Fatigue-Related Biomechanical Changes During a Half-Marathon Under Field Conditions Assessed Using Inertial Measurement Units

ound/Objectives: Running is one of the most popular physical activities world- wide and have been widely studied in relation to performance and injury prevention. In addition to measurements conducted under standardized laboratory conditions, in- ertial measurement units (IMUs) allow for the assessment of biomechanical parameters in real-world settings—particularly during endurance runs. The aim of this study was to investigate how running a half-marathon under field conditions affects exertion and various biomechanical parameters, as measured using IMUs.Methods: Twenty runners completed a half-marathon on a flat, even-surfaced walkway at a self-selected, constant pace corresponding to a brisk training run. In addition to lower limb biomechanics, heart rate (HR) and ratings of perceived exertion (REP) were also recorded.Results: A significant increase in both HR and RPE was observed toward the end of the half-marathon, indicating the presence of fatigue during the later stages of the run. The biomechanical results further demonstrate that this fatigue was associated with increased peak tibial acceleration, peak angular velocity in the sagittal plane of the foot, and peak rearfoot eversion

were also recorded.Results: A significant increase in both HR and RPE was observed toward the end of the half-marathon, indicating the presence of fatigue during the later stages of the run. The biomechanical results further demonstrate that this fatigue was associated with increased peak tibial acceleration, peak angular velocity in the sagittal plane of the foot, and peak rearfoot eversion velocity, while foot strike angle, stride frequency, and stride length remained unchanged. Furthermore, a progressive increase in ground contact time and a decrease in flight time were observed over the course of the run, resulting in an increased duty factor.Conclusions: These findings highlight the value of IMU-based assessments for detecting fatigue-related biomechanical changes during prolonged runs in real-world conditions, which may contribute to early identification of overload and inform injury prevention strategies. Keywords:running; half-marathon; biomechanics; inertial measurement unit; peak tibial acceleration; peak rearfoot eversion velocity 1. Introduction Running is one of the most widely practiced forms of physical activity and is asso- ciated with substantial health benefits. Beyond its health-promoting effects, the appeal of competition and performance comparison has become increasingly prominent, as evi- denced by the growing number of marathon or half-marathon participants worldwide [1–3]. However, running also carries a notably high incidence of overuse injuries, with prevalence Biomechanics2025,5, 101 https://doi.org/10.3390/biomechanics5040101

Biomechanics2025,5, 101 2 of 16 ranging from 19.4% to 92.4% among distance runners [4]. This is likely attributable in part to increased training volumes and intensities during competition preparation phases, al- though the relationship between training load and injury risk remains a subject of ongoing debate [5,6]. Given the multifactorial etiology of overuse injuries, attention has increasingly turned to biomechanical parameters that may act as modifiable risk factors. In this context, several studies have focused on biomechanical parameters in distance running particularly in relation to muscular fatigue, which naturally accumulates during prolonged endurance runs or at higher intensities, in healthy, recovered, and acutely symptomatic runners [7–23]. One of these biomechanical parameters is peak tibial acceleration (PTA), which pro- vides information on how the shock wave resulting from the forces at initial ground contact (IC) propagates through the lower extremities and allows inferences about the internal forces acting on the musculoskeletal system [24,25]. Thereby, excessive values in PTA have been associated with increased mechanical strain on the tibia and surrounding struc- tures, potentially predisposing athletes to bone stress injuries [11,26–30]. The influence of muscular fatigue on PTA has been examined primarily during running on treadmills or indoor running tracks, with inconsistent results. While some investigations suggest that muscular fatigue induces adaptations in running mechanics—reflected in increased PTA values [9,11]—other studies have not reported consistent evidence of such changes [7,12]. To date, apart from indoor running studies, only a few investigations have addressed PTA during endurance running in real-world settings, likewise with inconsistent results. For example, Morio et al. [18] reported no change in PTA during field running, whereas Ruder et al. [22] observed a decrease at the end of an endurance run. In addition to the impact loading parameter PTA, it is also relevant to consider the influence of muscular fatigue on kinematic measures such as peak eversion velocity (evVel), peak angular velocity in the sagittal plane (PAV), and foot strike angle (FSA). In this context, several studies have shown that with increasing muscular fatigue, evVel increases significantly [10,11,31], whereas FSA may decrease significantly [12,20,32]. No recent study could be identified

is also relevant to consider the influence of muscular fatigue on kinematic measures such as peak eversion velocity (evVel), peak angular velocity in the sagittal plane (PAV), and foot strike angle (FSA). In this context, several studies have shown that with increasing muscular fatigue, evVel increases significantly [10,11,31], whereas FSA may decrease significantly [12,20,32]. No recent study could be identified that examined the influence of muscular fatigue on PAV. However, this parameter is highly relevant as it reflects the foot rollover in the sagittal plane [33]. It may provide insights into the ability to actively decelerate plantarflexion after IC, a function that relies on the dorsiflexors and may be impaired by their fatigue, which in turn can influence shock absorption [34]. Furthermore, fatigue-related changes in spatiotemporal parameters—including step frequency (SF), ground contact time (tC), flight time (tF), and duty factor (DF)—are of particular interest, as they represent fundamental descriptors of running mechanics. They are generally stable at constant speeds but may show subtle changes under muscular fa- tigue, which can reflect compensatory strategies and influence running efficiency or injury risk [7–9,11,12,18] . However, the existing literature reports inconsistent findings regard- ing these adaptations, which may be partly explained by differences in methodological approaches such as running speed, exercise intensity, or running surface. As shown, most studies investigating the effects of muscular fatigue on the above- mentioned biomechanical parameters have been conducted on treadmills or indoor running tracks, often with inconsistent results [7–11,15,17]. Further investigations have been carried out under field conditions, likewise reporting inconsistent findings or not including relevant parameters of impact loading (e.g., PTA) and foot rollover kinematics (e.g., evVel) in conjunction with spatiotemporal parameters [13,14,19–21,23]. These inconsistencies point to a broader issue: muscular fatigue appears to elicit different biomechanical responses depending on the running context. Specifically, the

Biomechanics2025,5, 101 3 of 16 existing literature indicates that biomechanical adaptations in response to muscular fatigue may differ between treadmill and field running. This suggests that changes in parameters such as PTA are strongly influenced by the study design—particularly the running surface, speed, and intensity. Supporting this, higher PTA values have been reported during outdoor running compared to treadmill conditions [35–37]. Consequently, biomechanical variables associated with overuse injuries—such as medial tibial stress syndrome (MTSS) or tibial stress fractures (TSF)—identified through laboratory-based gait analysis may not accurately reflect their magnitudes during field running. To address this limitation, studies conducted in natural environments (e.g., on streets or sidewalks) are necessary to investigate how muscular fatigue influences biomechanical parameters under ecologically valid conditions [38]. Furthermore, a review by Xiang et al. [39] highlighted that, beyond impact parameters such as PTA or ground reaction forces, a holistic approach should be adopted by integrating additional biomechanical aspects, with inertial measurement units (IMUs) providing a promising tool for this purpose. In this context, IMUs, which combine accelerometers and gyroscopes, offer a particu- larly suitable solution for field-based research due to their compact size and lightweight design, as has been demonstrated in several field-based studies [13,18,33,36,37]. As demonstrated above, previous findings remain inconsistent, underscoring a crucial gap in knowledge regarding how fatigue affects running biomechanics under field condi- tions. This highlights the need for approaches that enable the comprehensive assessment of multiple biomechanical parameters outside the laboratory. Therefore, the aims of this study were twofold: (1) to investigate whether IMUs are suitable for capturing a broad range of biomechanical parameters during a half-marathon run under field conditions, extending beyond the more limited approaches used in previous research, and (2) to examine how such a run affects exertion and the above mentioned biomechanical parameters. It was hypothesized (I) that both heart rate and perceived exertion would significantly increase with the duration of the run. With respect to biomechanics, it was hypothesized (II) that PTA would significantly increase from the beginning to the end of the run as a result of acute muscular fatigue and a reduced capacity to

and the above mentioned biomechanical parameters. It was hypothesized (I) that both heart rate and perceived exertion would significantly increase with the duration of the run. With respect to biomechanics, it was hypothesized (II) that PTA would significantly increase from the beginning to the end of the run as a result of acute muscular fatigue and a reduced capacity to absorb impact forces. Furthermore, it was hypothesized (III) that foot kinematics—specifically, evVel and PAV—would increase over the course of the run due to progressive muscular fatigue in the lower extremities, whereas the FSA would decrease. Regarding spatiotemporal parameters, it was hypothesized (IV) that, despite a constant running speed, increasing fatigue during a half marathon would lead to a decrease in strLen and tF, while SF and tCwould increase. We also hypothesized (V) that exertion parameters are correlated with biomechanical parameters. This study aims to demonstrate the potential of IMU-based systems for continuous, field-based monitoring of running biomechanics under fatigue conditions during prolonged endurance runs. Understanding how biomechanical parameters change under real-world conditions provides valuable insights for injury prevention in endurance runners. In addition, the findings may inform training strategies and fatigue monitoring through real-time feedback to promote safer and more efficient running over longer distances. 2. Materials and Methods 2.1. Participants In summary, 20 runners without any injuries in the last six months were recruited for this study. Demographic and running-related characteristics of the runners are presented in Table. Participants ran an average of 36.8±21.8 km per week. During the half-marathon run, the average running speed was 11.4±0.9 km/h.

Biomechanics2025,5, 101 4 of 16 Table 1.Demographic and running-related characteristics of the runners (mean±SD). Age [Years] Height [cm] Weight [kg] Gender [Male/Female] Weekly Running Distance [km] Running Speed [km/h] 36.9±12.4 177.2±6.4 71.5±10.8 16/4 36.8 ±21.8 11.4 ±0.9 The runners were informed about the purpose and design of this study, signed an informed consent document, and completed a form with their personalized data. All procedures were performed in accordance with the recommendations of the Declaration of Helsinki. This study was approved by the Ethics Committee of the Chemnitz University of Technology (#101627696). 2.2. Experimental Setup and Procedures 2.2.1. Running Data acquisition took place on a straight, flat concrete sidewalk approximately 1.4 km in length. Following an individual warm-up period of 5 to 10 min, each participant com- pleted 15 laps on the sidewalk, totaling the official half-marathon distance of approximately 21.1 km. Running speed was self-selected and corresponded to each runner’s typical train- ing pace, as suggested in previous studies [10,18]. Participants were instructed to choose a running pace they could maintain continuously over the entire half-marathon distance with- out the need to stop; however, a minimum speed of 10 km/h was required. The running speed was constantly checked by the experimenter on a bicycle with a speedometer [33,36]. During the half-marathon run, participants used their own running shoes. Runners rated their perceived exertion (RPE) on a 15-point Borg scale (from 6 to 20) at the beginning of the run and after each lap [40]. Furthermore, participants wore a Garmin Forerunner ® 735XT (Garmin, Olathe, KS, USA) equipped with an external heart rate sensor (HRM-Pro™Plus, Garmin, Olathe, KS, USA) to monitor heart rate (HR). 2.2.2. Sensor Setup Based on Hill et al. [36], four small, lightweight inertial measurement units (IMU; ICM-20601, InvenSense, San Jose, CA, USA; weight: 4 g; sampling rate: 2000 Hz) combin- ing a tri-axial accelerometer (measurement range:±353 m/s 2 ) and a tri-axial gyroscope (measurement range:±4000 ◦ /s) were used to measure biomechanical data (FigureA). All IMUs were connected via cables to a data logger, which was secured on a belt around the participant’s waist. According to Kiesewetter

InvenSense, San Jose, CA, USA; weight: 4 g; sampling rate: 2000 Hz) combin- ing a tri-axial accelerometer (measurement range:±353 m/s 2 ) and a tri-axial gyroscope (measurement range:±4000 ◦ /s) were used to measure biomechanical data (FigureA). All IMUs were connected via cables to a data logger, which was secured on a belt around the participant’s waist. According to Kiesewetter et al. [41], one IMU was attached to the shaved medial aspect of each tibia, midway between the malleolus and tibial plateau, using double-sided adhesive tape, an elastic strap, and a compression sleeve to minimize sensor movement. The sensor axes were aligned with the longitudinal axes of the tibia. Addi- tionally, an IMU was mounted on the heel cap of each running shoe using double-sided adhesive tape and secured with additional elastic tape.

Biomechanics2025,5, 101 5 of 16 Figure 1.(A) Schematic illustration of the setup with inertial measurement units (IMUs), data logger, Garmin Forerunner ® , and heart rate sensor; (B) Vertical acceleration data from the heel-mounted IMU (black line) as well as from the tibia-mounted IMU (blue line), showing peak tibial acceleration (PTA); (C) Angular velocities from the heel-mounted IMU in the frontal plane (green line) and sagittal plane (red line), used to determine peak rearfoot eversion velocity (evVel) and peak angular velocity for foot rollover sagittal (PAV). Toe-off was defined as maximum angular velocity in the sagittal plane, >100 ms post IC (red triangle). A sample dataset was used for illustration. 2.3. Data Analysis Raw data from the IMUs were analyzed post-processing using MATLAB 2024a (Math- Works™, Natick, MA, USA). Prior to analysis, signals were filtered to reduce noise using a zero-lag, 4th-order Butterworth low-pass filter—applied at 80 Hz for accelerometer data and 50 Hz for gyroscope data [41]. Biomechanical Parameters For accurate detection of IC and stride segmentation, the vertical acceleration signal from the heel-mounted IMU was further processed using a zero-lag Butterworth high-pass filter at 80 Hz. The first prominent peak in the filtered signal was identified as IC, ensuring a valid detection of foot strike events for subsequent gait analysis [42]. The stride duration (strD) was then calculated as the time between two consecutive ICs. In addition, the SF was calculated as the inverse of the time between two consecutive ICs, multiplied by two (assuming that the right and left legs have the same step duration). According to Sabatini et al. [43], toe-off was identified as the point at which the angular velocity in the sagittal plane of the foot-mounted sensor reaches its maximum value, occurring at least

Biomechanics2025,5, 101 6 of 16 100 ms after IC (FigureC). Furthermore, t Cwas defined as the interval between IC and the corresponding toe-off event, while tFwas computed as follows (1): tF= strD−tC (1) The duty factor (DF), representing the proportion of time the foot is in contact with the ground during a stride, was used to assess mechanical work [20,44]. A higher DF indicates a longer tC, whereas a lower DF reflects longer tFand reduced tC. The following Equation (2) was used to calculate the relative DF: DF = tC(tF+ tC) −1 * 100% (2) The highest peak in the vertical acceleration signal of the tibia-mounted sensor was defined as PTA (FigureB). To assess evVel, the maximum angular velocity in the frontal plane of the shoe-mounted sensor was analyzed [45] (FigureC). For foot rollover, PAV was determined within 100 ms after IC, following the approach described by Bräuer et al. [33] (FigureC). To determine the FSA at IC, the orientation of the shoe in the sagittal plane was calculated using angular velocity data from the heel-mounted IMU [46]. This signal was integrated to obtain the foot angle (θ). To correct for integration drift and offset, a nulling algorithm was applied using two consecutive stance phases where the foot was flat on the ground, identified by minimal changes inθ. A linear offset correction between these points was subtracted fromθ. FSA was then defined as the corrected orientation angle at IC. A detailed description of this procedure can be found in Mitschke et al. [46]. The drift- and offset-correctedθsignal, along with the corresponding vertical and horizontal acceleration data, was used to calculate strLen between two consecutive ICs [46]. Due to technical issues (such as cable breakage or sensor failures) occurring during testing, all biomechanical parameters were calculated based on data from only one leg per runner. 2.4. Statistics Each runner completed a total of 15 laps. For each lap, means were calculated for all biomechanical parameters. Subsequently, for lap 1 as well as laps 8 and 15, group means and standard deviations (mean±SD) were calculated across all runners. The three laps represent

all biomechanical parameters were calculated based on data from only one leg per runner. 2.4. Statistics Each runner completed a total of 15 laps. For each lap, means were calculated for all biomechanical parameters. Subsequently, for lap 1 as well as laps 8 and 15, group means and standard deviations (mean±SD) were calculated across all runners. The three laps represent different race segments: the start (L1), running in a “non-fatigued state”, running in the middle of the race (L8), and running in a “fatigued state” (L15). This segmentation into three phases was based on the approach of Prigent et al. [20]. Statistical analyses were performed using IBM SPSS Statistics (IBM Corp., Armonk, NY, USA, Version 30.0). All data were visually inspected and tested for normal distribution using the Shapiro–Wilk test. In cases where normality was confirmed, a one-way repeated- measures ANOVA was conducted to compare the three laps (L1, L8, and L15), followed by a Bonferroni-corrected post hoc test. If the assumption of normality was violated, the Friedman test was applied for the same laps, with subsequent Dunn-Bonferroni tests used for post hoc analysis. Statistical significance was set atα= 0.05 for all analyses. In addition, effect size (Cohen’s d) was calculated to quantify the magnitude of differences when statistical significance was found. The coefficients were interpreted as trivial (d < 0.2), small (d < 0.5), medium (d < 0.8), or large effects (d≥0.8) [47]. Pearson’s correlation coefficient was used to assess linear relationships between exertion parameters (RPE and HR) and biomechanical parameters.

Biomechanics2025,5, 101 7 of 16 3. Results 3.1. Exertion Parameters Table lowest values were observed in L1 and the highest in L15. The ANOVA revealed significant differences across laps. Subsequent post hoc tests showed significant pairwise differences between all laps, with large effect sizes (d > 0.98). Table 2.Effects of running duration and muscular fatigue on heart rate (HR) and perceived exertion (RPE) (mean±SD). Statistically significant group differences are indicated with *, while significant differences between laps 1 (L1), 8 (L8), and 15 (L15) are marked with a, b, and c, respectively; effect sizes (Cohen’s d) are reported for the respective pairwise comparisons. L1 L8 L15 p-Value d HR [1/min] 133.8 ±13.8 a; b 148.4±15.8 a; c 156.6±19.2 b; c * < 0.001 a < 0.001 b < 0.001 c < 0.001 a = 0.98 b = 1.36 c = 0.47 RPE (6–20) 9.0 ±1.6 a; b 12.8±2.3 a; c 15.9±2.2 b; c * < 0.001 a < 0.001 b < 0.001 c < 0.001 a = 1.90 b = 3.53 c = 1.36 3.2. Biomechanical Parameters The biomechanical parameters for the three examined segments of the half-marathon are presented in Table. An increase in PTA over the course of the run was measured. ANOVA revealed significant differences between L1 and L15 (p< 0.001) with a medium effect size (d = 0.58). For both evVel and PAV, increases in angular velocities were observed from L1 to L15. While evVel showed only a trend toward higher velocities, PAV exhibited significant differences between all three laps (L1 vs. L8:p= 0.046; L1 vs. L15: p= 0.011; L8 vs. L15:p= 0.024), with trivial (L8 vs. L15: d = 0.012) to small effect sizes (L1 vs. L8: d = 0.31; L1 vs. L15: d = 0.42). No significant difference in FSA was found between the laps (p= 0.210). Table 3.Effects of running duration and muscular fatigue on peak tibial acceleration (PTA), peak eversion velocity (evVel), peak angular velocity in the sagittal plane (PAV), and foot strike angle (FSA) (mean±SD). Statistically significant group differences are indicated with *, while significant differences between laps 1 (L1), 8

This research highlights the impact of fatigue on running biomechanics.