Cardiovascular disease accounts for one in four deaths (~23%) in the United States (Murphy et al., 2021). Cancer, Alzheimer’s disease, diabetes, and hypertension also rank within the top 15 causes of death in the United States (Murphy et al., 2021). These chronic diseases have been linked to low levels of cardiorespiratory fitness (Defina et al., 2013; Zhang et al., 2014; Zaccardi et al., 2015; Sui et al., 2017; Robsahm et al., 2019; Lee, 2021). The Centers for Disease Control and Prevention and the American College of Sports Medicine currently recommend that individuals participate in moderate-intensity physical activity for 150 or more minutes per week for optimal health (Haskell et al., 2007; Garber et al., 2011). Improving cardiorespiratory fitness through increased physical activity can significantly reduce the risk of all-cause mortality (Blair et al., 1995; Brawner et al., 2017; Davidson et al., 2018). Although the majority of health benefits associated with high cardiorespiratory fitness are mediated by changes in traditional risk factors such as blood pressure, inflammatory markers, and blood lipids, roughly 40% of the beneficial effects of exercise cannot be explained by traditional risk factors (Mora et al., 2007; Joyner and Green, 2009). Furthermore, the cellular and molecular mechanisms underlying the salutary effects of exercise are not well understood. Therefore, inbred and genetically modified mice are frequently used to investigate the integrative physiological responses to exercise and the molecular mechanisms responsible for the beneficial adaptations to exercise training.

There are three commonly used paradigms for exercise training in rodents – swimming, voluntary wheel running, and “forced” wheel or treadmill running – and each has been used to study the molecular basis of responses to acute exercise and chronic exercise training. Treadmill running and wheel running induce adaptations in mice associated with endurance exercise training (Allen et al., 2001; Kemi et al., 2002; De Angelis et al., 2004; Waters et al., 2004; Massett and Berk, 2005; Chow et al., 2007). However, the two paradigms are inherently different (Poole et al., 2020) such that the correlation between treadmill running performance and voluntary wheel-running performance among mouse strains is nominal (Allen et al., 2001; Lightfoot et al., 2001, 2004). One advantage of treadmill running as an exercise paradigm is that the total amount of work performed among all mice can be established by the investigator through the selection of exercise testing and training parameters. Unlike for humans, there are no published well-accepted standards for exercise training paradigms or levels of activity required for optimal changes in exercise capacity or other training adaptations (Fuller and Thyfault, 2021). The published exercise testing and training paradigms are quite variable (Kemi et al., 2002; Billat et al., 2005; Hoydal et al., 2007; Marcaletti et al., 2011; Ayachi et al., 2016; Petrosino et al., 2016). Therefore, the purpose of this review was to characterize the variation in exercise training protocols in mice and determine key training parameters involved in adaptations to exercise training. This review focuses on treadmill running because the training parameters can be more easily quantified and any potential recommendations regarding these parameters could be incorporated into future research utilizing treadmill-based exercise training in mice.

The protocol for systematic reviews of animal studies was used as a guide for this review and meta-analysis (de Vries et al., 2015).

The following terms were used to search PubMed and EMBASE databases: (((((((““Inbred Mouse Strains”” OR ““Inbred Strain of Mice”” OR ““Inbred Strain of Mouse”” OR ““Inbred Strains of Mice”” OR ““Mice, Inbred Strains””[MeSH Terms] OR ““Mice, Inbred Strains”” OR ““Mouse, Inbred Strain””) OR (““Mice””[MeSH Terms] OR ““Mice”” OR ““Mice, House”” OR ““Mice, Laboratory”” OR ““Mouse”” OR ““Mouse, House”” OR ““Mouse, Laboratory”” OR ““Mouse, Swiss”” OR ““Mus”” OR ““Mus domesticus”” OR ““Mus musculus”” OR ““Mus musculus domesticus”” OR ““Swiss Mice””))) NOT ((transgenic OR knockout OR db/db OR ob/ob OR mdx OR ApoE)))) AND (((((““Aerobic Exercise”” OR ““Exercise””[MeSH Terms] OR ““Exercise”” OR ““Exercise Training”” OR ““Exercise, Aerobic”” OR ““Exercise, Physical”” OR ““Physical Activity”” OR ““Running””[MeSH Terms] OR ““Running””))) OR (((““High-Intensity Intermittent Exercise”” OR ““High-Intensity Interval Training””[MeSH Terms] OR ““High-Intensity Interval Training”” OR ““Sprint Interval Training””)))))) AND treadmill) AND (sedentary OR control).” Additional abstracts were obtained from reference lists of potentially eligible articles. The search was competed in February 2020.

The main findings of this systematic review and meta-analysis of mouse exercise training studies are: (1) a relatively small number of studies incorporating exercise training report a “classical” measure of training efficacy; (2) many studies do not report complete information regarding the exercise training protocol; (3) the majority of exercise training studies utilize male mice only; (4) exercise training significantly increases measures of training efficacy; and (5) exercise prescription parameters do not explain a significant amount of variation between studies when changes in exercise performance are used as a marker for training efficacy.

Our systematic review identified 164 full-text articles that included a treadmill training protocol with untreated mice assigned to either a sedentary control group or exercise training group. Of these, approximately 35% included a “classical” marker of training efficacy. Increases in skeletal muscle enzyme activity, mitochondrial DNA, and/or changes in skeletal muscle fiber types are possible markers for adaptations to endurance exercise training (Booth et al., 2010). An increase in peak or maximal oxygen consumption is often considered the gold standard in human-based endurance exercise training studies. In animal studies, changes in exercise performance are typically used as a surrogate for maximal oxygen consumption (Fuller and Thyfault, 2021). Therefore, only studies including these or other well-known markers for exercise training adaptations were included (Holloszy and Coyle, 1984; Hellsten and Nyberg, 2015). The majority of studies that were excluded for lack of such a marker utilized body weight differences between sedentary and exercise-trained groups as a general marker for exercise training. Although lower body weights in the exercise training group might be related to increased physical activity, body weight differences alone do not necessarily indicate that the exercise training elicited beneficial biochemical and/or cardiorespiratory fitness adaptations. For purposes of replication and thorough analysis of the responses to exercise, exercise training studies should include all relevant information regarding the training protocol such as frequency, intensity, and duration (Booth et al., 2010). All protocol information was included in 48% of the studies. Treadmill velocity (38%) and incline (39%) were the most frequently omitted variables. Most reported frequency, session time and duration. Exercise intensity was reported in 68% of studies, but the basis for qualifiers low, moderate, and high were unclear. Treadmill velocity and incline were frequently omitted when exercise intensity as a percentage of maximum was reported. Collectively, these results indicate that treadmill-based exercise training studies in mice frequently do not report all the components of the exercise training program or well-accepted adaptations to exercise training as indicators of training efficacy.

Mouse strain, sex, and age have been reported to influence exercise training responses. Overall, these moderators had limited effects on exercise training responses. When outcome variables were divided into performance-based and biochemical outcomes, sex and mouse strain significantly influenced biochemical responses to training (Figure 9). Male mice had significantly greater biochemical adaptations to exercise training than female mice. In contrast, performance-based outcomes were somewhat greater in females than males, but not significantly so (Figure 5). In a direct comparison, Kemi et al. (2002) reported that V˙O_2max was significantly greater in trained female mice than in similarly trained male mice. Similarly, exercise training-induced cardiac hypertrophy was greater in female mice compared with males (Foryst-Ludwig et al., 2011). However, less than 10% of the included studies utilized female mice and only three full-text articles included both male and female mice (Kemi et al., 2002; Foryst-Ludwig et al., 2011; Abadi et al., 2013). Therefore, additional studies are needed investigating the responses to endurance exercise training in female mice as well as studies directly comparing responses in mice of both sexes.

The influence of mouse strain was not significant overall (Table 2), but was significant in studies measuring biochemical markers of exercise training. For subgroup analyses, strains were coded as C57BL/6 or “Other.” The “Other” group included data from 27 strains. As with sex comparisons, only three articles included data from multiple mouse strains (Massett and Berk, 2005; Avila et al., 2017; Kim et al., 2020). Each of those publications reported significant strain-dependent changes in exercise capacity in response to exercise training. However, those findings were not supported by the results of the current study for performance-based outcomes. One possible explanation for this disparity is the “Other” strain category is composed of too many individual strains, leading to a high level of variation across subject populations and training protocols. However, the precision and dispersion of the effect estimates are similar for both C57BL/6 and “Other” subgroups suggesting that the variability in response to training is comparable. Thus, the strain-dependent differences in changes in exercise capacity with exercise training reported by Massett and colleagues (Massett and Berk, 2005; Avila et al., 2017; Kim et al., 2020) might be specific to the exercise training and testing paradigm used in those studies. Each of those studies utilized similar exercise training parameters with some strain-specific adjustments which facilitated direct comparisons with minimal variation between training protocols. Conversely, responses in C57BL/6 mice were significantly greater than other strains for biochemical markers of exercise training. This result implies that C57BL/6 mice show greater biochemical adaptations to exercise training than mice from other strains. This contrasts with performance-based outcomes where C57BL/6 mice have low to moderate responses to training compared with other strains (Massett and Berk, 2005; Avila et al., 2017; Kim et al., 2020). Future research investigating the effect of mouse strain on exercise training responses should consider including multiple strains within the same study design and measuring both performance-based and biochemical markers of training efficacy. Collectively, the findings regarding the contribution of sex and mouse strain on responses to exercise training suggest that direct comparisons within a given experimental design might yield results different from those obtained in a pooled analysis of the published studies utilizing individual mouse strains.

Overall, exercise training elicited significant increases in exercise training-associated outcomes. Heterogeneity was high for the combined analysis as well as for outcome-specific analyses. Therefore, subgroup analyses were performed for the combined data and for performance and biochemical outcomes separately. In the combined analysis, the greatest percentage of variation in the effect size was explained by exercise test protocol subgroups (Table 2). Exercise training parameters of frequency, treadmill velocity and incline, exercise session time, and training duration also were investigated to determine their contribution to the heterogeneity between studies/as potential moderator variables. In the overall analysis of 105 studies and in the separate analyses based on training outcome, subgroup analysis was significant for training duration (Table 2; Figures 5, 9). Studies utilizing a training duration of 5–8weeks had significantly greater outcomes than those incorporating longer or shorter periods. Typically, exercise training protocols include one or more weeks during which training time/intensity is progressively increased until the target parameters are reached. For studies ≤4weeks, the target workload might be sustained for too short a period (e.g., 2weeks) to elicit maximal training responses. Protocols longer than 8weeks showed effects comparable to those ≤4weeks, implying that longer duration training protocols might hinder adaptations to training. The mechanism for this is unclear, but declines in performance with prolonged training, especially at higher intensities, can be associated with overtraining syndrome (Pereira et al., 2012; Meeusen et al., 2013). These results suggest that regardless of training outcome a training duration of 5–8weeks appears optimal for adaptations to treadmill training in mice.

Significant subgroup differences for treadmill incline were observed in the combined analysis and for performance-based outcomes (Table 2; Figures 5, 6). Exercise training protocols utilizing an incline above 10° had the largest effect on training outcomes in the combined and performance-based analyses (Table 2; Figures 5, 6). The higher incline should require more work/greater effort and therefore, might elicit greater adaptations to training (Hoydal et al., 2007; Poole et al., 2020). Kemi and colleagues (Kemi et al., 2002; Hoydal et al., 2007) reported that the best estimates of mouse V˙O_2max are obtained during treadmill exercise at inclines between 15° and 35°. They demonstrated significant improvements in maximal oxygen consumption after 8weeks of training at 25° in male and female C57BL/6J mice (Kemi et al., 2002). Therefore, they recommended an inclination of 25° as optimal for exercise training. However, Petrosino et al. (2016) limited the treadmill incline to 15° in the development of their exercise testing protocol because they observed that mice had difficulty maintaining gait at inclinations above 15°. Although gait changes during treadmill running in rodents, including raising of the snout and lowering the hindquarters, can occur prior to exhaustion (Copp et al., 2009), it is unclear if similar gait changes occur at higher treadmill inclinations in the absence of fatigue. Only five studies utilized an incline >10° (Kemi et al., 2002; Kaurstad et al., 2012; Herbst et al., 2015; Alves et al., 2020) and no direct comparisons of treadmill incline on training responses were included in those articles. Therefore, additional research is required to confirm that treadmill inclination above 10° should be utilized for exercise training programs.

The other exercise training parameter showing subgroup differences was exercise time/session for biochemical-based outcomes. Exercise time per session varied from 30 to 90min in the biochemical outcome group. One study with an exercise time >60 and a biochemical outcome showed a very large effect of exercise training (Vihko et al., 1979), but a wide 95% CI. Exercise times ≤30min had a greater effect on training responses than those with training session times between 31 and 60min. The effect size for the ≤30min subgroup also was the largest in the combined group analysis, however, there were no significant subgroup differences between exercise times in the overall analyses. Given the relatively small number of studies per subgroup, additional research is required to confirm that shorter exercise training sessions might elicit greater biochemical training adaptations than longer individual exercise sessions.

Significant subgroup differences were observed for the exercise test used to assess changes in performance. The three most common testing protocols were increasing treadmill velocity at a fixed incline (ILT), increasing both treadmill velocity and incline at fixed intervals (GXT), and tests measuring maximal oxygen consumption (V˙O_2max). Time or distance were typically used to assess performance in the ILT and GXT tests. The largest effect of exercise training was observed for V˙O_2max tests (Figures 5, 8). This subgroup included six studies and had a relatively wide 95% CI (1.44–3.96). The testing protocols used to measure mouse V˙O_2max are like the protocols for ILT, increasing speed at a constant incline. But the criteria for reaching V˙O_2max, e.g., a plateau in V˙O₂ with increasing workload and respiratory exchange ratio above 1.0, is more clearly defined than those for ILT and GXT tests (e.g., time touching the shock grid or number of shocks; Poole et al., 2020). Therefore, V˙O_2max tests were placed in a separate category from ILT. The effect size for ILT was greater than that for GXT. The GXTs were primarily used by one group (Massett and Berk, 2005; Avila et al., 2017; Kim et al., 2020) and all the study protocols were 4weeks in length. In contrast, average training duration in studies utilizing ILT protocols was 7.4±2.8weeks. Shorter duration exercise training was associated with smaller responses to exercise training (Figures 5, 9) and thus, might explain some of the differences between studies utilizing ILT vs. GXT protocols. It is unclear whether differences between GXT and ILT test protocols would be observed if training programs were matched for duration. Furthermore, the combination of increasing treadmill incline and speed throughout the GXT test results in larger increases in exercise intensity at specific stages. In humans, GXT-type tests result in less uniform increases in physiological responses and more variable estimates of exercise capacity and/or oxygen consumption (Myers et al., 1991; Pescatello et al., 2014). Similar results in mice might lead to inaccurate exercise prescription and subsequently less than optimal responses to exercise training (Hoydal et al., 2007).

To further explore the contribution of moderator variables on exercise training effects, meta-regression was used to determine the role of individual variables as well as multiple variables on variation across studies. The models tested included frequency of training, treadmill velocity and incline, time per session, and training study duration to identify the exercise prescription variables most related to exercise training outcomes. The model including all these variables accounted for 0% of between-study variance when all studies were considered. When studies were divided by outcome variables, this same model did not explain any of the between-study variance for performance-based outcomes (R²=0.0). In contrast, this same model explained 100% of the variance for studies reporting a biochemical outcome for exercise training despite no individual variable having a p<0.05. The general recommendation for meta-regression is that 10 studies should be included for each moderator variable (Baker et al., 2009). The number of studies included in the meta-regression analysis for all studies met this recommendation. However, the number of studies included with biochemical outcomes was small and therefore, the strength of the association should be interpreted with caution. Nevertheless, these results suggest that biochemical outcome variables are more strongly related to exercise training program components than are performance-based outcomes. This association implies that biochemical measurements should be incorporated into exercise training studies to provide evidence of training efficacy. This recommendation was proposed previously (Booth et al., 2010) but comes with the caveat that many of these measurements are invasive and require terminal procedures (Handschin et al., 2010).

The measurement of exercise performance in mice, including V˙O_2max, is somewhat controversial. Versions of different protocols for measuring V˙O_2max in mice have been proposed in the literature, each with varying levels of evidence to support the protocol (Kemi et al., 2002; Marcaletti et al., 2011; Ayachi et al., 2016; Petrosino et al., 2016; Lemaire et al., 2017). In addition, the validity of surrogates for V˙O_2max (e.g., time to exhaustion) as estimates of exercise capacity have been questioned because of the subjective nature of the definition of volitional fatigue and/or exhaustion (Booth et al., 2010; Fuller and Thyfault, 2021) and issues with repeatability (Knab et al., 2009). Knab et al. (2009) speculated that repeatability of exercise performance measures during a maximal exercise test in mice might be related to the outcome variable and the investigator’s definition of maximum. In contrast, biochemical outcomes are laboratory-based measurements with quantitative outcomes which might lead to less subjective interpretation of the outcome variable. Although some variation is likely associated with biochemical markers (Lonbro et al., 2019), standardized measurement procedures could reduce intra- and inter-investigator variation. Therefore, changes in these variables might demonstrate more consistent responses to a specific exercise intervention.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

HK, CM, and MM reviewed the abstracts, titles, and full text, extracted and reviewed the data, and drafted, edited, and revised the manuscript. MM analyzed the data. All authors contributed to the article and approved the submitted version.

This work was supported by funds from the Texas Tech University Office of Vice President for Research.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.