Scand J Med Sci Sports 1996: 6: 323-328

C o p j r i g h t 0 Munksgaard 1996

Printed in Denmark ' A l l rights reserved

Scandinavian Journal of
M E D I C I N E & SCIENCE
I N SPORTS
lSSN 0905-7188

Viscoelastic stress relaxation during

static stretch in human skeletal muscle in
the absence of EMG activity
Magnusson SP, Simonsen EB, Dyhre-Poulsen P, Aagaard P, Mohr T, Kjaer M.
Viscoelastic stress relaxation during static stretch in human skeletal muscle in
the absence of EMG activity.
Scand J Med Sci Sports 1996: 6: 323-328.0 Munksgaard, 1996

The present study sought to investigate the role of EMG activity during passive static stretch. EMG and passive resistance were measured during static
stretching of human skeletal muscle in eight neurologically intact control subjects and six spinal cord-injured (SCI) subjects with complete motor loss.
Resistance to stretch offered by the hamstring muscles during passive knee
extension was defined as passive torque (Nm). The knee was passively
extended at 5"/s to a predetermined final position, where it remained stationary for 90 s (static phase) while force and integrated EMG of the hamstring
muscle were recorded. EMG was sampled for frequency domain analysis in a
second stretch maneuver in five control and three SCI subjects. There was a
decline in passive torque in the 90-s static phase for both control and SCI subjects, P<0.05. Although peak passive torque was greater in control subjects,
PxO.05, there was no difference in time-dependent passive torque response
between control (33%) and SCI (38%) subjects. Initial and final 5-s IEMG
ranged from 1.8 to 3.4 pV.s and did not change during a stretch or differ
between control and SCI subjects. Frequency domain analysis yielded similar
results in both groups, with an equal energy distribution in all harmonics,
indicative of 'white noise'. The present data demonstrate that no measurable
EMG activity was detected in either group during the static stretch maneuver.
Therefore, the decline in resistance to static stretch was a viscoelastic stress
relaxation response.

Despite the prevalence of stretching exercises in rehabilitation and sports, the mechanism behind its
proposed effect remains ambiguous. Although
stretching has been demonstrated to bring about
acute and chronic changes in maximal joint range of
motion in humans, the mechanism by which it
works has yet to be clearly demonstrated (1-8).
The acute mechanism of stretching has been explained in neurophysiological terms, which suggests
that the neural component of the target muscle is activated and consequently contributes significant resistance to stretch. Accordingly, the aim of the
stretching is to lower this neural component,
thereby decreasing the resistance to the stretch (9).
Paradoxically, studies on humans show that stretching exercises that produce the greatest improvement

S. P. Magnusson',
E. 8. Simonsen2,
P. Dyhre-Poulsen3, P. Aagaard3,
T. Mohr", M. Kjaer'
'Team Danmark Test Center, Rigshospitalet,
TTA, 'Institute of Medical Anatomy, 31nstitute of
Medical Physiology, Department of
Neurophysiology, 'Institute of Medical
Physiology, Panum Institute, University of
Copenhagen, 5Department of Rheumatology,
Bispebjerg Hospital, Copenhagen Muscle
Research Center, Copenhagen, Denmark

Key words: stress relaxation: EMG activity;

torque; stretch

S. Peter Magnusson, Team Danmark Test Center,

Rigshospitalet, Af. 2001, Blegdamsvej 9, 2100
Copenhagen, Denmark
Tel.: t45 35 45 2801. Fax: t45 35 45 2008
E-mail: P.MagnussonQmfi.ku.dk
Accepted for publication February 22,1996

in joint range of motion yield the largest EMG response (3, 8, 10). Therefore, the contribution and
the role of EMG activity of a muscle undergoing a
stretch remains questionable.
Alternatively, stretching has been described in
biomechanical terms (11). During passive stretch
the muscle-tendon unit is considered to respond
viscoelastically (1 1). Viscoelastic material when
stretched to a new constant length, analogous to the
static stretching technique, will decline in tension
over time. Viscoelastic stress relaxation of the muscle and tendon have been demonstrated in vitro (1 113) and in vivo in human skeletal muscle (14, 15).
Studies on human subjects have concentrated on
EMG activity and joint range of motion during
stretch, while the viscoelastic response has largely

323

Magnusson et al.
unteered to participate in the study (35.5k5.9 years,
78.7k12.9 kg, 1.79f0.07 m). All SCI subjects had
complete paralysis of the lower extremities. Three
subjects were sensory incomplete. The level of spinal cord lesion ranged from C6 to T5. The SCI subjects were 5-22 years (range) post-injury and performed stretching of the hamstring muscle group as
part of their activities of daily living.
Instrumentation

Fig. 1. Test apparatus and position for stretch maneuver. 1) Platform for distal thigh. 2) Thigh strap. 3 ) Pelvic strap. 4) KinCom
ankle attachment with load cell.

been limited to animal investigations. Few investigations have simultaneously examined the EMG response and the resistance to stretch in vivo (14, 15).
A lack of a relationship between EMG response and
viscoelastic stress relaxation during a static stretch
(14, 15), and the apparent incongruity between the
observed EMG response and the most effective
stretching technique (3, 8, 10) shows that the role of
the EMG response during stretching remains unanswered.
Spinal isolation is a useful electrically silent
preparation for studying muscular behavior in the
absence of EMG activity (16). Therefore, humans
who have sustained spinal cord injuries that result in
complete motor loss of the lower extremities may
serve as a model to investigate the effect of stretching in the absence of EMG compared to control subjects. The purpose of the present study was to examine EMG and resistance to stretch during passive
static stretching of human skeletal muscle, in vivo,
in spinal cord-injured (SCI) and neurologically intact control subjects.

Material and methods

Subjects

Eight neurologically intact control subjects volunteered to participate in the study (29.5k3.6 years,
75.1f3.8 kg, 1.79f0.05 m) (meanfSD). These subjects were free of lower extremity or low back pathology at the time of testing. Six SCI subjects vol-

324

Resistance to stretch was defined as the passive

torque (Nm) offered by the hamstring muscle group
during passive knee extension using a KinCom dynamometer (Kinetic Communicator, Chattecx
Corp., Chattanooga, TN) with a modified thigh pad.
All measurements of passive torque were gravity
corrected (17). Subjects were seated with the trunk
perpendicular to the seat for the stretch procedure.
The trunk was stabilized with Velcro straps. The
thigh rested on a specially constructed thigh pad, elevating it to 30 degrees from horizontal (Fig. l). For
the control subjects the trunk and thigh position ensured that none of the subjects tested could reach
maximal range of motion during the stretch maneuver irrespective of individual flexibility, i.e. they
were unable to reach complete knee extension.
Force (N) was detected by the load cell of the dynamometer and torque was calculated. The lever
arm attachment was placed 2 cm proximal to the
malleolus. The distal thigh and pelvis were firmly
secured with straps. The load cell was calibrated
with a 98 Newton load for 90 s to ensure that stress
relaxation of the load cell did not occur.
Prior to administration of the stretch maneuver
the final position for the control subjects during the
stretch maneuver was determined by passively extending the lower leg to a knee extension angle that
provoked a sensation similar to a static stretching
maneuver for the control subjects. Care was taken to
avoid a painful response during determination of final position. For the SCI subjects the final position
was complete knee extension. The leg was then immediately returned to the starting position. For the
stretch maneuver the dynamometer extended the lever arm and knee passively at 5/s (dynamic loading
phase) from the starting position of 70 degrees below horizontal (0 degrees) to the final position (negative values indicate a position above horizontal)
(Fig. 2). In the final position (static phase) the lever
arm remained stationary for 90 s. The stretch
maneuver was applied to the left leg of all subjects.
Control subjects were instructed to relax the leg
throughout the stretch maneuver and not to offer
any voluntary resistance. Previous test-re-test results had shown that resistance to stretch using this

Viscoelastic stress relaxation

Peak passive torque

60 -

r -90

- 50 E

40-

2 30-

.-

$ 2 0 a

phase2

phase1

-20

20

'

40

60

'

80

'

100

time (s)

Fig. 2. Graphic representation of passive torque (solid line) and

knee extension angle (dotted line) during an entire stretch
maneuver. During phase 1 the lever arm of the dynamometer
extended the knee passively at 5"/s from the starting position of
70 degrees below horizontal to the final position. Zero degrees
corresponds to horizontal leg position. During the static phase
(phase 2) the lever arm remains stationary in the final position
for 90 s. Passive torque during the stretch was evaluated in the
static phase: the instance the lever arm reached the final position (peak passive torque) and at the end of the 90-s stretch period (final passive torque).

tion), and stored on a PC for subsequent analysis.

Peak passive torque is the resistance to stretch the
instance the lever arm and lower limb reach the final
position (static phase). Final passive torque is the
resistance to stretch at the end of the static phase.
Delta passive torque is the difference between peak
and final passive torque, expressed as a percentage
of peak passive torque.
Initial 5-s IEMG is the IEMG activity in the first
5 s of the static phase. Final 5-s IEMG is the IEMG
activity in the last 5 s of the static phase. Prior to the
frequency domain analysis the EMG signal for the
second stretch maneuver was digitally highpass filtered at 10 Hz, lowpass filtered at 1000 Hz and detrended for any DC offset. A window function and a
Fast Fourier transform algorithm were employed to
obtain the power density spectrum, which allowed
for frequency domain analysis ( 1 8).
Wilcoxon and Mann-Whitney tests were used for
paired and unpaired statistical analysis (19). An alpha level of Pc0.05 was accepted as significant. Results are reported as meanfSEM.
Results

technique yielded correlation coefficients of ~ 0 . 9 9 Table 1 shows the peak, final and delta passive
with a coefficient of variation of 6.5% (15).
torque during the first stretch maneuver. There was
Gross electrical activity of the human hamstring
a significant decline in passive torque during the 90muscle group was measured with Ag/AgCl surface
s static phase in both the control and SCI subjects
electrodes (Medicotest, Type QN- 10-A, Denmark)
(P<0.05). Peak passive torque differed significantly
placed midway between the gluteal fold and the
between control and SCI subjects (P<0.05). Final
knee joint (14), with a 3-cm inter-electrode distance.
passive torque did not differ between the groups
Custom-made amplifiers with a frequency response
(P=0.09).There was no significant difference in the
of 20 Hz to 10 KHz and 1:l pre-amplifiers were
delta passive torque between the groups.
used for EMG signal sampling. The EMG signal
The initial 5-s IEMG was 1.8k1.2 pV. s for the
was full wave rectified, integrated, averaged (time
control and 2.4f1.5 pV. s for the SCI subjects. The
final 5 s IEMG was 2.9f1.6 pV. s for the control
constant 200 ms) and expressed as pV. s (IEMG)
and
3.4k1.5 pV. s for the SCI subjects. There was
(18).
To examine the EMG signal further during pasno difference between initial and final 5-s IEMG in
sive static stretch, a second stretch maneuver was
either group, nor were there any differences beadministered to the final position in six control and
tween the groups. The frequency domain analysis
three SCI subjects 10 min following the first stretch
for the second stretch is shown in Table 2. MVC for
maneuver. Additionally, the control subjects were
control subjects yielded power density spectrums
instructed to perform a 5-s maximal isometric hamwith a median frequency of 63-80 Hz. The initial
string contraction (MVC) in the final position at the
and final 2 s of the static phase yielded a median freend of this 90-s stretch. During this stretch maneuquency of 400-548 without a characteristic concenver EMG activity was sampled at 2048 Hz in the initial and final 2 s of 90-s static phase and during
MVC.
Table 1.Passive torque (Nm)
Data analysis

Passive force, joint angle, velocity and hamstring

EMG were recorded continuously over the entire
first stretch maneuver. Signals were sampled at 50
Hz, A/D converted (model 2801A; Data Transla-

Peak
Final
Delta (Yo)

Control subjects

SCI subjects

34.2k3.8b
23.2+3.Oa
33.2k2.5

12.2k3.Za
38.752.3

19.7k5.0

MeankSEM. Significantly different from peak passive torque, aP<0.05.

Significantly different from SCI subjects, bP<0.05.

325

Magnusson et al.
Table 2.Power density spectrum
Median frequency

Control subjects
A
8
C
D
E
SCI subjects
A
8
C

(Hz)

Initial 2 s

Final 2 s

MVC 2 s

542
546
514
526
400

531
548
507
514
477

65
77
80
68
63

540
513
500

501
478
488

Power density spectrum with median frequency (Hz) from the initial and final
2 s during the static phase of the second stretch for the control and SCI subjects. In addition the median frequency from the MVC in the control subjects is
shown.

tration of frequencies in the spectrum, and strikingly

low power. This equal distribution of energy in all
harmonics is characteristic of white noise. Corresponding analysis in the SCI subjects yielded similar results with median frequencies of 478-540 Hz
(Fig. 3).
The torque produced during MVC was 83.3k17.2
Nm with a corresponding EMG amplitude of
331.6k49.6 pV. s. The peak passive torque represented 5 3 f l l % of the torque produced during
MVC. The average EMG amplitude of the initial 5 s
during the stretch represented 0.45+0.32% of the
EMG amplitude during MVC.
The final position was sig icantly different between control (-3.5k4.5 deg, ;s) and SCI subjects
(-30.7fl .O degrees), P<O.Ol. A difference existed
in age between the groups, P<0.05, but not in height
or weight.

previous studies (14, 15) have reported IEMG values ranging from 76 to 194 pV. s. Furthermore, the
data show that the low level IEMG activity of the
human hamstring muscle does not change during
the 90-s static phase, which is in agreement with
previous findings using the same stretching technique (15). A significant decline in IEMG over a 45s static phase using a straight leg raise to a maximum tolerated joint range of motion has been
shown (14). It is possible that the disagreements between the studies are related to the definition of the
end point. Alternatively, the positioning and the stabilization may have influenced the results.
Further examination of the EMG signal by way of
frequency domain analysis during the MVC in the
control subjects demonstrated an accustomed response with a median frequency of 63-80 Hz. On
the other hand, the power density spectrum for both
the SCI and control subjects during the stretch was
consistent with white noise. This is evidence that
no physiologically meaningful EMG activity was

500000
400000

1I
400

800

1200

5000,

4000

3000
2000

1I

1000

Discussion

The aim of the present study was to examine

whether EMG plays a role in viscoelastic stress relaxation during passive static stretching in controls
and SCI subjects with complete motor loss of the
lower extremities. The main finding of this study
was that control and SCI subjects exhibited a 33%
and a 38% decline in passive torque despite the absence of any measurable EMG activity. In other
words, viscoelastic stress relaxation, a material response, was observed during a static stretch in both
groups.
Few studies have examined the influence of EMG
on resistance to stretch during the static phase of a
static stretch (14, 15). The present study demonstrated very low level 5-s IEMG during the static
stretch, ranging from 1.8 to 3.4 pV. s. In contrast,

326

400

5000

4000

2000
3000

1000

!I

800

1200

400

800

1200

Frequency (Hz)

Fig. 3. Frequency domain analysis obtained from the power

density spectrum during (a) MVC in a control subject, (b) initial
2 s of the static phase in a control subject, (c) initial 2 s of the
static phase in a SCI subject. Note the difference in power in the
MVC in contrast to initial 2 s of the static phase in control and
SCI subjects.

Viscoelastic stress relaxation

20

za,

18

16

,$

14

12

20

40

60

80

100

Time (s)

I
20

20

40

60

80

100

Time (s)

Fig. 4. Graphic representation of the passive torque (Nm) during the 90-s static phase of a static stretch in a SCI subject (a)
and in a control subject (b). Note the similar initial steep and
subsequent gradual relaxation component in the control and
SCI subjects despite the difference in peak passive torque.

recorded during the static phase for the control and

SCI subjects. With surface electrodes it cannot be
ruled out that motor units situated deep within the
muscle are active and therefore not registered. It has
been shown that to produce a torque comparable to
the passive torque decline requires an EMG activity
at least 10 times that of the initial EMG activity
(20). Further, since during the stretch the EMG amplitude was 0.4% and peak torque was 53% of that
during MVC, it is unlikely that EMG activity contributed significantly to passive peak torque. Accordingly, the time-dependent passive torque response was unlikely to have been influenced by neural input. It is widely believed that the various
stretching techniques, excluding ballistic stretching,
are designed to minimize EMG activity, thereby increasing its effectiveness (9). However, the role of
the EMG response during a stretch has never been
firmly established. It has recently been suggested
that EMG and passive torque remain unaffected by
proprioceptive neuromuscular facilitation (PNF),
but that stretch perception is modified (21). The
present data provide evidence that a tangible decline
(33-38%) in passive torque occurs over a 90-s static
phase in the absence of any EMG activity. This confirms that stretching of the human muscle-tendon
unit affects its material properties.
The present data have demonstrated viscoelastic

behavior in human muscle, in vivo, in response to

static stretch. Viscoelastic behavior implies a combination of viscous properties, where deformation is
rate dependent, and elastic properties, where deformation is load dependent (22). Viscoelastic material
when stretched to a new constant length, analogous
to the static stretching technique, will decline in tension in a non-linear fashion over time (11, 12). In
the present study it can be seen (Fig. 3) that in the
absence of EMG activity passive torque declined in
a non-linear fashion. Similar to data on whole muscle and collagen (12, 13) the non-linear stress relaxation in the present study had an initial steep component with a subsequent more gradual relaxation
component. Although previously unconfirmed, it
has been suggested that the decline in resistance to
stretch in humans is largely mechanical in nature
(15). The present study corroborates that the decline
in resistance to stretch, undiluted of EMG, in human
muscle in vivo is viscoelastic stress relaxation.
Peak passive torque was greater in control than in
SCI subjects and a similar trend existed for final
passive torque. It should be noted that the SCI subjects were brought to a greater joint range of motion
(final position), which results in a greater resistance
for a given individual. It is possible that the lower
passive torque in SCI subjects may reflect plastic
deformation of the hamstring muscle-tendon unit as
a result of the repetitive hamstring elongation required during activities of daily living. Alternatively, the difference in resistance to stretch may
represent atrophic changes in the properties of the
muscle-tendon unit from chronic immobilization in
SCI subjects. In the present study the circumference
of the thigh in the SCI subjects was visually markedly less than that of the controls. It has been suggested that cross-sectional area contributes to resistance to stretch (23). Despite the difference in absolute passive torque the comparable stress relaxation
response is noteworthy. Furthermore, qualitatively
the initial steep and subsequent gradual relaxation
component appears similar in the control and SCI
subjects (Fig. 4), which is further evidence that the
human muscle response to static stretching is a material property.
Since the main observation of the present study
was viscoelastic stress relaxation response in the absence of any measurable EMG activity, the structures responsible for the response is speculative. It
has been suggested that perimysium is the principal
component in the parallel elastic component, and its
chief function is to distribute stress evenly and to
prevent over-stretching (24, 25). The endomysium
has been suggested to be of significance in the series
elastic component, and its chief function is to transfer force from the contractile component to the ten-

327

Magnusson et al.
don and bone in series (24). It is likely that the
present stretch maneuver placed tension on both the
parallel and series elastic components. From an injury standpoint, conceivably, the viscoelastic stress
relaxation is an acute adaptation of the parallel elastic component to lower the imposed load across the
myotendinous junction, where injury is known to
occur (26-28).
In summary, the present study examined EMG
and resistance to stretch during passive static
stretching of human skeletal muscle, in vivo, in control subjects and SCI subjects with complete loss of
motor control in their lower extremity. During the
90-s static phase of the stretch, resistance declined
significantly in both control and SCI subjects. Peak
passive torque was significantly greater in control
subjects; however, the decline over the 90-s period
was similar for control (33%) and SCI (38%) subjects. No measurable EMG activity was detected in
either group during the stretch maneuver. Therefore,
the decline in resistance to static stretch is viscoelastic stress relaxation.
References
1. Etnyre BR, Lee JA. Chronic and acute flexibility of men
and women using three different stretching techniques. Res
tQ 1988: 59: 222-8.
2. Guissard N, Duchateau J, Hainaut K. Muscle stretching and
motoneuron excitability. Eur J Appl Physiol 1988: 58: 4752.
3. Moore MA, Hutton RS. Electromyographic investigation
of muscle stretching techniques. Med Sci Sports Exerc
1980: 12: 322-9.
4. Cornelius WL, Ebrahim K, Watson J, Hill DW. The effects
of cold application and modified PNF stretching techniques
on hip joint flexibility in college males. Res Q Exerc Sport
1992: 63: 311-14.
5. Cornelius WL, Hinson MM. The relationship between isometric contractions of hip extensors and susbquent flexibility in males. J Sports Med 1980: 20: 75-9.
6. Prentice WE. A comparison of static stretching and PNF
stretching for improving hip joint flexibility. Athletic Training 1983: Spring: 56-9.
7. Sady SP, Wortman M, Blanke D. Flexibility training: ballistic, static or proprioceptive neuromuscular facilitation?
Arch Phys Med Rehabil 1982: 63: 261-3.
8. Osternig LR, Robertson R, Troxel R, Hansen P. Muscle activation during proprioceptive neuromuscular facilitation
(PNF) stretching techniques. Am J Phys Med 1987: 66:
298-307.
9. Hutton RS. Neuromuscular basis of stretching exercise. In:
Komi PV, ed. Strength and power in sports. Oxford: Blackwell Scientific Publications, 1993: 29-38.
10. Osternig LR, Robertson RN, Troxel RK, Hansen P. Differ-

328

ential responses to proprioceptive neuromuscular facilitation (PNF) stretch techniques. Med Sci Sports Exerc 1990:
22: 106-111.
11. Taylor DC, Dalton JD, Jr., Seaber AV, Garrett WEJ. Viscoelastic properties of muscle-tendon units. The biomechanical effects of stretching. Am J Sports Med 1990: 18: 300-9.
12. Abbott BC, Lowy J. Stress relaxation in muscle. Proc R
SOCLond 1956: 146B: 281-8.
13. Sanjeevi R. A viscoelastic model for the mechanical properties of biological materials. J Biomech 1982: 15: 107-9.
14. McHugh MP, Magnusson SP, Gleim GW, Nicholas JA. Viscoelastic stress relaxation in human skeletal muscle. Med
Sci Sports Exerc 1992: 24: 1375-82.
15. Magnusson SP, Simonsen EB, Aagaard P, Gleim GW,
McHugh MP, Kjaer M. Viscoelastic response to repeated
static stretching in human skeletal muscle. Scand J Med Sci
Sport 1995: 5 : 342-7.
16. Roy RR, Baldwin KM, Edgerton VR. The plasticity of
skeletal muscle: effects of neuromuscular activity. Exerc
Sport Sci Rev 1991: 19: 269-312.
17. Beyer N, Voigt M. Passive knee joint resistance and its influence on gravity correction in isokinetic dynamometry.
FIMS 7th European Sports Medicine Congress 1993: 5-10
October : 182 (abstr). Nicosia, Cyprus: FIMS.
18. Basmajian JV, DeLuca CJ. Muscles alive. Their functions
revealed by electromyography, 5th edn. Baltimore, MD:
Williams & Wilkins, 1985: 94-7.
19. Conover WJ. Practical nonparametric statistics, 2nd edn.
New York: John Wiley & Sons, 1980.
20. Magnusson S, Simonsen EB, Aagaard P, Montz U, Kjaer
M. Contraction specific changes in passive torque in human
skeletal muscle. Acta Physiol Scand 1995: 155: 377-86.
21. Magnusson SP, Simonsen EB, Aagaard P, Dyhre-Poulsen P,
McHugh MP, Kjaer M. Mechanical and physiological responses to stretching with and without preisometric contraction in human skeletal muscle. Arch Phys Med Rehab
1996: 77: 373-8.
22. LeVeau BF. Biomechanics of human motion, 3rd edn. Philadelphia: W.B. Saunders Company, 1992: 29-59.
23. Chleboun GS, Howell JN, Giesey JJ, Brownfield D, Cassidy R, Cramer H, Hoyng A, McDonald D, Siwik D. An
analysis of two phases of the passive elastic stiffness of intact human muscle. Second World Congress of Biomechanics 1994: 76 (abstr). Amsterdam, the Netherlands.
24. Borg TK, Caulfield JB. Morphology of connective tissue in
skeletal muscle. Tissue Cell 1980: 12: 197-207.
25. Purslow PP. Strain-induced reorientation of an intramuscular connective tissue network: implications for passive
muscle elasticity. J Biomech 1989: 22: 21-31.
26. Garrett WE, Jr., Safran MR, Seaber AV, Glisson RR, Ribbeck BM. Biomechanical comparison of stimulated and
nonstimulated skeletal muscle pulled to failure. Am J
Sports Med 1987: 15: 448-54.
27. Noonan TJ, Best TM, Seaber AV, Garrett WEJ. Identification of a threshold for skeletal muscle injury. Am J Sports
Med 1994: 22: 257-61.
28. Taylor DC, Dalton JD, Jr., Seaber AV, Garrett WEJ. Experimental muscle strain injury. Early functional and structural
deficits and the increased risk for reinjury. Am J Sports
Med 1993: 21: 190-194.