bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024.

The copyright holder for this preprint

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

Original Article

Hamstrings are stretched more and faster during accelerative running compared to
speed-matched constant speed running

Reed D. Gurchiek1,2,*, Zachary Teplin2, Antoine Falisse2, Jennifer L. Hicks2, Scott L. Delp2,3

1
Department of Bioengineering, Clemson University, Clemson, SC 29634, USA
2
Department of Bioengineering, Stanford University, Stanford, CA 94305, USA
3
Department of Mechanical Engineering, Stanford University, Stanford, CA 94305, USA

*Please direct correspondence to:

Reed D. Gurchiek
204 Rhodes Annex
Clemson, SC 29634
Phone: 864-656-5552
Email: [email protected]
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

Abstract
Background: Hamstring strain injuries are associated with significant time away from sport and
high reinjury rates. Recent evidence suggests that hamstring injuries often occur during
accelerative running, but investigations of hamstring mechanics have primarily examined
constant speed running on a treadmill. To help fill this gap in knowledge, this study compares
hamstring lengths and lengthening velocities between accelerative running and constant speed
overground running.

Methods: We recorded 2 synchronized videos of 10 participants (5 female, 5 male) during 6

accelerative running trials and 6 constant speed running trials. We used OpenCap (a
markerless motion capture system) to estimate body segment kinematics for each trial and a 3-
dimensional musculoskeletal model to compute peak length and step-average lengthening
velocity of the biceps femoris (long head) muscle-tendon unit. To compare running conditions,
we used linear mixed regression models with running speed (normalized by the subject-specific
maximum) as the independent variable.

Results: At running speeds below 75% of top speed accelerative running resulted in greater
peak lengths than constant speed running. For example, the peak hamstring muscle-tendon
length when a person accelerated from running at only 50% of top speed was equivalent to
running at a constant 88% of top speed. Lengthening velocities were greater during accelerative
running at all running speeds. Differences in hip flexion kinematics primarily drove the greater
peak muscle-tendon lengths and lengthening velocities observed in accelerative running.

Conclusion: Hamstrings are subjected to longer muscle-tendon lengths and faster lengthening
velocities in accelerative running compared to constant speed running. This provides a
biomechanical explanation for the observation that hamstring strain injuries often occur during
acceleration. Our results suggest coaches who monitor exposure to high-risk circumstances
(long lengths, fast lengthening velocities) should consider the accelerative nature of running in
addition to running speed.
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

1. Introduction
Hamstring strain injuries are a major time-loss injury for athletes in many sports including
soccer1, track and field2, basketball3, American football4, and rugby5. A recent meta-analysis6 of
63 articles and more than 7 million exposure hours estimated that hamstring strains comprised
10% of all injuries in field-based sports. This analysis revealed relatively constant injury rates
over the last 30 years, while Ekstrand et al. reported increases in elite soccer7. Reinjury rates
are high8,9. Injury prevention is thus paramount, but knowledge about injury mechanisms is still
limited. A better understanding of the movement conditions that pose a high risk for injury could
lead to innovations in preventative screening and athlete monitoring, and interventions to
prevent injury.

The peak length of the hamstring muscle increases with running speed both at the muscle
fiber10 and muscle-tendon unit11 levels. High running speeds also present large hamstring
lengthening velocities under high loads12. These eccentric loads are associated with increased
muscle fiber strains13 and eccentric work, predictors of microdamage14–16. These biomechanical
features of the hamstrings during high-speed running (large peak lengths, lengthening
velocities, and eccentric loads) are thought to contribute to injuries12,17,18 and may help explain
the occurrence of injuries during high-speed running19–22. As a result, high-speed running is
often used as an index of load for athlete monitoring23,24.

The eccentric loading and injury incidence associated with running has motivated the
investigation of hamstring mechanics during high-speed running25. For example, kinematic
analyses have shown that hamstring lengths (normalized by the length in an upright posture)
peak in late swing phase and are greatest in the biceps femoris long head compared to the
other hamstring muscles due to moment arm differences11,26. Kinetic analyses have
demonstrated that hamstring loads may exceed 2000 N27 and are greater in swing phase (10.5-
26.4 N/kg) than in stance phase (4.6-11.6 N/kg)25. Moreover, eccentric work is done exclusively
in swing phase26,28 and hamstring muscle activity during stance phase is greatest at foot
contact29.

Previous investigations of hamstring injury mechanisms have primarily considered running at

relatively constant and high speeds25,30. However, it was recently shown that hamstring injuries
often occur when athletes are accelerating31 and not necessarily at high running speeds22.
Moreover, the increase in injury incidence in elite soccer players reported by Ekstrand et al.
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

between 2001-201432 may partly result from an evolution in the nature of running activity soccer
players perform. For example, the number of sprints, sprint distance, and “explosive” (high-
acceleration) sprints performed by soccer players all increased within a similar time frame
(2006-2013)33.

While the biomechanics of accelerative running have been explored34–37, the kinematics of the
hamstrings (muscle-tendon length and lengthening velocity) during acceleration are less
understood. The hamstrings are important for generating propulsive ground reaction forces
during acceleration35,36, with large hamstring muscle forces (6-8 bodyweights) during a sprint
that starts from stationary35. In a prospective study, Schuermans et al. (2017) identified
increased anterior pelvic tilt and increased frontal plane trunk bending as kinematic features
during the acceleration phase of sprinting of soccer athletes who sustained an injury compared
to those who did not38. Higashihara et al. (2018) found the biceps femoris was more active than
the semitendinosus during stance phase in accelerative running, whereas the semitendinosus
was more active in swing phase in both accelerative and top speed running39. However, this
study considered accelerative running at 15 m after the start of the sprint wherein athletes had
already obtained a relatively high speed (an average of 8.54 m/s, approximately 90% of the
average top speed of 9.52 m/s). Schuermans et al. (2017) observed a similar phase of the sprint
(15-25 m after the start), but running speeds were not reported. Thus, the effect of running
speed on hamstring mechanics (i.e., both kinematics and kinetics) during accelerative running
remains unclear. Furthermore, the mechanics of a runner accelerating from a relatively high
speed may not closely represent accelerative running activity during gameplay, which often is
initiated from submaximal running speeds (a “flying” sprint). Finally, hamstring lengths and
lengthening velocities during accelerative running have not been explored using a 3-
dimensional musculoskeletal model similar to the work of Thelen et al. (2005) for non-
accelerative, high-speed running on a treadmill11.

Given the lack of knowledge about hamstring lengths and lengthening velocities during
accelerative running, we aimed to compare these variables during accelerative running to
running at a constant speed across a range of speeds. To emulate in-game running activity, we
used a markerless motion capture technology (OpenCap40) that allowed analysis of overground
accelerative running in an unconstrained outdoor setting. We hypothesized the hamstrings
would be subject to greater peak muscle-tendon lengths and lengthening velocities in
accelerative running compared to constant speed running.
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

2. Material and Methods

2.1 Data Collection
Ten participants volunteered to participate in this study (Table 1). They were recreationally
active with an average self-reported exercise frequency of 5 days per week for 1 hour per day.
Following a self-directed warm-up, participants performed a 30-meter maximal effort sprint, 6
constant speed running trials, each at a different speed, and 6 accelerative running trials. The
30-meter maximal effort sprint was used to determine the subject-specific top running speed.
The accelerative running trials were flying sprints wherein participants began running at a
constant speed and, when given a verbal cue, accelerated forward as fast as possible. These
trials were designed to represent in-game accelerative running activity in team sports where
sprinting is often initiated from a moving start. The constant speeds that preceded the 6
accelerations were the same as those in the constant speed running trials. The first two running
speeds were self-selected (jogging and brisk running) and the last 4 were prescribed (3, 4, 5,
and 6 m/s, performed in that order). At each speed, subjects performed the constant speed
running trial first followed by the accelerative running trial and were given as much rest time
between trials as desired to avoid fatigue. The order of running speeds and conditions was not
randomized. All trials were performed overground on an outdoor running surface. All study
activities were approved by the Stanford University Institutional Review Board. Informed written
consent was obtained from all study participants.

For the jogging self-selected speed, participants were instructed to run at a comfortable jogging
speed. For the brisk-running self-selected speed, they were instructed to run at a brisk running
speed that was faster than the jogging speed, but not a maximal effort sprint. Participants
matched prescribed speeds by self-modulating their running speed such that they passed cones
positioned along the running lane at uniform distances (2.5 or 5.0 meters) to the beat of a
metronome. The metronome frequency was based on the distance between cones and the
prescribed speed (e.g., 60 beats per second with 5-meter cone spacing for a 5 m/s speed
prescription). Subjects practiced running at each speed before data was collected. It was not
imperative that speeds were matched exactly, but that a range of speeds were captured to allow
investigation of the effects of running speed.

Two mobile phones were used to record videos (60 frames per second) of the running trials.
The phones were attached to tripods placed such that the center of the capture volume was
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

approximately 35 meters from the start of all running trials. Each phone was positioned 4 meters
from this 35-meter mark at an angle from the direction of travel on each side of the running lane
ensuring both cameras could view the anterior-frontal plane along with the right-sagittal plane
for one and the left-sagittal plane for the other.

Table 1
Subject characteristics
Male Female All Sexes
Mean (SD) Mean (SD) Mean (SD)
Sample Size 5 5 10
Age (years) 23.3 (4.5) 21.8 (0.9) 22.6 (3.2)
Height (cm) 177.8 (8.6) 168.7 (8.2) 173.2 (9.3)
Mass (kg) 79.4 (16.0) 66.7 (8.7) 73.0 (13.9)
Top Speed* (m/s) 7.9 (0.5) 6.9 (0.5) 7.4 (0.7)
*As observed in a maximal effort 30-meter sprint

2.2 Data Analysis

Video recording, synchronization, and processing as well as the scaling and kinematic analysis
of a 3-dimensional musculoskeletal model41 were performed with OpenCap40 using OpenPose42
as the underlying pose estimation model. We reprocessed the video data to leverage
OpenPose’s high accuracy settings, which were shown to provide more accurate results than
the default settings at the expense of computational cost40. The model was scaled using video
data from a static standing calibration trial. We identified consecutive foot contact events to
segment kinematic data into individual steps. Foot contact events were identified as the instant
coincident with the negative-going zero crossing in the shank sagittal plane angular velocity
(lowpass filtered with a 2 Hz cutoff frequency for event detection only). This event follows a
relatively large-magnitude, positive signal associated with the swing phase43 and thus is reliably
identified algorithmically and facilitates a consistent step segmentation. We analyzed the 2
steps (1 full gait cycle) immediately following the initiation of acceleration in the accelerative
running trials and the 3 steps closest to the center of the capture volume in the constant speed
trials.

The length of the biceps femoris (long head) muscle-tendon unit (MTU) was calculated using
OpenSim44 and the subject-specific musculoskeletal model. The MTU length is dependent only
on the hip and knee joint angles, which were calculated within OpenCap and were lowpass
filtered with a zero-phase shift, 4th order Butterworth filter with 15 Hz cutoff frequency adjusted
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

for the backward pass45. The 15 Hz cutoff frequency was chosen because, on average, 90% of
the hip and knee flexion angle signal power was contained below 15 Hz across all running trials.
Hip and knee flexion speed were computed by numerically differentiating the hip and knee
flexion angles with respect to time using the 5-point central difference method. The MTU
lengthening velocity was calculated in OpenSim using the joint angles and speeds. For each
step from each trial, we used the peak MTU length and average lengthening velocity (on the
side exclusively in the swing phase during the step) as primary outcomes to compare running
conditions. We chose to explore the average lengthening velocity instead of the peak because it
is more robust to noise introduced from numerical differentiation. Nonetheless, we also explored
the peak lengthening velocity in a post-hoc analysis (results reported in the supplementary
material).

2.3 Data Quality Assessment

We identified and removed 61 poorly reconstructed steps (out of 300 total) leaving 239 available
for analysis (140 steps for constant speed running and 99 steps for accelerative running). This
was based on several data quality checks. First, trials were removed if the kinematic
reconstructions were unacceptable following visual inspection in the OpenSim graphical user
interface or if any foot contacts were not identified. Then, an individual step was removed if
either outcome measure (peak length or lengthening velocity) for the step was deemed an
outlier, called a discrete outlier, or if one of several other relevant kinematic time-series for the
step was deemed an outlier, called a time-series outlier. A discrete outlier was defined as a
value that was greater than 3 standard deviations away from the mean. For detection of time-
series outliers, all time-series were first resampled to express their values as a function of the
percentage of the step duration. A time-series was labeled an outlier if any of these discrete-
time values met the discrete outlier criterion. The time-series included for outlier detection were
the hip and knee flexion angle and speed and the MTU length and lengthening velocity. The
mean and standard deviation used for outlier detection were computed using data from all
participants, but specific to running condition (constant speed or accelerative) and running
speed (0-55%, 55-65%, 65-75%, 75-85%, and 85-100% of top speed).

2.4 Statistical Analysis

We used 2 separate linear mixed model regression analyses to compare the primary outcomes
(MTU peak lengths and MTU lengthening velocities) between running conditions with running
speed as the independent variable. To account for size variations across participants, we
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

normalized MTU lengths and lengthening velocities by the MTU length in the neutral
configuration (hereon referred to as the neutral length). To account for variability in top running
speed across participants, we normalized running speeds by the subject-specific maximum
observed in the 30-meter sprint. Each regression model had 4 fixed-effects coefficients
including 2 condition-specific intercepts and 2 condition-specific slopes (modeling the effect of
running speed) along with 4 random effects (condition-specific slopes and intercepts) per
participant. Running speeds were shifted (normalized running speed minus 1) for model fitting
such that the intercept corresponded to the response (peak length or lengthening velocity) at top
speed. We used the fitlme function in Matlab (R2021a, The MathWorks, Inc., Natick, MA) for
model fitting using restricted maximum likelihood estimation (REML) and estimated all elements
of the covariance matrix (i.e., a FullCholesky covariance pattern).

To explore the effects of different joint and segment kinematics on the primary outcomes, we
also fit 6 additional linear mixed regression models with the same structure as for the primary
outcomes, but with different response variables: hip flexion, knee flexion, pelvic tilt, and thigh tilt
(relative to the global vertical axis) angles at the instant of peak MTU length and the
contributions of the hip and knee flexion speed to the MTU lengthening velocity. The latter were
determined by first computing the product of the hip or knee flexion speed and the joint angle
specific MTU moment arm. The sample average across all products for which the MTU was
lengthening was used as the response. Pelvic and thigh tilt angles were calculated using an
Euler angle decomposition (ZXY for the pelvis, ZYX for the thigh) of the orientation (rotation
matrix) of each segment relative to the global coordinate system. The tilt angle for both
segments was the angle associated with the first rotation (about the Z-axis). Because the
motion is primarily sagittal, the tilt angle can be interpreted as the angular deviation of the
segment long axis relative to the global vertical axis. For example, the pelvic and thigh tilt angle
would both be 0° during a standing posture, 90° during a supine posture, and -90° during a
prone posture.

3. Results
The general pattern of the biceps femoris MTU length throughout the step was similar for
constant speed and accelerative running (Figure 1). Minimum MTU length occurred at
approximately 25% of the step cycle. The hamstrings then lengthened up to a peak length at
approximately 75-80% of the step cycle and terminated with a brief shortening phase just before
foot contact (Figure 1, top row).
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

Normalized Normalized Normalized Normalized Normalized

Running Speed: Running Speed: Running Speed: Running Speed: Running Speed:
1.2 0-0.55 1.2 0.55-0.65 1.2 0.65-0.75 1.2 0.75-0.85 1.2 0.85-1.00
(neutral lengths)
MTU Length

1.1 1.1 1.1 1.1 1.1

1 1 1 1 1

0.9 0.9 0.9 0.9 0.9

0.8 0.8 0.8 0.8 0.8

90 90 90 90 90
Hip Flex. (deg)

60 60 60 60 60

30 30 30 30 30

0 0 0 0 0

-30 -30 -30 -30 -30

120 120 120 120 120

Knee Flex. (deg)

90 90 90 90 90

60 60 60 60 60

30 30 30 30 30

0 0 0 0 0
0 50 100 0 50 100 0 50 100 0 50 100 0 50 100

% Step % Step % Step % Step % Step

Constant Speed Acceleration

contralateral ipsilateral
foot foot
contact contact

Figure 1. Biceps femoris MTU length (top row, units: neutral lengths), hip flexion angle (middle row, units:
degrees), and knee flexion angle (bottom row, units: degrees) as a percentage of the step duration
(contralateral foot contact to ipsilateral foot contact) for the accelerative (red) and constant speed (teal)
running conditions. Solid lines and shaded areas indicate the ensemble mean +/- standard deviation
across all participants within the same speed bin (indicated at the top of each column). Running speeds
are normalized by the subject-specific top speed.

Despite similarity in the general pattern of the MTU length trajectories, accelerative running
resulted in greater peak MTU lengths, with more pronounced differences for slower running
speeds (Figure 1). At top speed, the peak lengths were 111-112% of neutral length (11-12%
strain), and this was not different between conditions (p=0.431, Figure 2A). However, the peak
MTU length was greater (p<0.001, Figure 2A) when accelerating from slow speeds than in
speed-matched, constant speed running (i.e., below 75% of top speed based on 90% prediction
intervals in Figure 2A). For example, the regression analysis predicted that the peak MTU
length experienced when a person accelerates while running at only 50% of top speed is the
same as if running with a constant speed of 88% of top speed. (The linear mixed regression
results are also tabulated in Table S1 of the supplementary material.)
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

A B
Constant Speed
Acceleration
2

(neutral lengths per second)

MTU Lengthening Velocity
1.15
Peak MTU Length
(neutral lengths)

1.5
1.1

1
1.05

0.5
1
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Running Speed Running Speed

(normalized by top speed) (normalized by top speed)

Figure 2. Peak biceps femoris MTU length (A) and average lengthening velocity (B) vs. normalized
running speed during both constant speed (teal) and accelerative (red) running. Neutral length is the MTU
length in the neutral configuration (upright standing posture). Each data point (triangles for constant
speed, circles for acceleration) represents a single step. Solid lines depict the linear mixed model
regression lines, and the shaded area illustrates the 90% prediction interval.

The greater peak MTU lengths in accelerative running resulted from a more flexed hip (hip
flexion lengthens the hamstrings), despite a more flexed knee (knee flexion shortens the
hamstrings) compared to constant speed running (Figure 3). The MTU lengthening effect of hip
flexion was greater than the shortening effect of knee flexion because the hamstring moment
arm about the hip is greater than the knee. The differences between accelerative and constant
speed running in the hip and knee flexion angle at the instant of peak MTU length were greater
for slower running speeds (Figure 1, middle and bottom rows). For example, the hip was more
flexed (p<0.001) in accelerative than in constant speed running at speeds below 96% of top
speed (based on overlap of the 90% prediction intervals, Figure 3A). This greater hip flexion
angle at the instant of peak MTU length was due primarily to a more horizontal thigh (Figure 4B)
in addition to a more anteriorly tilted pelvis (Figure 4A). Based on the linear mixed model, the
hip was flexed 6° more in accelerative running than in constant speed running with every 10%
decrease in running speed, where 4° of the difference was due to a more horizontal thigh and 2°
to a more anteriorly tilted pelvis.
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

A B
Constant Speed
Acceleration
100 80

Knee Flexion (deg)

80
Hip Flexion (deg)

60

60 40

40 20

20 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Running Speed Running Speed

(normalized by top speed) (normalized by top speed)

Figure 3. Hip (A) and knee (B) flexion angle at the instant of peak MTU length vs. normalized running
speed during both constant speed (teal) and accelerative (red) running. Each data point (triangles for
constant speed, circles for acceleration) represents a single step. Solid lines depict the linear mixed
model regression lines, and the shaded area illustrates the 90% prediction interval.

A B
Constant Speed
Acceleration
60
30

25 50
Pelvis Tilt (deg)

Thigh Tilt (deg)

20
40
15
30
10
20
5

0 10
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Running Speed Running Speed

(normalized by top speed) (normalized by top speed)

Figure 4. Pelvis (A) and thigh (B) tilt angle at the instant of peak MTU length vs. normalized running
speed during both constant speed (teal) and accelerative (red) running. Positive pelvis and thigh tilt
angles both contribute to hip flexion. For example, at top speed (normalized running speed = 1) in
accelerative running, the pelvis (19°) and thigh (47°) tilt angles sum to 66°, the approximate hip flexion
angle in the same condition (Figure 3A). Each data point (triangles for constant speed, circles for
acceleration) represents a single step. Solid lines depict the linear mixed model regression lines, and the
shaded area illustrates the 90% prediction interval.
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

The average MTU lengthening velocity was greater during accelerative than in constant speed
running across all running speeds (Figure 2B). This resulted from a greater hip flexion speed,
rather than from a greater knee extension speed (Figure 5). In fact, below 78% of top running
speed, the MTU lengthening velocity resulting from the knee extension speed was greater in
constant speed than in accelerative running (based on 90% prediction interval overlap, Figure
5B).

A B
Constant Speed
Acceleration
1.5
(neutral lengths per second)

(neutral lengths per second)

1.5
MTU Lengthening Velocity

MTU Lengthening Velocity

From Knee Extension
From Hip Flexion

1 1

0.5 0.5

0 0

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Running Speed Running Speed

(normalized by top speed) (normalized by top speed)

Figure 5. Hip flexion (A) and knee extension (B) speed contributions to the average MTU lengthening
velocity vs. normalized running speed during both constant speed (teal) and accelerative (red) running.
Neutral length is the MTU length in the neutral configuration (upright standing posture). Each data point
(triangles for constant speed, circles for acceleration) represents a single step. Solid lines depict the
linear mixed model regression lines, and the shaded area illustrates the 90% prediction interval.

4. Discussion
Our results demonstrate that the hamstrings are stretched more and faster during accelerative
compared to constant speed running. Furthermore, the differences were more pronounced
when accelerating from slower running speeds. This provides a biomechanical explanation for
the observation that hamstring injuries often occur when athletes are accelerating31 (and not
necessarily at high running speeds22). Our results also have implications for athlete monitoring.
For example, while our data support the use of exposure to high running speeds (e.g., from
body-worn GPS units) as a proxy for increased hamstring length and lengthening velocity, they
also suggest the accelerative nature of running should be included to more comprehensively
characterize exposure to high-risk circumstances.
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

Our results illustrate how hip-knee coordination underlies the differences in accelerative and
constant speed running and the implications for exposing the hamstrings to long lengths. The
kinematics of accelerative running were such that the peak MTU length occurred at an average
hip flexion angle of 67° (Figure 3A), and this was not influenced by running speed. However, the
knee flexion angle at which the peak MTU length occurred was influenced by running speed.
For example, when accelerating from slower running speeds, the knee was more flexed at the
instant of peak MTU length than in speed-matched, constant speed running (approximately 3°
more knee flexion with every 10% of top speed decrease in running speed). Without this
increase in knee flexion, the peak MTU length would have been even greater when accelerating
from slower running speeds (i.e., when the capacity to accelerate is greater). Thus, a
coordinated motion wherein the hip and knee flex together during the late swing phase of
accelerative running reduces the peak length of the hamstrings. In contrast, an accelerative
running pattern wherein the knee does not flex synchronously with the hip exposes the
hamstrings to longer peak lengths.

The importance of hip-knee coordination in accelerative running is further supported by the

observed MTU lengthening velocities. Large hip flexion speeds drove the increase in MTU
lengthening velocity during accelerative running compared to constant speed running. This was
despite the knee extension speed contributing less to MTU lengthening velocity in accelerative
compared to constant speed running (Figure 5B). In fact, our linear mixed models predicted that
when accelerating from below approximately 40% of top speed, the knee contributed a net
average shortening velocity (negative vertical axis in Figure 5B). Therefore, a lack of
simultaneous hip and knee flexion during accelerative running would be two-fold detrimental. It
would result in greater peak MTU length and greater MTU lengthening velocity. An altered hip-
knee coordination in this way is generally consistent with others who have noted the potential
influence of altered coordination on injury risk30.

Cumulative eccentric MTU work has been proposed as a predictor of muscle damage and
would be a candidate proxy for hamstring injury risk. Eccentric work can be interpreted as the
average MTU lengthening velocity weighted by MTU force. This supports our use of average
MTU lengthening velocity as a kinematic surrogate for eccentric work. Nonetheless, future work
should compare muscle kinetics between accelerative running and constant speed running.
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

OpenCap allowed us to quantify the kinematics of accelerative running overground in outdoor

conditions. Its use of computer vision presents some limitations, especially regarding
measurement of the pelvis tilt angle. The OpenPose pose detection algorithm only identifies 2
key points on the pelvis (the 2 hip joint centers) rendering the pelvis tilt angle unobservable.
Observability is made possible following prediction of the positions of additional pelvis-fixed
anatomical landmarks from the OpenCap neural network, which is trained on a large dataset
that includes running motions. Confidence in our results is supported by similarity with previous
studies. Thelen et al. (2005) reported peak MTU lengths between 1.09-1.10 neutral lengths that
occurred at hip and knee flexion angles between 63-65° and 44-45°, respectively, during
constant speed running at 80-100% of top speed11. Under the same conditions, the linear mixed
models based on our data predict similar results for peak MTU length (1.09-1.11 neutral
lengths) and hip flexion (between 52-64°) but with less knee flexion (28-35°). We also observed
the same general pattern of the MTU length trajectories (e.g., compare the top row of our Figure
1 with the top row of Figure 2 from Thelen et al. (2005)). Schache et al. report peak biceps
femoris MTU strain of 12% at top speed26 which is comparable to our results (11%). The linear
models based on our data predict anterior pelvic tilt angles between 14-19° across both
accelerative and constant speed running at top speed, which is comparable to the range
reported by previous studies (e.g., 10-20°46 and 15-30°38). We report hip and knee angle
trajectories comparable to Higashihara et al.39 with slightly lower peak values which may arise
because the average top speeds of the sprinters in their study (9.52 m/s) is higher than ours
(7.4 m/s).

We considered only accelerative running where running speed was increased (positive
acceleration). However, hamstring injuries have also been observed in decelerative running22,31.
It is unclear how joint and hamstring mechanics differ between decelerative compared to
accelerative or constant speed running and should be examined in future work.

5. Conclusion
The hamstrings are stretched to greater peak lengths and are subject to greater lengthening
velocities in accelerative than in constant speed running. Greater hip flexion and hip flexion
speed during accelerative running was the primary driver of the greater hamstring lengths and
velocities. Thus, the accelerative nature of running should be considered in addition to running
speed for monitoring athlete exposure to high-risk circumstances. These high-risk
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

circumstances can be mitigated in part by coordinating the hip and knee motions during
accelerative running.

Acknowledgments
This work was supported by the Wu Tsai Human Performance Alliance at Stanford University
and the Joe and Clara Tsai Foundation, and the National Institutes of Health through Grant
P41EB027060. We thank Julie Muccini for her help running experiments.

Competing Interests
AF is the cofounder of Model Health, Inc., which supports the non-academic, commercial use of
the open-source software used for data collection in this study.

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

None to declare.

References
1. Ekstrand J, Hagglund M, Walden M. Injury incidence and injury patterns in professional football: the
UEFA injury study. Br J Sports Med. 2011;45(7):553-558. doi:10.1136/bjsm.2009.060582

2. Edouard P, Branco P, Alonso JM. Muscle injury is the principal injury type and hamstring muscle
injury is the first injury diagnosis during top-level international athletics championships between 2007
and 2015. Br J Sports Med. 2016;50(10):619-630. doi:10.1136/bjsports-2015-095559

3. Lu Y, Pareek A, Lavoie-Gagne OZ, et al. Machine Learning for Predicting Lower Extremity Muscle
Strain in National Basketball Association Athletes. Orthop J Sports Med.
2022;10(7):232596712211117. doi:10.1177/23259671221111742

4. Volpi A, Haselman W, Photopoulos C, Banffy M. Regular-Season Injury Rates in the National

Football League After an Attenuated Preseason Due to COVID-19. Orthop J Sports Med.
2022;10(11):232596712211337. doi:10.1177/23259671221133776

5. Brooks JHM, Fuller CW, Kemp SPT, Reddin DB. Incidence, Risk, and Prevention of Hamstring
Muscle Injuries in Professional Rugby Union. Am J Sports Med. 2006;34(8):1297-1306.
doi:10.1177/0363546505286022

6. Maniar N, Carmichael DS, Hickey JT, et al. Incidence and prevalence of hamstring injuries in field-
based team sports: a systematic review and meta-analysis of 5952 injuries from over 7 million
exposure hours. Br J Sports Med. 2023;57(2):109-116. doi:10.1136/bjsports-2021-104936

7. Ekstrand J, Bengtsson H, Waldén M, Davison M, Khan KM, Hägglund M. Hamstring injury rates have
increased during recent seasons and now constitute 24% of all injuries in men’s professional football:
the UEFA Elite Club Injury Study from 2001/02 to 2021/22. Br J Sports Med. 2023;57(5):292-298.
doi:10.1136/bjsports-2021-105407
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

8. Martin RL, Cibulka MT, Bolgla LA, et al. Hamstring Strain Injury in Athletes: Clinical Practice
Guidelines Linked to the International Classification of Functioning, Disability and Health From the
Academy of Orthopaedic Physical Therapy and the American Academy of Sports Physical Therapy of
the American Physical Therapy Association. J Orthop Sports Phys Ther. 2022;52(3):CPG1-CPG44.
doi:10.2519/jospt.2022.0301

9. Visser H de, Reijman M, Heijboer MP, Bos PK. Risk factors of recurrent hamstring injuries: a
systematic review. Br J Sports Med. 2012;46(2):124-130. doi:10.1136/bjsports-2011-090317

10. Fiorentino NM, Rehorn MR, Chumanov ES, Thelen DG, Blemker SS. Computational Models Predict
Larger Muscle Tissue Strains at Faster Sprinting Speeds. Med Sci Sports Exerc. 2014;46(4):776-786.
doi:10.1249/MSS.0000000000000172

11. Thelen DG, Chumanov ES, Hoerth DM, et al. Hamstring Muscle Kinematics during Treadmill
Sprinting: Med Sci Sports Exerc. 2005;37(1):108-114. doi:10.1249/01.MSS.0000150078.79120.C8

12. Chumanov ES, Heiderscheit BC, Thelen DG. The effect of speed and influence of individual muscles
on hamstring mechanics during the swing phase of sprinting. J Biomech. 2007;40(16):3555-3562.
doi:10.1016/j.jbiomech.2007.05.026

13. Fiorentino NM, Epstein FH, Blemker SS. Activation and aponeurosis morphology affect in vivo
muscle tissue strains near the myotendinous junction. J Biomech. 2012;45(4):647-652.
doi:10.1016/j.jbiomech.2011.12.015

14. Brooks SV, Zerba E, Faulkner JA. Injury to muscle fibres after single stretches of passive and
maximally stimulated muscles in mice. J Physiol. 1995;488(2):459-469.
doi:10.1113/jphysiol.1995.sp020980

15. Mair SD, Seaber AV, Glisson RR, Garrett WE. The Role of Fatigue in Susceptibility to Acute Muscle
Strain Injury. Am J Sports Med. 1996;24(2):137-143. doi:10.1177/036354659602400203

16. Butterfield TA, Herzog W. Quantification of muscle fiber strain during in vivo repetitive stretch-
shortening cycles. J Appl Physiol. 2005;99(2):593-602. doi:10.1152/japplphysiol.01128.2004

17. Danielsson A, Horvath A, Senorski C, et al. The mechanism of hamstring injuries – a systematic
review. BMC Musculoskelet Disord. 2020;21(1):641. doi:10.1186/s12891-020-03658-8

18. Yu B, Liu H, Garrett WE. Mechanism of hamstring muscle strain injury in sprinting. J Sport Health Sci.
2017;6(2):130-132. doi:10.1016/j.jshs.2017.02.002

19. Woods C. The Football Association Medical Research Programme: an audit of injuries in professional
football--analysis of hamstring injuries. Br J Sports Med. 2004;38(1):36-41.
doi:10.1136/bjsm.2002.002352

20. Heiderscheit BC, Hoerth DM, Chumanov ES, Swanson SC, Thelen BJ, Thelen DG. Identifying the
time of occurrence of a hamstring strain injury during treadmill running: A case study. Clin Biomech.
2005;20(10):1072-1078. doi:10.1016/j.clinbiomech.2005.07.005

21. Schache AG, Wrigley TV, Baker R, Pandy MG. Biomechanical response to hamstring muscle strain
injury. Gait Posture. 2009;29(2):332-338. doi:10.1016/j.gaitpost.2008.10.054

22. Kerin F, Farrell G, Tierney P, McCarthy Persson U, De Vito G, Delahunt E. Its not all about sprinting:
mechanisms of acute hamstring strain injuries in professional male rugby union—a systematic visual
video analysis. Br J Sports Med. 2022;56(11):608-615. doi:10.1136/bjsports-2021-104171
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

23. Beato M, Drust B, Iacono AD. Implementing High-speed Running and Sprinting Training in
Professional Soccer. Int J Sports Med. 2021;42(04):295-299. doi:10.1055/a-1302-7968

24. Buckthorpe M, Wright S, Bruce-Low S, et al. Recommendations for hamstring injury prevention in
elite football: translating research into practice. Br J Sports Med. 2019;53(7):449-456.
doi:10.1136/bjsports-2018-099616

25. Kenneally-Dabrowski CJB, Brown NAT, Lai AKM, Perriman D, Spratford W, Serpell BG. Late swing
or early stance? A narrative review of hamstring injury mechanisms during high-speed running.
Scand J Med Sci Sports. 2019;29(8):1083-1091. doi:10.1111/sms.13437

26. Schache AG, Dorn TW, Blanch PD, Brown NAT, Pandy MG. Mechanics of the Human Hamstring
Muscles during Sprinting. Med Sci Sports Exerc. 2012;44(4):647-658.
doi:10.1249/MSS.0b013e318236a3d2

27. Miller RH, Umberger BR, Caldwell GE. Limitations to maximum sprinting speed imposed by muscle
mechanical properties. J Biomech. 2012;45(6):1092-1097. doi:10.1016/j.jbiomech.2011.04.040

28. Chumanov ES, Heiderscheit BC, Thelen DG. Hamstring Musculotendon Dynamics during Stance and
Swing Phases of High-Speed Running. Med Sci Sports Exerc. 2011;43(3):525-532.
doi:10.1249/MSS.0b013e3181f23fe8

29. Haralabidis N, Serrancolí G, Colyer S, Bezodis I, Salo A, Cazzola D. Three-dimensional data-tracking

simulations of sprinting using a direct collocation optimal control approach. PeerJ. 2021;9:e10975.
doi:10.7717/peerj.10975

30. Bramah C, Mendiguchia J, Dos’Santos T, Morin JB. Exploring the Role of Sprint Biomechanics in
Hamstring Strain Injuries: A Current Opinion on Existing Concepts and Evidence. Sports Med.
Published online September 19, 2023. doi:10.1007/s40279-023-01925-x

31. Aiello F, Di Claudio C, Fanchini M, et al. Do non-contact injuries occur during high-speed running in
elite football? Preliminary results from a novel GPS and video-based method. J Sci Med Sport.
2023;26(9):465-470. doi:10.1016/j.jsams.2023.07.007

32. Ekstrand J, Waldén M, Hägglund M. Hamstring injuries have increased by 4% annually in men’s
professional football, since 2001: a 13-year longitudinal analysis of the UEFA Elite Club injury study.
Br J Sports Med. 2016;50(12):731-737. doi:10.1136/bjsports-2015-095359

33. Barnes C, Archer D, Hogg B, Bush M, Bradley P. The Evolution of Physical and Technical
Performance Parameters in the English Premier League. Int J Sports Med. 2014;35(13):1095-1100.
doi:10.1055/s-0034-1375695

34. Nagahara R, Matsubayashi T, Matsuo A, Zushi K. Kinematics of transition during human accelerated
sprinting. Biol Open. 2014;3(8):689-699. doi:10.1242/bio.20148284

35. Pandy MG, Lai AKM, Schache AG, Lin Y. How muscles maximize performance in accelerated
sprinting. Scand J Med Sci Sports. 2021;31(10):1882-1896. doi:10.1111/sms.14021

36. Morin JB, Gimenez P, Edouard P, et al. Sprint Acceleration Mechanics: The Major Role of
Hamstrings in Horizontal Force Production. Front Physiol. 2015;6. doi:10.3389/fphys.2015.00404

37. Rabita G, Dorel S, Slawinski J, et al. Sprint mechanics in world-class athletes: a new insight into the
limits of human locomotion. Scand J Med Sci Sports. 2015;25(5):583-594. doi:10.1111/sms.12389
 bioRxiv preprint doi: https://doi.org/10.1101/2024.03.25.586659; this version posted March 29, 2024. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-ND 4.0 International license.

38. Schuermans J, Van Tiggelen D, Palmans T, Danneels L, Witvrouw E. Deviating running kinematics
and hamstring injury susceptibility in male soccer players: Cause or consequence? Gait Posture.
2017;57:270-277. doi:10.1016/j.gaitpost.2017.06.268

39. Higashihara A, Nagano Y, Ono T, Fukubayashi T. Differences in hamstring activation characteristics

between the acceleration and maximum-speed phases of sprinting. J Sports Sci. 2018;36(12):1313-
1318. doi:10.1080/02640414.2017.1375548

40. Uhlrich SD, Falisse A, Kidziński Ł, et al. OpenCap: Human movement dynamics from smartphone
videos. Marsden AL, ed. PLOS Comput Biol. 2023;19(10):e1011462.
doi:10.1371/journal.pcbi.1011462

41. Lai AKM, Arnold AS, Wakeling JM. Why are Antagonist Muscles Co-activated in My Simulation? A
Musculoskeletal Model for Analysing Human Locomotor Tasks. Ann Biomed Eng. 2017;45(12):2762-
2774. doi:10.1007/s10439-017-1920-7

42. Cao Z, Hidalgo G, Simon T, Wei SE, Sheikh Y. OpenPose: Realtime Multi-Person 2D Pose
Estimation Using Part Affinity Fields. IEEE Trans Pattern Anal Mach Intell. 2021;43(1):172-186.
doi:10.1109/TPAMI.2019.2929257

43. Ben Mansour K, Rezzoug N, Gorce P. Analysis of several methods and inertial sensors locations to
assess gait parameters in able-bodied subjects. Gait Posture. 2015;42(4):409-414.
doi:10.1016/j.gaitpost.2015.05.020

44. Seth A, Hicks JL, Uchida TK, et al. OpenSim: Simulating musculoskeletal dynamics and
neuromuscular control to study human and animal movement. Schneidman D, ed. PLOS Comput
Biol. 2018;14(7):e1006223. doi:10.1371/journal.pcbi.1006223

45. Robertson DGE, Dowling JJ. Design and responses of Butterworth and critically damped digital
filters. J Electromyogr Kinesiol. 2003;13(6):569-573. doi:10.1016/S1050-6411(03)00080-4

46. Mendiguchia J, Castaño-Zambudio A, Jimenez-Reyes P, et al. Can we modify maximal speed

running posture? Implications for performance and hamstring injuries management. Int J Sports
Physiol Perform. 2021;17(3):374-383. doi:doi.org/10.1123/ijspp.2021-0107