lifting OR muscle contraction OR contraction, muscle OR contractions, muscle OR muscle contractions OR muscular contraction OR contraction, muscular OR contractions, muscular OR muscular contractions)) AND ((rats OR Rat OR rattus OR rattus norvegicus OR rats, norway OR rats, laboratory OR laboratory rat OR laboratory rats OR rat, laboratory OR mice OR Mus OR mouse OR mus musculus OR mice, house OR house mice OR mouse, house OR house mouse OR mus domesticus OR mus musculus domesticus OR domesticus, mus musculus OR mice, laboratory OR laboratory mice OR mouse, laboratory OR laboratory mouse OR mouse, swiss OR swiss mouse OR swiss mice OR mice, swiss OR wistar rats)).

| Inclusion and exclusion criteria
For inclusion of articles, the following criteria were followed: (a) samples composed of rats aged 2-13 months; (b) resistance training protocol performed on the ladderbased equipment (at least 1-m height); (c) outcomes that included quantification of the mCSA OR fCSA OR the mass of the skeletal muscles OR the quantification of the maximum carried load measured by muscular endurance test OR the quantification of volume differences of load trained between the first and last training sessions, and (d) having a control group not submitted to the training model.
We excluded all articles that investigated exercise effects on mice, training without additional load, genetically modified animals, interventions such as surgery, muscle unloading or electrical shock stimulation, use of any drug or food supplement, different types of animal diet and studies with insufficient data or that used old animals (above 16 months of age at the beginning of the intervention). Rat lineage was not stated as an inclusion criterion.

| Studies selection
The selection of studies was conducted by independent researchers (IL, WKN, LSPA, VMMO, VLG, and GHSF). After reading the titles and abstracts, a meeting determined the number of studies included for the full-text analysis. A week later, investigators met again to identify the final number of studies included and to resolve any conflict of opinion about the selection process. Upon completing the ultimate selection of studies, an analysis of quality and risk of bias was initiated. When necessary, the corresponding author of the study was contacted to request further information.

| Analysis of data quality, assessment of risk of bias, and publication bias
We assessed the risk of bias of the included studies using a questionnaire described elsewhere (Hooijmans et al., 2014;Hooijmans et al., 2012). We based these criteria on the possible presence of selection bias (questions 1, 2, and 3), performance bias (questions 4 and 7), detection bias (questions 5, 6, and 8), and attrition bias (questions 9 and 10). The quality analysis and risk of bias were independently assessed by two reviewers (IL and WKN), using predefined judging criteria (Hooijmans et al., 2012). The scores "Yes" indicate a low risk of bias; the score "No" indicates a high risk of bias, "Unclear" indicates an unknown risk of bias. To detect publication bias, funnel plots were created.

| Data extraction
We extracted data about rodent lineage, age, gender, training parameters (MLCC and training protocols), primary outcomes, and main results.

| Outcomes
The primary outcome was muscle hypertrophy [mCSA and fCSA]. The secondary outcomes were individual muscle mass (MM) and maximum load-carrying capacity (MLCC).

| Data synthesis and meta-analysis
Systematic review data were organized in Tables 1 and 2. For the meta-analysis, the mean and standard deviation values were extracted from each outcome. Studies that investigated the effects of LRT on more than one muscle type per result had a sequential number added to their identification in the forest plots (i.e., Padilha et al., 2019). The number of muscle samples analyzed in each study was added to the forest plots as a sample number. Meta-analysis was applied for fCSA, MM, and MLCC. mCSA data was insufficient to run meta-analysis. For statistical analysis, we used review manager software 5.3 to calculate the standardized mean difference ([SMD], the mean of the experimental group minus the mean of the control group divided by the pooled SD of the two groups), 95% confidence interval (95% CI) and heterogeneity by the I 2 , Chi 2 , and Tau 2 values. We used I 2 to assess heterogeneity between studies using randomeffect models (I 2 values <50% indicate low heterogeneity, 50%-75% moderate heterogeneity, and >75% high level of heterogeneity). Analysis of subgroups was applied as necessary. For the overall effect, p ≤ .05 was considered statistically significant.

| Description of the included studies
After the initial search, we identified 1,574 articles titles. From this point, independent evaluators read all titles and abstracts, selecting 87 articles for full-text analysis. After inclusion and exclusion criteria, 34 papers were included for systematic review (Figure 1).
Tables 1 and 2 present the overall data and main results of each selected article. Sprague-Dawley, Wistar, and Fisher 344 rats were the rodent lineages used within the studies. A total of 544 rats (434 male and 110 female) were included summing all studies. Rodent sample number per group varied from 4 to 18 rats. Twenty-three articles described the rat's initial body mass (Antonio-Santos et al., 2016;Cassilhas et al., 2013;Deus et al., 2012;Deschenes et al., 1994;Deschenes, Sherman (Deschenes et al., 1994;Lim et al., 2018), flexor digitorum profundus [FDP] (Chi et al., 2020;Kwon et al., 2018), flexor digitorum longus [FDL] (Cassilhas et al., 2013), tríceps brachialis (Nascimento et al., 2013) and bíceps brachialis (Souza et al., 2014). Figure 2 shows the average results of the risk of bias assessment. In all, 73.53% of the studies stated that the allocation of experimental units to treatment groups was randomized. However, only one study mentioned the randomization method (Padilha et al., 2019); nevertheless, all studies presented the division of groups in a similar way. None of the included articles described whether the allocation of groups during the randomization process was hidden or whether the caregivers knew which groups the animals were from. Only five studies reported having blinded the evaluation of results (Cassilhas et al., 2013;Deschenes et al., 2000Deschenes et al., , 2015Luciano et al., 2017;Tibana et al., 2017). Also, 73.5% of studies did not describe whether there was randomization in the investigation of outcomes between groups. No study reported whether there was any sample loss throughout the training intervention. The quality scores varied between 4 and 6 points ("Yes" score). Only four studies scored 6 points (Cassilhas et al., 2013;Deschenes et al., 2015;Luciano et al., 2017;Tibana et al., 2017).

| Quality of reporting, risk of bias, and publication bias
The presence of publication bias was assessed for the outcomes fCSA, MM, and MLCC. Funnel plots demonstrated asymmetries for all three issues analyzed (Figures 3-5).
About EDL, one article showed fCSA hypertrophy (Lim et al., 2018), while others did not .
Due to the high degree of heterogeneity, subgroup analyses on individual skeletal muscles fCSA and training duration were applied.
About TA, two articles showed no change in MM (Lee et al., 2016;Neves et al., 2019).
Sensitivity analysis was applied to each skeletal muscle ( Figure 10). Subgroup analysis revealed that soleus (seven studies, SMD 1.32, 95% CI [0.11, 2.54], p = .03, I 2 = 86%) and FHL (seven studies, SMD 1.92, 95% CI [1.00, 2.85], p < .0001, I 2 = 71%) presented significant overall effect. However, no other muscle had a significant overall effect. The degree of heterogeneity for the included studies that assessed the MM of the TA and quadriceps femoris was considered low (I 2 = 0%). For all others, the degree of heterogeneity remained moderate-high.

| Maximum load carrying capacity
The MLCC protocol most described within the studies used an initial load equals to 75% of the rodent bodyweight with 30 g increases until failure (2 min interval between climbs). However, considerable variability of training protocols was found here. The most common training protocol was described by Hornberger and Farrar (2004) [progressive loading increases from 50%, 75%, 90%, and 100% body weight with subsequent 30 g increases until failure; 2 min interval; and training frequency of 3 days/week]. Training duration varied from 6 to 36 weeks between studies.

| DISCUSSION
This systematic review demonstrated the following main results: (a) the methodological quality of the studies that investigated the effects of LRT on muscle hypertrophy in rodents needs to be improved; (b) LRT provides sufficient mechanical stimulation to induce increases in mCSA and fCSA in most types of skeletal muscle; however, some muscle types with specific morphological and biochemical characteristics may not be hypertrophied through the LRT (e.g., soleus); (c) in general, the chronic response of LRT over skeletal MM seems to vary between different types of muscles; and (d) MLCC increases considerably after a period of LRT, regardless of its duration and the characteristics of the protocols.
The quality analysis of the included studies was considered moderate in this systematic review. Most studies reported that groups were randomized during interventions; however, only one study cited how this process was conducted (Padilha et al., 2019). We can assume that in many laboratories, it is common to randomize the groups only after the MLCC tests. This attitude aims to subject to training only animals that were able to climb the ladder. In this way, randomization is done for convenience. However, the lack of adequate information on this criterion may raise questions about the sample selection bias. Therefore, we suggest that the authors fully describe how to sample randomization processes are being conducted.
The lack of description regarding blinding analysis of muscle tissue samples also raises some degree of concern. Only five studies reported that the histological slides were F I G U R E 8 Forest plots of data examining the effect of ladder resistance training duration on mean muscle fiber cross-sectional area [fCSA] (produced in the review manager 5.3 software) coded for blind analysis of the results (Cassilhas et al., 2013;Deschenes et al., 2000Deschenes et al., , 2015Luciano et al., 2017;Tibana et al., 2017). The absence of blinding of the samples demonstrates a critical bias. Also, funnel plots showed moderate publication bias through asymmetries in the three outcomes investigated in this study. Specifically, in the fCSA result, three studies influenced this issue (Luciano et al., 2017;Prestes et al., 2012;Tibana et al., 2017). The use of a few rodents by groups and small measures of dispersion may have influenced this question.
The increase in MM and CSA, as a chronic response to resistance training, is not a new outcome in the literature. However, many questions were asked about the efficiency and effectiveness of training equipment and protocols, used in research with rodents, to stimulate significant changes in the structure and composition of skeletal muscles (Cholewa et al., 2013;Krause Neto et al., 2016). Here, we demonstrated that LRT is efficient to induce significant increases in mCSA and fCSA. Besides, we show that from the quantification of the total mCSA, it is possible to affirm that the total muscular cross-sectional area is larger in the groups of rodents submitted to LRT than in the control groups. Despite this, few skeletal muscles were quantified using this measure [FHL and plantaris]. In the studies in question, the total mCSA was estimated from calculations that took into account individual skeletal muscle mass, muscle fiber length, and muscle density (Hornberger & Farrar, 2004;Lee & Farrar, 2003). This outcome is capable of providing, at least indirectly, an adequate measure to estimate muscle hypertrophy in rodents. However, our analysis identified that not all muscle types show greater fCSA. This fact led us to indicate that, as is done in humans, the idea is to quantify both mCSA and fCSA. Also, few studies have investigated the effects of LRT on the type of muscle fibers alone (Deschenes et al., 2000(Deschenes et al., , 2015. This situation can also converge to an interpretative error since it is possible to measure larger mCSA without a uniform change in the typology of muscle fibers (Bjørnsen et al., 2019).
The capacity for muscle hypertrophy depends fundamentally on the amount of mechanical stimulation imposed on skeletal muscles. Here, we demonstrate that there is great variability in the types of resistance training protocols being used by researchers. Interestingly, it seems that rats, like humans, are more susceptible to the volume of training (series × reps × load) than the level of effort imposed by the session (Lasevicius et al., 2019;Luciano et al., 2017;Lacerda et al., 2019;Tibana et al., 2017). Tibana et al. (2017)  greater fCSA compared to the control group. However, the group with the highest volume (8 climbs) had the largest fCSA averages [Control = 1,800 ± 30; RT4 = 2,650 ± 60; RT8 = 3,050 ± 125]. Recent evidence has shown that LRT can stimulate muscle hypertrophy by increasing the phosphorylation of proteins such as mTOR (mammalian target of rapamycin), p70S6k (p70S6 kinase 1) and MyoD (myoblast determination protein 1) of the gastrocnemius (Ribeiro et al., 2017). Also, these same authors reported that in addition to the increase in cell signaling pathways for anabolism, there was also a reduction in the phosphorylation of proteins associated with muscle catabolism. Corroborating these data, Luciano  (2017) demonstrated that larger total loads (volume × intensity) are probably necessary to stimulate the greatest mean increases in fCSA, phosphorylation of mTOR, and their respective regulatory enzymes.
When analyzing the chronic response of each muscle, we demonstrated that the soleus, EDL, and plantaris do not seem to respond with the same magnitude of muscle hypertrophy as other muscles. One of the probable explanations for this fact can be directly associated with the particularities of each training protocol. Deschenes et al. (2015) failed to demonstrate a substantial increase in fCSA of soleus and plantaris muscles after seven weeks of LRT, 3x/week, F I G U R E 1 1 Forest plots of pre-post training data examining the effect of ladder resistance training on maximum load-carrying capacity [MLCC] (produced in the review manager 5.3 software) using a protocol with 10 submaximal climbs. On the other hand, Krause Neto and Gama (2017) found higher averages of fCSA of the soleus and EDL muscles compared to the control group (5×/week, six climbs, 16 weeks). These divergences lead us to raise the hypothesis that some types of muscle may need higher training loads, while others need more significant volumes of training to hypertrophy. When analyzing these two cited studies, it is possible to verify that Deschenes et al. (2015) trained the rats for 21 sessions, while Krause Neto and Gama (2017) submitted their animals to 80 training sessions. Taking into account, it is plausible to suggest that some types of muscle may need more time to show higher hypertrophy levels than others. Recently, Padilha et al. (2019) demonstrated that the soleus muscle was responsive only to the protocol with the highest number of climbs (8-16 climbs/session).
On the other hand, muscles, such as plantaris and FHL, similarly hypertrophy in both types of training volume (high vs. moderate), showing a significant increase in protein synthesis. These data lead us to suggest that the results obtained here may have been influenced by the predominance of the type of muscle fiber in each muscle. However, due to the small number of studies that quantified the different types of muscle fibers, we were unable to investigate further each of these relationships. Despite this, Ribeiro et al. (2017) suggest that the lack of effect on soleus muscle hypertrophy could be directly linked to the inability of LRT to stimulate significant increases in muscle anabolism, even with a reduction in catabolism pathways. Also, weekly frequency, number of climbs, relative intensity, and duration of training are variables that can, in a certain way, affect the response of each type of muscle individually.
The mass of each isolated muscle is currently used as a marker of muscle hypertrophy in experimental models. However, our review results demonstrated that there is not necessarily a general relationship between the mass of individual muscles and the increase in CSA of muscle fibers. Despite this, the FHL muscle appears to demonstrate significant improvements in both fCSA and mCSA, in addition to its muscle mass (Hornberger & Farrar, 2004;Lee et al., 2016Lee et al., , 2018Lee & Farrar, 2003). On the other hand, muscles like soleus seem to demonstrate a significant increase in their muscle mass, without necessarily affecting fCSA. This fact can, in part, be explained by a probable edematous muscle swelling induced by resistance training (Damas et al., 2016). Still, an inverse relationship can be seen regarding the TA and quadriceps femoris muscles, whose muscle mass is not different from those found in the control groups, but demonstrate more significant hypertrophic responses in their fCSA.
Finally, we demonstrated that MLCC increases regardless of the protocol used and the duration of the training. This interesting fact is easily explained since the animals have a high degree of sedentary lifestyle during the period of accommodation. Thus, by placing the animal under the physical F I G U R E 1 2 Forest plots of pre-post training data examining the effect of ladder resistance training duration on maximum load-carrying capacity [MLCC] (produced in the review manager 5.3 software) stress of training, they obtain a rapid and significant functional gain. Corroborating, Deus et al. (2012) demonstrated that the LRT, without any additional load, is already sufficient to increase the load-carrying capacity by the rodent, having its function enhanced by increasing the training load. In addition, morphological adjustments to peripheral nerves may explain, in part, the increase in muscle strength without necessarily increasing the size of the muscle or its fibers (Carbone et al., 2017).

| CONCLUSION
The results obtained in this study led us to the following conclusions: (a) LRT is efficient in inducing hypertrophy of skeletal muscles, although this effect varies between different types of skeletal muscles, and; (b) the ability of rodents to carry load increases regardless of the nature and duration of the protocol used.