Clinical Biomechanics 20 (2005) 97104

www.elsevier.com/locate/clinbiomech

Altered hamstring-quadriceps muscle balance in patients with knee

osteoarthritis
Tibor Hortobagyi
a

a,*

, Lenna Westerkamp a, Stacey Beam a, Jill Moody a, Joseph Garry b,

Donald Holbert c, Paul DeVita a

Biomechanics Laboratory, East Carolina University, 332A Ward Sports Medicine Building, Greenville, NC 27858, USA
b
Departments of Family Medicine, East Carolina University, Greenville, NC 27858, USA
c
Departments of Biostatistics, East Carolina University, Greenville, NC 27858, USA
Received 19 March 2004; accepted 18 August 2004

Abstract
Background. To compare hamstring to quadriceps muscle coactivity during level walking, stair ascent, and stair decent between
individuals with and without knee osteoarthritis.
Methods. In a cross-sectional study, subjects with grade II knee osteoarthritis (n = 26), healthy age- and gender-matched (n = 20)
and healthy, young adults (n = 20) performed three activities of daily living. During the stance phase of these activities surface electromyography was measured. Two coactivity ratios were computed, the biceps femoris to vastus lateralis ratio and the ratio of the
biceps femoris EMG activity relative to the EMG activity measured during contraction- and velocity-specic maximal voluntary
biceps femoris contraction, i.e., biceps femoris to maximal biceps femoris activity.
Findings. Subjects with knee osteoarthritis had signicantly higher coactivity than age-matched healthy adults and young adults
and healthy adults had more coactivity than young adults regardless the type of coactivity ratio. The biceps femoris to vastus lateralis ratio yielded 25% higher coactivity value than the biceps femoris to maximal biceps femoris ratio (P < 0.0001). The EMG
activity of the vastus lateralis relative to maximal vastus lateralis EMG activity was 92% in subjects with knee osteoarthritis,
57% in age-matched controls, and 47% in young adults (P < 0.0001).
Interpretation. Patients with knee osteoarthritis revealed increased hamstring muscle activation while executing activities of daily
living. Altered muscle activation at the knee may interfere with normal load distribution in the knee and facilitate disease progression. Therapeutic interventions should focus not only on quadriceps strengthening but also on improving muscle balance at the
knee.
2004 Elsevier Ltd. All rights reserved.
Keywords: Hamstring coactivity; Antagonistic muscle activity; Gait; ADL; Aging

1. Introduction
Human gait is the result of force production by coordinated activation of muscle groups. In normal gait
agonist and antagonist muscle pairs at each lower
extremity joint contract in an alternating pattern with
low levels of concurrent activity between the main acti*

Corresponding author.
E-mail address: hortobagyit@mail.ecu.edu (T. Hortobagyi).

0268-0033/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.clinbiomech.2004.08.004

vation bursts (Grasso et al., 2000; Ivanenko et al., 2004;

McFayden and Winter, 1988). In the intact knee joint,
activation of the agonist quadriceps muscle generates
an anterior shear force on the tibia relative to the femur
and activation of the antagonistic hamstring muscle
group counteracts this force, producing joint stability
(Bernardi et al., 1997; Ebenbichler et al., 1998; Hagood
et al., 1990; Solomonow et al., 1987). When the quadriceps is activated to generate power in activities of daily
living (ADLs), central command concurrently also

T. Hortobagyi et al. / Clinical Biomechanics 20 (2005) 97104

98

activates the hamstrings (Hansen et al., 2002) and spinal

reexes, including reciprocal inhibition, modulate the
amount of coactivity.
Knee osteoarthritis (OA) is associated with reduced
knee joint stability due to impaired quadriceps strength,
pain, and an altered joint structure (Hurley, 2003;
McAlindon et al., 1993; OReilly et al., 1997; Sharma,
2001). These factors lead to a diminished force output
from one muscle group on one side of the knee and create joint instability (Sharma, 2001). To maintain joint
function and reduce instability, subjects with knee OA
must generate compensatory muscle activity. Indeed,
the pattern of joint torques in ADLs reveals an enhanced torque generation at the hip and reduced torque
production at the knee in OA compared with healthy
adults (DeVita et al., 2002; Kaufman et al., 2001; Pai
et al., 1994). It is thus reasonable to hypothesize that
muscle activation pattern at the knee is altered in knee
OA so that OA subjects execute ADLs with heightened
hamstring muscle coactivity compared with healthy
adults. This hypothesis has not been tested in the handful of studies that used electromyography (EMG) to
examine muscle activation patterns in diverse aspects
of knee OA (Brucini et al., 1981; Hinman et al., 2002;
Marks et al., 1994; Wretenberg et al., 1993). The clinical
implication of this hypothesis is the need to use muscleenhancing exercises in rehabilitation that target not only
quadriceps weakness but also improve the control of the
quadriceps and hamstrings and the balance between
them. The purpose of the study was to compare hamstring muscle coactivity between subjects with and without knee OA during ADLs.

2. Methods
This is a non-randomized casecontrol study. We
compared the muscle activation patterns between patients with unilateral knee OA and age- and gendermatched group of old adults and a gender-matched
group of young adults. These data are part of a large
study that aims to determine how low- and high-inten-

sity exercise produces positive changes in pain and function and modies gait kinematics, kinetics, and muscle
activation in individuals with knee OA.
2.1. Subjects and design
Table 1 shows the characteristics of the 66 subjects included in this cross-sectional study. OA subjects (n = 26)
were recruited through referrals from family physicians
and advertisements in local newspapers. Potential subjects were rst interviewed on the telephone and those
who appeared to meet the inclusion criteria were interviewed in person. Using a detailed medical questionnaire, one of the investigators conrmed that the
participants, except for knee OA, were ostensibly
healthy and free from known neuromuscular-skeletal illnesses or injuries. All knee OA subjects had their diagnosis conrmed by a physician and were included with
grade II or higher bilateral knee OA in the tibiofemoral
compartment (Kelgren and Lawrence, 1957). OA subjects exhibited chronic and stable knee pain, radiographic signs of hypertrophic changes, marginal spur
formation, subchondral sclerosis or cyst formation,
non-uniform joint space narrowing, and had diculty
while rising from a chair, and ascending or descending
stairs. Exclusion criteria were medical conditions that
precluded safe participation (e.g. heart disease, stroke,
insulin dependent diabetes, osteoporosis), rheumatoid
arthritis, using medication that caused dizziness, unstable medication schedule, history of falls or other motor
decits, severe recent modications of diet, inability to
walk up and down a ight of stairs, corticosteroid injection within the past 30 days, inability to comprehend
and follow instructions, and an apparent lack of commitment to participation. The OA participants were
remarkably healthy except for knee OA.
Table 1 also shows the characteristics of the healthy
comparison group (n = 20). Except for knee OA, knee
pain, and poorer mobility, the subjects in this comparison group met the same exclusion and inclusion criteria
as the OA patients. To control for the eect of age on
muscle coactivity, we also included a group of healthy

Table 1
Subject characteristics
Variable

Age, years

Height, m

Mass, kg

Mean

SD

Mean

SD

Mean

SD

OA subjects
Men
Women
Healthy
Men
Women
Young
Men
Women

26
7
19
20
7
13
20
10
10

58.4
62.4
56.9
58.5
65.4
58.5
21.4
22.3
20.4

8.8
13.9
8.8
10.1
10.7
10.1
1.9
1.3
2.0

1.67
1.75
1.64
1.70
1.75
1.67
1.73
1.78
1.67

0.08
0.08
0.07
0.08
0.10
0.05
0.08
0.05
0.07

87.1
96.9
83.4
68.9
81.0
62.4
70.1
74.4
65.8

18.3
17.7
17.5
12.3
11.9
6.0
14.1
12.2
15.1

T. Hortobagyi et al. / Clinical Biomechanics 20 (2005) 97104

young adults (n = 20). All subjects signed a written informed consent form that was approved by the University Institutional Review Board.
2.2. Testing procedures
2.2.1. General testing protocol
Subjects visited the laboratory on 3 days. On their
rst visit, subjects were interviewed, familiarized with
the testing equipment and environment, and signed the
consent form. On the second visit surface EMG activity
from specic muscles were recorded during level walking, stair ascent, and stair descent. On the third visit
EMG activity was recorded during maximal isokinetic
knee extension and exion. The order of these test sessions was alternated between subjects. Each testing session lasted about 1 h. Twelve OA subjects were re-tested
to assess reliability of the dependent variables. For these
subjects, the test 1 data were used in the analyses.
Subjects dressed in a T-shirt and form-tting, spandex cycling shorts. Height and weight were measured
and in OA subjects the level of knee pain was recorded.
The skin over the bula head, vastus lateralis, the biceps
femoris, tibialis anterior, and gastrocnemius lateralis
was shaved and alcohol washed on the involved side in
OA and dominant side in healthy subjects. Two singleuse diagnostic ECG electrodes (ConMed Inc., Utica,
NY, USA) were attached to each muscle belly with a
2.5 cm center-to-center inter-electrode distance to detect
surface EMG activity. The ground electrode was placed
on the skin over the bula head. The electrode cables
were placed inside the spandex shorts to avoid interference with movement of the leg.
As a general warm-up, subjects rode a bicycle ergometer at 60 RPM for 5 min at 0.52.0 kg resistance and
performed 3 min of lower extremity stretching. Measurements were performed on the leg for which the subject
reported more pain (OA group) or on the dominant
leg (control groups), determined by kicking a ball.
2.2.2. Knee pain
Subjects verbally rated their knee pain in each leg
immediately before testing on a 5-point scale (0 = no
pain, 1 = slightly painful, 2 = moderately painful,
3 = very painful, 4 = excruciatingly painful) during level
walking, walking up stairs, walking down stairs, and rising from a chair. The mean of the pain scores recorded
on the two testing days was the dependent measure.
2.2.3. ADLs
EMG activity was measured during level walking,
stair ascent, and stair descent. For level walking, participants walked through the experimental area for several
minutes until they were relaxed and comfortable. A
starting point was selected so that the more painful foot
(OA) or the dominant foot (control subjects) would con-

99

tact the force platform in a normal stride. Subjects of

both groups walked at 1.3 m/s on 20-m level walkway
that was instrumented with a force platform (AMTI ,
Model LG6-4-1, Newton, MA, USA) and an infrared
timing system (Lafayette Instruments, model 63520,
Lafayette, IN, USA) to measure the time required for
each subject to walk a 4-m interval over the platform.
Five successful trials were collected for each subject.
Subjects ascended and descended on a custom-built
four-step wooden stairway that had a force platform
(model OR6-6, AMTI ) embedded in the center of the second step. We have previously described the details of
this set-up and protocol (Hortobagyi et al., 2003). The
stairway was equipped with railings, but the data reported here do not include trials during which the subjects touched the railings. During ascent and descent,
subjects synchronized their foot contact with the beat
of a metronome set at 80 beats per minute. This frequency produced a slow to moderate pace that was comfortably performed by all subjects. Subjects stepped on
the force platform with their involved leg or the dominant leg (controls). As a warm-up, subjects walked up
and down the four-step stairway for 1 min. Five successful trials were collected for each subject. There were
5 min of seated rest between ADLs and no participants
reported fatigue nor requested rest while being tested
for a specic ADL. We used the force platforms to
measure the vertical ground reaction forces to identify
the stance phase of gait for EMG analysis.
2.2.4. Maximal EMG activity
To normalize EMG activity measured during ADLs,
subjects performed unilateral maximal voluntary isokinetic contractions on a dynamometer (Kin-Com,
500H, Chattecx, Inc., Chattanooga, TN, USA). Subjects
sat on the dynamometer seat with a knee and hip joint
angle of 90 and arms folded in front of the chest. The
anatomical zero was set at a knee angle of 180. Two
crossover shoulder harnesses, a lap belt, a thigh strap,
and an ankle cu limited extraneous movements of the
upper body and the test leg. The transverse axis of
the knee joint was aligned with the transverse axis of
the dynamometers power shaft. The length of the lever
arm was individually determined. Subjects performed
maximal eort concentric and eccentric contractions
with the quadriceps and hamstrings at 90/s. We selected
this velocity because preliminary data suggested that the
knee joint angular velocity during ADLs would vary between 120 and 120/s. The data analysis of the maximal EMG activity is provided in the next section.
2.3. Data analysis
We collected EMG data with the TeleMyo telemetric
hardware system (Noraxon USA, Inc., Scottsdale, AZ,
USA). As detailed previously Hortobagyi and DeVita,

100

T. Hortobagyi et al. / Clinical Biomechanics 20 (2005) 97104

2000; Hortobagyi et al., 2003), this system contains a

battery-powered transmitter tied to the subjects waist
with a cloth belt so that the subject can freely move
while performing the ADLs. The dierential ampliers
have a gain of 2000, an input impedance of 10 MX,
and a common mode rejection ratio of 130 dB. The
EMG signals have a bandwidth of 3 dB at 16 to
500 Hz. The root-mean-square (RMS) of the direct
EMG signal was obtained by using a 20-ms smoothing
window. The peak RMS EMG value (in mV) was determined between two cursors; the rst cursor was set
200 ms before heel strike and the second cursor was set
at toe-o. During the isokinetic contractions the peak
RMS EMG value was also determined.
One hamstring coactivity ratio was computed as the
quotient of biceps femoris RMS EMG activity divided
by vastus lateralis RMS EMG activity multiplied by
100, abbreviated as BF/VL. A second hamstring coactivity ratio was computed as the quotient of biceps femoris
RMS EMG activity divided by the maximal biceps femoris EMG activity measured during maximal voluntary
isokinetic contractions and multiplied by 100, abbreviated as BF/BFmax. We decided to compute both of these
ratios because the BF/BFmax ratio may be necessary to
correctly interpret the BF/VL ratio. A misleadingly high
BF/VL ratio can occur, especially in OA, due to poor
quadriceps activation and not due to improper activation of the hamstrings. To compute BF/BFmax in level
walking and stair descent, we used peak RMS EMG
value measured during maximal concentric voluntary
contraction of the hamstrings. To compute BF/BFmax
in stair ascent, we used peak RMS EMG value measured during maximal eccentric voluntary contraction
of the hamstrings. These ratios are widely used in the literature but it is unknown if they give similar values for
muscle coactivity (Bernardi et al., 1997; De Vito et al.,
2003; Hortobagyi and DeVita, 2000; Solomonow
et al., 1987). Fig. 1 shows representative examples of direct EMG recordings during level walking, stair ascent,
and stair descent in an OA and a healthy adult subject.
Vastus lateralis EMG activity during ADLs was also
expressed as its normalized value measured during maximal isokinetic eccentric (level walking and stair descent)
and concentric (stair ascent) voluntary contraction of
the quadriceps. The quadriceps activation ratio is abbreviated as VL/VLmax. Finally, we also determined the
gastrocnemius lateralis coactivity relative to the activity
of tibialis anterior (GA/TA) during level walking to see
if the hypothesized modications in coactivity are specic to the knee joint or are also present during ankle
plantar exion.
We also expressed EMG coactivity by computing the
average EMG activity during the stance phase of the
three ADLs and normalizing these values to maximal
EMG activity measured during maximal voluntary contractions. There was an r = 0.82 correlation between the

Fig. 1. Representative trials of direct surface EMG activity of the

vastus lateralis (upper tracing in each panel) and biceps femoris
(middle tracing) and the vertical ground reaction force (lower tracing)
during level walking (A, B), stair ascent (C, D), and stair descent (E, F)
in one subject without (A, C, E) and one with knee OA (B, D, F). Note
the virtual reversal of vastus lateralis and biceps femoris EMG activity
in the subject with knee OA during level walking (B) compared with
the healthy control subject (A). Note the heightened (A, C, E)
compared with the diminished activity of the vastus lateralis in the
subject with knee OA (B, D, F). Increased biceps femoris muscle
coactivity is present in all three tasks in the OA subject. The analysis of
the EMG data was done in a window that started 200 ms before heel
strike (rst upward deection in the force signal) and ended at toe-o
(return of the force signal to baseline). Vertical calibration bars are
0.25 mV for EMG and 1 kN for force. Horizontal calibration bar is
200 ms.

coactivity values computed from the peak EMG and

from the average EMG data, suggesting that the two
methods give essentially the same results. Therefore,
we report here only the coactivity data computed from
peak EMG activity.
2.4. Statistical analyses
Reliability of the dependent variables was estimated
with intraclass correlation coecients. The key analysis
was a Group (OA, healthy, young) by Task (level walking, ascent, descent) by Ratio (BF/VL, BF/BFmax), a 3
by 3 by 3, analysis of variance with repeated measures
on Task and Ratio. A Tukeys post hoc contrast was
used to determine signicant dierences between the
mean values obtained. The signicance for the statistical
tests was set at P < 0.05.

3. Results
Reliability intraclass correlation coecient of BF/VL
and BF/BFmax was R = 0.79 for level walking, R = 0.94

T. Hortobagyi et al. / Clinical Biomechanics 20 (2005) 97104

for stair ascent, and R = 0.78 for stair descent. The test
retest mean dierences in coactivity ratio were 1%
(level walking), 4% (stair ascent), and 3% (stair descent). On a scale from 0 to 4, OA patients reported pain
of 1.83 (SD 0.29). These data indicate that the OA subjects in this study had pain levels similar to those in previously reported studies (Al-Zahrani and Bakheit, 2002;
Hurley, 2003; McAlindon et al., 1993; OReilly et al.,
1997). Immediately after testing, knee pain was 1.62
(SD 0.16), suggesting that the testing did not increase
pain compared with rest. Healthy adults and young subjects did not report any knee pain.
Table 2 shows the BF/VL and BF/BFmax coactivity
ratios during level walking, stair ascent, and stair descent. The top portion of Table 2 shows the cell means
that make up the Group by Task by Ratio 3-way interaction. This interaction was not signicant (F = 1.4, P =
0.3294). There was a signicant Ratio main eect
(F = 13.7, P < 0.0001), indicating that the BF/VL ratio
yielded 25% higher coactivity value than the BF/BFmax

101

ratio. The signicant Group by Ratio interaction

(F = 11.7, P = 0.0073) revealed that OA had higher
coactivity than healthy and healthy had higher coactivity than young subjects regardless the type of ratio.
Using the BF/VL ratio, OA had 1.6-fold greater coactivity than healthy and healthy had also 1.6-fold greater
coactivity than young subjects (all P < 0.05, Tukeys
post hoc). Using the VL/VLmax ratio, OA had 1.9-fold
greater coactivity than healthy and healthy had 1.5-fold
greater coactivity than young adults.
Table 3 shows the VL/VLmax data. There was no signicant Group by ADLs interaction but there was a signicant group main eect (F = 39.6, P = 0.0001). The
VL/VLmax was almost 2-fold greater in OA than in
controls. On the average, OA subjects executed ADLs
at about 90% of maximal VL EMG activity. The GA/
TA ratio was statistically higher in OA (78, SD 50,
Tukeys post hoc, P < 0.05) than in healthy adults
(38%, SD 33) and young adults (25%, SD 12) and the
ratio was also higher (P < 0.05) in healthy adults than

Table 2
Group data of the coactivity ratios
Task

Level walking

Stair ascent

Stair descent

All tasks

Group

BF/VL

OA
Healthy
Young
OA
Healthy
Young
OA
Healthy
Young
OA
Healthy
Young
All groups

BF/BFmax

Mean

SD

Mean

SD

105
64
46
48
31
14
78
49
31
77a
48b
30

39
19
23
19
14
11
42
19
18
33
17
18

72
25
17
58
36
26
57
34
23
62a
32b
22

47
12
9
36
21
8
45
20
11
43
18
9

52

23

39c

23

Values are percent biceps femoris to vastus lateralis (BF/VL) and biceps femoris EMG activity measured during ADL relative to maximal EMG
activity measured during a maximal voluntary contraction (BF/BFmax). The Group by ADL by Ratio 3-way interaction was not signicant (top
portion). The Group and the Ratio main eect were both signicant (bottom portion). OA = subjects with knee osteoarthritis.
a
Compared with Healthy and Young without knee OA (P < 0.05).
b
Compared with Young (P < 0.05).
c
Compared with BF/VL ratio (P < 0.05).

Table 3
Group data for vastus lateralis EMG activity
Task

Level walking
Stair ascent
Stair descent
All tasks

OA

Healthy

Young

Mean

SD

Mean

SD

Mean

SD

75
118
83
92*

44
51
32
42

35
71
64
57

22
36
26
27

21
74
48
47

16
29
36
26

Values are percent, expressing the vastus lateralis EMG activity measured during ADLs relative to the EMG activity measured during a maximal
voluntary contraction. See Section 2.3 more details. OA, osteoarthritis; Healthy, healthy adults matched for age with OA; Young, young adults.
*
Signicantly dierent from Healthy and Young (P < 0.05).

102

T. Hortobagyi et al. / Clinical Biomechanics 20 (2005) 97104

in healthy young adults (one-way ANOVA, F = 8.3,

P = 0.0010).

4. Discussion
The main nding in this study was that hamstring
muscle coactivity during ADLs was greater in OA subjects compared with age- and gender-matched healthy
adults and young adults. We also found that subjects
with knee OA performed ADLs with a substantially
higher relative EMG activation of the quadriceps muscle
than control subjects. Thus, a new nding was that subjects with knee OA execute ADLs with an altered muscle
activation pattern compared with healthy adults.
It is well documented that knee OA is associated with
quadriceps weakness (Hurley, 2003; McAlindon et al.,
1993; OReilly et al., 1997; Slemenda et al., 1997). One
consequence of a reduced contribution of torque from
the agonist muscles to the net knee joint torque is a decrease in knee joint stability (Solomonow et al., 1987).
Indeed, most but not all (Al-Zahrani and Bakheit,
2002) studies found that OA subjects ambulate on level
surface and stairs with less knee and more hip exion
and reduced knee and increased hip torques (DeVita
et al., 2002; Kaufman et al., 2001; Pai et al., 1994). When
healthy adults walk with more hip exion the EMG
activity of the hip extensors increase (Grasso et al.,
2000). The present results for all three ADLs agree with
this activation pattern. We observed 1.6 (BF/VL) and
1.9-fold (BF/BFmax) higher hamstring muscle coactivity
in OA compared with healthy adults and 2.6 and 2.8fold more coactivity compared with young adults. The
functional interpretation of these data is that OA subjects increase hip extensor muscle activity in a compensation for impaired quadriceps function and shift the
locus of power generation to the hip joint unaected
by OA (DeVita et al., 2002; Pai et al., 1994). Additional
roles of an increased hamstring muscle activity in knee
OA are to stabilize the knee by increasing the compressive force, improve the eccentric control of loading,
and increase ligament protection during stance phase
(Bernardi et al., 1997; Ebenbichler et al., 1998; Hagood
et al., 1990).
The hamstrings are two-joint muscles and act as exors at the knee and function as extensors at the hip. It is
dicult to separate hip and knee function of the hamstrings in gait. The current analysis revealed that the
hamstring activity was heightened in OA at least
200 ms before heel strike and through the duration of
the stance phase, the entire duration of the analysis window. This increased hamstrings coactivity is probably
used as much in positioning and accelerating the limb
prior to heel strike as it is in supporting and propelling
body mass during the stance phase and increasing overall knee stability.

We expressed hamstring muscle activity as BF/VL

and BF/BFmax and both ratios are conceptually acceptable (Aagaard et al., 2000; Enoka, 1983; Hortobagyi
and DeVita, 2000; Kellis, 1998). The benet of BF/VL
is that it functionally reects the ongoing pattern of
coactivity of the two muscles on the opposite side of
the knee. Its disadvantage is that a disproportionately
low activation of the quadriceps would produce articially high BF/VL ratios, a highly likely scenario in subjects with knee OA. Indeed, we observed a 105% BF/VL
value and 72% BF/BFmax value in OA during level walking. The data in Table 3 conrm the suspicion that the
source of these high BF/VL coactivity values in knee
OA subjects is the impaired quadriceps activation
(Hurley, 2003; Marks et al., 1994). We found that OA
subjects executed the three ADLs by activating their
vastus lateralis muscle 92% of their maximum activity
measured during velocity- and contraction-specic maximal voluntary contraction, whereas the activation level
was 57% and 47%, respectively, for age- and gendermatched healthy adults and young adults. The interpretation of these data is that due to quadriceps weakness
the BF/VL ratio is probably of limited clinical value
and the BF/BFmax ratio more faithfully describes the
muscle activation patterns around the knee in subjects
with knee OA. This is because maximal voluntary contraction force and the associated EMG activity of the
hamstrings are not aected by OA-related inhibition
(Slemenda et al., 1997). The signicantly greater BF/
BFmax ratio in knee OA compared with controls suggests that hamstring muscle coactivity increased even
when it was normalized to BFmax and not to VL. These
data indicate an intrinsic change in neural control of
muscle balance in knee OA mediated by impaired quadriceps and heightened hamstring activity.
The altered muscle balance observed at the knee in
subjects with knee OA was not restricted to the muscles
around the knee. The activation ratios of the gastrocnemius to tibialis anterior were also statistically higher
in knee OA than in healthy adults and healthy young
adults, with large variations between OA subjects. The
interpretation of these data is that changes in the neural
properties that control muscle balance in the lower
extremity are not specic to the involved joint. The
two heads of the gastrocnemius muscle cross the knee
joint and contribute to the stabilization of the knee in
stance phase of the gait. The clinical interpretation is
that muscle-enhancing exercises (see below) should target all muscles that cross the knee and also involve the
hip and the ankle joints.
Muscle activation patterns change with age and there
is evidence that old adults execute voluntary movements, including ADLs, with higher coactivity of
the lower extremity muscles (Hakkinen et al., 1998;
Hortobagyi and DeVita, 2000; Macaluso et al., 2002).
Because it is unknown at what age this change in muscle

T. Hortobagyi et al. / Clinical Biomechanics 20 (2005) 97104

balance occurs, we used a healthy age-matched and a

young group to control for the eect of age on coactivity
even though the OA group was relatively young, averaging 60 years old. Somewhat surprisingly, we found that
hamstring coactivity while performing ADLs was about
1.6-fold higher in the 60-year-old controls compared
with young adults at the knee (Table 2) and were also
higher at the ankle joint. Increased ankle coactivity during gait in OA may serve to stien the limb and compensate for the instability and weakness at the knee. These
data suggest that fundamental changes occur in neuromuscular control of human movement at a relatively
young age and that the use of an age-matched control
group in interventions studies in knee OA may be
necessary.
There are some limitations in the present study. We
recorded EMG activity only from the VL and BF to represent the quadriceps and the hamstrings. This limitation probably does not aect our conclusions because
several studies showed synchronized activity between
the vastus lateralis and medialis and between the biceps
femoris and semitendinosus during gait (Burden et al.,
2003; Ivanenko et al., 2004) and isokinetic knee extension (Pincivero et al., 2003). Using a dierent muscle
from the quadriceps and hamstrings may produce
slightly dierent results but we expect the magnitude
of coactivity measures to be similar to the current results
and the group ordering to remain identical.
We matched the type and speed of the voluntary contraction with the muscle action during ADLs, we did not
match the knee joint positions in these two conditions
where the EMG was measured. Under these conditions
exact matching for position is used infrequently in research settings and it is rarely considered in clinical settings where an MVC is done mostly at an arbitrary
position. This is of course a compromise between the
two methods. However, we suspect that using an
EMG activity value even 1015 away from the location
of maximal force, would not change the normalized values dramatically. Another limitation is that we did not
quantify the magnitude of cross-talk. Because during
maximal voluntary contractions the cross-correlation
coecients indicated a vastus lateralis to biceps femoris
cross-talk of about 68% (Aagaard et al., 2000), considering the large dierences in coactivity between groups
in ADLs that were executed at a submaximal level, it
is likely that cross-talk did not aect our conclusions.
In addition, although conceptually higher cross-talk is
associated with thicker subcutaneous fat, as is the case
in OA compared with healthy subjects, recent studies
seem to refute this possibility De Vito et al., 2003;
Solomonow et al., 1987).We also analyzed the data only
in the amplitude domain and OA subjects may reveal
compensatory adaptations in the time domain as well.
For example, Fig. 1B shows that the timing and durations of the VL EMG activity are dierent in the OA

103

subject compared with control subject Hinman et al.,

2002). OA subjects in the present study had moderate
not severe OA and thus the ndings are limited to this
segment of the OA population. Finally, healthy and
OA subjects walked at the same speed of 1.3 m/s during
level walking but at dierent speeds during stairway
locomotion that could aect EMG. Level and stair
walking speeds in both self selected and maximum speed
gaits were between 20% and 30% slower in OA versus
healthy. Therefore, we tested OA subjects at somewhat
higher relative speeds compared to the healthy group.
However, the test speeds were well within the capabilities of the OA group and all OA subjects walked comfortably in level and stair gait. No OA subjects
complained about the gait conditions even when asked
about them during the tests.
The ndings reported here have important clinical
implications. We interpret the increased hamstring muscle coactivity in knee OA as a natural compensatory
adaptation to quadriceps weakness, pain, and the altered local environment (Hurley, 2003; Sharma, 2001).
We do not consider this as a maladaptation rather
as a coping strategy to preserve function and possibly
manage pain. The data indicate that the conventionally
used quadriceps strengthening exercises should be supplemented with muscle-enhancing interventions (Hurley,
2003; Sharma et al., 2003). These protocols should be
designed not only to improve maximal quadriceps
strength but also to enhance the ability to accurately
control the recruitment and decruitment of synergistic
muscles and antagonistic muscles pairs in single and
multi-joint movements (Hurley, 2003; Sharma et al.,
2003). An initial step in this process could be to improve
OA subjects ability to produce force smoothly and
accurately with the muscles around the knee (Hortobagyi et al., 2004) because OA subjects impaired ability
to scale force accurately may underlie the impaired muscle balance reported here. Such approaches to treatment
of knee OA may slow disease progression.
In conclusion, subjects with knee OA executed ADLs
with heightened hamstring muscle activity compared
with age- and gender-matched healthy and young
adults. The source of the increased coactivity was reduced quadriceps activation and increased hamstring
muscle activity. Altered muscle activation at the knee
may interfere with normal load distribution in the knee
and facilitate disease progression. Therapeutic interventions should focus not only on quadriceps strengthening
but also on improving muscle balance at the knee.

Acknowledgments
This study was supported in part by an NIA
AG16192 grant (to T.H.) and by a research and creative
activity grant from East Carolina University Faculty

104

T. Hortobagyi et al. / Clinical Biomechanics 20 (2005) 97104

Senate (P.D.). We thank graduate students Jovita Jolla,

Chris Mizelle, and Kim Smith for their assistance with
data collection and analysis.
References
Aagaard, P., Simonsen, E.B., Andersen, J.L., et al., 2000. Antagonist
muscle coactivation during isokinetic knee extension. Scand. J.
Med. Sci. Sports. 10, 5867.
Al-Zahrani, K.S., Bakheit, A.M., 2002. A study of the gait characteristics of patients with chronic osteoarthritis of the knee. Disabil.
Rehabil. 24, 275280.
Bernardi, M., Solomonow, M., Baratta, R.V., 1997. Motor unit
recruitment strategy of antagonist muscle pair during linearly
increasing contraction. Electromyogr. Clin. Neurophysiol. 37, 312.
Brucini, M., Duranti, R., Galletti, R., et al., 1981. Pain thresholds and
electromyographic features of periarticular muscles in patients with
osteoarthritis of the knee. Pain 10, 5766.
Burden, A.M., Trew, M., Baltzopoulos, V., 2003. Normalisation of gait
EMGs: a re-examination. J. Electromyogr. Kinesiol. 13, 519532.
DeVita, P., Moody, J., Beam, S., et al., 2002. Neuromuscular
responses to knee osteoarthritis during stairway locomotion. In:
4th World Congress of Biomechanics, Calgary, Canada.
De Vito, G., McHugh, D., Macaluso, A., et al., 2003. Is the
coactivation of biceps femoris during isometric knee extension
aected by adiposity in healthy young humans? J. Electromyogr.
Kinesiol. 13, 425431.
Ebenbichler, G.R., Kollmitzer, J., Glockler, L., et al., 1998. The role
of the biarticular agonist and cocontracting antagonist pair in
isometric muscle fatigue. Muscle Nerve. 21, 17061713.
Enoka, R.M., 1983. Muscular control of a learned movement: The
speed control hypothesis. Exp. Brain Res. 51, 135145.
Grasso, R., Zago, M., Lacquaniti, F., 2000. Interactions between
posture and locomotion: motor patterns in humans walking with
bent posture versus erect posture. J. Neurophysiol. 83, 288300.
Hagood, S., Solomonow, M., Baratta, R., et al., 1990. The eect of
joint velocity on the contribution of the antagonist musculature to
knee stiness and laxity. Am. J. Sports Med. 18, 182187.
Hakkinen, K., Kallinen, M., Izquierdo, M., et al., 1998. Changes in
agonist-antagonist EMG, muscle CSA, and force during strength
training in middle-aged and older people. J. Appl. Physiol. 84,
13411349.
Hansen, S., Hansen, N.L., Christensen, L.O., et al., 2002. Coupling of
antagonistic ankle muscles during co-contraction in humans. Exp.
Brain Res. 146, 282292.
Hinman, R.S., Bennell, K.L., Metcalf, B.R., et al., 2002. Delayed
onset of quadriceps activity and altered knee joint kinematics
during stair stepping in individuals with knee osteoarthritis. Arch.
Phys. Med. Rehabil. 83, 10801086.
Hortobagyi, T., DeVita, P., 2000. Muscle pre- and coactivity during
downward stepping are associated with leg stiness in aging.
J. Electromyogr. Kinesiol. 10, 117126.

Hortobagyi, T., Mizelle, C., Beam, S., et al., 2003. Old adults perform
activities of daily living near their maximal capabilities.
J. Gerontol. 58A, M453M460.
Hortobagyi, T., Garry, J., Holbert, D., et al., 2004. Aberrations in the
control of quadriceps muscle force in patients with knee osteoarthritis. Arthritis. Rheum. Care 51, 562569.
Hurley, M.V., 2003. Muscle dysfunction and eective rehabilitation of
knee osteoarthritis: what we know and what we need to nd out.
Arthritis Rheum. 49, 444452.
Ivanenko, Y.P., Poppele, R.E., Lacquaniti, F., 2004. Five basic muscle
activation patterns account for muscle activity during human
locomotion. J. Physiol. 556, 267282.
Kaufman, K.R., Hughes, C., Morrey, B.F., et al., 2001. Gait
characteristics of patients with knee osteoarthritis. J. Biomech.
34, 907915.
Kelgren, J., Lawrence, J., 1957. Radiological assessment of knee
osteoarthritis. Ann. Rheum. Dis. 16, 494501.
Kellis, E., 1998. Quantication of quadriceps and hamstring antagonist activity. Sports Med. 25, 3762.
Macaluso, A., Nimmo, M.A., Foster, J.E., et al., 2002. Contractile
muscle volume and agonistantagonist coactivation account for
dierences in torque between young and older women. Muscle
Nerve. 25, 858863.
Marks, R., Percy, J.S., Semple, J., et al., 1994. Comparison between
the surface electromyogram of the quadriceps surrounding the
knees of healthy women and the knees of women with osteoarthrosis. Clin. Exp. Rheumatol. 12, 1115.
McAlindon, T.E., Cooper, C., Kirwan, J.R., et al., 1993. Determinants of disability in osteoarthritis of the knee. Ann. Rheum. Dis.
52, 258262.
McFayden, B.J., Winter, D.A., 1988. An integrated biomechanical
assessment of stair ascent and descent. J. Biomech. 21, 733744.
OReilly, S., Jones, A., Doherty, M., 1997. Muscle weakness in
osteoarthritis. Curr. Opin. Rheumatol. 9, 259262.
Pai, Y.C., Chang, H.J., Chang, R.W., et al., 1994. Alteration in
multijoint dynamics in patients with bilateral knee osteoarthritis.
Arthritis Rheum. 37, 12971304.
Pincivero, D.M., Dixon, P.T., Coelho, A.J., 2003. Knee extensor
torque, work, and EMG during subjectively graded dynamic
contractions. Muscle Nerve 28, 5461.
Sharma, L., 2001. Local factors in osteoarthritis. Curr. Opin.
Rheumatol. 13, 441446.
Sharma, L., Dunlop, D.D., Cahue, S., et al., 2003. Quadriceps
strength and osteoarthritis progression in malaligned and lax
knees. Ann. Intern. Med. 138, 613619.
Slemenda, C., Brandt, K.D., Heilman, D.K., et al., 1997. Quadriceps
weakness and osteoarthritis of the knee. Ann. Intern. Med. 127,
97104.
Solomonow, M., Baratta, R., Zhou, B.H., et al., 1987. The synergistic
action of the anterior cruciate ligament and thigh muscles in
maintaining joint stability. Am. J. Sports Med. 15, 207213.
Wretenberg, P., Arborelius, U.P., Weidenhielm, L., et al., 1993. Rising
from a chair by a spring-loaded ap seat: a biomechanical analysis.
Scand. J. Rehabil. Med. 25, 153159.