Effects of running-induced fatigue on joint kinematics and kinetics during overground running: a systematic review and meta-analysis

ABSTRACT

This study evaluated the acute fatigue-inducing effect of distance running on kinematics and kinetics during overground running. Standardised mean differences (SMD) with 95% confidence intervals (95% CI) were used to pool data across 16 studies. Effects during consistent (pre- and post-fatigue running speed within ± 5%) versus varied speed running (difference of >5% between running speeds) were analysed separately. There was strong evidence that running-induced fatigue significantly increases ground contact times at consistent running speeds (SMD 0.52 [95% CI 0.22, 0.82]) and moderate evidence that step length shortens at varied running speeds (SMD −1.27 [95% CI −1.79, −0.75]). There was strong evidence that fatigue does not change peak: hip and knee flexion angles, hip adduction angle, hip and knee internal rotation angles, hip and knee extension moments, hip and knee abduction moments, knee abduction angle, knee flexion and extension moments, knee adduction moment, rearfoot eversion angle, and plantarflexion moments, or knee flexion and plantarflexion range of motion during stance. Running-induced fatigue increases contact times and reduces step length, whereas lower-body joint angles and moments are unchanged. Minimising changes in stride parameters could provide a mechanism for reducing the effects of fatigue on running performance.

KEYWORDS:

• Biomechanics
• gait analysis
• distance running
• performance
• injury

Introduction

Running-related injuries (RRIs) are prevalent, with studies reporting anywhere between 19.4% and 79.3% of long distance runners sustaining an RRI each year (Van Gent et al., 2007). Common RRIs include Achilles tendinopathy (10.3% of RRI incidence), medial tibial stress syndrome (9.4%), patellofemoral pain syndrome (6.3%), and plantar fasciitis (6.1%) (Kakouris et al., 2021). Risk factors associated with the development of RRIs include previous injury (Saragiotto et al., 2014; Van Gent et al., 2007) and lower-extremity joint kinematics and kinetics during running (Willwacher et al., 2022). Atypical running biomechanics can induce altered or abnormal tissue loads that accumulate beyond the material capabilities of the tissue, causing damage and degradation to these structures (Hoenig et al., 2022). Fatigue can cause changes in biomechanics in as little as 10 min of continuous running at a high intensity (Rabita et al., 2011). Identifying whether fatigue induced by continuous running causes changes in running biomechanics associated with RRIs could be an important step in understanding mechanisms underpinning RRIs.

Previous researchers (Apte et al., 2021; Darch et al., 2022; Zandbergen et al., 2023) have summarised the effects of running-induced fatigue on biomechanics, with contrasting findings. Apte et al. (Apte et al., 2021) summarised the effects of running-induced fatigue on biomechanics during treadmill and overground running. The authors did not perform a meta-analysis of the included studies; instead, they reported trends on the direction of change in biomechanical outcomes that emerged across studies. A limitation of this approach is that it does not provide pooled estimates of effect size to indicate the magnitude or significance of any changes. The authors reported a decreasing trend for stride length, peak vertical ground reaction forces (GRFs), leg stiffness, peak knee flexion angle during swing and plantarflexion angle at contact, as well as an increasing trend for contact times during overground running. Contradicting trends were identified for various outcomes (e.g., stride length) depending on whether they were measured during treadmill or overground running. This supports the need to analyse the effects of running-induced fatigue on spatiotemporal parameters and lower-body joint angles and moments collected during treadmill and overground running separately (Sinclair et al., 2013; Van Hooren et al., 2020).

Darch and colleagues (Darch et al., 2022) performed a meta-analysis to synthesise the fatiguing effect of continuous running (i.e., excluding repeated sprint protocols) on impact loads during running, irrespective of whether biomechanical data were collected during treadmill or overground running. The authors analysed 25 studies reporting outcomes related to GRFs, tibial acceleration, and outcomes derived from external loads (e.g., limb or body stiffness and shock attenuation). Leg stiffness, which describes the ratio of the vertical ground reaction force to leg-spring compression during mid-stance (Girard et al., 2013,) was the only outcome to reduce after running-induced fatigue. Finally, Zandbergen et al. (Zandbergen et al., 2023) conducted the most recent meta-analysis on the fatigue-inducing effect of long distance running (defined as >3 km) on sagittal plane kinematics and vertical and leg stiffness during running, combining overground and treadmill running in their analyses. The authors analysed 33 articles and 19 biomechanical outcomes and found that runners flexed their knee more at initial contact and during swing, and decreased leg stiffness, in response to running-induced fatigue. The authors did not perform a subgroup analysis to compare effects during treadmill versus overground running because only six of the included studies analysed overground running biomechanics. The authors acknowledged, however, that peak knee flexion angle during stance tended to decrease or show no change in overground measurements, whereas it tended to increase or show no change in treadmill measurements. Additionally, the hip angle at midstance decreased in overground measurements, whereas it remained the same or increased in treadmill measurements. In summary, examining the pooled effect of running-induced fatigue on biomechanics collected during treadmill and overground running studies does not accurately reflect changes in overground running biomechanics with fatigue. Further, it remains unclear how fatigue affects lower-body joint moments and frontal and transverse plane joint kinematics.

Understanding fatigue-induced biomechanical changes in the frontal and transverse planes may be important given that joint mechanics in these anatomical planes have been prospectively linked with the development of RRIs (Becker et al., 2018; Dudley et al., 2017; Noehren et al., 2013; Stefanyshyn et al., 2006). For example, distance runners who developed medial tibial stress syndrome during a two-year follow-up period demonstrated significantly greater contralateral pelvic drop (ES = 1.06), and greater peak amounts (ES = 1.42) and durations (ES = 2.52) of rearfoot eversion during stance compared with the uninjured group at baseline (Becker et al., 2018). Distance runners have been found to increase peak rearfoot eversion (ES = 2.37) in as little as 20 min of running at the speed corresponding with lactate threshold (Clansey et al., 2012). Thus, fatigue induced by distance running could increase biomechanical risk factors associated with the development of medial tibial stress syndrome. Runners who developed iliotibial band syndrome during a two-year follow-up study demonstrated significantly greater peak hip adduction, knee internal rotation, and rearfoot eversion during stance compared with healthy controls at baseline (Noehren et al., 2007). Altered frontal and transverse plane kinematics are believed to increase strain and torsional loads on the tissues of the knee joint including the iliotibial band (Hamill et al., 2008). Increased peak external knee adduction moment (Dudley et al., 2017) and peak knee abduction impulse (Stefanyshyn et al., 2006) have also been prospectively linked with the development of various RRIs, suggesting that abnormal lower-limb joint moments could contribute to RRIs. Additionally, there is also a positive relationship between lower-body sagittal-plane joint torques and running speed (Dorn et al., 2012; Weyand et al., 2000). Given the role of frontal and transverse plane kinematics and lower-body joint moments in the aetiology of common RRIs, and the relationship between lower-body joint kinetics and running speed, understanding whether fatigue is associated with changes in these biomechanical parameters could have important implications for both injury and performance.

Therefore, the purpose of this study was to (1) systematically evaluate the acute fatigue-inducing effect of distance running on GRFs, spatiotemporal parameters, and trunk, pelvis, and lower-body joint angles and moments (in all three anatomical planes) during overground running; and (2) where appropriate, conduct meta-analyses to obtain estimates of pooled effect size to help resolve uncertainty surrounding these outcomes. We hypothesise that acute running-induced fatigue will significantly affect spatiotemporal parameters, increase peak knee and dorsiflexion angles, and decrease peak lower limb joint moments, during overground running.

Materials and methods

This systematic review and meta-analysis was designed using a prospective online protocol, registered at Center for Open Science (DOI: https://doi.org/10.17605/OSF.IO/5CJRB) and reported according to the Preferred Reporting Items for Systematic review and Meta-Analysis Protocols (PRISMA-P) (Moher et al., 2015).

Eligibility criteria

Studies were eligible for inclusion in this review if they satisfied the following criteria: (i) population: participants with running experience; (ii) intervention: acute fatigue induced by a continuous prolonged run (≥90 min) or a continuous exhaustive run (e.g., time trial, competitive race) or a continuous running protocol of ‘maximal’ intensity (refer to Table 1) (Norton et al., 2010) performed overground or on an electric treadmill; (iii) comparator: baseline (start of the test or non-fatigued state); (iv) outcomes: discrete GRFs, spatiotemporal parameters, and trunk, pelvis, and lower-body joint angles and moments (in all three anatomical planes) during level overground running; and, (v) study design: repeated-measures studies.

Table 1. Categorisation of exercise intensity (Norton et al., 2010).

Intensity	% VO2max	% HRmax	RPE (C)	RPE(C-R)
Rest-Light	<40	<55	≤10	≤2
Moderate	40 < 60	55 < 70	11–13	3–4
Severe	60 < 85	70 < 90	14–16	5–6
Maximal	≥85	≥90	≥17	≥7

VO₂max maximal aerobic capacity; HR_max maximum heart rate; RPE (C) Borg rating of perceived exertion scale (6–20); RPE (C-R) Borg rating of perceived exertion category ratio scale (0–10).

Studies were excluded from this review if any of the following exclusion criteria were met: (i) population: non-human studies or studies where an injury or medical condition could affect participants’ ability to run; (ii) intervention: studies where the effects of running-induced fatigue could not be isolated, studies where the fatiguing protocol involved sprinting or was <3 km duration (Zandbergen et al., 2023), or studies where the fatiguing protocol was predominantly incline running; (iii) comparator: studies that only reported change scores, or studies where the baseline was measured at a timepoint > 10% into the fatiguing protocol; and, (v) outcomes: studies where the outcomes were measured during sprinting or a non-fatigued state (such as following a recovery period). Studies involving fatiguing protocols that were predominantly incline running (e.g., trail running competitions) were excluded because grade-specific biomechanical differences, neuromuscular adaptations, and physiological responses occur during uphill and downhill running (Vernillo et al., 2017).

Search strategy

The databases MEDLINE, MEDLINE Complete, CINAHL Complete, SPORTDiscus, Embase, and Scopus were first searched on the 8^th of March 2022 with a second search occurring on the 18^th of August 2023. Within each database, journal article titles, abstracts and keywords were searched using the following terms and Boolean operators: (run* OR ran) AND (fatigue* OR exhaust* OR exert* OR prolong*) AND (biomechanic* OR mechanic* OR kinematic* OR kinetic* OR spatiotemporal* OR ground reaction force*). This search strategy was adapted for individual databases. No publication date limit was set. Full-text journal articles reporting in non-English and conference proceedings were excluded during each database search. No additional filters or search limitations were used.

Study selection

All titles and abstracts initially identified through the database searches were imported into EndNote X9.2 (Clarivate Analytics, Philadelphia, PA, USA) then batch imported into Covidence (Veritas Health Innovation, Melbourne, Victoria, Australia) where duplicates were automatically removed.

Each title and abstract were independently evaluated for inclusion against the eligibility criteria by two members of the review team (DT & CKD). Titles and abstracts were retained for full text review if there was insufficient information in the title or abstract to determine exclusion. Any discrepancies between the two reviewers were resolved during a consensus meeting with a third reviewer (JB). Full texts were retrieved for the retained studies, then full texts were independently assessed for inclusion by two members of the review team (DT & CKD) against the eligibility criteria. Any discrepancies between the two reviewers were resolved during a consensus meeting with a third reviewer (JB). The reference lists of the articles obtained were also searched manually to identify additional studies not identified electronically. If a reference list identified any potentially relevant studies, then the study’s abstract was screened before obtaining the full-text version if appropriate.

Data items

Participant data available for extraction from each study included the sample size, age, sex, body mass, height, running training status (e.g., maximal aerobic capacity [VO₂max], hours per week, kilometres per week), and habitual foot strike pattern. When studies compared healthy versus injured participants, only data pertaining to healthy participants were included in the meta-analyses. If studies analysed two groups of runners separately (e.g., low versus high calibre, male versus female, rearfoot versus forefoot strikers) then data for both groups were independently included in the meta-analyses. If studies analysed one group of runners under two separate conditions (e.g., dominant versus non-dominant limb, specialised shoes versus neutral shoes, self-paced versus restricted pacing) then data pertaining to only one condition were included in the meta-analyses and this is noted in Table 2 (e.g., dominant limb only, neutral shoe condition only, self-paced run only). Intervention data extracted from each study included the type of fatigue protocol (e.g., run to exhaustion, race, time trial), running surface (i.e., overground or treadmill), running duration, running speed, total distance, and measures of intensity (e.g., rating of perceived exertion (RPE), heart rate). Comparison data were extracted at baseline (start of the test or a non-fatigued state ≤10% into the fatiguing protocol) and post-fatigue (end of the test or immediately following the fatiguing protocol). Any time delays between the fatiguing run (i.e., a treadmill run) and measurement of overground running biomechanics are reported when possible. When outcomes were measured at multiple times then the initial and final timepoints were included in the meta-analyses only. Outcome data extracted from each study included discrete GRFs, spatiotemporal parameters, and trunk, pelvis, and lower-body joint angles and moments (in all three anatomical planes) that were measured during level overground running at baseline and post-fatigue. All joint measures are reported as internal joint angles or moments. The effects of running-induced fatigue on biomechanics during consistent versus varied speed running (between baseline and follow-up) are analysed separately. In the context of this review, a ‘consistent running speed’ occurred when the follow-up running speeds were within ± 5% of baseline running speeds. Running speed could remain steady (i.e., unchanging) throughout the fatiguing protocol or running speed could change throughout the fatiguing protocol provided that the follow-up testing speed was within ± 5% of the baseline testing speed. Studies that did not meet these criteria were considered to analyse ‘varied running speeds’ because their baseline and follow-up testing speeds had a greater than ± 5% difference.

Table 2. Study characteristics.

Study	Study Population	Fatigue Protocol	Running Analysis	Outcome Measures
Bazett-Jones et al. (2013)	10 M/9F (Control group) 24.3 + 4.3 yrs 1.74 + 0.09 m 70.2 + 11.3 kg Physically active (10.0 + 7.56 hr/wk)	Treadmill run to exhaustion at a self-selected pace 9.5 + 1.8 km/hr 36.3 + 22.7 min 16.6 + 1.3 RPE 86.1 + 10.7% HRmax	10 m overground runway 4.0 + 0.5 m/s Delay: ‘Immediately’ following fatigue protocol	3D analysis of trunk, pelvis, hip, and knee joint angles and moments during stance
Brown et al. (2014)	20F RFS 29 ± 6 yrs 1.60 ± 0.10 m 56.8 ± 5.2 kg Ran ≥15 miles/wk	Treadmill run to exhaustion at a self-selected pace. 24.6 ± 7.0 min ≥ 17 RPE	25 m overground runway 3.35 m/s ± 10% Delay: 7 ± 0.002 min	3D analysis of hip, knee, and ankle joint angles and moments during stance Dominant limb only
Brown et al. (2016)	20F (Control group) RFS 28.9 ± 6.1 yrs 1.60 ± 0.09 m 56.8 ± 5.2 kg Ran ≥15 miles/wk	Treadmill run to exhaustion at a self-selected pace. 24.7 ± 7.0 min 3.2 ± 0.29 m/s ≥ 17 RPE	30 m overground runway 3.6 ± 0.22 m/s Delay: 7 ± 0.002 min	Peak hip adduction and internal rotation angles during stance Peak hip abductor moment during stance Peak hip external rotation moment during swing
Elliot et al. (1981)	8 M 19–28 years National level runners	10 km overground track race at Australian National Championships 29:19.53–31:24.30 min	400 m overground track 780 m: 5.52 ± 0.15 m/s 9580 m: 5.04 ± 0.32 m/s	Spatiotemporal parameters 2D analysis of trunk angles
Girard et al. (2013)	12 M 21.4 ± 2.6 yrs 1.81 ± 0.06 m 70.2 ± 6.4 kg VO2max: 62.9 ± 8.3 mL/kg/min vVO2max: 19.0 ± 1.2 km/h	5 km overground time trial 17:31 min (15:58–20:42 min)	200 m overground track 200 m: 5.19 ± 0.54 m/s 4800 m: 4.89 ± 0.17 m/s	Spatiotemporal parameters Vertical oscillation GRFs (C)
Hamzavi et al. (2021)	15–20 km/wk 14 M/16F RFS 21.56 ± 2.28 yrs 1.67 ± 0.08 m 61.43 ± 11.57 kg 15 M/15F FFS 19.73 ± 1.68 yrs 1.71 ± 0.08 m 65.7 ± 13.45 kg	Incremental treadmill run to exhaustion ≥ 17 RPE or ≥ 80% HRmax	16 m overground runway Participants ran barefoot 3.3 m/s ± 5% Delay: 2 min cool down at self-selected pace	Contact time
A. Jafarnezhadgero et al. (2019)	26F RFS with foot pronation 24.1 ± 5.6 yrs 1.66 ± 0.10 m 64.2 ± 12.1 kg	Incremental treadmill run to exhaustion. 18.1 ± 0.8 RPE 177.6 ± 7.4 bpm	18 m overground runway Neutral unfatigued: 3.29 ± 0.03 m/s Neutral fatigued: 3.29 ± 0.04 m/s Delay: ‘Immediately’ following fatigue protocol	3D analysis of hip, knee, and ankle joint angles and moments during stance and swing Neutral shoe condition only
A. A. Jafarnezhadgero et al. (2019)	22F 24.1 ± 5.6 yrs 1.66 ± 0.10 m 63.21 ± 12.05 kg	Incremental treadmill run to exhaustion. ≥ 17 RPE or ≥ 80% HRmax 177 ± 7 bpm	18 m overground runway Neutral unfatigued: 3.28 ± 0.04 m/s Neutral fatigued: 3.27 ± 0.04 m/s Delay: ‘Immediately’ following fatigue protocol	Vertical loading rate*, peak vertical GRFs *, time to peak force*, knee adduction moment*, and peak rearfoot eversion during stance Neutral shoe condition only
Matta et al. (2020)	12 M/4F 38.6 ± 11.3 yrs 1.74 ± 0.07 m 71.5 ± 12.2 km/wk	6 hr self-paced ultramarathon 58.9 ± 9.4 km > 17 RPE	400 m overground track Lap 1: 3.76 ± 0.43 m/s 6 hr: 3.09 ± 0.37 m/s Self-paced run only	2D analysis of spatiotemporal parameters (C)
Melaro et al. (2021)	13 M RFS 32.1 ± 9.7 yrs 1.80 ± 0.01 m 73.6 ± 11.9 kg 72.6 ± 21.2 km/wk	Long run on a treadmill 18.2 ± 5.3 km 90.6 ± 21.5 min	25 m overground runway Preferred long run speed ± 5% Delay: 2–3 min	Hip, knee, and ankle extension excursion during stance Peak hip extension, knee flexion, and plantarflexion moments during stance
Paavolainen et al. (1999)	19 M 23.8 ± 3.6 yrs 1.79 ± 0.06 m 68.5 ± 6.2 kg 9.1 ± 3.2 yrs training	10 km overground time trial 37.8 ± 2.2 min	200 m overground track Constant velocity lap (4.5 m/s) at 200 and 9800 m	Contact times (ND) GRFs (ND)
Paquette et al. (2017)	12 M RFS 33 ± 10 yrs 1.80 ± 0.08 m 74.4 ± 12.0 kg 76.6 ± 21.0 km/wk	Long run on a treadmill 19.1 ± 5.3 km 92.8 ± 21.5 min	25 m overground runway Preferred long run speed ± 5% 3.4 ± 0.4 m/s Delay: ‘Immediately’ following fatigue protocol	3D analysis of hip, knee, and ankle joint angles and moments during stance Vertical loading rate
Quan et al. (2021)	50F RFS 23.89 ± 1.27 yrs 1.63 ± 0.15 m 65.39 ± 22.47 kg 5–10 km/wk	Treadmill run to fatigue at 14.4 km/h > 17 RPE or ≥ 80% HRmax	15 m overground runway 3.3 ± 5% m/s Delay: NR	3D analysis of hip, knee, and ankle joint angles and moments during stance (C)
Rabita et al. (2011)	6 M/3F 23.2 ± 3.2 yrs 1.82 ± 0.02 m 68.3 ± 5.4 kg VO2max 71.5 ± 6.5 mL/kg/min vVO2max 5.1 ± 0.3 m/s	Overground run to exhaustion at 95% of vVO2max 5.1 ± 0.3 m/s 10.7 ± 2.6 min TTE 3258 ± 819 m	200 m overground track at 95% of vVO2max Measured at 10% and 100% of the run.	Spatiotemporal parameters Knee flexion angle at initial contact GRFs
Schena et al. (2014)	13 M/5F 40.7 ± 4.9 yrs 1.73 ± 0.08 m 66.0 ± 9.4 kg VO2max 52.9 ± 6.0 mL/kg/min	60 km overground simulation at 100 km race pace 285 ± 26.5 min 3.53 ± 0.33 m/s	400 m overground running track 0 km: 3.69 ± 0.33 60 km: 3.52 ± 0.37 Delay: 40 secs	Spatiotemporal parameters
Tam et al. (2017)	12 M well-trained 27.9 ± 3.1 yrs 1.80 ± 0.68 m 73.8 ± 10.9 kg 10 km PB: 42.0 ± 2.9 min 10 M Trained 25.1 ± 4.9 yrs 1.80 ± 0.68 m 78.5 ± 8.7 kg 10 km PB: 51.2 ± 5.0 min	10 km treadmill run Well-trained: 19.0 ± 1.1 RPE Trained: 19.5 ± 0.6 RPE	40 m overground running track at 10 km race pace Well-trained: 4.4 m/s Trained: 3.5 m/s Delay: ≤ 2 min	3D analysis of knee and ankle joint angles and moments during stance GRFs (ND)
Winter et al. (2021)	23 M/14F (Healthy; C) 44.78 ± 12.46 yrs 1.71 ± 0.09 m 65.35 ± 10.55 kg 52.89 ± 18.84 km/wk	8 km overground time trial Elite: 29.23–32.24 min Advanced: 32.27–35.47 min Intermediate: 35.55–39.35 min Slow: 40.21–48.40 min	400 m overground track Baseline (730 m) Elite: 4.32 ± 0.19 m/s Advanced: 3.88 ± 0.25 m/s Intermediate: 3.54 ± 0.16 m/s Slow: 3.13 ± 0.23 m/s Follow up (7130 m) Elite: 4.34 ± 0.22 m/s Advanced: 3.88 ± 0.14 m/s Intermediate: 3.54 ± 0.12 m/s Slow: 3.02 ± 0.20 m/s	Spatiotemporal parameters (C)

M male; F female; 3D three-dimensional; RPE rate of perceived exertion; HRmax maximum heart rate; RFS rearfoot strike; GRFs ground reaction forces; C data obtained from correspondence with authors; FFS forefoot strike; bpm beats per min; * Data digitised from a figure; 2D two-dimensional; ND no data available for inclusion; NR not reported; VO₂max maximum oxygen uptake; vVO₂max velocity at VO₂max; PB personal best run t.

Data extraction

Data were extracted from all included studies independently by one reviewer (DT). Data extraction was repeated by the reviewer after three months to check for transcribing errors. For each outcome, point estimates and measures of variability (standard deviation, standard error, confidence intervals) were extracted from studies at baseline and following the fatiguing run. Change scores were not accepted. Where point estimates and measures of variability were unavailable in text or table format, the corresponding author of the study was contacted via email addresses on publications or by searching online for current contact details. Follow-up emails were sent two weeks later following the original correspondence. If the corresponding author did not respond within one month of the follow-up email or they were unable to provide the point estimates and measures of variability then, where available, data were extracted from graphs using Web Plot Digitizer 4.6 (Ankit Rohatgi, Pacifica, CA, USA). If point estimates and measures of variability were unavailable in text, tables, and graphs then the study was excluded from the analyses. Table 2 records these data as ‘correspondence’ (C) for unpublished data obtained via contact with authors, ‘digitised’ (D) for data extrapolated from graphs, and ‘no data’ (ND) following unsuccessful attempts to contact authors.

Study risk of bias assessment

Methodological risk of bias was assessed using a modified version of the Downs and Black Quality Index as per the Cochrane non-randomised studies method group recommendations (Reeves et al., 2022). The original scale comprises of four subscales (reporting, external validity, internal validity: bias, internal validity: confounding) and 26 questions with scoring criteria described for each question. The maximum score on the original 26 question scale is 32 points. The original appraisal tool has high internal consistency (KR-20: 0.89), test–retest reliability (r = 0.88), and inter-rater reliability (r = 0.75) (Downs & Black, 1998). The modified version of the Downs and Black Quality Index consists of 15 questions from the original four subscales. The maximum score on the modified 15 question scale is 16 points. Two independent reviewers evaluated the quality of the included studies (DT & CKD) against the modified Downs and Black Quality Index. Outcomes were discussed in a team meeting and discrepancies were resolved by consulting a third reviewer (JB). Studies scoring ≤5 were rated as low-quality, studies with scores of 6-11 were rated as medium-quality, and studies with scores of ≥12 were rated as high-quality.

Effect measures

Standardised mean differences (SMD) with 95% confidence intervals (95% CI) were calculated for studies and, where possible, meta-analyses were performed by pooling the results on each outcome using Review Manager 5.4.1 (The Cochrane Collaboration, Copenhagen, Denmark). The SMD used was the effect size known as Hedges’ (adjusted) g. Hedge’s g is similar to Cohen’s d but includes an adjustment for small sample size. Meta-analyses were performed using a random-effects model (inverse variance method) to calculate the pooled and weighted mean SMD (SMD_p) and 95% CI between baseline and follow-up for each outcome. Statistical homogeneity was established using the Higgins I² statistic (Higgins et al., 2003). As proposed by Higgins, values of <25%, 25–50% and >50% were used to indicate small, medium, and large heterogeneity among studies, respectively. Sensitivity analyses were performed when it seemed possible that the same participants could have been included in multiple studies for an outcome (e.g., Brown et al., 2014, 2016). Effect sizes were interpreted according to Hopkins’s scale (Hopkins, 2000), where effect size values <0.20 indicate trivial, 0.20–0.59 indicate small, 0.60–1.19 indicate moderate, and values ≥1.20 indicate large effects.

Strength of evidence

The strength of evidence for each outcome was categorised based on the quantity and Downs and Black quality rating of the included studies using a scale developed by Neal et al (Neal et al., 2016). and adapted from Van Tulder et al (Van Tulder et al., 2003).

Strong evidence

Pooled results from three or more studies, including at least two high-quality studies that are statistically homogeneous.

Moderate evidence

Pooled results from multiple studies that are statistically heterogeneous, including at least one high-quality study; or pooled results from multiple medium-quality or low-quality studies that are statistically homogeneous.

Limited evidence

Results from one high-quality study; or pooled results from multiple medium- or low-quality studies that are statistically heterogeneous.

Very limited evidence

Results from one medium-quality study or one low-quality study.

No evidence

Non-significant pooled results from multiple studies, regardless of quality, that are statistically heterogeneous.

Results

Search strategy

The first database search identified a total of 5558 potentially relevant journal articles (Figure 1) and the second database search identified an additional 672 articles. A total of 3666 duplicate references were removed. Screening of titles and abstracts for inclusion and exclusion criteria resulted in 90 studies. The full-text versions for these studies were screened and an additional 73 studies were removed resulting in 17 studies being included in this systematic review and 16 studies included in the meta-analyses (due to insufficient reporting of data for inclusion in one study) (Paavolainen et al., 1999).

Figure 1. Study selection flow chart.

Data extraction

The 17 studies included in this review were published between the years 1981–2021, and study characteristics are provided in Table 2. A total of 383 participants completed follow-up testing and the average sample size was 22.5 ± 14.1 (range: 8–60) participants. One hundred and seventy-nine men (46.7%) and 204 women (53.3%) participated in these studies. There were four studies (Bazett-Jones et al., 2013; Quan et al., 2021; Tam et al., 2017; Winter et al., 2021) that included novice runners, ten studies (Brown et al., 2014, 2016; Girard et al., 2013; Hamzavi & Esmaeili, 2021; A. Jafarnezhadgero et al., 2019; A. A. Jafarnezhadgero et al., 2019; Paavolainen et al., 1999; Schena et al., 2014; Tam et al., 2017; Winter et al., 2021) included recreational runners, and five studies (Elliot & Ackland, 1981; Matta et al., 2020; Melaro et al., 2021; Paquette & Melcher, 2017; Rabita et al., 2011) included experienced runners. There were 2 studies (Tam et al., 2017; Winter et al., 2021) that analysed runners of different abilities separately.

For the fatiguing protocol, seven studies (Bazett-Jones et al., 2013; Brown et al., 2014, 2016; Hamzavi & Esmaeili, 2021; A. Jafarnezhadgero et al., 2019; A. A. Jafarnezhadgero et al., 2019; Quan et al., 2021) implemented a treadmill run to fatigue, 1 study (Tam et al., 2017) implemented a 10 km time trial on a treadmill, and two studies (Melaro et al., 2021; Paquette & Melcher, 2017) implemented a long run on a treadmill. One study (Schena et al., 2014) implemented a 60 km overground long run. Other studies implemented overground time trials of 5 km (Girard et al., 2013), 8 km (Winter et al., 2021), 10 km (Paavolainen et al., 1999), and a maximum distance trial of 6 hours (Matta et al., 2020). One study (Rabita et al., 2011) implemented an overground run to exhaustion at 95% of the velocity at VO₂max (vVO₂max). One study (Elliot & Ackland, 1981) analysed data from the 10,000 m final at the Australian Track and Field Championships (1979).

Overground running biomechanics were analysed at baseline and follow-up at a variety of consistent running speeds: 3.3 m/s (Hamzavi & Esmaeili, 2021; A. Jafarnezhadgero et al., 2019; A. A. Jafarnezhadgero et al., 2019; Quan et al., 2021), 3.4 m/s (Melaro et al., 2021; Paquette & Melcher, 2017), 3.5 m/s (Tam et al., 2017), 3.6 m/s (Brown et al., 2014, 2016), 4.0 m/s (Bazett-Jones et al., 2013), 4.4 m/s (Tam et al., 2017), 4.5 m/s (Paavolainen et al., 1999), and at 95% vVO₂max (5.1 ± 0.3 m/s) (Rabita et al., 2011). There were 3 studies (Matta et al., 2020; Schena et al., 2014; Winter et al., 2021) that analysed overground running biomechanics at varied running speeds (i.e., the researchers did not control baseline or follow-up running speeds).

There were 6 studies (Elliot & Ackland, 1981; Girard et al., 2013; Matta et al., 2020; Rabita et al., 2011; Schena et al., 2014; Winter et al., 2021) that analysed running biomechanics during the fatiguing run. There were 4 studies (Hamzavi & Esmaeili, 2021; Melaro et al., 2021; Schena et al., 2014; Tam et al., 2017) that reported a delay of <3 min and 2 studies (Brown et al., 2014, 2016) that reported a delay of 7 min between the fatiguing protocol and the overground running assessment. The biomechanical assessment was performed ‘immediately’ following the fatiguing protocol in 4 studies (Bazett-Jones et al., 2013; A. Jafarnezhadgero et al., 2019; A. A. Jafarnezhadgero et al., 2019; Paquette & Melcher, 2017), but these studies did not report a length of time. One study did not comment on the delay between the fatiguing protocol and overground running assessment (Quan et al., 2021).

There were 75 unique outcomes. There were 7 spatiotemporal outcomes, 7 trunk angles, 2 pelvis angles, 7 hip joint angles and 6 hip joint moments, 11 knee joint angles and 6 knee joint moments, 12 ankle joint angles and 6 ankle joint moments, and 11 GRF outcomes. Forest plots are presented for outcomes with strong evidence only (see Section 2.8.1). All studies included in the forest plots were of high-quality, and the included studies are arranged from highest to lowest methodological quality. Heterogeneity was high for two outcomes: peak hip internal rotation moment and peak dorsiflexion angle during the stance phase at consistent speed running. There was medium heterogeneity for ground contact times and flight times during varied speed running.

Methodological quality and risk of bias

The Downs and Black scores obtained by each study are available in Table 3. The mean methodological quality of the studies was 11.18 ± 1.88 out of 15, with scores ranging from 6 to 13. Twelve studies were categorised as high-quality, and five studies were medium quality. All studies provided estimates of random variability for main outcomes (item 7), clearly described, or reported no losses to follow‐up (item 9) and used appropriate statistical tests to assess the main outcomes (item 18). Eight studies clearly described all relevant characteristics (age, height, weight, gender; running level or training history; foot strike pattern) of the participants included in the study (item 4). Ten studies reported actual probability values for the main outcomes (item 10). No studies attempted to blind those measuring the main outcomes of the intervention (item 15). Only three studies included participants recruited from the wider community (item 11) and only two studies included an a priori power or sample size calculation (item 27).

Table 3. Downs and black quality ratings.

Study	Downs and Black Item															Total/15	Quality
	1	2	3	4	6	7	9	10	11	15	16	17	18	20	27
Bazett-Jones et al. (2013)	1	1	1	1	1	1	1	1	1	0	1	1	1	1	0	13	H
Brown et al. (2014)	1	1	2	1	1	1	1	0	0	0	1	1	1	1	0	12	H
Brown et al. (2016)	1	1	2	1	1	1	1	1	0	0	1	1	1	1	0	13	H
Elliot et al. (1981)	0	0	0	0	1	1	1	0	1	0	0	1	1	0	0	6	M
Girard et al. (2013)	1	1	1	0	1	1	1	0	0	0	1	1	1	1	0	10	M
Hamzavi et al. (2021)	1	1	1	1	1	1	1	1	0	0	1	0	1	1	0	11	H
A. Jafarnezhadgero et al. (2019)	1	1	2	1	1	1	1	1	1	0	1	0	1	1	0	13	H
A. A. Jafarnezhadgero et al. (2019)	1	1	1	1	1	1	1	0	0	0	1	0	1	1	0	10	M
Matta et al. (2020)	1	1	1	1	1	1	1	0	0	0	1	1	1	1	0	11	H
Melaro et al. (2021)	1	1	2	0	1	1	1	1	0	0	1	1	1	1	1	13	H
Paavolainen et al. (1999)	1	0	1	0	0	1	1	0	0	0	1	1	1	1	0	8	M
Paquette et al. (2017)	1	1	2	0	1	1	1	1	0	0	1	0	1	1	1	12	H
Quan et al. (2021)	1	1	2	1	1	1	1	1	0	0	1	0	1	1	0	12	H
Rabita et al. (2011)	1	1	1	1	1	1	1	1	0	0	1	1	1	1	0	12	H
Schena et al. (2014)	1	1	1	1	1	1	1	0	0	0	1	1	1	1	0	11	H
Tam et al. (2017)	1	1	1	1	1	1	1	1	0	0	1	1	1	1	0	12	H
Winter et al. (2021)	1	1	1	0	1	1	1	1	0	0	1	1	1	1	0	11	M

H high-quality; M medium-quality.

Findings

Strong evidence at consistent running speeds

Running-induced fatigue has a statistically significant and small effect on increasing ground contact times (Figure 2a). Running-induced fatigue has no effect on peak hip flexion angle (Figure 2b), peak hip extension moment (Figure 2c), peak hip adduction angle (Figure 2d), peak hip abduction moment (Figure 3a), and peak hip internal rotation angle (Figure 3b) during stance. Excluding either study by Brown et al (Brown et al., 2014, 2016). (due to the possibility of duplicating results from the same participants) does not affect findings in relation to the hip. Running-induced fatigue has no effect on knee flexion angle at initial contact (Figure 3c). Running-induced fatigue also has no effect on peak knee flexion angle (Figure 3d), knee flexion range of motion (Figure 4a), peak knee flexion moment (Figure 4b), peak knee extension moment (Figure 4c), peak knee abduction angle (Figure 4d), peak knee abduction moment (Figure 5a), peak knee adduction moment (Figure 5b), and peak knee internal rotation angle (Figure 5c) during stance. Excluding either study by Jafarnezhadgero et al. (A. Jafarnezhadgero et al., 2019; A. A. Jafarnezhadgero et al., 2019). (due to the possibility of duplicating results from the same participants) does not affect findings in relation to the knee. Running-induced fatigue has no effect on plantarflexion range of motion (Figure 6a), peak plantarflexion moment (Figure 6b), and peak rearfoot eversion (Figure 6c) during stance.

Figure 2. Forest plots showing individual study and pooled effects (i.e., Hedges’ g) of running-induced fatigue on biomechanical outcomes. Heterogeneity statistics and tests for overall average effect are also presented for each variable. Squares with horizontal lines indicate the SMD and 95% CI between pre-fatigue and post-fatigue for each study. Diamonds represent the SMD_p and 95% CI for that outcome from all studies in the meta-analyses. A positive SMD favours an increase in the biomechanical outcome, and the magnitude of the SMD represents the size of the effect. All studies included in the meta-analyses are high-quality. std. standardised; IV inverse variance method; CI confidence interval; RFS runners with rear-foot strike; FFS runners with forefoot strike.

Figure 3. Forest plots showing individual study and pooled effects (i.e., Hedges’ g) of running-induced fatigue on biomechanical outcomes. Heterogeneity statistics and tests for overall average effect are also presented for each variable. Squares with horizontal lines indicate the SMD and 95% CI between pre-fatigue and post-fatigue for each study. Diamonds represent the SMD_p and 95% CI for that outcome from all studies in the meta-analyses. A positive SMD favours an increase in the biomechanical outcome, and the magnitude of the SMD represents the size of the effect. All studies included in the meta-analyses are high-quality. std. standardised; IV inverse variance method; CI confidence interval.

Figure 4. Forest plots showing individual study and pooled effects (i.e., Hedges’ g) of running-induced fatigue on biomechanical outcomes. Heterogeneity statistics and tests for overall average effect are also presented for each variable. Squares with horizontal lines indicate the SMD and 95% CI between pre-fatigue and post-fatigue for each study. Diamonds represent the SMD_p and 95% CI for that outcome from all studies in the meta-analyses. A positive SMD favours an increase in the biomechanical outcome, and the magnitude of the SMD represents the size of the effect. All studies included in the meta-analyses are high-quality. std. standardised; IV inverse variance method; CI confidence interval.

Figure 5. Forest plots showing individual study and pooled effects (i.e., Hedges’ g) of running-induced fatigue on biomechanical outcomes. Heterogeneity statistics and tests for overall average effect are also presented for each variable. Squares with horizontal lines indicate the SMD and 95% CI between pre-fatigue and post-fatigue for each study. Diamonds represent the SMD_p and 95% CI for that outcome from all studies in the meta-analyses. A positive SMD favours an increase in the biomechanical outcome, and the magnitude of the SMD represents the size of the effect. All studies included in the meta-analyses are high-quality. std. standardised; IV inverse variance method; CI confidence interval.

Figure 6. Forest plots showing individual study and pooled effects (i.e., Hedges’ g) of running-induced fatigue on biomechanical outcomes. Heterogeneity statistics and tests for overall average effect are also presented for each variable. Squares with horizontal lines indicate the SMD and 95% CI between pre-fatigue and post-fatigue for each study. Diamonds represent the SMD_p and 95% CI for that outcome from all studies in the meta-analyses. A positive SMD favours an increase in the biomechanical outcome, and the magnitude of the SMD represents the size of the effect. All studies included in the meta-analyses are high-quality. std. standardised; IV inverse variance method; CI confidence interval.

Moderate evidence at consistent and varied running speeds

There is moderate evidence that running-induced fatigue has no effect on step/stride length and frequency, lower-body joint angles and moments, and vertical loading rates at consistent running speeds (Table 4). Moderate evidence indicates that step length is reduced following a fatiguing run at varied running speeds (Table 4).

Table 4. Pooled effects of running-induced fatigue on biomechanical outcomes with moderate levels of evidence.

Outcome Measure	Total n	I^2 (p-value)	SMDp	[95% CI]
Consistent running speeds
Step/stride length (Rabita et al., 2011; Schena et al., 2014)	27	0% (p = 0.70)	−0.45	[−1.00, 0.09]
Step/stride frequency (Rabita et al., 2011; Schena et al., 2014)	27	0% (p = 0.44)	0.15	[−0.39, 0.68]
Hip flexion range of motion during stance (Melaro et al., 2021; Quan et al., 2021)	63	0% (p = 0.80)	0.34	[−0.01, 0.69]
Peak hip flexion moment during stance (A. Jafarnezhadgero et al., 2019; Quan et al., 2021)	76	0% (p = 0.63)	−0.03	[−0.35, 0.29]
Peak hip internal rotation moment (Bazett-Jones et al., 2013; A. Jafarnezhadgero et al., 2019)	45	0% (p = 0.006)	−0.23	[−1.41, 0.94]
Peak hip external rotation moment during swing (Brown et al., 2016; A. Jafarnezhadgero et al., 2019)	46	0% (p = 0.43)	−0.19	[−0.60, 0.22]
Peak knee external rotation moment during stance (Bazett-Jones et al., 2013; A. Jafarnezhadgero et al., 2019)	45	0% (p = 0.66)	0.22	[−0.20, 0.63]
Plantarflexion angle at initial contact (Quan et al., 2021; Tam et al., 2017)	72	53% (p = 0.12)	−0.24	[−0.81, 0.32]
Peak plantarflexion angle during stance (Quan et al., 2021; Tam et al., 2017)	72	0% (p = 0.76)	−0.03	[−0.36, 0.30]
Peak dorsiflexion angle (Brown et al., 2014; Quan et al., 2021)	70	80% (p = 0.02)	−0.23	[−1.06, 0.60]
Peak dorsiflexion moment during stance (A. Jafarnezhadgero et al., 2019; Quan et al., 2021)	76	0% (p = 0.36)	−0.04	[−0.36, 0.27]
Peak ankle internal rotation moment during stance (Brown et al., 2014; A. Jafarnezhadgero et al., 2019)	46	0% (p = 0.43)	0.19	[−0.22, 0.60]
Vertical loading rates (A. A. Jafarnezhadgero et al., 2019; Paquette & Melcher, 2017)	34	0% (p = 0.49)	0.48	[−0.00, 0.97]
Varied running speeds
Ground contact time (Girard et al., 2013; Matta et al., 2020; Winter et al., 2021)	65	70% (p = 0.005)	0.41	[−0.26, 1.08]
Flight time (Girard et al., 2013; Matta et al., 2020; Winter et al., 2021)	65	57% (p = 0.04)	−0.07	[−0.62, 0.47]
Step length (Elliot & Ackland, 1981; Girard et al., 2013; Matta et al., 2020)	36	0% (p = 0.52)	−1.27	[−1.79, −0.75]
Step frequency (Elliot & Ackland, 1981; Girard et al., 2013; Matta et al., 2020; Winter et al., 2021)	73	44% (p = 0.09)	−0.34	[−0.80, 0.11]

n number; I² Heterogeneity statistic; p-value probability value; SMD_p pooled standardised mean difference; CI confidence interval.

Limited evidence at consistent running speeds

There is limited evidence that running-induced fatigue has no effect on flight time, vertical oscillation, trunk, pelvis and lower-body joint angles and moments, and GRFs and impulses at consistent running speeds (Electronic Supplementary Material 1).

Very limited evidence at varied running speeds

There is very limited evidence that running-induced fatigue has no effect on swing time, total stride duration, vertical oscillation, trunk angles, and GRFs and impulses at varied running speeds (Electronic Supplementary Material 1).

Discussion and implications

This systematic review and meta-analysis summarises the acute fatigue-inducing effect of distance running on GRFs, spatiotemporal parameters, and trunk, pelvis, and lower-body joint angles and moments during overground running. There is strong evidence that running-induced fatigue has a small effect on increasing ground contact times at consistent running speeds. Ground contact times increased by 5.5% on average after fatigue. There is also moderate evidence that running-induced fatigue has a large effect on reducing step length at varied running speeds. Step lengths decreased when running-induced fatigue slowed overground running speed by over 5% (Elliot & Ackland, 1981; Girard et al., 2013; Matta et al., 2020). The reduction in step length causes this decrease in running speed and could be a target for future training interventions. There were no changes in the other spatiotemporal variables. These findings could indicate that the muscles responsible for forcefully pushing on the ground (i.e., the plantarflexors) fatigue earlier than other lower-limb muscles (e.g., the hamstrings or quadriceps). Running-induced fatigue does not change lower-body joint angles and moments during overground running. Thus, our hypothesis that spatiotemporal parameters are affected by fatigue is accepted, whereas we reject our hypothesis that lower limb joint angles and moments would change following a period of running-induced fatigue.

The absence of any fatigue-related changes in sagittal-plane hip kinematics is consistent with findings from a recent meta-analysis (Zandbergen et al., 2023). However, our meta-analysis is the first to provide strong evidence that running-induced fatigue does not change peak hip joint moments or frontal and transverse plane hip joint angles. We found strong evidence that running-induced fatigue does not affect knee flexion angle at initial contact, which contrasts with past meta-analytical findings that running-induced fatigue increases knee flexion at initial contact by 1.64° (Zandbergen et al., 2023). While these authors reported a significant effect on knee flexion at initial contact, the 1.64° difference is small and likely falls within the typical error of measurement. Differences between meta-analysis protocols could also explain these divergent findings. Zandbergen et al. (Zandbergen et al., 2023). pooled data for treadmill and overground running, and two out of three of their included studies measured biomechanical outcomes during treadmill running. The authors acknowledged that knee flexion angles during treadmill and overground running respond differently to running-induced fatigue. Knee flexion typically decreases or shows no change in overground measurements following fatigue, whereas it tends to increase or show no change in treadmill measurements (Apte et al., 2021; Zandbergen et al., 2023). Our meta-analysis is also the first to provide strong evidence that running-induced fatigue does not change peak knee joint moments or frontal and transverse plane hip joint angles during overground running. Past research has found increased peak hip adduction angle (Noehren et al., 2007), peak external knee adduction moment (Dudley et al., 2017), and peak knee abduction impulse (Stefanyshyn et al., 2006) among runners that develop an RRI, such as iliotibial band syndrome or patellofemoral pain. Our findings provide assurance that running under fatigue does elicit undesirable changes in running mechanics that have been prospectively linked to injuries.

We found strong evidence that running-induced fatigue does not affect peak plantarflexion moments during stance at consistent overground running speeds. However, all studies showed a trend for small decreases in peak plantarflexion moments and additional evidence may reveal a statistically significant effect. The ankle plantarflexors have the largest relative contribution to vertical support forces and stride length during running (Dorn et al., 2012). Thus, the effects of running-induced fatigue on plantarflexor moments and running performance warrants further investigation. We also found strong evidence that fatigue does not affect peak rearfoot eversion during consistent speed running. This finding is important because greater peak rearfoot eversion angles have been prospectively associated with risk of developing RRIs (Ceyssens et al., 2019), such as tibial stress injuries (Kuhman et al., 2016) and patellofemoral pain (Powers et al., 2017). Running-induced fatigue does not appear to consistently affect ankle mechanics in the frontal plane and may be unrelated to the development of RRIs.

It was not possible to pool data for any outcomes related to the trunk or pelvis because outcomes are only analysed by one study each. A treadmill run to exhaustion did not affect peak trunk flexion, contralateral trunk bending, ipsilateral trunk bending, anterior pelvic tilt, or contralateral pelvic drop among physically active participants during overground running (Bazett-Jones et al., 2013). There were also no changes in trunk flexion angles among highly skilled male long-distance runners throughout a 10,000 m National Final (Elliot & Ackland, 1981). Outcomes related to the trunk and pelvis should not be overlooked in future analyses on the effects of running-induced fatigue on overground running mechanics, particularly since greater peak contralateral pelvic drop and forward trunk lean have been associated with RRIs (Becker et al., 2018; Bramah et al., 2018). For every 1° increase in pelvic drop, researchers have found an 80% increase in the likelihood of being classified as injured (Bramah et al., 2018). Additionally, a 3–4° increase in trunk flexion can shift the body’s centre of mass anteriorly by 0.02–0.03 m, increasing loads on the back extensors and gluteus maximus to control larger thorax flexion-extension moments (Preece et al., 2016).

There were insufficient studies to pool evidence related to peak GRFs and impulses; however, moderate evidence suggests that running-induced fatigue increases vertical loading rates during consistent speed running (SMD_p 0.48 [−0.00, 0.97]) (A. A. Jafarnezhadgero et al., 2019; Paquette & Melcher, 2017). A previous meta-analysis (Darch et al., 2022) found moderate evidence that fatigue has no effect on vertical average loading rate when pooling treadmill and overground running studies (SMD_p 0.24 [−0.08, 0.56]). This suggests that fatigue-induced changes in loading rates may be influenced by running surface (Apte et al., 2021). Prospective studies provide conflicting evidence on whether GRF metrics are associated with the development of RRIs (Matijevich et al., 2019). There is also a lack of evidence showing that GRF metrics reflect the loads experienced by internal musculoskeletal structures during running (Matijevich et al., 2019); thus, acute fatigue-induced increases in vertical loading rates are unlikely to lead to RRIs.

Limitations

Similar to previous systematic reviews and meta-analyses on this topic (Apte et al., 2021; Darch et al., 2022), runners of all abilities and varying fatigue protocols were combined in this meta-analysis to overcome the limited number of studies for each outcome and to ensure large enough sample sizes for computing meaningful pooled findings. Past researchers have shown that there are no differences between fast and slow distance runners’ biomechanics with the onset of fatigue (Paavolainen et al., 1999; Siler & Martin, 1991). There also does not appear to be a correlation between weekly running mileage (i.e., running experience) and biomechanical changes following running-induced fatigue to justify a subgroup analyses based on running experience (Paquette & Melcher, 2017).

Nearly half (44%) of the included studies performed the fatiguing protocols on a treadmill, most likely due to the increased control and convenience of treadmill running in comparison to overground running. A limitation of these studies is the discontinuity and time delay between the fatiguing treadmill run and the collection of overground running biomechanics. When collecting three-dimensional motion capture data, tracking markers may need to be reapplied if they fell off due to high perspiration during the fatigue protocol. This rest period may alleviate some of the effects of fatigue and contribute to the lack of statistically significant findings. There is preliminary evidence that kinematic changes persist following a relatively short submaximal treadmill run (3.9 ± 1.6 km) despite a four-min rest period (Gruber et al., 0000; Melaro et al., 2021). However, it is possible that the effects of fatigue could be alleviated in studies reporting a longer rest period (Brown et al., 2014, 2016) or for studies that did not report the length of time between the end of the fatigue protocol and the follow-up biomechanical assessment (Bazett-Jones et al., 2013; A. Jafarnezhadgero et al., 2019; A. A. Jafarnezhadgero et al., 2019; Paquette & Melcher, 2017; Quan et al., 2021). Future research should be conducted in a laboratory with the aim to minimise rest periods when transitioning from the treadmill to the overground biomechanical assessment, and rest periods must be clearly reported.

Finally, it was not possible to include more competition-based studies to improve the ecological validity of these findings. Competition-based studies were commonly excluded because (i) the effects of fatigue could not be isolated (e.g., authors did not report running speed during biomechanical measurements so the effect of fatigue could not be interpreted); (ii) there was no true baseline measurement or the first measure was > 10% into the race; (iii) the fatiguing protocol was predominantly incline running (e.g., trail races); and, (iv) there was no outcome variable of interest (e.g., waveform data only; tibial accelerations).

Conclusion

There is strong evidence that fatigue induced by distance running has a small effect on increasing ground contact times (+5.5%) at consistent running speeds. Moderate evidence indicates that step length is reduced following a fatiguing run when running speed slows by over 5%. Minimising changes in these stride parameters could provide a mechanism for reducing the effects of fatigue on running performance. Running-induced fatigue does not affect trunk, pelvis, or lower-limb joint angles and moments during overground running. Thus, running under fatigue does not elicit undesirable changes in running mechanics that have been prospectively linked to injuries.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14763141.2024.2353390

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.