Mechanical Properties of Treadmill Surfaces Compared to Other Overground Sport Surfaces

The mechanical properties of the surfaces used for exercising can affect sports performance and injury risk. However, the mechanical properties of treadmill surfaces remain largely unknown. The aim of this study was, therefore, to assess the shock absorption (SA), vertical deformation (VD) and energy restitution (ER) of different treadmill models and to compare them with those of other sport surfaces. A total of 77 treadmills, 30 artificial turf pitches and 30 athletics tracks were assessed using an advanced artificial athlete device. Differences in the mechanical properties between the surfaces and treadmill models were evaluated using a repeated-measures ANOVA. The treadmills were found to exhibit the highest SA of all the surfaces (64.2 ± 2; p < 0.01; effect size (ES) = 0.96), while their VD (7.6 ± 1.3; p < 0.01; ES = 0.87) and ER (45 ± 11; p < 0.01; ES = 0.51) were between the VDs of the artificial turf and track. The SA (p < 0.01; ES = 0.69), VD (p < 0.01; ES = 0.90) and ER (p < 0.01; ES = 0.89) were also shown to differ between treadmill models. The differences between the treadmills commonly used in fitness centers were much lower than differences between the treadmills and track surfaces, but they were sometimes larger than the differences with artificial turf. The treadmills used in clinical practice and research were shown to exhibit widely varying mechanical properties. The results of this study demonstrate that the mechanical properties (SA, VD and ER) of treadmill surfaces differ significantly from those of overground sport surfaces such as artificial turf and athletics track surfaces but also asphalt or concrete. These different mechanical properties of treadmills may affect treadmill running performance, injury risk and the generalizability of research performed on treadmills to overground locomotion.


Introduction
Treadmills are widely used in different settings including sports training, exercise testing, rehabilitation and research [1]. Although it is frequently assumed that locomotion on a treadmill is a surrogate for ground locomotion, there is controversy as to the comparability of the biomechanical, physiological, perceptual or performance outcomes between the two conditions [1][2][3].
Insufficient familiarization and a lack of air resistance can make treadmill running differ from running overground [4][5][6]. However, there is recent meta-analytical evidence that differences can still be found between the two conditions independent of previous familiarization [3] and that the effect of air resistance becomes a signifcant confounder only at relatively high running speeds-approximately above 16 km/h, which is actually faster than the speeds used in most studies in the feld [1]. Factors other than familiarization or air resistance might thus be involved. In this regard, the role of the belt dimensions and intra-belt speed fuctuations remains largely unclear but might be relatively small for modern treadmills with strong driving mechanisms that provide minimal intra-stride belt speed variability, including high-quality research-based treadmills [3]. On the other hand, the controversy in the feld regarding the comparison of treadmill vs. overground running could also be caused by dissimilarities in the mechanical properties of the running surfaces used in the different studies [2,3,7,8]. Indeed, treadmills' mechanical properties have an important infuence-and in fact, greater than that of the lack of air resistance-on physiological responses [2,9] and can also affect running biomechanics [3], since athletes adjust their leg stiffness and dynamics when running on surfaces with different mechanical properties [10][11][12][13].
Although the mechanical properties of many sport surfaces (e.g., artifcial turf pitches, athletics tracks, sports hall foors, tennis courts and gymnastic crash mats) are frequently assessed to ensure they meet the criteria established by sport international federations and other governing bodies [14], this is not the case for treadmill surfaces, for which there are yet no standardized criteria. In this sense, current regulations (both European and American) defne constructive and general safety aspects without any mention of the mechanical properties of the surface [15][16][17]. The same limitation applies to the bulk of scientifc research comparing treadmill and overground locomotion [3].
Assessing the mechanical properties of treadmill surfaces is therefore an important issue, not only in sports but also from a clinical perspective. Indeed, treadmill surfaces' mechanical properties have a signifcant infuence on peak plantar forces and metabolic energy consumption [8,18], and treadmill running has been associated with a lower risk of developing tibial stress fractures but an increased risk of overload injuries at the Achilles tendon compared to overground running [19][20][21], due to altered lower-extremity kinetics and kinematics.
Generally, regulations require that the three main mechanical properties of sports surfaces-shock absorption (SA), vertical deformation (VD) and energy restitution (ER)-are evaluated [22,23]. However, the few studies that have characterized treadmills' mechanical properties in any way have mainly focused on surface stiffness [18,24]. Although stiffness is closely related to VD, it provides little information regarding SA and ER. In this context, and given that the mechanical properties of treadmills remain largely unknown, the main purpose of this study was to characterize SA, VD and ER among different treadmill models designed for ftness, research and rehabilitation purposes, and to compare the results with those obtained for other man-made surfaces typically used in sports-artifcial turf and athletics track surfaces. In addition, the relationship between the different mechanical properties can provide a more comprehensive understanding of the behavior of the surface and its infuence on athletes. Although these relationships have been previously studied in overground surfaces, they remain largely unknown for treadmills. Therefore, a second aim was to assess the relationship between SA, VD and ER and whether this relationship remained consistent across surfaces.

Sample
A total of 77 treadmills, 30 artifcial turf pitches and 30 track and feld tracks were included in the study. The treadmills comprised 70 conventional fat treadmills from ftness centers (ft-TR), 6 non-instrumented treadmills from different research laboratories (lab-TR), and one curved non-motorized treadmill (NM-TR) ( Table 1). Artifcial turf and track samples were selected randomly from a database of feld tests performed by a certifed laboratory.

Procedures
We assessed SA, VD, and ER with an advanced artifcial athlete (AAA) device (Wireless Value; Emmen, The Netherlands) that consists of a mechanical drop test simulating the support of an athlete's foot on the ground. The characteristics of the apparatus are thoroughly described in Section 12 of current FIFA standards [23], the model used here being a wireless handheld device that provided ease of operation and simple and fast measurements. Artifcial turf and track surfaces were assessed at different locations in accordance with current FIFA and World Athletics protocols, respectively [23,25]. For that, we performed three repetitions of the drop test at each test location, with intervals of 30 ± 5 s. We discarded the results of the frst test and calculated the mechanical properties of each location as the mean values of the second and third tests. The treadmills were assessed at three points as described elsewhere [26], performing only one drop test per location. For each surface included in the study, we calculated the SA, VD, and ER as the mean values of all the test locations.

Statistical Analysis
Data are presented as means and standard deviations (SDs). We used the Kolmogorov-Smirnov and Levene's test to check the normality of the data distribution and homogeneity of variances, respectively. We compared mechanical properties across the three types of surfaces (ft-TR, artifcial turf and athletics track) with a one-way analysis of variance (ANOVA) test, with the Bonferroni test used for post hoc pairwise comparisons. We used the same approach to compare the mechanical properties within the different ft-TR models. We calculated the effect size for the group effect (ES) with 2 the partial Eta-squared (η p 2 ) value with the following interpretation: small (η p = 0.01-0.059), medium 2 2 (η p = 0.06-≥ 0.14) and large effects (η p > 0.14). Finally, we also calculated the Pearson's correlations between the three mechanical properties within each type of surface. We used the statistical software SPSS V24.0 for Windows and set the level of signifcance at p < 0.05.

Results
We excluded lab-TR and NM-TR data from the analyses, as they did not follow the premises of normal distribution and homogeneity of variances. The results for these treadmills are shown for information in the graphical analysis ( Figure 1).

Results
We excluded lab-TR and NM-TR data from the analyses, as they did not follow the premises of normal distribution and homogeneity of variances. The results for these treadmills are shown for information in the graphical analysis ( Figure 1).
When comparing the overall differences in the mechanical properties across the three types of surfaces (fit-TR, artificial turf, and track and field) we found a significant group (i.e., "type of surface") effect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower in track than in the other two surfaces (p < 0.001 vs. both fit-TR and artificial turf) and lower in artificial turf than in fit-TR (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf, respectively) and lower in fit-TR than in artificial turf (p < 0.001). By contrast, ER was higher in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf) and also lower in artificial turf than in fit-TR (p = 0.002). Data are mean (±) SD. Abbreviations: ER, energy restitution; SA, shock absorption; VD, vertical deformation. Symbols: Ŧ p < 0.05 vs. treadmill; * p < 0.05 vs. artificial turf. Of note, World Athletics states that the artificial athlete (AA) device should be used instead of the advanced artificial athlete (AAA) to assess the mechanical properties of track surfaces. The equivalence between both test apparatus has been previously described [22]. Thus, the above reported values for track surfaces (which were obtained using the AAA) would be equivalent to SA and VD values of ≈35.5% and ≈ 1.73 mm, respectively, when assessed with the AA. Table 3 shows the differences between the six fit-TR models, revealing a significant group effect for SA, VD and ER. The treadmill models of the brand Life Fitness (LF97T and LFDX) displayed higher values of SA, VD and ER compared to the other treadmills (p < 0.01 for all cases), while the Precor model (PRE956I) showed the lowest values of VD and ER (p < 0.05 for all cases), with no significant differences in SA compared to the Technogym models.
Sensors 2020, 20, x FOR PEER REVIEW

Results
We excluded lab-TR and NM-TR data from the analyses, as they did not follow the pre normal distribution and homogeneity of variances. The results for these treadmills are sh information in the graphical analysis ( Figure 1).
When comparing the overall differences in the mechanical properties across the three surfaces (fit-TR, artificial turf, and track and field) we found a significant group (i.e., " surface") effect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower than in the other two surfaces (p < 0.001 vs. both fit-TR and artificial turf) and lower in artif than in fit-TR (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.00 TR and artificial turf, respectively) and lower in fit-TR than in artificial turf (p < 0.001). By ER was higher in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf) lower in artificial turf than in fit-TR (p = 0.002).  Table 3 shows the differences between the six fit-TR models, revealing a significant grou for SA, VD and ER. The treadmill models of the brand Life Fitness (LF97T and LFDX) displaye values of SA, VD and ER compared to the other treadmills (p < 0.01 for all cases), while th model (PRE956I) showed the lowest values of VD and ER (p < 0.05 for all cases), with no sig differences in SA compared to the Technogym models.

Results
We excluded lab-TR and NM-TR data from the analyses, as they did not follow the premises of normal distribution and homogeneity of variances. The results for these treadmills are shown for information in the graphical analysis ( Figure 1).
When comparing the overall differences in the mechanical properties across the three types of surfaces (fit-TR, artificial turf, and track and field) we found a significant group (i.e., "type of surface") effect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower in track than in the other two surfaces (p < 0.001 vs. both fit-TR and artificial turf) and lower in artificial turf than in fit-TR (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf, respectively) and lower in fit-TR than in artificial turf (p < 0.001). By contrast, ER was higher in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf) and also lower in artificial turf than in fit-TR (p = 0.002). Data are mean (±) SD. Abbreviations: ER, energy restitution; SA, shock absorption; VD, vertical deformation. Symbols: Ŧ p < 0.05 vs. treadmill; * p < 0.05 vs. artificial turf. Of note, World Athletics states that the artificial athlete (AA) device should be used instead of the advanced artificial athlete (AAA) to assess the mechanical properties of track surfaces. The equivalence between both test apparatus has been previously described [22]. Thus, the above reported values for track surfaces (which were obtained using the AAA) would be equivalent to SA and VD values of ≈35.5% and ≈ 1.73 mm, respectively, when assessed with the AA.

Results
We excluded lab-TR and NM-TR data from the analyses, as they did not follow the prem normal distribution and homogeneity of variances. The results for these treadmills are sho information in the graphical analysis ( Figure 1).
When comparing the overall differences in the mechanical properties across the three t surfaces (fit-TR, artificial turf, and track and field) we found a significant group (i.e., " surface") effect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower than in the other two surfaces (p < 0.001 vs. both fit-TR and artificial turf) and lower in artifi than in fit-TR (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.001 TR and artificial turf, respectively) and lower in fit-TR than in artificial turf (p < 0.001). By c ER was higher in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf) a lower in artificial turf than in fit-TR (p = 0.002).

Results
We excluded lab-TR and NM-TR data from the analyses, as they did not follow the premises of normal distribution and homogeneity of variances. The results for these treadmills are shown for information in the graphical analysis ( Figure 1).
When comparing the overall differences in the mechanical properties across the three types of surfaces (fit-TR, artificial turf, and track and field) we found a significant group (i.e., "type of surface") effect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower in track than in the other two surfaces (p < 0.001 vs. both fit-TR and artificial turf) and lower in artificial turf than in fit-TR (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf, respectively) and lower in fit-TR than in artificial turf (p < 0.001). By contrast, ER was higher in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf) and also lower in artificial turf than in fit-TR (p = 0.002). Data are mean (±) SD. Abbreviations: ER, energy restitution; SA, shock absorption; VD, vertical deformation. Symbols: Ŧ p < 0.05 vs. treadmill; * p < 0.05 vs. artificial turf. Of note, World Athletics states that the artificial athlete (AA) device should be used instead of the advanced artificial athlete (AAA) to assess the mechanical properties of track surfaces. The equivalence between both test apparatus has been previously described [22]. Thus, the above reported values for track surfaces (which were obtained using the AAA) would be equivalent to SA and VD values of ≈35.5% and ≈ 1.73 mm, respectively, when assessed with the AA. Table 3 shows the differences between the six fit-TR models, revealing a significant group effect for SA, VD and ER. The treadmill models of the brand Life Fitness (LF97T and LFDX) displayed higher Sensors 2020, 20, x FOR PEER REVIEW

Results
We excluded lab-TR and NM-TR data from the analyses, as they did not follow the pre normal distribution and homogeneity of variances. The results for these treadmills are sh information in the graphical analysis ( Figure 1).
When comparing the overall differences in the mechanical properties across the three surfaces (fit-TR, artificial turf, and track and field) we found a significant group (i.e., " surface") effect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower than in the other two surfaces (p < 0.001 vs. both fit-TR and artificial turf) and lower in artif than in fit-TR (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.00 TR and artificial turf, respectively) and lower in fit-TR than in artificial turf (p < 0.001). By ER was higher in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf) lower in artificial turf than in fit-TR (p = 0.002). Data are mean (±) SD. Abbreviations: ER, energy restitution; SA, shock absorption; VD, vertical defo Symbols: Ŧ p < 0.05 vs. treadmill; * p < 0.05 vs. artificial turf. Of note, World Athletics states that the athlete (AA) device should be used instead of the advanced artificial athlete (AAA) to assess the m properties of track surfaces. The equivalence between both test apparatus has been previously descr Thus, the above reported values for track surfaces (which were obtained using the AAA) would be e to SA and VD values of ≈35.5% and ≈ 1.73 mm, respectively, when assessed with the AA. Table 3 shows the differences between the six fit-TR models, revealing a significant grou for SA, VD and ER. The treadmill models of the brand Life Fitness (LF97T and LFDX) displaye

Discussion
Our results show differences between the mechanical properties of treadmill surfaces, artificial turf pitches and athletics tracks. Taken together, artificial turf surfaces comply with the international standards for both football [23] (SA, 55-70%; VD, 4-11 mm; ER, N/A) and rugby [27] (SA, 55-70%; VD, 5.5-11.0 mm; ER, 20-50%), and the track surfaces meet the criteria established by World Athletics when assessed with the AA [25] (SA, 35-50%; VD, 0.6-2.5 mm; ER: N/A). When compared to these surfaces, treadmills show statistically significant differences in all mechanical properties. Thus, treadmills have the highest SA ability of all the surfaces, while their VDs and ERs range between those of the artificial turf and the track, being much closer to the first. When compared to other surfaces such as asphalt or concrete-with SA values below 2%, and VDs and ERs close to 0 [7,28]these differences are even higher. This suggests that, despite having been conceived for running and walking, the mechanical behavior of treadmill surfaces differs remarkably from that of other surfaces used for similar purposes such as tracks or asphalt roads. By contrast, treadmill surfaces seem to better reproduce the mechanical properties of the artificial turf.
Our results are in line with those of previous studies reporting that treadmill surfaces are usually more compliant than overground running surfaces [13] and also with those reporting that treadmill surfaces overall have a less compliant-here indicated by a lower VD-and higher damping behavior-here indicated by a higher ER-than artificial turf surfaces [9,29]. However, our findings regarding the mechanical behavior of treadmills cannot be generalized since there are large differences between treadmill models, even within the same brand. Indeed, our results show significant differences between the treadmills commonly used in fitness centers (fit-TR) of up to 6%, 3.1 mm and 25% in SA, VD and ER, respectively. These findings suggest that fit-TR may not be considered as homogeneous surfaces in terms of mechanical properties and that each treadmill model When comparing the overall differences in the mechanical properties across the three types of surfaces (ft-TR, artifcial turf, and track and feld) we found a signifcant group (i.e., "type of surface") effect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower in track than in the other two surfaces (p < 0.001 vs. both ft-TR and artifcial turf) and lower in artifcial turf than in ft-TR (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.001 vs. ft-TR and artifcial turf, respectively) and lower in ft-TR than in artifcial turf (p < 0.001). By contrast, ER was higher in track than in the other two surfaces (p < 0.001 vs. ft-TR and artifcial turf) and also lower in artifcial turf than in ft-TR (p = 0.002). cluded lab-TR and NM-TR data from the analyses, as they did not follow the premises of tribution and homogeneity of variances. The results for these treadmills are shown for in the graphical analysis ( Figure 1). comparing the overall differences in the mechanical properties across the three types of it-TR, artificial turf, and track and field) we found a significant group (i.e., "type of ffect for SA, VD and ER (Table 2). In post hoc pairwise comparisons, SA was lower in track other two surfaces (p < 0.001 vs. both fit-TR and artificial turf) and lower in artificial turf R (p = 0.001). VD was also lower in track than in the other two surfaces (p < 0.001 vs. fitificial turf, respectively) and lower in fit-TR than in artificial turf (p < 0.001). By contrast, her in track than in the other two surfaces (p < 0.001 vs. fit-TR and artificial turf) and also tificial turf than in fit-TR (p = 0.002). an (±) SD. Abbreviations: ER, energy restitution; SA, shock absorption; VD, vertical deformation. < 0.05 vs. treadmill; * p < 0.05 vs. artificial turf. Of note, World Athletics states that the artificial device should be used instead of the advanced artificial athlete (AAA) to assess the mechanical f track surfaces. The equivalence between both test apparatus has been previously described [22]. ove reported values for track surfaces (which were obtained using the AAA) would be equivalent D values of ≈35.5% and ≈ 1.73 mm, respectively, when assessed with the AA.
p < 0.05 vs. treadmill; * p < 0.05 vs. artifcial turf. Of note, World Athletics states that the artifcial athlete (AA) device should be used instead of the advanced artifcial athlete (AAA) to assess the mechanical properties of track surfaces. The equivalence between both test apparatus has been previously described [22]. Thus, the above reported values for track surfaces (which were obtained using the AAA) would be equivalent to SA and VD values of ≈35.5% and ≈ 1.73 mm, respectively, when assessed with the AA.
Sensors 2020, 20, 3822 5 of 9 Table 3 shows the differences between the six ft-TR models, revealing a signifcant group effect for SA, VD and ER. The treadmill models of the brand Life Fitness (LF 97T and LF DX ) displayed higher values of SA, VD and ER compared to the other treadmills (p < 0.01 for all cases), while the Precor model (PRE 956I ) showed the lowest values of VD and ER (p < 0.05 for all cases), with no signifcant differences in SA compared to the Technogym models.  Figure 1 shows the product-moment correlations between the mechanical properties of each surface, taking all of the ft-TR models as a single group. All the surfaces showed a strong positive correlation between the SA and VD, this association being slightly weaker for the ft-TR. As for the SA vs. ER and the VD vs. ER relationships, artifcial turf and track surfaces showed a strong negative correlation in both cases, whereas positive correlations (moderate and strong, respectively) were found for ft-TR.

Discussion
Our results show differences between the mechanical properties of treadmill surfaces, artifcial turf pitches and athletics tracks. Taken together, artifcial turf surfaces comply with the international standards for both football [23] (SA, 55-70%; VD, 4-11 mm; ER, N/A) and rugby [27] (SA, 55-70%; VD, 5.5-11.0 mm; ER, 20-50%), and the track surfaces meet the criteria established by World Athletics when assessed with the AA [25] (SA, 35-50%; VD, 0.6-2.5 mm; ER: N/A). When compared to these surfaces, treadmills show statistically signifcant differences in all mechanical properties. Thus, treadmills have the highest SA ability of all the surfaces, while their VDs and ERs range between those of the artifcial turf and the track, being much closer to the frst. When compared to other surfaces such as asphalt or concrete-with SA values below 2%, and VDs and ERs close to 0 [7,28]-these differences are even higher. This suggests that, despite having been conceived for running and walking, the mechanical behavior of treadmill surfaces differs remarkably from that of other surfaces used for similar purposes such as tracks or asphalt roads. By contrast, treadmill surfaces seem to better reproduce the mechanical properties of the artifcial turf.
Our results are in line with those of previous studies reporting that treadmill surfaces are usually more compliant than overground running surfaces [13] and also with those reporting that treadmill surfaces overall have a less compliant-here indicated by a lower VD-and higher damping behavior-here indicated by a higher ER-than artifcial turf surfaces [9,29]. However, our fndings regarding the mechanical behavior of treadmills cannot be generalized since there are large differences between treadmill models, even within the same brand. Indeed, our results show signifcant differences between the treadmills commonly used in ftness centers (ft-TR) of up to 6%, 3.1 mm and 25% in SA, VD and ER, respectively. These fndings suggest that ft-TR may not be considered as homogeneous surfaces in terms of mechanical properties and that each treadmill model should be tested individually in order to characterize its mechanical behavior. Moreover, our results suggest that differences may exist between treadmill brands, as previously suggested [30], although the small sample of brands and models included in this study precludes the ability to draw general conclusions.
While keeping in mind that lab-TR could not be included in the statistical analyses, our results suggest that differences across lab-TR could be even greater than those reported for ft-TR. In this regard, some studies have shown that differences in the mechanical properties of treadmill surfaces can affect the metabolic cost and ground reaction forces during running [18,31], and others have reported that the varying mechanical properties of the running surface may result in premature fatigue or undesirable challenge during a certain task [32,33]. Collectively, these fndings suggest that researchers, clinicians and athletes using a lab-TR for specifc purposes must carefully choose the model to be used, since this may affect the generalizability of clinical assessments or research performed on the treadmill, potentially leading to erroneous research fndings [3,13,18,31,34]. For example, our fndings imply that marked differences in mechanical properties between treadmill and overground surfaces could critically affect footwear studies using treadmills to assess the effects of running shoes on running economy and running biomechanics [35][36][37], since the optimal footwear on a treadmill may not necessarily be the optimal footwear on an overground surface. Therefore, researchers using treadmills to reproduce overground conditions in research or clinical settings should attempt to use a treadmill whose surface mimics as closely as possible the mechanical properties of the specifc overground surface, since the comparability between both conditions will vary depending on the treadmill platform [18]. We therefore encourage the persistent testing and reporting of the mechanical properties of the surfaces to allow reliable comparisons to be made in this context, especially in research that aims to investigate the relationship between treadmill and overground locomotion, or where there is the need to reproduce overground conditions for specifc purposes-e.g., to investigate the effects of footwear.
Our results show a greater dispersion of treadmills' mechanical properties compared to those of artifcial turf and track surfaces ( Figure 1). Our fndings on the relationship between SA, VD and ER in artifcial turf and track surfaces support previous studies reporting that an increased compliance (i.e., higher VD) in overground surfaces is associated with a reduction in the re-utilization of elastic energy (i.e., a lower ER) [38][39][40], which would lead to an increased metabolic cost and alterations in running kinematics. However, as opposed to overground surfaces, both SA and VD are directly proportional to ER in treadmills, meaning that treadmills with more shock-absorbing and compliant surfaces would increase energy return to the runners. This supports previous research pointing that the metabolic cost of running is greater for treadmills with stiffer running platforms [18,23], contrary to what is encountered overground [7]. Moreover, the fact that the ER of some lab-TR is drastically lower than that of track surfaces could also explain previous fndings reporting that the metabolic cost at low [32] and submaximal speeds (with controlled air resistance) [2] is signifcantly higher on a treadmill compared to that on track surfaces. The increase in the treadmill ER as VD increases will most likely be due to the materials and structural components forming their surfaces, which determine their viscoelastic (or damping) properties relevant during the unloading phase. The latter may have relevant implications in terms of muscle activity and injury risk, as well as in terms of performance outcomes and the reproducibility of kinematic patterns when comparing treadmill to overground locomotion. In this sense, it has been reported that stiffer surfaces lead to increased muscle activity [41] and that surfaces providing increased mechanical cushioning affect running kinematics [11]. Nevertheless, the implications for performance and injury risk of surfaces with comparable stiffnesses but different damping properties remain unclear.
Overall, the present fndings support the importance of regulating the mechanical properties of treadmill surfaces because (1) the mechanical properties of all sports surfaces are considered to be important determinants of performance and injury risk, and (2) our results indicate that the mechanical properties of treadmills vary across models and do not match those of other surfaces that are often used for walking and running. Moreover, since treadmills with very similar VD (which is an indicator of their stiffness) may differ strongly in SA and ER, our results also indicate that assessing and regulating only stiffness in treadmill surfaces may not suffice for fully characterizing their mechanical behavior. Similarly, relating research results to surface stiffness could potentially lead to misleading conclusions. Further research in this area may help manufacturers to design treadmills with surface properties that match those of specifc overground surfaces, or treadmills with surface properties specifcally designed to achieve certain purposes such as enhancing athletic performance or decreasing injury risk. Additionally, future research should assess whether mechanical properties of treadmill surfaces could correlate with other variables such as a treadmill's usage time, temperature or kilometers traveled, which is something that the present research failed to investigate due to a lack of data.

Conclusions
The mechanical properties (shock absorption, vertical deformation and energy restitution) of treadmill surfaces differ signifcantly from those of commonly used overground sport surfaces such as artifcial turf and athletics tracks. Our results also suggest that, unlike overground surfaces, treadmills with more shock-absorbing and compliant surfaces would be expected to increase energy return to the athletes. Moreover, our results show remarkable differences between different treadmill models, suggesting that treadmills will most likely vary in their comparability to overground surfaces depending on the mechanical properties of their platforms.