Learning effects in over-ground running gait retraining: A six-month follow-up of a quasi-randomized controlled trial

ABSTRACT

This study evaluated learning and recall effects following a feedback-based retraining program. A 6-month follow-up of a quasi-randomized controlled trial was performed with and without recall. Twenty runners were assigned to experimental or control groups and completed a 3-week running program. A body-worn system collected axial tibial acceleration and provided real-time feedback on peak tibial acceleration for six running sessions in an athletic training facility. The experimental group received music-based biofeedback in a faded feedback scheme. The controls received tempo-synchronized music as a placebo for blinding purposes. The peak tibial acceleration and vertical loading rate of the ground reaction force were determined in a lab at baseline and six months following the intervention to assess retention and recall. The impacts of the experimental group substantially decreased at follow-up following a simple verbal recall (i.e., run as at the end of the program): peak tibial acceleration:-32%, p = 0.018; vertical loading rate:-34%, p = 0.006. No statistically significant changes were found regarding the retention of the impact variables. The impact magnitudes did not change over time in the control group. The biofeedback-based intervention did not induce clear learning at follow-up, however, a substantial impact reduction was recallable through simple cueing in the absence of biofeedback.

KEYWORDS:

• Motor learning
• biofeedback
• biomechanics
• motor control
• music

Introduction

The notion of gait retraining in human running has recently been studied in relation to the objective of impact reduction (Cheung et al., 2018; Clansey et al., 2014; Futrell et al., 2021) and, relatedly, to injury risk management (Chan et al., 2018; Morris et al., 2020). A clinical trial conducted by Chan et al. (2018) focused on gait retraining for impact reduction. The trial included an evaluation of running-related injuries up to one year after the retraining program. Novice runners were initially screened on the vertical loading rate, which is an impact measure and represents the steepness of the force-time waveform in the early stance phase of running gait. Those with a moderate-to-high vertical loading rate were studied further. In a randomized controlled design, the experimental group received a gait instruction (i.e. “run softer”) and real-time visual biofeedback on their vertical ground reaction force. This feedback gave them information about a modifiable biomechanical metric and is provided with the goal of being able to manipulate it at will. By the end of the program, the experimental group had successfully lowered their vertical loading rate. One year later, this group showed a 62% decrease in total running-related injury risk compared with the control group. Although the mechanisms for this reduction in overuse injuries are unclear and multiple ex vivo studies refute the idea of higher loading rates playing a key role in the pathophysiology of biological tissues (Edwards, 2018; Loundagin et al., 2018; Zitnay et al., 2020), the results of this clinical trial suggest that retraining for lower-impact running can have a positive effect on injury risk. While these findings are intriguing, Chan and colleagues did not document the if the impacts were reduced at follow-up. So, it is unknown if the motor system adapted in time. Without a persistent change in mechanics, implying long-term retention of the gait modifications, injury risk is not likely to be altered according to Bowser et al. (2018). Therefore, it is warranted to explore gait retraining for impact reduction in distance running from a motor learning perspective.

Wearable technology may help learn motor skills like lower impact running outside of a lab. We can use body worn equipment to identify the axial peak tibial acceleration (PTA_a). PTA_a has been successfully correlated with the maximum vertical loading rate at multiple speeds in level running (Van den Berghe et al., 2019). Runners can reduce PTA_a with the simple use of real-time feedback on PTA_a (Clansey et al., 2014; Van den Berghe, Derie, et al., 2022). In one study, PTA_a has been reduced by about 30% following a 3-week retraining program in a laboratory setting (Clansey et al., 2014). More recently, a feedback-based retraining program has been completed in a controlled training environment (Van den Berghe, Derie, et al., 2022), which also resulted in a clear reduction in the PTA_a upon its completion. However, there hasn’t been much research on the learning effects of using real-time biofeedback without gait instruction to reduce impacts (Clansey et al., 2014).

Gait-retraining studies often focus implicitly or explicitly on the immediate effects of impact reduction and on variables potentially related to injury risk such as ground reaction forces (Chan et al., 2018; Clansey et al., 2014; Crowell & Davis, 2011; Derie et al., 2022; Van den Berghe et al., 2021). This approach is useful for examining the acute sensitivity of running-related variables to feedback interventions, but neglects long-term motor learning that may occur. Altering a motor pattern that has been reinforced over millions of cycles might be possible if guidance and practice are provided (Bowser et al., 2018; Davis & Futrell, 2016). Data from a recent report on gait retraining, including the use of visual feedback on axial tibial acceleration and an instruction to run softer, suggests that impact reduction can be achieved and maintained up to one year (Bowser et al., 2018); however, the retraining program was performed in a lab environment and without control group. More recently, feedback-based gait retraining targeting a reduction in PTA_a in a controlled training centre has been studied in a quasi-randomized controlled trial (Van den Berghe, Derie et al., 2022). Furthermore, the impact reduction achieved while receiving the real-time, music-based feedback on PTA_a was relatively reproducible through simple verbal cueing without requiring this feedback (Derie et al., 2022; Van den Berghe; Derie et al., 2022). Although the studies performed by (Van den Berghe and Derie et al). have been performed in a controlled trial design (Derie et al., 2022; Van den Berghe; Derie et al., 2022), they are limited to the short-term influence of the biofeedback. Thus, an evaluation should be carried forward to examine if the impact reduction persists over time and whether lower impact running is recallable.

In this study, we took an extended approach to evaluating learning effects and asked whether lower impact running persists and is recallable half a year after completing the biofeedback-based retraining program. This question is well suited to the use of retention testing, which is our primary method for assessing the long-term retention of motor skill learning. Researchers conducting motor learning studies often include retention testing as a means to determine the durability and persistence of the learned skills over time (Bowser et al., 2018; Bramah et al., 2019; Clansey et al., 2014; Futrell et al., 2020). Using retention testing provides information about the persistence of the motor adaptation, and helps advance our understanding of the learning process and potential interventions for skill retention. The retention testing phase typically occurs after an initial training or learning phase, where participants undergo practice to acquire or improve a specific motor skill or task. During retention testing, participants may perform the previously learned motor skill without any additional training or biofeedback (Davis & Futrell, 2016). The performance is then compared to their past performance level (Bowser et al., 2018; Clansey et al., 2014). Here, the main objective of retention testing is to determine whether the motor adaptation is retained over time or if it diminishes. Furthermore, we evaluated whether the impact reduction is recallable through a simple reminder. Based on the few studies on this topic that included some form of retention testing (Bowser et al., 2018; Clansey et al., 2014), we hypothesized that motor learning occurs and the impact reduction would persist without and with recall.

Methods

Experimental design

This follow-up of a parallel, quasi-randomized controlled trial is part of a study series (Derie et al., 2022; Van den Berghe; Derie et al., 2022). Hence, the study cohort is identical to preceding studies in which the immediate effects on impact reduction and running biomechanics have been documented (Derie et al., 2022; Van den Berghe; Derie et al., 2022). Reporting of the study followed the CONSORT statement (Figure 1). The institutional ethics committee reviewed and approved the experimental procedure. We carried out the methods following their guidelines and regulations. Interested volunteers completed an online questionnaire and made an appointment with the research team.

Figure 1. CONSORT 2010 diagram: flow chart of study participants and conditions.

Participants enrolled in the first (screening) phase were invited to participate in the second (intervention and follow-up) part of the study if they fulfilled the following inclusion criteria: aged 18–60 years, no contraindication to performing running activity, typically running 15 km per week or more spread over at least two weekly sessions. To fully qualify, participants were required to attend a brief running impact screening in a sports laboratory in which only tibial acceleration was collected while running at ~3.2 m/s. Extended methods have been made available and further describe the data collection of the screening phase performed (Van den Berghe, Derie et al., 2022). We iteratively invited the subject with the highest PTA_a for further participation in the study and the first twenty who agreed to participate were enrolled in the intervention phase (Derie et al., 2022; Van den Berghe; Derie et al., 2022). These participants were assigned to either the retraining group (n = 10, age: 32.1 ± 7.8 years, body mass: 71.5 ± 18.3 kg, body height: 1.74 ± 0.11 m, weekly running volume: 27 ± 10 km, self-reported training speed: 2.9 ± 0.3 m/s, males/females: 5/5) or the control group (n = 10, age: 39.1 ± 10.4 years, body mass: 69.9 ± 12.8 kg, body height: 1.74 ± 0.11 m, weekly running volume: 37 ± 18 km, self-reported training speed: 2.9 ± 0.4 m/s) and were blinded to the group assignment. The sample size was chosen so that short-term effects in impact reduction could be observed in a quasi-randomized controlled design through the application of music-based biofeedback (Derie et al., 2022; Van den Berghe; Derie et al., 2022).

Two lab visits were scheduled, prior to and after the intervention, during the 6^th month of the follow-up. Both tests were performed while running continuously on an indoor track (Figure 2) at a pace of 2.9 ± 0.2 m∙s⁻¹ in mimicked pairs of the participants’ habitual running footwear. An overview of the subjects’ habitual footwear and matching laboratory footwear has been made available (Derie et al., 2022). The first visit comprised a 5-min warm-up that served as a familiarization to the setup, and a continuous run of about 25 minutes to determine baseline values. The second visit also comprised a warm-up which was followed by two running conditions to assess learning: the retention condition and the recall condition. In case of retention testing, participants may intentionally produce the desired gait pattern during assessments, making it difficult to discern if motor learning and retention have truly occurred (Futrell et al., 2020). Therefore, to avoid response bias, participants were intentionally distracted in the retention condition. They performed a cognitive distraction task each time when passing the lab’s measurement zone. Specifically, a modified Stroop test started each time a participant would enter the measurement zone (Figure 2, video S1). We projected a written name of a colour, in which the name of the colour and the colour of the text not necessarily matched (video S1). They were instructed to say the colour of the text out loud, but not the word itself. Participants were asked to say the word accurately and as fast as possible. For the recall condition, the participants were asked to reproduce the running technique from the intervention six months earlier. We used a simple verbal instruction for all subjects: “Try running as you did in the final session of the retraining program. “The instructed running speed was in agreement with that of the running program (Derie et al., 2022).

Figure 2. Experimental setup. The running course is illustrated on top. The picture shows a participant during the retest condition at follow-up. A stroop test was presented to participants in front of the runway while they ran laps on an indoor track. The projection screen was positioned approximately 15 m away from the centre of the measurement zone. The associated video clip is supplemented (video S1).

Running program

Both groups were subjected to a 3-week running program in a supervised training facility (Van den Berghe, Derie et al., 2022). The 3-week schedule and the six sessions were similar to the program design made by Clansey et al. (2014). Each running session consisted of twenty minutes of running with music. Participants were told to run with specifically selected music tailored to them. The music-only group listened to music tracks that were suited for synchronization to the individual’s running cadence. As such, the smart music player adjust and could mirror the beats per minute of the music to the steps per minute of the runner. The experimental group additionally received real-time feedback on PTA_a. This feedback was music-based wherein the music was distorted at medium and high PTA_a (Van den Berghe, Derie et al., 2022; Van den Berghe et al., 2021). This impact measure was converted into noise that is perceptible by the runner, wherein the conversion is done based on a predefined relationship between perceived distortion levels and imposed distortion levels (Lorenzoni et al., 2019). Specifically, pink noise was superimposed on the music if the peak tibial acceleration was ≥ 70% of the runner’s starting PTA_a (Van den Berghe, Derie et al., 2022). This starting value was determined during the 5-minute warming-up run without any musical feedback (Van den Berghe, Derie et al., 2022). The limb exhibiting the highest mean PTA_a in this period was considered dominant and used further. It has been concluded that a faded feedback approach is superior to continuous feedback for retaining motor skills (Winstein, 1991). Fading of the feedback encourages internalization of the altered running form (Davis & Futrell, 2016), implying the motor skill learning may benefit from fading the feedback on PTA_a in time. Therefore, we implemented a two-phased feedback scheme desired to stimulate motor learning (Van den Berghe, Derie et al., 2022). The fading of the biofeedback prevents the reliance on the feedback and enhances the internalization, and thus learning, of the new motor pattern (Davis & Futrell, 2016). In the first (acquisition) phase, feedback was continuously provided which helps to develop the connection between the extrinsic feedback and the internal sensory cues associated with the desired target behaviour (Davis & Futrell, 2016). So, the first two sessions of running comprised of 20 min of continuous biofeedback. In the second (transfer) phase, the feedback was systematically removed, meaning the time of biofeedback provision gradually decreased in the last four sessions. We did not inform the experimental group about this faded feedback scheme in which the volume of the feedback varied. The running speed was steered to approximately 2.9 m.s⁻¹ throughout these sessions (Van den Berghe, Derie et al., 2022), which corresponded to the mean self-reported endurance training speed of the participants. Please see reference (Van den Berghe, Derie et al., 2022) for a detailed description of the retraining protocol. In this 3-week program, participants were free to choose whether to maintain the gait modifications during their regular training routine. Following completion of the program, participants were advised to adopt their new gait pattern in their regular running practice (Chan et al., 2018).

Materials and lab-based measurements

Participants were equipped with a wearable sensor system to determine PTA_a in the lab visits. The main components of this system were two accelerometers (LIS331, Sparkfun, Colorado, USA, sampling rate: 1000 Hz/axis) wired to a microcontroller. This microcontroller was connected to a 7-inch tablet strapped to a stripped backpack (Van den Berghe et al., 2019, 2020). Sensor weight and the method of attachment plays an important role in data quality according to a systematic review (Norris et al., 2014). The accelerometers were lightweight with a mass of less than 3 grams per unit. Our method of attachment was similar to an approach that has been shown to be reliable within and between sessions of over-ground, level running (Van den Berghe et al., 2019). Each sensor was firmly taped to the skin at eight centimetres above the malleolus medialis on the anteromedial aspect of a lower limb. Prior to this action, the skin in the vicinity of the selected sensor location got pre-stretched with sports tape to ensure a rigid coupling. Then, a test leader visually aligned the vertical axis of each accelerometer with the longitudinal axis of the tibia. Thus, we assumed the acceleration collected corresponds to the acceleration of the tibial bone. The body-worn sensors permitted participants to run continuously along a 30-m instrumented running track (Derie et al., 2022). Two pairs of photo gates were positioned in the measurement zone 6 metre apart. Two force platforms (AMTI, Watertown MA, USA; dimensions: 2.1·0.5-m and 1.2·1.2-m; sampling rate: 2000 Hz) were situated in the centre of the measurement zone. These platforms had covers on top which were flush with the running track. Acceleration and force plate data were synchronized in time up to millimetre accuracy by applying a synchronization protocol described elsewhere (Van den Berghe et al., 2019).

Statistical analysis

Baseline demographic data were analysed with independent t-tests to determine presence of differences between the experimental and control groups. Running variables of interest were PTA_a and the vertical loading rate. The PTA_a was defined as the maximal positive axial acceleration of the lower leg while the foot contacted the ground. The events of foot-ground contact were derived from the vertical ground reaction force. This force was filtered using a 2^nd order, zero-lag low-pass Butterworth filter with a 60 Hz cut-off frequency. The force threshold was set at 20 N. The vertical loading rate was calculated as the maximum value of the first derivative of the ground reaction force in the first 0.050s of stance. These variables were analysed via linear mixed model with an α = 0.05 (JASP 0.16.3). The fixed effect variables we entered were Group (experimental, control) and Conditions (baseline, retention, recall) and the subjects as random effects grouping factor. In the case of a significant group × time interaction, post-hoc comparisons were conducted with a Bonferroni correction. Cohen’s d was calculated to determine the magnitude of the effect (d < 0.2 – very small, 0.2 ≤ d < 0.5 – small, 0.5 ≤ d < 0.8 – medium, d ≥ 0.8 – large). Handling missing data is an important, yet difficult and complex task when analysing results of randomized trials. Therefore, in a post-hoc analysis, missing values were replaced using single imputation of the mean (Jakobsen et al., 2017). In simple mean imputation, missing values are replaced by the mean for the variable of interest.

Results

Two experimental subjects were lost because of an injury or due to a personal reason after the retraining phase was completed. The follow-up happened after 183 ± 9 days (mean ± standard deviation). There were no significant differences between groups in the running variables of interest at baseline. There were significant group × time interactions for PTA_a (F = 3.369, p = 0.047) and vertical loading rate (F = 4.345, p = 0.021). Subsequent pairwise comparisons revealed only a significant change from Baseline-Recall for the experimental group: PTA_a (Meandifference = −2.43 g or −31%, t = 3.905, p = 0.007, d = 1.295) and vertical loading rate (Meandifference = −30.5 BW/s or −34%, t = 4.231, p = 0.003, d = 1.195) (Figure 3, Table S1). For the control group, no significant changes occurred at 6-month follow-up in either the Retention or Recall compared with Baseline levels. The post-hoc analysis with imputed missing data revealed that the change between Baseline and Retention in the experimental group became statistically significant for PTA_a (|Mean difference| = −1.96 g, t = 3.429, p = 0.023, d = 1.065) (Table S1).

Figure 3. Peak axial tibial acceleration and vertical loading rate for the experimental and control groups at each of the time points. The retention and recall measurements were conducted within the same lab visit six months after the baseline measurement. Error bars represent 95% confidence interval. * indicates a statistical difference from the post-hoc comparisons.

Discussion

The current study evaluated retention and recall effects of a biofeedback-driven program of gait retraining that stimulated impact reduction in running. The hypothesis was that motor learning would occur, which was only partially supported. First, we addressed learning from the perspective of retention. We found no significant change in impact magnitudes after six months. Contrary to our expectation, the reductions in PTA_a and vertical loading rate following completion of the retraining program did not persist in this small cohort. Our original research data challenge the idea that just three weeks of gait retraining can permanently change people’s running form, and countermand a recent systematic review concluding that real-time tibial acceleration feedback can reduce PTA_a for periods to 12 months when the feedback has been removed (Li et al., 2023). We will discuss several potential reasons for this discrepancy later. Second, from the perspective of recall and immediately after the retention condition we asked the experimental group asked to reproduce the running form from the last retraining session. They did achieve a significant reduction in impact, which suggests that the changes needed for impact reduction are recallable long time after completing a 3-week intervention. The reduction in impact at six months (mean difference in PTA_a: −2.43 g; vertical loading rate: −30.5 B/W) aligns well with the impact reduction observed post intervention (mean difference in PTA: −2.55 g; vertical loading rate: −31.5 B/W) in the same testing environment (Derie et al., 2022). Although the new motor pattern was not fully ingrained yet, it is encouraging that participants were able to recall it with just a simple reminder. This suggests that verbal reminders given by a practitioner can help runners remember the self-discovered gait modifications a long time after finishing the program. The changes in impact that we observed in the follow-up condition were likely because people consciously tried to modify their running form, not because a new skill became persistent. Looking at the group who did not receive the biofeedback, the impact magnitudes remained stable after half a year. This finding supports a previous observation that impact magnitudes of a group of runners during running on level ground are reliable when tested again later (Van den Berghe et al., 2019).

Our results contradict the results reported by Bowser et al. (2018) who suggested that impact reduction can be achieved and maintained up to one year following running gait retraining. A few possible explanations for this discrepancy are given. A possible explanation lies in the difference in training and testing environments. Previous studies conducted the intervention and retention testing in the same environment, resulting in participants spending a considerable amount of time in an artificial lab setting during the retraining. As a result, when these participants reappeared in the same lab for the retention testing, there is a higher chance of instinctive recall and response bias. In our study, the retraining program and learning evaluation were conducted at different locations, meaning the participants performed the learning tests in an environment different from their practice sessions. Furthermore, a cognitive distraction task (i.e., modified Stroop test) was added to our retention testing to limit the possibility of performance bias during the data collection (Futrell et al., 2020). It is unlikely that this task has influenced the impact magnitudes because previous research has detected impact reduction during distracted running immediately post-retraining (Cheung et al., 2018). Another possible explanation is that participants did not have enough practice time to achieve clear motor learning. The statistically non-significant reduction in impact magnitudes at retention suggests that participants may need more sessions than anticipated. Our study design was largely based on that of Clansey et al. (2014) wherein participants receive only six practice sessions, while studies reporting impact reduction at follow-up involved eight practice sessions (Bowser et al., 2018; Crowell & Davis, 2011). Based on our observations, we recommend tailoring the number of sessions in the retraining program to each individual’s needs. It is plausible that participants entered the transfer phase too soon and could have benefitted further from more continuous feedback as provided in the acquisition phase. So, we suggest to start fading the feedback only if the participant has shown significant improvement for a while. This reduction can be qualitatively analysed by a supervisor closely monitoring the participant or quantitatively analysed through a change-point analysis (Van den Berghe et al., 2020). Additionally, occasional refresher training or recall sessions may be beneficial for achieving motor learning eventually.

These results have relevance to the users, coaching staff and sports scientists currently interested in the effects of wearable devices that stimulate less PTA_a in running at a comfortable running speed. Real-time biofeedback has been found to be the most effective way of reducing impact, and runners tend to employ a distal strategy unless given specific cues (Napier et al., 2015). The distal strategy for reducing PTA_a has been found in this sample of habitual rearfoot striking runners who used the music-based biofeedback device (Derie et al., 2022), with plantar pressure analysis showing it spans the entire foot strike spectrum. More specifically, eight out of ten kept landing on their heel region, one switched to midfoot striking and another one to forefoot striking following the running retraining intervention (Derie et al., 2022). The vast majority kept rearfoot striking and increased the foot strike angle by + 5° (i.e., more dorsiflexion at touchdown) on average. This finding is in agreement with our previous cross-sectional studies (Van den Berghe, Breine, et al., 2022; Van den Berghe, Warlop, et al., 2022). Namely, at a comfortable steady-state running speed on level ground, heel-toe runners who experience relatively low PTA_a during level running are more likely to perform an obvious rearfoot strike. Although the major movement changes due to the biofeedback have been studied, participant characteristics potentially influencing how easily humans motorically adapt (e.g., motor coordination, training history) were not included here. Crowell and Davis (2011) disclosed that all participants in their running retraining program reported that the new way of running felt natural after six practice sessions. Although participants had more practice time during the retraining sessions of the current study, it is plausible that the changes in running style for reducing impact did not feel natural after our six-session program. Furthermore, it has been shown that endurance training volume can affect running biomechanics (Boyer et al., 2014; Verheul et al., 2017), and we acknowledge that the cohort examined in the present study demonstrated variation in weekly training volume and presumably also in experience of endurance training prior to enrolment. The running mileage following completion of the retraining intervention was also not tracked, which is another variable that may influence the retention. In future research, it would be valuable to ask participants how they perceive the modified gait pattern (e.g., level of difficulty, confidence, etc.) and to track their training profile because it may help explaining why people chose not to use it on a typical run but can recall it.

A clear relationship between PTA_a and injury risk in running is yet to be established (Sheerin et al., 2019). The use of these kinds of biofeedback devices as a potential tool for injury risk management stands or falls with their effect on the running-related injury risk eventually. For this reason, efficacy trials are needed to determine whether an intervention (e.g., impact reduction) produces the expected result (e.g., an altered injury risk) under ideal circumstances. Randomized controlled clinical trials would be helpful to determine if reducing PTA_a alters the injury risk in very large groups of runners or not, and if the large-scale intervention would result in a persistent change in running mechanics. Some might wonder whether reducing impacts with large effect sizes leads to changes in general indicators of musculoskeletal loading. To address this, our research unit has conducted comprehensive gait analysis on the experimentals group of heel-toe runners that was beyond the scope of the present study. In this group of small sample size, we observed no statistically significant change in the maximum vertical ground reaction force, the peak moment and power of the ankle and knee joints, or the muscular work done at these joints (Derie et al., 2022). Based on the data at hand, practitioners should not assume that a change in the commonly used impact metric PTA_a corresponds to a similar change in the aforementioned indicators of musculoskeletal loading. Wearable sensor systems that interact with runners to reduce PTA_a may still be useful in incorporating variation into a training regime of distance runners.

The current study is limited by its small sample size with missing data. Simulating experiments in small samples have shown that the results can be inconsistent compared to the original large samples, leading to both false-positive and false-negative outcomes (Oakes, 2017). The issue of incomplete data is common in longitudinal study designs. If we applied mean imputation, the interpretation of the results would be different. In such case, we could have detected a retention effect for the feedback variable. However, using single imputation often underestimates variability because unobserved values are given equal weight to observed values in the analysis (Jakobsen et al., 2017). The decrease in variability was noticeable in the small experimental group, indicating the need to increase the sample size in future research post-COVID pandemic. Another limitation is that the intervention took place in a training environment (Van den Berghe, Derie et al., 2022), but the follow-up testing was conducted in a laboratory setting. Unsupervised follow-up testing in the field with wearables monitoring larger samples is warranted to validate the results of these gait-retraining interventions.

Conclusion

The biofeedback intervention did not induce a clear learning effect after six months. Interestingly, a substantial impact reduction was recallable at the six-month follow up through simple verbal cueing. Participants could recall and re-enact the impact reduction in the absence of feedback on PTA_a. In the absence of additional feedback sessions between the end of the intervention and follow-up, a verbal recall by a supervisor or coach appears effective to evoke the impact reduction achieved during the intervention.

Confirmation of ethical compliance

All applicable international, national, and/or institutional guidelines to study humans were followed. The study was approved by the UZ Gent ethics committee.

Patient involvement statement

Study participants were not involved in the design, conduct, interpretation, or translation of the current research.

Data sharing statement

Data are available upon reasonable request.

Acknowledgments

Mizuno Corporation provided product support. We would like to acknowledge Prof. Dr. Marc Leman and IPEM for the assistance with the feedback device, and ing. Davy Spiessens for his assistance.

Disclosure statement

The authors declare that the research was conducted in the absence of any financial relationships that could be construed as a potential conflict of interest. One of the authors declared that he was an editorial member of the journal at the time of submission. This had no impact on the peer review process and the final decision.

Supplementary material

Supplemental data for this article can be accessed online https://doi.org/10.1080/02640414.2024.2323849

Additional information

Funding

This study was funded by Interreg EU [Nano4Sports-0217] and the Research Foundation–Flanders [FWO.3F0.2015.0048.01]. Mizuno Corporation provided product support. The manuscript was published with support of the University Foundation from Belgium (‘Uitgegeven met steun van de Universitaire Stichting van België’, WA-0473 to P.V.d.B).