Study on the effects of shoe cushioning on trail-running: perception, bench test and biomechanical approach

Abstract This work aimed to investigate the association between the perceived cushioning of a set of trail-running shoes with both their mechanical response to the impact test and the acceleration of lower limb segments while running indoors and outdoors. Earlier studies have typically focussed on the effect of road-running shoe cushioning, whereas very few have examined the perception of trail runners and outdoor trail-running conditions. Seven trail runners were trained to become reliable in evaluating the level of cushioning on a scale ranging from 0 to 100 and then asked to rate the cushioning of eight trail-running shoes while running. Shoe specimens were mechanically characterized through an impact test. In addition, the effect on running biomechanics was tested by wearing two accelerometers on the tibia and foot. The in-lab impact test revealed that the perceived cushioning was inversely associated with the force amplitude and directly associated with the impact duration. The running test showed that the median frequency of the tibial acceleration during the first 25% of the stance phase decreases with increasing cushioning during both indoor (p = 0.02, rs = −0.83) and outdoor conditions (p < 0.001, rs = −0.79). In conclusion, the perceived cushioning was quantitatively associated with the outcomes of impact and running tests.


Introduction
In the last few years, trail-running has become a very popular sport activity because of its accessibility and health benefits (Barton and Pretty, 2010). Trail-running surfaces are usually irregular, so significant joint coordination variability and dynamic stability are required (Bean, 2018). Because the physical demand and the movements derived from trail-and road-running are different, the two related running populations are expected to look for different features in the shoes they adopt. The spreading popularity of trailrunning is leading the footwear industry and R&D in sports engineering to create new products with specific features. A huge effort has been made to identify optimal cushioning, which is generally known as the capacity of the shoe to absorb shocks during a foot-to-ground collision (Baltich et al., 2015;Malisoux et al., 2017). Previous authors argued that an optimal cushioning system is expected to increase the duration and reduce the magnitude of an impact at heel strike, thus reducing the impulsive effects of the foot strike on the musculoskeletal system. However, optimal cushioning has not yet been found for either road-running or trail-running shoes. Indeed, footwear design and its influence on both running biomechanics and injury risk deserve further investigation (Knapik et al., 2014;Ryan et al., 2011).

Shoe cushioning assessment: analysis of current approaches
In studies dealing with road-running shoes, the footwear cushioning is usually quantified by three different approaches: in-lab mechanical tests, user perception scores, and in-field running biomechanical variables.

In-lab mechanical tests
This approach is commonly used by shoe manufacturers to measure the properties of their products and prototypes during controlled experimental conditions. In particular, to assess shoe cushioning, the most used standards are the impact test ASTM 2006 (procedure A or B) and the DIN EN 12743. These methodologies aim to accurately simulate the time course of the load at the heel strike during running (Schwanitz et al., 2010). In short, the ASTM F 1614 standard, procedure A, consists of lifting an 8.51 kg mass above the heel part of the shoe and releasing it for a free fall from a height of 50 mm towards the innersole. The impact is measured by a load cell or a single-axis accelerometer. The ASTM F 1614, procedure B, and the DIN EN 12742 are two hysteresis tests.

User perception scores
One of the possible ways to assess cushioning is by collecting the runners' opinions. However, the identification of a suitable vocabulary and the training of the testers represent two necessary steps to exploit the inherent capabilities of runners to reliably asses cushioning. It is interesting to observe that in research fields (Sokolowsky and Fischer, 2012) and other consumer product evaluations (Verriele et al., 2012), researchers have taught a group of people (the sensory-trained panel) how to discriminate a specific property in a reliable and repeatable fashion. This approach has also been adopted successfully to recognize and rate differences in cushioning among road shoe models (Delattre and Cariou, 2018).

In-field running biomechanical variables
The influences of cushioning on running biomechanics are still debated in the literature because some of the reported findings appear controversial (Shorten, 2002). Some studies documented a reduced impact force associated with increased shoe cushioning (O'Leary et al., 2008), whereas others noticed an opposite trend (Baltich et al., 2015). The evaluation of heel cushioning through the analysis of the first peak of the vertical ground reaction force (GRF), the one related to the initial ground contact, may be masked by the superimposition of low-frequency 'active' components, higher frequency 'impact' components, and forces acting on other parts of the foot (Shorten, 2002). The most relevant alternative to GRF and plantar pressure measurements for outdoor testing consists of analyzing the shock-related accelerations of lower limb body segments in both the time and frequency domains (Shorten and Winslow, 1992). These signals are usually collected by low-weight accelerometers, i.e. wearable sensors well suited for outdoor acquisitions. Peak tibial acceleration (PTA) was commonly used to assess the loading impact during running (Fortune et al., 2014;Giandolini et al., 2016). It has been demonstrated that shocks transmitted to the lower limb segments account for both low-frequency components (4-8 Hz), probably representing the voluntary motion, and the vertical acceleration of the centre of mass during the stance phase, and high-frequency components (10-20 Hz), mostly related to tissue resonance (Hamill et al., 1995). In this respect, Horvais and colleagues suggested that the analysis in the frequency domain correlates with different shoe cushioning conditions (Horvais et al., 2019).

Aim of the study
Very few of the cited studies examine the perceptions of trail runners and outdoor trail-running conditions. Conversely, this study aimed to expressly investigate the association between the perceived cushioning of a set of trail-running shoes and both their mechanical response to the impact test and the acceleration of lower limb segments while running indoors and outdoors. Users' perception was adopted as the reference metric to rank the tested shoes based on the evidence in the literature that the perceived cushioning highly correlates with several biomechanical factors, such as ground reaction force parameters (Milani et al., 1997), peak plantar pressure, and with shoe mechanical properties (Goonetilleke, 1999) affecting the gait performance.
It was therefore hypothesized that: experienced trail runners can discriminate different levels of trail-running shoe cushioning after training; the in-lab impact test reflects the users' perceived level of cushioning; the acceleration of lower limb body segments during trail running depends on the shoe cushioning.

Methods
In the present work, a set of commercially available trailrunning shoes was tested. For each of the specimens, the shoe cushioning was evaluated in terms of three main outcomes: in-lab mechanical testing, user perception, and effects on running biomechanics. For each of these three issues, a specific protocol was designed.

Shoe specimens
Eight sample shoes were chosen after market research and consultations with expert athletes to cover the range of currently available levels of nominal cushioning among trailrunning products ( Figure 1). Table 1 summarizes all the models ordered from a nominal minimum to a nominal maximum cushioning property. The mass and heel-to-toe drop (i.e. the difference in thickness between the heel and the toe part of the sole) are also reported according to manufacturers' specifications for each pair of shoes.

Subjects
Seven subjects were recruited for this study. The inclusion criteria were: trail-running practitioner, running on average at least 15 km per week, aged between 18 and 50 years old, and having a 270/275 mm foot size. Before participating in the experiments, all subjects were given an oral and written explanation of the test protocol and they gave their written informed consent. The experiments were approved by the innovation and research manager of Ober Alp SpA, as officers responsible for the safety and ethics review of research activities in the company. The demographic and anthropometric information of the participants are collected in Table 2. Six participants were free from injuries to the lower extremities and lumbar spine over the last 6 months before participation. The seventh participant had suffered a posterior tibial muscle contracture and patellofemoral chondropathy on the right leg, but no influence on running motion was claimed.

Impact test: experimental setup and protocol
The first approach of the cushioning evaluation consisted of measuring the response of the sole to the impact of a striker. To replicate the impact between the outsole and the ground during running, a reverse experimental setup was used, as illustrated in Figure 2. A mechanical support hosted a plastic foot shape worn by the shoe to be tested and was connected to the mainframe through a load cell. A magnetic releasing system could be manually deactivated to let the striker fall freely and perpendicularly towards the heel part of the outsole. The striker had a mass of 8.51 kg and was slightly convex on the impact surface to neutralize any tilting movement during the fall. A plastic foot was used to emulate a heel-shaped distribution of sole strains and deflections.
The impact test protocol was conceived with three different starting heights for the striker, i.e. 3 cm, 9 cm, and 15 cm. The values of the heights were chosen to match the typical range of energies involved in running (2.50 J, 7.51 J, and 12.52 J, respectively).
For each shoe specimen and each height (i.e. 3 cm, 9 cm, 15 cm), 5 repetitions were performed. The force signals were collected by a load cell (HBM U10M, force transducer for static and dynamic measurements up to 2.5 MN) through a universal amplifier (HMB QuantumX MX440B). Data were digitally sampled at 1200 Hz. An offset removal was needed due to the difference in shoe weights, while a notch filter at 50 Hz and multiples were sufficient for noise removal. Data were analyzed off-line to estimate the following impact-related features: impact peak force magnitude; impact time duration.

User perception test protocol
The main goal of this test consisted of scoring the level of cushioning of shoe specimens according to the end-user perception. To achieve this task, participants were preliminarily trained to discriminate different levels of cushioning, as reported by Delattre and Cariou (2018).
In more detail, participants were first asked to agree on a common definition of cushioning for footwear. In this respect, they focussed on the shock absorption feature only, although it was acknowledged that the cushioning can affect other aspects during trail-running, such as impact resistance, stability, and the reactivity of the sole. Then, participants donned the eight shoe specimens and tested them during several motor tasks, such as walking, running, or jumping freely. Accordingly, they updated the definition of cushioning. After the training, participants were expected to be able to discriminate different levels of cushioning along a 0-100 visual analog scale (i.e. 0 min cushioning, 100 max cushioning), as reported in the literature (Mills et al., 2010;M€ undermann et al., 2002). They were, hence, asked to score the shoe specimens after each running trial, both indoors and outdoors. The median rating among all participants was then converted into a ranking of the shoe cushioning.

Running test protocol
During the running experiments, participants were asked to run both indoors and outdoors while the 3D linear acceleration of the foot and tibia was recorded by two commercial inertial measurement units (IMU; GaitUpPhysilog 5). The sampling frequency and the acceleration range of the IMUs were 256 Hz and 16 g, respectively. The sensors were mounted on the most distal part of the body segment to avoid the smoothing effect on accelerations of both the ankle and knee joints. The indoor testing consisted of 1 min of running on a treadmill at three different cadences: 155, 175, and 195 steps/min set by a metronome. The outdoor testing included 20 s of level and 20 s of downhill running on a soft and gravel road with a cadence of 175 steps/min. For both sessions, a 5-10 min acclimatization period was undertaken before starting the data recording. For each of the experimental sessions (i.e. indoor and outdoor tests), participants wore the shoe specimens in random order and were asked to refer to them according to the attached number (Figure 1). After each trial, they were asked to rate the perceived level of cushioning. The IMUs were firmly fixed on the shoelaces and on an elastic band tied around the distal part of the tibia, as illustrated in Figure 3. They synchronously saved data streams on onboard memories. Accelerations were off-line band-pass filtered (high-pass and low-pass cut-off frequencies at 0.5 Hz and 50 Hz, respectively; zero-lag, Butterworth filter) to remove both low-frequency and high-frequency artefacts. The lowpass frequency was set to limit the resonance frequency of the attachment (shoelaces or elastic band), typically ranging from 60 Hz to 100 Hz (Martyn R Shorten and Winslow, 1992).
After extracting the central 70% of the record (to avoid the acceleration and deceleration running phases), the following features were computed off-line for both the 3D acceleration and the components longitudinal to either the tibia or the foot: for each stride: negative and positive acceleration peaks (AP) (Zhang et al., 2016);^n egative and positive acceleration peaks in the first 25% of the stance (AP25) (Zhang et al., 2016); median frequencies in the first 25% of the stance part of the signal (MD25) (Martyn R Shorten and Winslow, 1992); average power of spectrum (PS) (Gruber et al., 2014). for the entire record: median frequencies (MD) (Martyn R Shorten and Winslow, 1992).
These features were then averaged across strides to provide a representative value for each subject and each trial.

Statistical analysis
The median and interquartile range were used as descriptive statistics to refer to the independent variables, i.e. perception scores, impact test features, and biomechanical variables.
The Kruskal-Wallis test was preliminarily carried out to test the hypothesis that the participants' perceived cushioning scores for both indoor and outdoor tests were comparable among shoe specimens.
The one-way analysis of variance (ANOVA) was used to test the hypothesis that the averaged outcomes of the impact test (i.e. impact peak force magnitude and time duration) were equal across shoe specimens. This analysis was followed by the correlation analysis (Spearman's rank correlation) to investigate the association between the outcomes of the impact test and those resulting from the user perception test in the indoor condition.
Biomechanical variables (i.e. AP, AP25, MD, MD25, PS) were preliminarily checked to test the hypotheses that their distributions were normal (Lilliefors test) and that the variance across shoe specimens was homogeneous (Bartlett test). Because both hypotheses were confirmed, the one-way ANOVA with repeated measures was used to test the hypothesis that the impact-related acceleration variables were comparable among shoes. The least significant difference (LSD) posthoc analysis was carried out to compare paired shoes in the case of a significant outcome of the repeated measures ANOVA. Biomechanical variables found to be different across shoe specimens were correlated with the perception ranking (Spearman's rank correlation) to investigate the possible association between the quantitative and subjective assessment. For all the above analyses, a significance equal to 0.05 was adopted.

User perception rating
The perceived level of cushioning in indoor and outdoor conditions for each shoe specimen is reported in Table 3, where shoes are ranked from minimum to maximum perceived cushioning. In both experimental conditions, the statistical analysis revealed a significant effect (p < 0.001) of the shoe specimen on the perception ratings of the subjects (Kruskal-Wallis test). Noticeably, the perceived cushioning differed slightly between the indoor and outdoor experimental sessions, mainly for lower values of cushioning. Figure 4 shows a representative time course of the impact force, as registered during the 3-cm-height-condition test, for all shoe specimens. Noticeably, the curves are ordered and coloured on a grey scale according to the perceived cushioning during the indoor test (Table 3).

Outcomes of the impact test
For all experimental conditions related to the impact test (i.e. 3-cm-height, 9-cm-height, and 15-cm-height), the statistical analysis (ANOVA) revealed a significant effect (p < 0.001) of the shoe specimen on the peak force amplitude ( Figure 5). As expected, the peak amplitude increased with the height of the striker. A negative correlation was observed relating the peak force amplitude and shoe specimens ordered in accordance with the indoor perceived cushioning (3-cm-height: rs ¼ À0.98, 9-cm-height: rs ¼ À0.81, and 15-cm-height: rs ¼ À0.83; Figure 5), except for the C and B models.
A significant effect of the shoe specimen on the impact time duration was detected as well (3-cm-height: p < 0.001, 9-cm-height: p ¼ 0.03, and 15-cm-height: p ¼ 0.02; Figure 6). This impact test feature showed a positive correlation with the indoor perceived level of cushioning, as documented by Spearman's correlation coefficient being greater than 0.88 for all the experimental conditions.

Outcomes of the running tests
The outcomes of the running tests, both indoor and outdoor, consisted of a set of features (i.e. AP, AP25, MD, MD25, and PS) extracted from the acceleration signals (i.e. 3D vector and longitudinal component) recorded by the IMUs located on the tibia and foot. However, according to the aim of our study, we only retained those showing a significant association with the perceived cushioning, as assessed by Spearman's correlation coefficient. The average and dispersion of the remaining features across shoe specimens are reported in the Supplementary Material for the sake of completeness.  Figure 4. Time-force profile for each shoe specimen during the 3-cmheight impact test. Curves are ordered and coloured on a gray scale according to perceived cushioning during the indoor test (i.e. light and dark grey refer to the minimum and maximum rate, respectively). Figure 5. Force peak amplitude across experimental conditions (3-cm, 9cm, and 15-cm height) for all shoe specimens. Shoes are ordered according to perceived cushioning during the indoor test (i.e. A and H sample shoes refer to the minimum and maximum rate, respectively). Figure 6. Impact time duration across experimental conditions (3-cm, 9cm, and 15-cm height) for all shoe specimens. Shoes are ordered according to perceived cushioning during the indoor test (i.e. A and H sample shoes refer to the minimum and maximum rate, respectively).

Indoor running test
The features showing a significant association with the perceived cushioning during the indoor running test were the MD25 extracted from signals related to the IMU on the tibia (3D vector) at slow and medium cadences, as reported in Figure 7 (panels A and B). Each panel also shows bars pinpointing shoe specimens found to be significantly different based on the posthoc statistical analysis. Noticeably, the shoe specimens are ordered according to the perceived cushioning in the indoor condition (Table 3). In more detail, a significant effect of the shoe model on MD25 related to the 3D tibial acceleration was found through the repeated measures ANOVA test (p ¼ 0.02 at medium cadence, p ¼ 0.01 at slow cadence). Indeed, Spearman's rank correlation coefficient revealed a strong association between MD25 and indoor perceived cushioning (rs ¼ À0.83 at medium cadence, rs ¼ À0.87 at slow cadence). Specifically, the MD25 becomes lower as the perceived cushioning becomes higher.

Outdoor running test
The features showing a significant association with the perceived cushioning during the outdoor running test were the MD25 during both downhill and level running conditions, related to the longitudinal tibial acceleration (Figure 8; noticeably, shoe specimens are ordered according to the perceived cushioning in the outdoor condition, as reported in Table 3).
In more detail, a significant effect of the shoe model was observed on the MD25 related to the longitudinal acceleration of the tibia both downhill (p < 0.001) and in level running (p < 0.001). The correlation analysis (Spearman's rank correlation coefficient) revealed a strong and negative association between these metrics and the perceived outdoor cushioning (rs ¼ À0.79 for downhill running, rs ¼ À0.90 for level running).

Discussion
This study was aimed at investigating the association between the perceived cushioning of a set of trail-running shoes and both their mechanical response to the impact test and the acceleration of lower limb segments while running indoors and outdoors. To achieve this twofold goal, we first investigated whether the cushioning properties of the chosen trail-running shoes could be consistently ordered based on the participants' perceptions. Then, we tested the hypothesis that the perceived cushioning properties, as reported by the users in the previous test, can be ascribed to the outcomes of both the impact test and the acceleration of lower limb segments during indoor and outdoor running tests.
The results confirmed that after the preliminary training, the enrolled participants could rate different trail-running shoe models by testing them both indoors and outdoors. Some differences in the perception ratings were observed between indoor and outdoor conditions, but the only noteworthy case was represented by shoe model B, which was perceived as less cushioned in outdoor conditions than in indoor conditions. As a general comment, participants reported that cushioning differences can be better perceived during downhill running compared to level running where they perceived a softer terrain that partially mitigated the impacts.
Concerning the in-lab impact tests, the outcomes showed a significant correlation with the perceived shoe cushioning. More specifically, the findings are consistent with the definition given by the American Society of Testing and Materials (ASTM, 1986), according to which shoe cushioning is 'the reduction of peak force by increasing the time over which the force is applied'. The countertrend response of the third-and the fourth-ranked shoes with higher impact energies (9-cm and 15-cm-height condition) is probably related to their lighter weight (thirdranked Shoe C: 244 g and fourth-ranked Shoe B: 224 g, Table 1) involving less material in the sole and, consequently, leading to a weaker shock absorption capacity. Regarding these models, it is also noteworthy to remember that participants gave lower values to the perceived level of cushioning of Shoe B in outdoor conditions than in indoor sessions (Table 3), confirming a lower cushioning for uneven terrains.
A significant correlation between the perceived shoe cushioning and the outcomes of the running tests were also observed. Specifically, a lower median frequency of the acceleration is associated with a higher perceived cushioning, suggesting that higher-ranked shoe specimens induce a smoother heel-strike impact. In this respect, our findings corroborate those already reported in the literature and reveal that higher perceived cushioning reflects a reduction of the impact-related frequencies in the range of 10-20 Hz (Hamill et al., 1995). Thus, it is confirmed that users can perceive a smoother impact that may also influence their proprioception reaction.
The perceived cushioning was associated with the 3D tibial acceleration during indoor tests and the longitudinal component of tibial acceleration during outdoor tests. This finding highlights that different experimental conditions, that is, indoors on a treadmill vs. outdoor overground tests, can involve different biomechanical constraints that can, in turn, modify the overall control of the neuro-muscular system (Lindsay et al., 2014).
Finally, we observed no significant correlation between the perceived cushioning and running test outcomes during indoor sessions when participants were asked to run at the faster cadence (195 steps/min), although a significant correlation was observed at slower cadences. We believe that this result can, in part, be ascribed to the neuromuscular adaptations occurring at faster speeds, which can more predominantly absorb the foot-to-ground impact over the mechanical properties of the foot sole. However, this hypothesis deserves to be further investigated.

Limitations of the study
The main limitation of this study consisted of a small number of enrolled subjects resulting in a limited strength of the statistical analysis. However, it is worth noticing that our criteria of recruitment allowed us to include only experienced trail-running practitioners, to reduce, as much as possible, the inherent variability across subjects. Extensive tests on larger samples are required to confirm our findings.

Conclusions
This study revealed that the perceived cushioning was consistent across participants. In this respect, the designed protocol including the panel training can also be adapted for the analysis of other footwear features that may be investigated in future works both regarding road-and trailrunning shoes. Moreover, the mechanical test protocol was conceived to assess in-lab the shoe shock absorption response to different impact energies: the outcomes were found to be well correlated with the perception scores of the athletes. Finally, the analysis of biomechanical features, as assessed by a pair of IMUs fixed on the foot and the tibia, showed a significant association with the perception ranking of the shoes, practically giving more value to the other outcomes also.
The authors' suggestions are, therefore, to perform a first in-lab cushioning evaluation for a rapid selection of the best materials and to secondly validate the perception and the running outcomes.
All in all, the most relevant outcomes of our study can be summarized as follows: trained subjects can discriminate different levels of trail-running shoe cushioning; in-lab impact peak force and the time duration match the perception of the subjects; median frequency of the tibial acceleration during the first 25% of the stance relates to footwear cushioning; this parameter representing running impacts decreases with more cushioned shoes in both indoor and outdoor running conditions.