A total of 10 marathon-experienced recreational runners (demographic information: age, 25.63 ± 2.88 years; height, 171.88 ± 3.64 cm; weight, 64.73 ± 5.68 kg; BMI, 21.99 ± 2.75 kg/m²; training status: running experiences, 4.25 ± 1.81 years; running frequency, 3.13 ± 0.64 times/week; running distance: 7.88 ± 3.76 km per time) were recruited to participate in this study. All of them have regularly joined in the half or full marathon race within 3 years and they were free from any lower limb and foot injuries at least 6 months before this experiment. All runners prefer a rear-foot strike at the beginning of long-distance running and the dominant leg was confirmed to be the right one based on the kicking test. Informed written consent was obtained and this study was approved by the Ethics Committee in Ningbo University (RAGH20201013).

A Vicon motion capture system (Oxford Metrics Ltd., Oxford, United Kingdom) with 8 infrared cameras was applied to collect the foot inter-segment kinematics during running at 200 Hz. A total of 30 spherical reflective markers with 9 mm in diameter were attached to the corresponding bony landmarks through the hole cut in the shoes based on a previously established protocol (Song et al., 2020), and three markers (LD1M, LMMA, and LCPA) were removed after the static calibration trials (Figure 1). GRFs data and gait cycle were obtained using an AMTI force platform (Advance Mechanical Technology Inc, Watertown, United States) embedded in the middle of a 20-m indoor walkway at a frequency of 1000 Hz. The Vicon Nexus software (Version 1.8.5, Oxford Metrics Ltd., Oxford, United Kingdom) was used to record the kinematics and kinetics data synchronously. The same pants and running shoes were provided and participants were required to wear them while running at least 1 week before the test for adaptation.

The baseline data were first collected during the test day after a 10-min basic warm-up and environment familiarization. Participants were instructed to run on the 20-m indoor track at 3.3 m/s (equal to 12 km/h), with the above motion capture system and force platform used to record data and a timing gate to monitor the running speed. After that, participants were asked to run for 10 km on a treadmill at 12 km/h. This running protocol was chosen because a running distance of 10 km is long enough to initiate changes, and 12 km/h corresponds to the average moderate running speed for recreational runners (Kim et al., 2018). The marker and GRFs data were again measured immediately after 5 km of running. After the interval test, participants kept running for another 5 km and did the final 10-km test.

Demographic information and training experience were collected and calculated before the test. Gait data were first labelled and run in Vicon Nexus. An experienced technician further checked the traces and removed all the inconsistent trials. Five successful trials for each participant were extracted for further analysis. The dominant foot inter-segment kinematics, including forefoot with respect to hindfoot angles (FFHFA) and hindfoot with respect to tibia angles (HFTBA) in the sagittal, frontal, and transverse planes, as well as hallux with respect to forefoot angle (HXFFA) in the sagittal plane, were measured. Angle values at initial contact (IC) and toe-off (TO), as well as peak values and range of motion (ROM) in the stance phase, were then derived. For GRFs parameters, the 1st and 2nd vertical GRFs, peak propulsive, and breaking GRFs in the stance phase, together with the 1st and 2nd VALR, were extracted. The peak propulsive and breaking GRFs refer to the peak positive and negative GRF in X-axis (anteroposterior direction), and VALR were calculated by dividing the corresponding vertical GRF with the time from IC. All kinetic parameters were normalized to body weight for further analysis.

For peak variables analysis, one-way repeated-measures analysis of variance (ANOVA) through SPSS 17.0 software (SPSS, Chicago, IL, United States) was taken to examine the significance of foot inter-segment kinematics and GRFs at critical points between the baseline condition, 5 km, and 10 km of running. The Shapiro-Wilk test was first performed to assess data normality and data were presented as mean ± SD (standard deviation). The mean differences (confidence intervals (95%CI)) among groups were also calculated. Moreover, ANOVA of one-dimensional statistical parametric mapping (SPM1d) was also conducted to further observe foot kinematics changes over the stance phase by using MATLAB 2019b software (The MathWorks, Natick, MA, United States). The statistical significance level was set at p < 0.05.

The time-series data of foot inter-segment kinematics and the corresponding SPM1d analysis during the stance phase among three conditions are shown in Figure 2. Table 1, Table 2, Table 3 exhibits test statistics for all kinematic parameters at critical points.

The group average and statistics of GRFs during the stance phase are presented in Table 4. No significant differences were found among conditions except peak propulsive, with its value decreasing significantly after 10 km compared to 5 km of running.

By exploring the forefoot motion with respect to rearfoot, it is possible to gain further insight into the plantar dynamic changes during long-distance running. In the sagittal plane, the plantarflexion angle at TO was found to significantly decrease both after 5 and 10 km of running. Although the differences seem relatively small at first glance (2.45° between baseline and 5 km, 2.01° between baseline and 5 km), several studies have revealed that the intricate foot inter-segment movement could have a significant effect on the plantar fascia function (Ferber and Benson, 2011; Michael E Graham et al., 2011; Chang et al., 2014). For instance, a tiny distance change between the forefoot and rearfoot (< 1 mm) would contribute to a 34.8% change of plantar fascia strain (Michael E Graham et al., 2011). Also, a change of 1° in arch angle can lead to a plantar fascia tension change of 0.4–0.7 times of body weight during the early stance phase (Ferber and Benson, 2011). Thus, we speculated that the decreased forefoot plantarflexion motion during the stance phase may result in the increased stress and strain of plantar aponeurosis and consequently higher injury risks. In addition, smaller forefoot dorsiflexion angles were found during the early (15–20%) stance phase at 10 km of running compared to baseline and 5 km of running, which indicates that runners may tend to a relative midfoot strike pattern after long-distance running, probably because of compensating for local muscle fatigue (Kim et al., 2018).

Concerning forefoot motion in the frontal plane, decreased supination angles were observed throughout the stance phase after 10 km of running compared to 5 km, which may also correlate with the changes of foot posture after long-distance running (Mei et al., 2019). Specifically, it has been proved that long-distance running would result in a more pronated foot posture and a redistributed forefoot plantar load with increased pressure under the 2nd and 3rd metatarsal while decreasing under 4th and 5th metatarsals (Bisiaux and Moretto, 2008; Mei et al., 2019). Moreover, the significantly reduced peak forefoot adduction angle and ROM, as well as relatively smaller forefoot adduction during early (15–21%) and mid-stance (37–51%) phase at 10 km of runnimg, could be another possible explanation for the above findings since forefoot adduction is a part of foot supination during the propulsion phase (Levinger et al., 2010). Similar kinematic changes in frontal and transverse planes were also found in individuals with flat arch during walking (Hunt and Smith, 2004). It has been demonstrated that the pronated foot posture after long-distance running could further lead to a reduced arch height (Fukano et al., 2018), which could let the foot move with a relatively “flat arch” pattern and may induce plantar pain during running.

Compared to previous data, a relatively different result of hallux motion was observed in this study. Generally, rearfoot-strike runners would present great hallux dorsiflexion during the early and last stance phase of running because of the rollover mechanism (Samson et al., 2014). However, our findings showed that, although there was no significant difference among conditions, the hallux exhibited considerably smaller dorsiflexion angle both at IC and TO. A possible explanation for this difference may be the narrow forefoot part of footwear due to modern aesthetic needs. Previous research has demonstrated that the insufficient forefoot space (width and height) may limit the ambulatory function of toes, affect its kinematic performance during locomotion, and potentially lead to foot injuries (such as bruised toenails) because of repetitive friction (Runners Connect, 2013; Wallden, 2016; Xiang et al., 2018). Two studies measuring the foot inter-segment kinematics while walking or running barefoot presented normal hallux dorsiflexion during the stance phase (Sun et al., 2018; Xiang et al., 2020a). Therefore, more comparisons concerning the effects of forefoot space of footwear on hallux biomechanics are warranted. In addition, the strike pattern may also contribute to this difference. Although all participants in this study preferred rearfoot strike at the beginning of the long-distance running test, it was demonstrated above that runners may shift to a relative midfoot landing strategy as indicated by the smaller forefoot dorsiflexion angles from early to mid-stance phases. Compared to baseline and 5 km of running, it was also found that the plantarflexion angles decreased during the early stance phase at 10 km of running. Together with the smaller peak propulsive force found after 10 km of running, it was speculated that the toes’ dynamic control function, such as gripping, may gradually reduce after long-distance running, which was also consistent with previous findings (Kim et al., 2018; Mei et al., 2018). Nevertheless, more studies concerning the hallux biomechanics during long-distance running are recommended.

Most of the distinct effects concerning foot inter-segment kinematics were observed in rearfoot motion. As for the sagittal plane, higher dorsiflexion angles at IC and during the early (0–5%) and mid-late (56–75%) stance phase were presented after 5 km of running, and a higher peak dorsiflexion angle and ROM were presented after both 5 and 10 km of running. Although stride length was not measured in this study, it was previously demonstrated that the increase in rearfoot dorsiflexion resulted in a greater stride length (Sun et al., 2018). Similarly, the decreased dorsiflexion angle at IC and increased peak plantarflexion angle, and larger plantarflexion during the early (15–18%) stance phase may indicate the stride length was reduced again after 10 km of running when compared to 5 km. As muscular fatigue may happen during long-distance running, the lower limb is likely to shift its gait to a shorter but quicker step frequency (Millet et al., 2009; Kim et al., 2018), which could be a possible explanation for the stride length change at 10 km of running.

In the frontal plane, higher eversion angles throughout the stance phase were found after 5 km of running compared to the other two conditions. Previous research provided evidence that increased rearfoot eversion motion would be an injury indicator, which may potentially increase the risk of plantar fasciitis because of excessive use (Ryan B Graham et al., 2011; Chang et al., 2014). Moreover, excessive rearfoot eversion would further lead to excessive midfoot dorsiflexion, and this abnormal kinematic coupling foot motion has been proved a risk factor for lower limb RRI (Takabayashi et al., 2018). In addition, greater internal to external rotation, especially during the mid-late stance (34–75%) phase and consequently increased ROM in the transverse plane, were also observed at 5 km of running. An increase in the rearfoot rotation was demonstrated to be associated with an increase in tibial rotation and then would further affect the proximal joints’ function (such as the knee joint) because of the coupling motion (Lundberg et al., 1989; Williams et al., 2001; Levinger et al., 2010). This could also serve as an underlying mechanism for knee injuries after long-distance running, such as iliotibial band syndrome (Aderem and Louw, 2015).

A comparison with previous data also revealed some different results in GRFs in our study. Kim et al. (2018) conducted a systematic review to investigate the effect of long-distance running on lower-limb biomechanical parameters in healthy runners. They summarized that lower vertical GRFs and loading rate would be presented after long-distance running due to the increased mechanical stress with decreased musculoskeletal capacities. However, no significant differences among conditions were found in this study except a lower peak propulsive force at 10 km of running compared to 5 km. The participants’ heterogeneity between studies would be the primary explanation for these differences since the fatigue-related changes were speculated to initiate at 10 km of running according to our findings in kinematics. Investigating longer-distance running (such as 20 km) based on participants involved in this study will be performed for further verification. Moreover, the inconsistent GRF results of this study may also be due to the equipment accuracy and future studies using more precise devices (e.g., insole-type pressure mat) could add insight (Lung et al., 2008; Jung et al., 2014).

Some limitations of this study should be considered. Firstly, the main purpose of the current study was to investigate the effect of long-distance running on foot adaptions, thus many other parameters, such as running surface, running shoes, and gender were not addressed. Secondly, the plantar pressure differences and the repeated stress loading on soft tissue stiffness before and after long-distance running were not measured, which may add more conclusive explanations for the findings of this study (Pu et al., 2018; Duan et al., 2021). Moreover, the muscular fatigue after 10 km of running was only speculated based on previous research results and further investigation about muscle activity before and after running is warranted for verification. In addition, distinct individuals may show different foot kinematic responses during running and a cluster analysis may help find the particular baseline patterns among them (Watari et al., 2021). Lastly, we only investigated the foot kinematic differences during the stance phase as it is the only time for foot–ground interaction. However, the swing phase kinematics shall present more preparatory foot adaptations during long-distance running (Dixon et al., 2011). Further large sample size studies concerning these aspects would reveal more profound knowledge.