1. Introduction
There is an ongoing debate on whether the foot strike pattern of long-distance runners plays a role in defining performance and injury risk in this population [
1,
2,
3]. Experienced long-distance runners are able to change their foot strike pattern during a competition [
4] or if they are asked to [
5]. Their ability to adopt a different foot strike pattern has been often interpreted as a sign of adaptability. The concepts of “adaptability” and ”ability to adopt different execution patterns”, however, are not equivalent [
6]. Adaptability refers to the level of organisation embodied by the human locomotor control system [
7]; it represents the richness of motor behaviours that equally can accomplish the task-goal [
8]. In contrast, the ability to adopt different execution patterns refers to the ability to change joint kinematics (and kinetics) without necessarily meeting the task-goal. It is unknown if runners who adopt different execution patterns (i.e., rearfoot strikers versus forefoot strikers) have developed a different level of adaptability.
During the stance phase of running, the ankle plays a dominant role in storing and generating energy for propulsion [
9,
10]. The mode of foot/ground initial contact may affect the subsequent joint angle time course and the associated joint stiffness. According to Günther and Blickhan [
11], the foot strike angle, stiffness and running velocity are crucial parameters for coordination of body movement dynamics. The concept of dynamic joint stiffness [
12,
13], defined “quasi-stiffness” by Latash and Zatsiorsky [
14], can be used to characterize the ankle behaviour during the stance phase of running [
15]. Here, the ankle exhibits a first loading state in which the internal plantarflexor moment rises during dorsiflexion, and the periarticular joint structures absorb energy. It follows an unloading state in which the plantarflexion moment decreases while the joint plantarflexes, and the periarticular joint structures produce energy. The level of stiffness (that is the variation of joint moment per unit of joint angle variation) can depend on both (i) structural adaptations of the muscle–tendon units surrounding this joint and (ii) neural adaptations that control instantly the characteristics of these muscle–tendon units [
16,
17,
18]. For instance, long-term adaptations in muscle and tendon architecture in the lower limb, such as shorter gastrocnemius medialis fascicles [
19], thicker Achilles tendon [
20] and stiffer foot arch [
21], were found in habitual forefoot strikers, who usually land with a plantar-flexed ankle. Such adaptations could lead to a different load distribution in the muscle–tendon unit [
22], in which the role of the elastic components is increased and the muscle fibres contract at a slower rate, which is advantageous for maximal power output and efficiency [
23]. Together, both the structural and the neural adaptations contribute to defining the dimensionality of the system (degrees of freedom of the neural control system), that is the number of structures (muscles) that can be actively controlled and can be used to regulate the ankle dynamic stiffness efficiently, according to the mechanical requirements [
24].
Ankle dynamic stiffness can be computed as the slope of the tangent to the moment–angle curve [
12]. Using similar approaches, previous studies investigated dynamic ankle stiffness during running [
9,
10,
11]. To our knowledge, Hamill and Gruber [
5] were the only researchers testing change in ankle joint stiffness in two groups of runners with distinct foot strike patterns. Participants were classified as either rearfoot or forefoot strikers based on the presence of an impact peak on the vertical ground reaction force and on the ankle angle at landing. Although using these criteria runners may have been misclassified [
25], according to Hamill and Gruber [
5], habitual forefoot strikers exhibited a more compliant ankle and absorbed more (negative) work than habitual rearfoot strikers when running with their preferred foot strike pattern (forefoot); however, no differences were found with habitual rearfoot strikers running with a forefoot strike pattern (nonpreferred mode).
It is common for studies concerning running and ankle stiffness to simplify the loading phase of the moment–angle loop by representing the linear slope from initial foot contact to peak moment [
5,
15,
26,
27,
28] (
Figure 1, dashed line). This approach overlooks the potentially meaningful details occurring within the loading phase. For instance, at initial foot contact, the ankle joint moment increases nonlinearly with the change in angle (red portion in
Figure 1). This state represents the ankle joint response to foot/ground initial loading. Thereafter, the ankle dorsiflexes slowly, while the ankle moment increases fast (blue portion in
Figure 1).
This is a less compliant condition, representing the loading of the passive structures of the muscle–tendon units, and can be seen as a different state [
29]. Another interesting phase that deserves attention is the unloading phase (light-blue portion in
Figure 1), where the movement of the joint reverses to plantarflexion and the joint moment decreases. To the best of our knowledge, no studies have investigated the stance phase of running by considering these three task-relevant subphases of the moment–angle loop, which we expect to yield a more sensitive insight of the differences between habitual rearfoot and forefoot strikers.
The aim of this study was to investigate if foot strike loading technique has an effect on the ankle moment–angle dynamics during the stance phase of running. We expected forefoot strikers to have lower dynamic stiffness during the loading phase, based on previous findings [
5]. We also expected forefoot strikers to have a higher proportion of negative work relative to positive work because of their loading technique that allows them to store and return elastic potential energy in the foot–ankle structure. To test whether habitual foot strike loading technique compromises the control of ankle stiffness, shoes with different assistive constructs were considered. We expected differences in ankle stiffness and work ratio between groups to be greater in minimally assistive shoes, due to the unfamiliarity of rearfoot strikers to this condition. Because we expected forefoot strikers have a greater intrinsic foot–ankle adaptability to external loading (i.e., greater ability to control ankle stiffness), we expected their (forefoot strikers) ankle stiffness to have a higher time-dependency. That is, ankle stiffness during unloading will depend more on ankle stiffness during loading in forefoot strikers.
3. Results
No main effect of group was found for K
ankle (
p = 0.164; η
p2 = 0.105), but the main effects for shoe type (
p = 0.008; η
p2 = 0.272) and slope (
p < 0.001; η
p2 = 0.889) were obtained (
Table S1). Posthoc analysis revealed that K
ankle was 12% higher in med-MI compared with high-MI shoes (
p = 0.007;
g = 0.706).
Table S2 and
Figure 2 show mean and SD for K
ankle in the three subphases of stance and among the three shoe conditions. Significant differences were found among all subphases: ERP–LRP (0.176 ± 0.01; 0.215 ± 0.01 Nm/kg/°∙100)
p = 0.001;
g = 3.9; ERP–DP (0.176 ± 0.01; 0.091 ± 0.01 Nm/kg/°∙100)
p < 0.001;
g = 9.5; LRP–DP (0.215 ± 0.01; 0.091 ± 0.01 Nm/kg/°∙100)
p = 0.001;
g = 12.4. Overall K
ankle was highest when wearing med-MI shoes (although not statistically different from low-MI shoes;
p = 0.246); K
ankle was highest during the late rising phase (LRP) and lowest during the unloading phase (DP).
Runners in high-MI shoes exhibited a lower stiffness (more compliant ankle) during the impact phase (ERP) and late rising phase (LRP); during the unloading phase (DP), low-MI shoes allowed the most compliant ankle. There was a
Shoe by
Slope interaction effect (
p = 0.008; η
p2 = 0.221;
Table S1) for K
ankle (
Figure 2,
Table S2). Pairwise multiple comparisons showed that during the impact phase (ERP), K
ankle in high-MI shoes was lower compared with that in both low-MI and med-MI shoes (−15%,
p = 0.013,
g = 1.2; and −16%,
p = 0.003,
g = 1.25, respectively). During the late rising phase (LRP), K
ankle was the highest in med-MI shoes (0.227 ± 0.01 Nm/kg/°∙100), but only statistically different from high-MI shoes (+12%,
p = 0.011,
g = 1.58). During the unloading phase (DP), differences between shoes were only significant for low-MI compared with med-MI shoes (−6%,
p = 0.009,
g = 0.5).
Figure 3 compares mean moment–angle loops for RFS and FFS. While curves are similar in low-MI shoes, (
Figure 3, top) the base (ankle range of motion) is shifted toward the left for FFS. This is also true for medium-MI (
Figure 3, middle), and high-MI shoes (
Figure 3, bottom).
Overall, runners exhibiting high K
ankle during the late rising phase (LRP) also have high K
ankle during the unloading phase (DP) (
Table 1). For FFS, the correlation between K
ankle in the impact phase (ERP) and in late rising phase (LRP) increased with shoes’ MI, with the highest correlation (r
s = 0.95;
p < 0.01) in high-MI shoes. A similar trend was reported for correlations between K
ankle in impact phase (ERP) and in unloading (DP), and between K
ankle in late rising phase (LRP) and in unloading (DP), with highest values in the high-MI condition (r
s = 0.84,
p < 0.01; r
s = 0.89,
p < 0.01, respectively). Values were only significant in high-MI shoe conditions; this means that FFS in high-MI shoes with high K
ankle during impact phase will also have high K
ankle during the loading and unloading phases. For RFS, correlations between K
ankle in impact phase (ERP) and in late rising phase (LRP) and correlations between K
ankle in impact phase (ERP) and in unloading (DP) vary irrespectively to the shoe condition. The correlation between K
ankle in late rising phase (LRP) and in unloading (DP) increased with shoes’ MI, with the highest correlation (r
s = 0.92;
p < 0.01) in high-MI shoes. This means, K
ankle during impact has less of an effect on the subsequent subphases in RFS; instead, the late rising phase plays a central role.
In low-MI shoes, both groups presented low regression values (r
2 ≤ 0.26, insets in
Figure 3). In medium-MI shoes, K
ankle of RFS during the loading phase (LRP) explained 49% of the K
ankle variance during the unloading phase (DP), while for FFS, only 22% was explained. K
ankle of FFS in high-MI shoes depended on the stiffness in the previous phase: that is, stiffness during the impact phase (ERP) explained 60% of the stiffness variance during the late rising phase (LRP) and 65% of the stiffness variance during the unloading phase (DP); likewise, stiffness during the late rising phase (LRP) explained 63% of the stiffness variance during the unloading phase (DP).
We found a main effect of shoes for W
abs and W
prod (
p = 0.001, η
p2 = 0.425;
p < 0.001, η
p2 = 0.517) but no main effect of group (
p = 0.105;
p = 0.716) or interaction effects for
Groups by
Shoes were found (
p = 0.051;
p = 0.097) (
Table S1).
Figure 4 shows that W
prod increased significantly from low-MI to med-MI shoes (7%,
p = 0.004,
g = 0.578) and from med-MI to high-MI shoes (11%,
p = 0.017,
g = 0.657); while W
abs decreased as an inverse function of shoe MI index, reaching highest values in high-MI shoes (−32.58 ± 1.71 Nm/kg/°∙100). The latter was significantly lower than W
abs in low-MI (−19%,
p = 0.002,
g = 0.839) and med-MI shoes (−14%,
p = 0.009,
g = 0.674). RFS exhibited higher W
net compared with FFS (24.99 ± 1.25 versus 19.47 ± 1.25;
p = 0.006,
g = 4.420); W
net increased with shoe MI index, with runners in low-MI shoes exhibiting statistically lower W
net (−12%;
p = 0.007,
g = 0.456) compared with those in med MI-shoes, and compared with those in high-MI shoes (−20%;
p = 0.028,
g = 0.728).
Rear foot strikers in high-MI shoes had the highest net work values (27.8 ± 8 Nm/kg/°∙100) associated to increased work absorbed (+28% from LOW,
p < 0.001
g = 2.09; +16% from MED,
p < 0.001,
g = 1.60) and produced (+30% from LOW,
p < 0.001,
g = 2.17; +21% from MED,
p < 0.001,
g = 1.17) (
Figure 4); however, the work ratio (absorbed/produced) for RFS was statistically lower than for FFS (0.55 vs. 0.59,
g = 0.533). FFS increased positive work going from LOW to MED (+5%;
p < 0.001,
g = 0.465) and from MED to HIGH (+6%;
p < 0.001,
g = 0.319); while negative work was not statistically different from LOW (28.84 ± 5.8 Nm/kg/°∙100) and MED (29.21 ± 6.0 Nm/kg/°∙100;
p = 0.327), but in HIGH, negative work was higher than in both LOW (+9%;
p < 0.001,
g = 0.342) and MED (+8%;
p < 0.001,
g = 0.299); however, net work in HIGH (20.4 ± 5.5 Nm/kg/°∙100) was similar (
p = 0.781) to MED (20.2 ± 5.0 Nm/kg/°∙100) and LOW (18.8 ± 6 Nm/kg/°∙100).
As for the correlation between energetic (work) measures (
Table 1), FFS exhibited high negative correlations values between W
abs and W
prod in all shoe conditions (r
s ≤ −0.69), meaning that the more work they absorbed during loading, the less work they needed to produce during the unloading phase. RFS did not show such correlations; instead, they exhibited high positive correlations (r
s ≥ 0.60) between W
prod and W
net, meaning that the net work increased as the produced work increased.
4. Discussion
The purpose of this study was to explore the effect of foot strike modes and footwear features on the dynamic control of the ankle dynamics stiffness. There was no group main effect for ankle stiffness, contrary to our hypothesis that FFS had a lower ankle stiffness than RFS. Hamill, Gruber [
5] investigated stiffness during the phase of stance that corresponds most closely to the LRP region of our study. By examining a main effect of group within the LRP region (ignoring ERP and DP), we have also confirmed a statistically higher (+14%;
p = 0.005,
g = 0.725) ankle stiffness in the RFS group. However, within the LRP, there was not a main effect of
Shoe on ankle stiffness (
p = 0.163). Previous studies found that changing shoe support altered the level of joint stiffness [
26,
33]; where ankle dynamic stiffness increased as the shoe hardness decreased [
34]. While increasing stiffness may be functional in preventing excessive joint movement [
35], it has been identified as a possible risk of Achilles tendon injuries in runners [
36].
The rearfoot strike loading technique generated more positive (produced) work by the ankle joint. This confirms our hypothesis and is consistent with previous studies that found ankle plantar flexor muscles to store more elastic energy (negative work) during the loading phase of fast running (i.e., forefoot strike) compared with positive work during unloading [
37,
38]. The RFS group in our study exhibited 34% higher net work compared with FFS (
Table S2 and
Table 1), which correlated strongly with the work produced (
Figure 3); indicating that there was more muscle energy produced compared with elastic energy stored [
39]. Efficient running is achieved by efficiently storing and releasing elastic energy at each step; our results are in line with previous literature data that found FFS to store and return more elastic energy than RFS [
40,
41,
42]. Despite this energetic advantage, FFS is consistently reported to be energetically inefficient [
43,
44], probably because storing energy in passive structures requires muscle contraction [
45]. Therefore, it may be concluded that saving and releasing energy in the plantarflexor muscles may not significantly reduce the whole-body metabolic cost of running with a forefoot strike pattern [
46].
The FFS group demonstrated a time-dependent ankle stiffness across the stance phase, especially for the high-MI shoe condition, fulfilling our hypothesis. Furthermore, within the same shoe condition, the FFS group had strong correlations between ankle stiffness (K
ankle) during both impact and loading phases and net work (W
net). By controlling ankle stiffness, the work around the ankle was modulated, probably to achieve a functional redistribution of loading along the lower limb joints [
10,
47]. Furthermore,
Figure 3 indicates that the K
ankle of FFS running in minimally supportive shoes is constant through the impact, late rising (loading) and unloading subphases, suggesting that foot strike at landing is a determinant for ankle dynamic stiffness not only at impact, but also during the loading and unloading phases. A similar correlation has been found between the initial joint stiffness and maximal stiffness during the stance phase of hopping [
48]. One possible explanation for a constant ankle stiffness is that in that configuration (ankle plantarflexion with minimal support) the ankle–foot complex can express its spring-like function [
49,
50,
51]; while increasing shoe support may introduce a level of instability that requires a trade-off between the task-goals of energy recycling and stable locomotion [
52].
Shoe characteristics influenced the control of ankle dynamic stiffness. Both groups were able to reduce ankle dynamic stiffness during impact and loading phase when wearing high-MI shoes (
Figure 2,
Table S2). However, both groups also increased the work produced and absorbed, so that the total net work done around the ankle during stance increased as a function of the shoe MI index (
Figure 4,
Table S2). Control and modulation of these loads need a certain level of adaptability of both the musculoskeletal and neuronal systems [
53]. This may explain the high risk of certain injuries when changing from low- to high-MI shoes [
54] or from RFS to FFS patterns [
55].
Limitations
We acknowledge that the energy absorbed or produced at the foot/ankle is not only associated to flexion/extension, but also to foot/shoes deformation [
56]. In addition, in this study, analysis was limited to the ankle joint. Indeed, adding analysis on the work done around knee and hip would have validated our assumption on leg-level force stabilization. Other limitations are the assumed symmetry between dominant and nondominant leg. The modulation of joint dynamic stiffness and the redistribution of joint work may vary if significant asymmetry exist [
57].