Clinical Biomechanics 17 (2002) 705712

www.elsevier.com/locate/clinbiomech

Eect of hamstrings muscle action on stability of the

ACL-decient knee in isokinetic extension exercise
Takashi Yanagawa a, Kevin Shelburne a,*
, Frank Serpas b, Marcus Pandy b

a
Steadman-Hawkins Sports Medicine Foundation, Vail, CO 81657, USA
b
Department of Biomedical Engineering, University of Texas, Austin, TX 78712, USA
Received 8 November 2001; accepted 1 August 2002

Abstract
Objective. To quantify the eect of hamstrings muscle action on stability of the anterior cruciate ligament decient knee during
isokinetic exercise at various speeds.
Design. Mathematical modeling and forward-dynamics computer simulation were used to study the interactions between knee-
extension speed, hamstrings co-contraction activity, and anterior tibial translation in the intact and anterior cruciate decient knee.
Background. There is much experimental evidence available to believe that hamstrings co-contraction can reduce anterior tibial
translation in the anterior cruciate decient knee. Little is known, however, about the level of hamstrings activation needed to keep
anterior tibial translation within normal limits during functional activity.
Methods. Isokinetic knee-extension was simulated with a sagittal-plane model used previously to study load sharing between the
muscles, ligaments, and bones during isometric knee-extension exercise, isokinetic exercise, and squatting exercise.
Conclusions. Some amount of hamstrings activation is needed to stabilize an anterior cruciate decient knee irrespective of how
fast the knee extends. The level of hamstrings co-contraction needed to stabilize an anterior cruciate decient knee is inversely
related to extension speed. Hamstrings co-contraction is more eective in reducing anterior tibial translation than low-resistance
extension exercise.

Relevance
Excessive anterior tibial translation during knee-extension exercise may lead to damage of the meniscus and other passive
structures inside the knee. If anterior cruciate decient patients can be trained to co-contract their hamstrings during isokinetic knee-
extension, then this exercise is appropriate for maintaining strength of the thigh muscles without compromising the anterior stability
of the knee.
2002 Elsevier Science Ltd. All rights reserved.

Keywords: Hamstrings co-contraction; Knee modeling; Rehabilitation exercise

1. Introduction quadriceps muscle strength [14]. Even though weight-

bearing exercises are emphasized in most ACL rehabil-
The primary function of the anterior cruciate liga- itation protocols, isokinetic knee-extension exercise is
ment (ACL) is to limit anterior translation of the tibia still commonly used in the eight weeks to six months
relative to the femur. Injury to the ACL often results in post-injury and post-operative periods [14]. Unfortu-
loss of thigh muscle strength, so much so that open- nately, knee-extension exercise causes relatively large
chain (i.e., knee-extension) and closed-chain (i.e., weight forces to be transmitted to the ACL in the intact knee,
bearing) exercises are routinely prescribed to maintain and yields much larger-than-normal anterior tibial
translation (ATT) when the ACL is absent [57]. Some
studies have found that the risk of injury to passive
structures such as the menisci, which provide secondary
*
Corresponding author. Address: 181 West Meadow Drive, Suite
restraint to ATT, is greatly increased [810]. ACL in-
1000, Vail, CO 81657, USA. suciency may also lead to degenerative changes inside
E-mail address: kevin.shelburne@shsmf.org (K. Shelburne). the knee consistent with osteoarthritis [11,12]. As a
0268-0033/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 8 - 0 0 3 3 ( 0 2 ) 0 0 1 0 4 - 3
706 T. Yanagawa et al. / Clinical Biomechanics 17 (2002) 705712

result, rehabilitative methods have been sought which (3) Which factor has a greater eect on ATT in the
limit ATT in the ACL-decient knee during isokinetic ACL-decient knee, knee-extension speed or ham-
knee-extension to levels present in the intact joint strings co-contraction force?
[1,13,14]. (4) In the range of extension speeds currently prescribed
A large number of in vivo [1520], in vitro [5,21], and for rehabilitation, what level of hamstrings activa-
modeling studies [6,2224] have shown that hamstrings tion is needed to stabilize an ACL-decient knee?
co-contraction can reduce ATT in the ACL-decient
knee. Hamstrings activation may reduce ATT since Mathematical modeling and forward-dynamics com-
these muscles insert on the back of the tibia, and they puter simulation were used to address these questions. A
may therefore apply a posterior pull to the leg. Imran forward-dynamics simulation approach is necessary to
and OConnor [23] used a two-dimensional model to study the relationship between hamstrings muscle acti-
study the eectiveness of hamstrings co-contraction vation and ATT because the level of hamstrings acti-
force on ATT during isometric extension. Their calcu- vation (and ultimately hamstrings co-contraction force)
lations showed an inverse relationship between ATT and could be controlled more directly in the model. Also,
hamstrings force, with hamstrings co-contraction being high muscle forces generated during isokinetic extension
least eective in reducing ATT near extension. These exercise makes the use of in vitro measurement methods
ndings are well supported by experimental measure- impractical for determining the dependence of ATT on
ments reported by Hirokawa et al. [21]. Relatively little quadriceps and hamstrings force.
is known, however, about the amount of hamstrings
activation needed to keep ATT within normal limits
during functional activity. Liu and Maitland [24] esti- 2. Methods
mated the amount of hamstrings activation needed to
stabilize the ACL-decient knee during level walking; The model presented in Shelburne and Pandy [27] has
however, their analysis was quasi-static and their nd- been expanded to permit simulation of an isokinetic
ings apply only to a single instant of the gait cycle. To knee-extension activity conducted in a sitting position
our knowledge, no study has reported on the eect of with the hip exed to 60. The only structures not xed
hamstrings muscle action on stability of the ACL-de- to the ground were those distal to the femur. The lower
cient knee during dynamic rehabilitation exercises like leg was modeled as a rigid body consisting of a shank
isokinetic extension. and a foot; the ankle was held xed in the neutral
It is also not clear whether hamstrings co-contraction (standing) position throughout the activity. Because all
is more eective in reducing ATT than is low-resistance motion was conned to the sagittal plane, the lower legs
extension exercise (i.e., knee-extension at high speeds). orientation relative to the thigh was dened by three
Wilk and Andrews [25] measured ATT in ACL-decient generalized coordinates (Fig. 1(A)): anteriorposterior
patients for speeds ranging from 60 to 300 deg/s. They translation (q1 ), proximaldistal translation (q2 ), and the
reported a linear drop in ATT as knee-extension speed angle between the long axis of the femur and the long
increased, but no comparison of the eect of hamstrings axis of the tibia (i.e., the knee-exion angle) (q3 ). Tibio-
co-contraction was made. femoral contact, which is assumed to be rigid and
The main objective of the present study was to frictionless, was enforced via a holonomic constraint.
investigate the eect of hamstrings muscle action on The tibial plateau, modeled as a at surface sloping 8
stability of the ACL-decient knee during isokinetic posteriorly, made single-point contact with a surface
exercise at various speeds. The analysis was based on represented by a two-dimensional quintic spline tted to
a sagittal-plane model used previously to study load the mid-parasagittal section of the lateral femoral con-
sharing between the muscles, ligaments, and bones dyle (Fig. 1(A)).
during isometric knee-extension [6,2628], isokinetic The anterior and posterior cruciate ligaments, the
exercise [7], and squatting [22]. The limit of knee-joint medial and lateral collateral ligaments, and the posterior
stability was dened as the peak ATT calculated for capsule were modeled by a total of 11 elastic bundles,
maximum isometric contractions of the quadriceps in each consisting of a straight-line segment joining the
the normal knee in the absence of hamstrings co-con- three-dimensional attachment sites obtained from ex-
traction. Four specic questions were addressed: periment (Fig. 2(A)). The force developed in each liga-
ment bundle varied nonlinearly with ligament length
(1) What is the relationship between ATT and speed for and linearly with ligament velocity (the rate of change of
isolated contractions of the quadriceps during iso- ligament length); parameters assumed for the model
kinetic exercise? ligaments are given in Serpas et al. [7].
(2) What is the relationship between ATT and ham- The motion of the lower leg was inuenced by the
strings muscle activation during isokinetic knee- action of 11 musculotendinous units (Fig. 1(B)). Each
extension? musculotendinous unit was modeled as a three element
 T. Yanagawa et al. / Clinical Biomechanics 17 (2002) 705712 707

Fig. 1. (A) The orientation of the lower leg was determined by three
generalized coordinates: anteriorposterior translation (q1 ) and prox-
imaldistal translation (q2 ) were measured in a tibia-xed frame rela-
tive to the approximate center of curvature of the posterior femoral
condyles; knee exion angle (q3 ) was dened as the angle between the
long axes of the femur and tibia. (B) The model simulates an isokinetic
knee-extension activity conducted in a sitting position with the hip
exed to 60 (i.e., the angle between the femur and tibia is 120). The
ankle is held xed in the neutral (standing) position throughout Fig. 2. (A) The passive structures of the tibiofemoral joint were
the activity. Three musclesrectus femoris (RF), vastus medialis modeled by a total of eleven elastic bundles, each consisting of a
(VMED), and vastus intermedius/lateralis (VINT)were maxi- straight-line segment joining the three-dimensional attachment sites.
mally activated for the duration of the activity. Muscles exing the The line segments shown represent the anterior (1) and posterior (2)
kneebiceps femoris long head (BFLH) and short head (BFSH), bundles of the ACL, the anterior (3) and posterior (4) bundles of the
semimembranosus (MEM), semitendinosus (TEN) and gastrocnem- posterior cruciate ligament, the anterior (5), intermediate (6), and
iuswere activated as described in the text. The three remaining posterior (7) bundles of the supercial medial collateral ligament, the
musclesgracilis, tensor fasciae latae (iliotibial tract), and sartorius anterior (8) and posterior (9) bundles of the deep medial collateral
contributed only passive forces during the simulations; these muscles ligament, the lateral collateral ligament (10), and the posterior capsule
are not shown in the diagram. (11). (B) Gastrocnemius, semimembranosus, and semitendinosus were
constrained to follow the shortest distance along appropriately placed
via cylinders. This requirement ensured both that the actuator forces
were applied in the proper orientation and that the muscle lengths were
muscle in series with a linearly elastic tendon. The calculated correctly. Gastrocnemius (GAS) originates at a, wraps
around one of the cylinders, and touches the tibia at b. Semi-
mechanical behavior of muscle was represented by a Hill-
membranosus (MEM) wraps around the same cylinder and inserts at
type contractile element that modeled muscles force c. Semitendinosus (TEN) wraps around a dierent cylinder and
lengthvelocity property, a series-elastic element that inserts at d.
modeled muscles active stiness, and a parallel-elastic
element that modeled muscles passive stiness. For the
duration of each simulated exercise, the quadriceps was F_ MT f lMT ; vMT ; F MT ; a; where 0 6 a 6 1 1
fully activated; all other muscles except the hamstrings
were fully deactivated, thus supplying only passive force. A straight, origin-to-insertion line of action was em-
The activation level of the hamstrings was xed at either ployed for each muscle, except where the muscle con-
0%, 25%, 50%, or 100%. (0% represented complete de- tacted and wrapped around bone or another muscle.
activation and 100% represented maximum activation.) The paths of the gastrocnemius, semimembranosus, and
For each musculotendinous unit, muscle activation (a), semitendinosus muscles were constrained to follow the
musculotendon length (lMT ), musculotendon velocity shortest distance along appropriately placed via cylin-
(vMT ), musculotendon force (F MT ), and the time rate of ders (Fig. 2(B)). The paths of the quadriceps muscles
change in musculotendon force (F_ MT ) were related by a across the patellofemoral joint required special atten-
nonlinear, rst-order, dierential equation of the form: tion. As with the tibiofemoral joint, rigid and frictionless
708 T. Yanagawa et al. / Clinical Biomechanics 17 (2002) 705712

single-point contact was maintained between a at sur- at the ankle, and the net torque required to maintain the
face (the patella facet) and a two-dimensional quintic machine arm motion at the designated isokinetic speed.
spline (tted to the mid-parasagittal section of the Requiring tibiofemoral contact and prescribing the
femoral groove) (see Fig. 1(A)). The patella (which was motion of the machine arm eectively eliminated 2 of
treated as massless) experienced a proximal force from the 3 degrees of freedom, leaving ATT as the only in-
the quadriceps tendon and a distal force from the pa- dependent generalized coordinate. ATT present during
tellar tendon. The patellar tendon was treated as an the simulations was a function of the net shear force
inextensible element. Static equilibrium dictates that the imparted to the lower leg by the muscles, the distal re-
magnitude and direction of the aforementioned forces straint, tibiofemoral contact, and inertia. Gravity was
and the orientation of the patella be related by a single turned o in the dynamical model to avoid the need
nonlinear equation that can be solved iteratively. Details to account for torque caused by the weight of the lower
of the model used to simulate patellofemoral mechanics leg, and to simulate the conditions under which iso-
are given in Appendix A of Shelburne and Pandy [27]. kinetic knee-extension exercises are actually performed.
The parameters assumed for the model muscles are
given by Serpas et al. [7].
The lower leg was distally constrained by its con- 3. Results
nection to the massless arm of a simulated isokinetic
exercise machine. This machine arm rotated about an ATT was inversely related to extension speed in both
axis coincident with the approximate center of curvature the intact and ACL-decient knee (Fig. 3(A) and (B)).
of the posterior femoral condyles. The interface between The eect of speed on ATT was uniform whether or
the lower leg and the machine consisted of a massless not the ACL was present (Fig. 4). For example, the
cu connected to the arm by a pin joint located 43 cm
from the arms rotation axis. This distance was chosen
to ensure that the restraining force was applied near the
ankle. The cu was free to translate relative to the lower
legs long axis, subject to a small degree of damping.
Equations of motion for the system were obtained
through the use of a computer software package called
SD/Fast. For a given simulation, the angular velocity of
the machine arm was prescribed in SD/Fast with an
inverse tangent function having asymptotes at 0 deg/s
(for the start of the activity) and the desired isokinetic
speed. The function was chosen such that the accelera-
tion to isokinetic speed took place between knee-exion
angles of 105 and 100, before the range to be studied
(905) was reached. Knee-exion angles less than 5
were not included in the simulated range because, in (A)
practice, people do not bring their knees to full exten-
sion during isokinetic exercise.
The dynamical equations of motion for the muscu-
loskeletal knee model can be represented in the form:
x_ f x; t 2
where the state vector contains 28 elements: 3 general-
ized coordinates, q, which dene the position and ori-
entation of the leg relative to the thigh; 3 generalized
speeds, q_ , dened as the time rate of change of the
generalized coordinates; 11 musculotendon forces, F MT ;
and 11 muscle activations, a. Simulations were run by
choosing an isokinetic speed of 30, 90, 180, 300, or 360
deg/s and integrating the equations using a Runge (B)
Kutta, xed-step integrator. Choosing the activations
Fig. 3. (A) ATT calculated for the ACL-intact knee without ham-
and prescribing the motion of the machine arm gave the
strings co-contraction for isokinetic knee-extension at 0, 30, 90, 180,
time-history of the generalized coordinates, the muscu- and 300 deg/s. (B) ATT calculated for the ACL-decient model knee
lotendon forces, the ligament forces, the tibiofemoral without hamstrings co-contraction for isokinetic knee-extension at 0,
and patellofemoral contact forces, the restraining force 30, 90, 180, and 300 deg/s.
 T. Yanagawa et al. / Clinical Biomechanics 17 (2002) 705712 709

(A)

Fig. 4. Peak ATT versus isokinetic speed without hamstrings co-con-

traction for the ACL-intact and ACL-decient knee model. Also
shown are data from Wilk and Andrews [25] for ACL-decient test
subjects.

dierence in ATT between the intact and ACL-decient

knee for isometric extension and for extension at 360
deg/s was around 6.5 mm (compare 0 and 360 deg/s in
Fig. 4). Furthermore, no matter how fast the knee was
extended, peak ATT occurred at 10 of exion for the
ACL-decient knee, and ATT dropped to zero at
around 70 of exion in both the intact and ACL-de-
cient knee (Fig. 3(A) and (B), 10 exion and 7075 (B)
exion, respectively). The uniform drop in peak ATT
with increasing speed found for the model ACL-de- Fig. 5. (A) Peak ATT versus hamstrings co-contraction level for iso-
cient knee is consistent with experimental measurements kinetic knee-extension at 0, 30, 90, 180, and 300 deg/s in the ACL-
intact knee. (B) Peak ATT versus hamstrings co-contraction level for
reported by Wilk and Andrews [25] (Fig. 4). isokinetic knee-extension at 0, 30, 90, 180, and 300 deg/s in the ACL-
ATT was inversely related to hamstrings co-contrac- decient knee. The horizontal line drawn at 7.5 mm represents the
tion level at all extension speeds (Fig. 5). For all speeds peak ATT calculated for isometric contractions of the quadriceps
in the intact joint, ATT dropped by 5 mm when ham- without hamstrings co-contraction; this is the amount of ATT used to
strings was fully activated (compare ATT at 0% and dene the limit of knee-joint stability in the ACL-decient model.
100% hamstrings activation at all speeds in Fig. 5(A)).
In the ACL-decient knee, by comparison, ATT de-
creased by roughly 12 mm at all speeds when the ham- remain stable; in contrast, when the model knee was
strings were fully turned on (compare ATT at 0% and extended at a rate of 360 deg/s, only 10% of maximum
100% hamstrings activation at all speeds in Fig. 5(B)). hamstrings activation was needed to limit the amount of
The decline in ATT was more rapid at lower levels of ATT and to keep the knee stable (compare 0 and 360
hamstrings activation than at higher levels, irrespective deg/s in Fig. 6).
of knee-extension speed (compare Fig. 5(A) and (B) at
025% of hamstrings activation).
Some amount of hamstrings activation was needed at 4. Discussion
all speeds to stabilize the model ACL-decient knee.
Although peak ATT decreased as extension speed in- Perhaps the most signicant limitation of the analysis
creased (Fig. 4), the decrease was insucient to prevent is that the model itself is comparatively simple. The
excessive ATT (knee instability), even at the highest model knee accounts only for relative movements of the
speeds of extension (Fig. 6, unstable region spans all bones in the sagittal plane. Furthermore, the model
speeds up to 360 deg/s). However, relatively low levels of bones are assumed to be rigid, so that there is only point
hamstrings activation were needed to stabilize the ACL- contact between the femur and tibia and between the
decient knee at the highest speeds. For isometric ex- femur and patella. Other simplications incorporated in
tension, the hamstrings must be activated at least 35% of the model include modeling the tibial plateau and pa-
maximum in order for the model ACL-decient knee to tellar facet as at surfaces; neglecting the inuence of the
710 T. Yanagawa et al. / Clinical Biomechanics 17 (2002) 705712

However, when the medial meniscus is excised subse-

quent to sectioning the ACL, anterior laxity increases
considerably [30]. This nding suggests that the menis-
cus plays an important role in providing anterior sta-
bility to the ACL-decient knee. To compensate for the
absence of a meniscus in the model, the stiness of the
secondary ligaments were adjusted until anterior laxity
in the ACL-decient model matched in vitro data re-
ported in the literature (see Fig. 4(A) in [6]). The cal-
culated values of ATT for the ACL-decient model are
therefore presumed to approximate the values of ATT
present in the real ACL-decient knee for isokinetic
extension exercise (see also Fig. 5(B) in [6]).
The aforementioned limitations notwithstanding, the
Fig. 6. Amount of hamstrings activation needed to stabilize the ACL- predictions of the model are consistent with trends re-
decient knee in isokinetic extension exercise. The limit of knee-joint ported previously by others. Nisell et al. [31], Kaufman
stability was dened as the peak ATT calculated for maximum iso- et al. [32], Baltzopoulos [33], Escamilla et al. [34], and
metric contractions of the quadriceps (without hamstrings co-con- Toutoungi et al. [35] all found peak tibial shear forces to
traction) in the ACL-intact knee (see horizontal line drawn at 7.5 mm
be less than 1 body weight (BW) for isokinetic knee-
in Fig. 5(B)).
extension exercises ranging in speed from 30 to 300 deg/s
in the intact knee. There is also a consensus among these
menisci and cartilage; assuming that the knee ligaments studies that the magnitude of the peak tibial shear force
and muscle tendons are elastic structures; approximat- decreases as extension speed increases. For example,
ing the paths of the knee ligaments as straight lines, and Nisell et al. [31] and Baltzopoulos [33] predicted peak
neglecting the eects of contact between the ligaments tibial shear forces of around 0.9 BW when the knee
and bones; and nally, assuming that the external re- extended at 30 deg/s, and peak shear forces ranging from
straining force acts at a single point on the leg. Many of 0.7 to 0.8 BW when knee-extension speed exceeded 200
these limitations have previously been discussed and deg/s. For the model used in this study, peak tibial shear
justied [7,27,28]. forces decreased from 0.7 BW at 30 deg/s to 0.4 BW
The model bones were conned to move in the sag- when the knee extended at 300 deg/s.
ittal plane, which excludes axial rotation experienced by The calculated values of ATT also agree with in vivo
the shank during extension of the knee in vivo. The ef- measurements of ATT obtained for the intact and ACL-
fect of this rotation is to decrease the force borne by the decient knee. Kizuki et al. [36] found peak ATT for
ACL, which means that the model predictions of ATT knee-extension performed at 30 deg/s to be around 10
may be somewhat higher than those experienced in the mm for the intact knee and around 15 mm for the ACL-
living knee. The reasoning is as follows. Markolf et al. decient knee. The model calculations show peak ATT
[29] (see Fig. 4 in their paper) found that both external at 30 deg/s for the intact and ACL-decient knee to be
and internal torques applied to the tibia increase ACL 6.8 and 13.5 mm, respectively. The model also correctly
force near full extension. Now, as the intact knee ex- predicts that peak ATT should occur somewhere be-
tends, the tibia rotates externally relative to the femur; tween 10 and 20 of exion in both the intact and ACL-
at full extension, the tibia can be rotated by as much as decient knee. Wilk and Andrews [25] measured ATT in
20 externally. This means that if the tibia was con- vivo for isokinetic knee extension at speeds ranging
strained to move in the sagittal plane, an internal torque from 60 to 300 deg/s. Their results showed that ATT
would have to be applied to the tibia in order to bring it decreased linearly as extension speed increased in the
to the neutral position at full extension. Markolfs re- ACL-decient knee, with peak ATT being 16.1 mm
sults show that internal torques increase ACL force. when the knee was extended at 60 deg/s, and decreasing
Thus, the eect of constraining the tibia to move in the to 11 mm when the knee was extended at a much faster
sagittal plane is to increase ACL force, which means rate of 300 deg/s. Fig. 4 indicates that the model simu-
that a sagittal-plane model of the knee would over- lations reproduce the uniform decrease in ATT with
estimate ACL force. Because an increase in ACL force is increasing knee-extension speed, and that the rate of this
ultimately the result of an increase in ATT, we may decrease is also consistent with that found from exper-
conclude that a sagittal-plane model of the knee over- iment (compare slopes of experimental and model data
estimates ATT as well. for the ACL-decient knee).
The results of in vitro experiments performed on the The inverse relationship between ATT and extension
ACL-intact knee indicate that anteriorposterior laxity speed for both the intact and ACL-decient knee is ex-
does not change much when the menisci are removed. plained by the forcevelocity property of the quadriceps
 T. Yanagawa et al. / Clinical Biomechanics 17 (2002) 705712 711

muscles in the model. ATT decreased as extension speed co-contraction is more eective than low-resistance ex-
increased (Fig. 4) because the forces developed by the tension exercise for the conservative treatment of an
quadriceps are a function of speed. As extension speed ACL-decient knee. The model simulations show that a
increases, the contraction velocities of the quadriceps 1 mm reduction in ATT requires only a 5% increase in
increase, so that (for a given exion angle) the resultant hamstrings activation, but that extension speed must be
force transmitted to the quadriceps tendon (and thus to increased by roughly 100 deg/s before a similar reduc-
the patellar tendon) decreases, and the anterior shear tion in ATT can be obtained (compare slopes of curves
force applied to the leg is lower [7]. for ACL-decient knee in Figs. 4 and 5). Given that
The inverse relationship between peak ATT and lower speeds yield higher quadriceps forces, these nd-
hamstrings activation is explained by the ability of these ings suggest that for quadriceps strengthening it is more
muscles to apply a posterior pull to the tibia at all benecial to exercise at lower extension speeds, provided
exion angles, except near extension. Peak ATT de- that the hamstrings can be contracted with sucient
creases as hamstrings activation increases because the force to keep ATT within an allowable limit (Fig. 6).
posterior shear force provided by these muscles over- Finally, we found that although peak ATT decreased
whelms the anterior shear forces supplied by the patellar- as extension speed increased, some amount of ham-
tendon force and the tibiofemoral compressive force. strings co-contraction was always needed to stabilize the
Even at small exion angles (between 10 and 15), for model ACL-decient knee (Fig. 6). If a rehabilitation
which the hamstrings insert steeply on the back of the program is to include isokinetic exercise as an adjunct
tibia, co-contraction of these muscles is still eective in to thigh muscle strengthening, it will be important to
reducing ATT because hamstrings develop their maxi- ensure that the ACL-decient patient is capable of co-
mum forces when the knee is near extension [27]. contracting his/her hamstrings during this exercise;
Hamstrings co-contraction decreases peak ATT in otherwise, excessive ATT over time may cause damage
both the intact and ACL-decient knee, but for any to the menisci and other passive structures, resulting
given speed the eect is greater when the ACL is absent. in further degeneration of the joint. Hamstrings co-
The results of Fig. 5 show that activating the model contraction is also critical for the patient with an ACL-
hamstrings to 30% of its maximum value reduces ATT reconstructed knee, because excessive ATT in this case
by 6.5 mm in the ACL-decient knee, compared with will risk damage to the immature graft and/or xation.
only 2.5 mm in the intact knee (compare ATT for 030% Some studies have shown that anterior knee stability can
hamstrings activation in Fig. 5(A) and (B)). Hamstrings be improved in an ACL-decient population by training
co-contraction is more eective in reducing peak ATT these patients to alter their muscle recruitment patterns
when the ACL is absent because the shear force applied during activity [37,38]. Further research is needed to
by the patellar tendon is smaller in this case [6]. The understand the eects of changes in muscle coordination
patellar-tendon shear force is smaller because the angle versus muscle strength training on ATT in the ACL-
between the patellar tendon and the tibial long axis is decient knee.
smaller, which occurs because tibiofemoral translation is
greater in the ACL-decient knee.
Despite the emphasis on weight-bearing exercises in
References
most ACL rehabilitation protocols, the use of isokinetic
knee-extension exercises to train or assess quadriceps [1] Irrgang JJ. Modern trends in anterior cruciate ligament rehabil-
muscular strength is still common in the eight weeks to itation: nonoperative and postoperative management. Clin Sports
six months post-injury and post-operative periods [14]. Med 1993;12:797813.
Thus, it is clinically relevant to determine the eect of [2] Mangine RE, Noyes FR. Rehabilitation of the allograft recon-
hamstrings muscle action and isokinetic speed on the struction. J Orthop Sports Phys Ther 1992;15:294302.
[3] Shelbourne KD, Nitz P. Accelerated rehabilitation after anterior
stability of the ACL-decient knee during isokinetic cruciate ligament reconstruction. J Orthop Sports Phys Ther
exercise. First, we found, in agreement with experi- 1992;15:25664.
mental results reported by others [25], that slow-speed [4] Wilk KE, Andrews JR. Current concepts in the treatment of
isokinetic exercise (30 deg/s) produces higher peak ATT anterior cruciate ligament disruption. J Orthop Sports Phys Ther
1992;15:27993.
values than fast speeds (300 deg/s and above). Clinically,
[5] Li G, Rudy TW, Sakane M, Kanamori A, Ma CB, Woo SL. The
this nding suggests that high-speed or low-resistance importance of quadriceps and hamstring muscle loading on knee
extension exercise can be used during the early phases of kinematics and in-situ forces in the ACL. J Biomech 1999;32:395
a rehabilitation training program for the ACL-recon- 400.
structed knee, at least until adequate graft maturation [6] Pandy MG, Shelburne KB. Theoretical analysis of ligament and
and strength have occurred. extensor-mechanism function in the ACL-decient knee. Clin
Biomech 1998;13:98111.
Secondly, if isokinetic extension protocols are to be [7] Serpas F, Yanagawa T, Pandy MG. Forward-dynamics simula-
designed to increase joint stability by reducing ATT, tion of knee-ligament forces developed during isokinetic dyna-
then the results of Figs. 4 and 5 indicate that hamstrings mometry. Comput Meth Biomech Biomed Eng 2002;5:3343.
712 T. Yanagawa et al. / Clinical Biomechanics 17 (2002) 705712

[8] Caborn DN, Johnson BM. The natural history of the anterior [23] Imran A, OConnor JJ. Control of knee stability after ACL injury
cruciate ligament-decient knee. A review. Clin Sports Med or repair: interaction between hamstrings contraction and tibial
1993;12:62536. translation. Clin Biomech 1998;13:15362.
[9] McCarroll JR, Shelbourne KD, Patel DV. Anterior cruciate [24] Liu W, Maitland ME. The eect of hamstring muscle compen-
ligament injuries in young athletes. Recommendations for treat- sation for anterior laxity in the ACL-decient knee during gait.
ment and rehabilitation. Sports Med 1995;20:11727. J Biomech 2000;33:8719.
[10] Mizuta H, Kubota K, Shiraishi M, Otsuka Y, Nagamoto N, [25] Wilk KE, Andrews JR. The eects of pad placement and angular
Takagi K. The conservative treatment of complete tears of the velocity on tibial displacement during isokinetic exercise. J Orthop
anterior cruciate ligament in skeletally immature patients. J Bone Sports Phys Ther 1993;17:2430.
Joint Surg Br 1995;77:8904. [26] Shelburne KB, Modeling the mechanics of the intact and
[11] Suter E, Herzog W, Leonard TR, Nguyen H. One-year changes in reconstructed knee. Ph.D. dissertation in Mechanical Engineering.
hind limb kinematics, ground reaction forces and knee stability The University of Texas at Austin, 1997.
in an experimental model of osteoarthritis. J Biomech 1998;31: [27] Shelburne KB, Pandy MG. A musculoskeletal model of the knee
5117. for evaluating ligament forces during isometric contractions.
[12] Noyes FR, Mooar PA, Matthews DS, Butler DL. The symptom- J Biomech 1997;30:16376.
atic anterior cruciate-decient knee. Part I: the long-term func- [28] Pandy MG, Shelburne KB. Dependence of cruciate-ligament load-
tional disability in athletically active individuals. J Bone Joint ing on muscle forces and external load. J Biomech 1997;30:101524.
Surg Am 1983;65:15462. [29] Markolf KL, Gorek JF, Kabo JM, Shapiro MS. Direct measure-
[13] Maltry JA, Noble PC, Woods GW, Alexander JW, Feldman GW, ment of resultant forces in the anterior cruciate ligament. An
Tullos HS. External stabilization of the anterior cruciate ligament in vitro study performed with a new experimental technique.
decient knee during rehabilitation. Am J Sports Med 1989;17: J Bone Joint Surg Am 1990;72:55767.
5504. [30] Levy IM, Torzilli PA, Warren RF. The eect of medial menisc-
[14] Kvist J, Karlberg C, Gerdle B, Gillquist J. Anterior tibial ectomy on anterior-posterior motion of the knee. J Bone Joint
translation during dierent isokinetic quadriceps torque in ante- Surg Am 1982;64:8838.
rior cruciate ligament decient and nonimpaired individuals. [31] Nisell R, Ericson MO, Nemeth G, Ekholm J. Tibiofemoral joint
J Orthop Sports Phys Ther 2001;31:415. forces during isokinetic knee extension. Am J Sports Med
[15] Steele JR, Roger GJ, Milburn PD. Tibial translation and 1989;17:4954.
hamstring activity during active and passive arthrometric assess- [32] Kaufman KR, An KN, Litchy WJ, Morrey BF, Chao EY.
ment of knee laxity. The Knee 1995;1:21723. Dynamic joint forces during knee isokinetic exercise. Am J Sports
[16] Baratta R, Solomonow M, Zhou BH, Letson D, Chuinard R, Med 1991;19:30516.
DAmbrosia R. Muscular coactivation. The role of the antagonist [33] Baltzopoulos V. Muscular and tibiofemoral joint forces during
musculature in maintaining knee stability. Am J Sports Med isokinetic concentric knee extension. Clin Biomech (Bristol, Avon)
1988;16:11322. 1995;10:20814.
[17] Kalund S, Sinkjaer T, Arendt-Nielsen L, Simonsen O. Altered [34] Escamilla RF, Fleisig GS, Zheng N, Barrentine SW, Wilk KE,
timing of hamstring muscle action in anterior cruciate ligament Andrews JR. Biomechanics of the knee during closed kinetic chain
decient patients. Am J Sports Med 1990;18:2458. and open kinetic chain exercises. Med Sci Sports Exerc 1998;30:
[18] Osternig LR, Caster BL, James CR. Contralateral hamstring 55669.
(biceps femoris) coactivation patterns and anterior cruciate [35] Toutoungi DE, Lu TW, Leardini A, Catani F, OConnor JJ.
ligament dysfunction. Med Sci Sports Exerc 1995;27:8058. Cruciate ligament forces in the human knee during rehabilitation
[19] Solomonow M, Baratta R, Zhou BH, Shoji H, Bose W, Beck C, exercises. Clin Biomech 2000;15:17687.
DAmbrosia R. The synergistic action of the anterior cruciate [36] Kizuki S, Shirakura K, Kimura M, Fukasawa N, Udagawa E.
ligament and thigh muscles in maintaining joint stability. Am J Dynamic analysis of anterior tibial translation during isokinetic
Sports Med 1987;15:20713. quadriceps femoris muscle concentric contraction exercise. The
[20] Walla DJ, Albright JP, McAuley E, Martin RK, Eldridge V, El- Knee 1995;2:1515.
Khoury G. Hamstring control and the unstable anterior cruciate [37] Sinkjaer T, Arendt-Nielsen L. Knee stability and muscle coordi-
ligament-decient knee. Am J Sports Med 1985;13:349. nation in patients with anterior cruciate ligament injuries: An
[21] Hirokawa S, Solomonow M, Luo Z, Lu Y, DAmbrosia R. electromyographic approach. J Electromyograph Kinesiology
Muscular co-contraction and control of knee stability. J Elect 1991;1:20917.
Kinesiol 1991;3:199208. [38] Hewett TE, Lindenfeld TN, Riccobene JV, Noyes FR. The eect
[22] Shelburne KB, Pandy MG. Determinants of cruciate-ligament of neuromuscular training on the incidence of knee injury in
loading during rehabilitation exercise. Clin Biomech 1998;13:403 female athletes. A prospective study. Am J Sports Med 1999;27:
13. 699706.