545

Knee Recurvatum
Biomechanics
D. Casey Kemkan,

in Gait: A Study of Associated Knee

MD, Lynn C. Deming, MS, PT, Maureen

ABSTRACT.
Kerrigan
C, Deming
LC. Holden
MK.
Knee
recurvatum
in gait: a study
of associated
knee biomechanics.
Arch
Phys Med Rehabil
1996;77:645-SO.

Objectives: To quantitatively
evaluate
peak knee extensor
torque
values
imparted
to the posterior
knee structures
during
gait in patients
with knee
recurvatum
compared
with
torque
values observed
in control
subjects,
and to assess the predictive
value of the degree
of knee hyperextension
and other clinical
factors
in estimating
peak knee extensor
torque.
Design: A retrospective
analysis
of clinical
and quantitative
gait data obtained
from
patients
and control
subjects.
Settings: A gait laboratory.
Subjects: Forty-one
consecutive
patients
with neurologically
based impairments
presenting
with knee hyperextension
during
gait (52 limbs)
and 46 able-bodied
control
subjects.
Main Outcome Measure: Peak knee extensor
torque
during
the stance period
of the gait cycle.
Results: Although
overall,
the patient
average
peak extensor
torque
was significantly
greater
(p < .oOl)
than the control
subjects
average
value,
knee extensor
torques
were within
or
below
a + I standard
deviation
range
for control
subjects
in
25% (I 3) of limbs tested.
Peak knee hyperextension
angle was
a poor predictor
of peak extensor
torque;
there was statistical
signifcance
(coefficient
.06 I, p < ,001) only for hyperextension
angles
of ~4.
Multiple
regression
incorporating
hyperextension angle
and other
clinical
variables
to predict
peak knee
extensor
torque
resulted
in an adjusted
r of 33.
Conclusion: Patients
with
knee recurvatum
have
variable
peak extensor
torque
values
associated
with their knee hyperextension.
Knowledge
of knee
hyperextension
angle
and other
clinical
factors
arc only partially
useful in predicting
a patients
peak knee extensor
torque
imparted
to the posterior
knee structures during
walking.
0 I YY6 by the American
and the American
Academy
tution

Congrcw
ojfhysical

r$Rehobilifution
Medicine

ond

Medicine
Kehuhili-

YNAMIC
KNEE
RECURVATUM,
defined
as hypercxtension
of the knee during
the stance
period
of the gait
cycle,
is common
in patients
with
a variety
of neurologically
based impairments.
It has been reported
in nearly
one half of
patients
with stroke
or traumatic
brain injury
in some clinical

K. Holden, PhD, PT

study samples,
and in patients
with cerebral
palsy
and poliomyelitis.
The disorder
is typically
ascribed
to a combination
of quadriceps
weakness,
ankle
plantar
flexor
spasticity.
heel
cord contracture,
quadriceps
spasticity,
and/or
gastrocsoleus
weakness.
The dynamic
knee recurvatum
may be advantageous in providing
a mechanism
to control
an otherwise
unstable limb during
the stance period
of the gait cycle.
However,
a
concern
for patients
with this disorder
is that hyperextension
may produce
an increased
external
extensor
torque
across
the
knee.
placing
the capsular
and ligamentous
structures
of the
posterior
aspect
of the knee at risk for injury.
Injury
to these
tissues
can cause pain, ligamentous
laxity,
or bony deformity.
which
may lead to functional
gait deficits.

Despite
the pervasive
nature of knee recurvatum
and concern
for injury
and secondary
complications,
there is no quantitative
information
regarding
the peak amounts
of knee extensor
torque
associated
with rccurvatum.
We became
interested
in this problem because
of preliminary
observations
from clinical
gait laboratory
analysis
studies
that although
some patients
with
knee
recurvatum
had high peak knee extensor
torque
values
associated with their rccurvatum.
other patients
had surprisingly
small
torque
values.
To date, there have been no reports
regarding
how peak knee extensor
torque
values
might
vary with peak
hyperextension
angle
or other
clinical
factors.
or how
these
peak extensor
torque values
compare
with values
found in ablcbodied subjects
who do not have knee recurvatum.
Finally,
there
is no quantitative
infommation
regarding
the effect
of treatments
such as an ankle-foot-orthosis
(AFO)
specifically
aimed
to reduce the knee extensor
torque
and risk for posterior
knee structural injury.
Modern
three-dimensional
quantitative
gait analysis
allows
for measurement
of knee joint extensor
torque
using force plate
information
combined
with inverse
dynamic
techniques
applied
to kinematic
data.,X Normally
there is first an external
flexor,
extensor.
and then flexor
torque about the knee during
the stance
period
of the gait cycle..
We have observed,
among
patients
referred
to our gait laboratory
for clinical
gait analysis,
that
patients
with knee hyperextension
during
the stance period
may
or may not have excessive
peak extensor
torque
values
while
the knee is hyperextended.
Patients
with dynamic
knee recurvaturn who have only a small associated
peak knee extensor
torque
appear
to have minimal
stretch
forces
imparted
to the posterior
knee structures,
implying
that the posterior
structures
arc not at
risk for injury.
For these patients,
knee hyperextension
could
he considered
a reasonable
compensation,
and treatments
spccifically
aimed IO reduce
knee hyperextension,
such as specific
physical
therapy
techniques,-
electrogoniometric
fcedhack,.
I or bracing
across
the knee
and/or
ankle.,
may
not he indicated.
Other
patients
may have large knee extensor
torque
values,
implying
that a large,
detrimental
stretch
force
is imparted
to the posterior
knee structures
during
walking.
The purposes
of this study were IO ( I) quantitatively
evaluate
the distribution
of peak knee joint extensor
torque
values associated with knee recurvatum
during
gait in patients
(without
and
with an AFO
if normally
used)
compared
to those values
obtained
in able-bodied
(control)
subjects,
and (2) preliminarily
evaluate
the predictive
value of peak knee hyperextension
angle

Arch

Phys Med Rehabil

Vol77,

July

1996

646

BIOMECHANICS

OF KNEE

and other clinical variables including age, gender, neurological

diagnosis, time since diagnosis, gait velocity, primary cause,
and timing of hyperextension in estimating peak knee extensor
torque.
METHODS
The kinematic and kinetic walking data from 41 consecutive
patients with knee recurvatum (52 limbs) secondary to neurological injury presenting to our Gait Laboratory were retrospectively analyzed. Data from a group of 46 able-bodied subjects
(31 women and 15 men, age 29.6 2 8.7 years) were used for
controls. These control subjectshad been excluded for neurological or musculoskeletal pathology. The retrospective analysis
and protocol for the control subjects were approved by our
Institutional Review Board. Patient subjects had neurological
injury secondary to stroke, traumatic brain injury, cerebral
palsy, multiple sclerosis, poliomyelitis, or lower motor neuron
spinal cord injury and had been referred for clinical evaluation
of their gait. The criteria for inclusion of patient subjects in the
present study were (1) knee hyperextension noted by motion
analysis to be greater than 0 beyond extension occurring during
some portion of the stance period, (2) knee recurvatum attributed to neurological injury confirmed by observational gait analysis performed by both a physiatrist and a physical therapist,
and (3) force plate data obtained for the involved limb. Fortyfour patients fit the first two criteria; however, three were excluded becauseforce plate data could not be obtained secondary
to the other limb striking the force plate during the samerelative
time frame. The mean age of the patient subjects was 35.9 years
with a standard deviation of 18.2 years; there were 24 men and
17 women. There were 32 right-involved lower extremities and
20 left-involved lower extremities (this included 11 subjects in
whom both extremities were involved). There were no significant differences in demographics or in any of the variables
evaluated, including peak hyperextension angle and peak extensor torque and velocity, between the patients with unilateral
involvement and patients with bilateral involvement (smallest
p value per unpaired t test was .23). Thus, the data from patients
with one and two knees involved were analyzed together, with
a correction (noted below) to account for the fact that not all
the data points for the patients with bilateral knee involvement
were independent.
For testing, each patient and control subject was instructed
to first stand and then walk barefoot at his or her own comfortable speed. Data from 3 walking trials per patient (the protocol
number of trials collected per each walking condition for clinical
gait laboratory analysis in our laboratory) and 5 trials per control
subject were averaged for the analysis. Data from an additional
3 trials while wearing an AFO were obtained in each patient
who normally wore one. An optoelectronic camera system was
used to measure the three-dimensional coordinates of 1.5-cmhemispherical, infrared reflective markers attached to the patients and control subjects skin over the following bony landmarks: the lower prominence of the sacrum, bilateral posterior
superior iliac spines, lateral femoral condyles, lateral malleoli,
and fifth metatarsals. Additional markers rigidly attached to
wands were placed over the lateral femoral condyles, the anterior tibia1 shafts, and the forefeet. Three-dimensional marker
position was collected at a sampling rate of lOOHz.. Four
video cameras were used with two cameras placed posterolaterally on each side of the subject. The accuracy of measure within
the working volume of a 2-m height, a 3-m length, and a l-m
width, was calibrated and predetermined before each patient
and control subject sessionto be within 0.2 C 2mm per 200mm
distance.
Ground reaction forces were measured synchronously with
Arch

Phys Med Rehabil

Vol77,

July

1996

RECURVATUM,

Kerrigan

the kinematic data at a sampling rate of 1OOHzusing two force

plates staggered along the walkway. The locations of the force
plates in the global reference plane were predetermined by acquiring coordinates of markers placed on their corners. A commercialized protocol, termed SAFLo (Servizio di Analisi della
Funzionalita Locomotoria), developed by Pedotti and Frigo,
was used to calculate the kinematics and kinetics. The following
anthropometric measurements were taken according to the
SAFLo protocol to calculate the kinematics and kinetics: body
weight, pelvic width and height, thigh, foot, and lower leg
length, and intracondylar and intramalleolar width. Kinetics
were calculated using the force plate data and inverse dynamic
techniques described by Winter and Eng*Oper the SAFLo protocol. Torques were normalized for body weight and height and
are reported as external, in Newton meters per kilogram meters
(Nm/kgm). Gait velocities were obtained utilizing the kinematic
and force plate data and are reported in meters per second
(rdsec).
The timing of occurrence of knee recurvatum during the
stance period was recorded on the basis of the kinematic data
for each patient to be early if it occurred primarily in the first
third of the stance period, late if it occurred primarily in the
last third of the stance period, and continuous if it occurred
primarily in the middle third or was continuous throughout the
stance period. The most likely or primary cause for the knee
recurvatum, ie, quadricep weakness, ankle plantar flexor weakness, ankle plantar flexor spasticity, heel cord contracture, or
mixed, was determined for each patient. The primary cause
determination was made on the basis of the gait and clinical
evaluation, which included static manual muscle testing and
range of motion assessmentperformed by and agreed upon by
both the physiatrist and the physical therapist.
The patients velocity, peak knee hyperextension angle, and
extensor torque values were compared to control subject values
using unpaired t tests. Both the peak knee hyperextension angle
and torque values were compared with and without an AFO
(23 limbs) using paired t tests. Peak knee hyperextension angle
was plotted against peak knee extensor torque. The variables
of age, gender, neurological diagnosis, time since diagnosis,
velocity, peak knee hyperextension angle, primary cause, and
stance phase timing of hyperextension were evaluated as possible predictors of peak knee extensor torque using a piecewise
linear multiple regression analysis. The categorical variables of
gender, neurological diagnosis, primary cause, and timing of
recurvatum during stance were converted to dummy variables
for use in the regression. Since 11 of 52 subjects data were
gathered on bilateral knees and, thus, not all data points were
independent, the HuberAVhite formula was used to supply robust standard errors for the coefficients in the significance tests
and confidence interval estimations.*~**
Several models were developed; at first all eight variables
and all levels were utilized. The final model reported in the
Results section does not include the four variables of age, gender, time since diagnosis, and gait velocity as predictors of
peak knee extension torque, because they were found to be
nonsignificant contributors to the model. Overall, peak knee
hyperextension angle also was initially found to be a nonsignificant contributor in predicting peak knee hyperextension
torque. Visual inspection of the graph of torque versus angle
(see fig 4), however, suggested a nonlinear, possible quadratic
function. Thus, peak hyperextension angle was converted into
two variables using a piecewise linear spline function, with a
breakpoint at 4. Four degrees was chosen as the breakpoint
because it was the point that had the smallest mean square error.
A statistically significant correlation was found only for those
patients with hyperextension angles of 54. The neurological

BIOMECHANICS

OF KNEE

RECURVATUM.

647

Kerrigan

10

FLX

IQlm/Lr*m

81

EMT
1

% sait Cycle

0.6

Fig 1. Knee flexor/extensor

momsnt
versus
percentage
of gait cycle:
Typical knee flexor/extensor
moment
pattern throughout
the gait cycle.
The solid curve represents
a typical patients
data,and the dashed curve
represents
the mean of control data. The horizontal
axis represents
the
percentage
of one gait cycle from 0 to 100%. The vertical axis represents
the moment,
expressed
in Newton
meters
per kilogram
meters.
The
vertical
line represents
the division
between
the patients
stance phase
and swing phase. FLX, flexor; EXT. extensor.

vELocllY
w-2

06
o4
02

Bo

categorical
variable was collapsed to two levelsupper motor neuron disorders and lower motor neuron disorders. For the categorical variable timing of hyperextension during the gait cycle, continuous timing was used as the reference
group, with early and late timing as the dummy variables. For
the categorical variable primary cause, plantarflcxor weakness
was used as the reference group. The I test analyses and descriptive statistics were performed using the statistical software True
Epistat, and the regression calculations were performed using
the statistical software Stata. Data are reported as means 2
one standard deviation. Significance was defined at p < .OS.
diagnostic

RESULTS
Figure 1 shows a typical knee flexor/extensor moment pattern
throughout the gait cycle from one representative patient compared with averaged control data. Figure 2 displays the distribution of peak knee extensor torque values for the patient subjects
along with the mean + one standard deviation for both the

Fig 2. Peak knee extensor

torque
cal axis, no. of patients;
horizontal
one standard
deviation
for patient

distribution
for patient subjects
lvertiaxis, torque in Nmlkgm).
The mean 2
and control subjects
is also illustrated.

0.5
0.4
0.3
0.2
01

0
wmwa

AFO

WthrVO

Fig 3. IA) Patient versus

control
data for peak knee hyperextension
angle, where the vertical
axis represents
angle in degrees.
lB) Average
walking
velocity
for patient versus control
data, where the vertical
axis
represents
velocity
in meters per second. (C) Patient data of peak external extensor
torque, with versus without
an AFO, where the vertical axis
represents
torque
in Newton
meters
per kilogram
meters.
All graphs
represent
the mean (0 in A and B) - one standard
deviation.

patient and control subject populations. The patient average

peak extensor torque, .27 t .I8 Nmlkgm, was significantly
greater (/, < .OOl) than the control subjects average value of
.I3 2 .06 Nmlkgm. In 39 patient limbs (75%), peak knee extensor torques were higher than the control z one standard deviation range, in 10 limbs (19%) extensor torques were within this
range, and in 3 limbs (6%) extensor torques were actually lower
than this range.
Figure 3 displays the patient versus control subjects for peak
knee hyperextension
angle (fig 3A) and gait velocity (fig 3B).
For the patients, the peak knee hyperextension ranged from - 1
to - 18 with a mean of -5.9 -C 4.7. This was statisticallv
different @ < .OOl) from the average peak extension of 4.9 1
3.9 observed in the control subjects. The average walking velocity was signiticantly slower for the patients (.42 z .21 m/
set) compared to the control subjects (.77 -t .18m/sec, p <
.OOl). The differences in torque values with and without an
AFO arc shown graphically
in figure 3C. In those limbs for

Arch

Phys Med

Rehabil

Vd

77, July

1996

BIOMECHANICS

Table

1: Means

of Peak Extensor

OF KNEE

cord

Peak Extensor Torque

iNm/kgm), Mean (SD)

injury

06

,238 t.178)
,431 t.240)
.415 t.182)

6.286 (3.817)
4.091 (3.265)
7.565 (5.558)

7
12
21
9
3

,286
,328
,198
,343
,360

.382
,404
.421
.500
,285

6.143
6.667
4.905
8.000
3.333

Fig 4. Peak knee external

extensor
torque versus peak hyperextension
angle in patient subjects.
Torque is reported
in Newton
meters per kilogram meters; hyperextension
angle is reported
in degrees.

Rehabil

Vol77,

July

1996

.364
,291
,627
,644
,460
,363
,523

c.247)
t.160)
c.166)
c.125)
C.079)

t.207)
t.118)
1.218)
i.206)
t.240)
t.131)
t.142)

4.6
7.0
6.0
6.4
2.5
5.5
7.0

t.149)
f.215)
t.199)
f.283)
f.134)

(5.060)
(5.080)
(5.196)
(4.930)
t.500)
(5.577)
(2.944)

(5.429)
(5.433)
(4.182)
(4.583)
(3.215)

extensor torque. Patients with lower motor neuron disorder had

significantly less knee extensor torques using upper motor neuron disorder as the reference group (p < .OOOl). The multiple
regression model using these variables resulted in an 2 of .61
and an adjusted ? of .53, p < .OOl.
DISCUSSION
In the patient limbs studied, the average peak extensor torque
across the knee was twice the control average value; however,
there was considerable overlap in the patient and control populations. Seventy-five percent of the patient limbs had torque values significantly higher than the control ? one standard deviation range, and 25% of patient limbs had peak knee extensor
torque values within or below the control + one standard deviation range. The higher torque values are of concern because an
increase in the extensor moment at the knee is believed to
increase the risk for damage to the posterior passive structures
of the knee. The lower torque values may not be of concern,
and imply that for these patients knee hyperextension
in the
involved limb is a reasonable and safe compensation to maintain
stability in stance; treatments aimed specifically to improve the
knee recurvatum may not be indicated.
Patients were asked to walk at their own comfortable speed,
which was found to be significantly slower on average compared with that of the control subject population. If a patient
were to walk faster, he or she would have a greater extensor
torque across the knee. This fact is probably not clinically
Table

Phys Med

t.167)
t.140)
t.184)
t.230)
i.080)
t.152)
t.076)

Peak Hyperextension
Angle (7, Mean (SD)

,411 1.189)
,176 t.141)
,326 f.154)

. .

Arch

Walking Velocity
(m/set), Mean (SD)

of Diagnosis,

7
22
23

421

as a Function

,277
,286
.453
.304
.360
.I23
,320

Angle

10
11
3
5
2
6
4

which the patient normally wore an AFO (n = 23), the knee

extensor torque was significantly less with the AFO (.20 + .I2
Nm/kgm) than without the AFO (.28 2.16 Nm/kgm, p < .OOl).
Table 1 shows the means of peak extensor torque, walking
velocity, and peak extension angle as a function of diagnosis,
timing of hyperextension, and primary cause. Age, gender, velocity, and time since diagnosis were not statistically significant
predictors of peak knee hyperextension torque. Figure 4 demonstrates a plot of peak knee hyperextension
angle versus peak
knee extensor torque. Hyperextension
angle was a statistically
significant predictor of knee extensor torque only for those patients with hyperextension angles of 54 (regression coefficient
.061, p < .OOl). A summary of the final regression model is
shown in table 2. Timing of hyperextension was a statistically
significant predictor of knee extensor torque (p for simultaneous
test of all categories < .OOOl). In particular, patients with late
stance phase timing had significantly less extensor torque using
continuous timing of hyperextension
as the reference group.
Another statistically significant predictor of peak knee extensor
torque was primary cause @ = .0034). In particular, patients
with quadriceps weakness had significantly
greater extensor
torque values using plantarflexor
weakness as the reference
group. Diagnosis, collapsed into upper and lower motor neuron
disorders, was another statistically significant predictor of knee

07
I

Kerrigan

Torque, Walking
Velocity,
and Peak Hyperextension
Timing of Hyperextension,
and Primary Cause
N

Diagnosis
Traumatic
brain injury
Cerebrovascular
accident
Cerebral palsy, diplegia
Cerebral palsy, hemiplegia
Multiple sclerosis
Lower motor neuron spinal
Polio
Timing
Early
Late
Continuous
Primary causes
Plantarflexor
spasticity
Plantarflexor
contracture
Plantarflexor
weakness
Quadriceps
weakness
Mixed

RECURVATUM,

2: Final Regression

Model

Predictor
Hyperextension
angle
For angles 5 4
Timing of hyperextension
(continuous
timing as
reference)
Late stance phase timing
Primary cause (plantafflexor
weakness
as reference)
Quadriceps
weakness
Diagnosis
(upper motor neuron
injury as reference)
Lower motor neuron injury

to Predict
Regression
Coefficient

Peak Knee Extensor

95%
Confidence
Intervals

Torque
P

Valuef
.0008

,061

(.030,.092)
<.OOOl

-.I70

l.243.T,095)
.0034

,162

(.078,.246)
<.OOOl

-.I58

(-.229,-,088)

r2 = .61, adjusted
r2 = .53.
* For each of the predictor
variables
represented
by 2 or more categories,
the reported
p value is for the simultaneous
test of all categories.

BIOMECHANKZS

OF KNEE

significant
because
an individual
normally
walks
at his or her
most
comfortable
speed.
Thus,
the measured
peak extensor
torque
during
comfortable
walking
speed is likely
a true reflection of the extensor
torque
imparted
to the posterior
structures
of the knee during
usual walking
activity.
Although
overall,
the peak knee extensor
torque
values
were
less with an AFO than without
an AFO
for those patients
who
routinely
used an AFO,
some patients
had little or no reduction
in peak knee extensor
torque
with the AFO (fig 30.
Conceivably, in these cases the AFO may have acted primarily
to reduce
the energy
requirement
of walking
rather
than to reduce
the
extensor
moment
at the knee. Quantitative
gait analysis
may be
useful
to assess the benefits
of an AFO with respect
to both knee
extensor
torque
and overall
biomcchanical
gait performance.
Similarly,
gait analysis
may be useful
for evaluating
the effects
of any other
treatments,
such as stretching
or strengthening
exercises,
specifically
aimed
to improve
knee hypcrcxtcnsion
and peak knee extensor
torque.
The degree
of peak knee hyperextension
angle by itself was
not a useful
predictor
of peak knee extensor
torque.
Moreover,
peak knee hyperextension
angle combined
with other
clinical
variables
were only partly
helpful
in predicting
the peak knee
extensor
torque
for a particular
patient.
The clinical
variables
of age, gender,
gait velocity,
and time since neurological
injury
were not useful
in predicting
knee extensor
torque.
Although
the variables
knee hyperextension
angle, diagnosis,
timing.and
primary
cause helped
to predict
the knee extensor
torque,
these
factors
explained
only
about
half the variance
of the torque
values.
The
relationship
between
peak
knee
hyperextension
angle and peak knee extensor
torque
was statistically
significant
only for those hyperextension
angles
of 54. This statistically
significant
relationship
is likely
not clinically
significant
because it is probably
not possible
to ascertain
the hyperextension
angle within
the range of 0 to 4 from observational
or video
analysis
alone.
The poor relationship
between
the clinical
variables
and knee
cxtcnsor
torque
could
be due in part to a small
sample
size,
given
the number
of variables
analyzed.
However,
we believe
a more likely
reason
for the poor relationship
is the fact that
there
arc complex
interactions
and compensations
about
the
trunk.
hip,and
ankle,
making
it difficult
to predict
the knee
extensor
torque.
For instance,
the extensor
torque
could
bc rcduccd
in the face of a large amount
of knee hypcrcxtension
if
foot contact
is relatively
calcaneal
or if the trunk
is relatively
hyperextended
during
the knee recurvatum.
Conversely,
relative
forefoot
contact,
a relatively
plantarflexed
ankle,
and/or
a relatively
flexed
trunk
at the time of knee hyperextension
would
tend to increase
the peak extensor
torque
across
the knee. The
interactions
of these
potential
events
arc likely
important
in
determining
the peak knee extensor
torque,
yet they arc diflicult
to evaluate
from clinical
evaluation
alone. These
findings
tend
to support
the role for quantitative
gait analysis
to study
knee
extensor
torque
on a routine
basis for patients
with knee hyperextension
to evaluate
an individual
patients
risk for posterior
knee structural
injury.
The complex
interactions
about
multiple
joints
might
also
explain
some
of the statistically
significant
findings
as well.
such as why patients
with quadriceps
weakness
as the primary
cause for the knee recurvatum
had greater
knee extensor
torque
values
compared
with the reference
group of patients
with gastrocsoleus
weakness.
A plausible
reason
for this statistically
significant
finding
could
be that those patients
with gastrocsoleus weakness
tend to have relatively
calcaneal
foot contact5
compared
to patients
with quadriceps
weakness.
A relative
calcaneal
foot contact
would
tend to reduce
the knee extensor
torque.
The finding
that hyperextension
occurring
late rather

RECURVATUM,

Kenigan

649

than early or continuous

during
the stance period
was associated
with less peak knee extensor
torque
can be attributed
to similar
complex
interactions.
At terminal
stance,
the hip is hypercxtending
and the ankle
begins
to dorsiflcx.
The trunk
weight
becomes
more anterior
to the knee, but the overall
weight
imparted
to the limb reduces
as weight
is being transferred
to the
other limb.
It was not expected
that patients
with lower
motor
neuron
disorder
would
have statistically
significant
lower peak extensor
torques
than patients
with upper motor neuron
disorder.
Patients
with upper motor
neuron
disorders
commonly
have an extensor
synergy
pattern
with
quadriceps
overactivity
resulting
in recurvatum.
While
this synergy
may affect
the hyperextension
angle, it may or may not affect
the knee extensor
torque.
Commonly,
patients
with upper motor
neuron
injury
have weakness
and/or
contracture
in muscle/tendon
groups
both proximal
and
distal to the knee. For example,
it is typical
to see a combination
of hip flexor
and heel cord contractures
in patients
with upper
motor
neuron
injury.
Both
of these factors
place
the ground
reaction
force
more
anterior
to the knee,
thus increasing
the
external
extensor
moment
at the knee. Although
patients
with
upper motor
neuron
injury
had relatively
higher
knee extensor
torques,
perhaps
these patients
have some protection
from the
knee extensor
torque.
The extensor
torque
is countered
by the
posterior
structures
of the knee. This includes
the posterior
capsule and the posterior
cruciate
ligament.
It also includes
the
long hamstring
and gastrocnemius
muscles
and tendons.
These
muscles
are typically
active
during
initial
to midstance
and
midstancc
to terminal
stance respectively.
Many
patients
with
upper motor
neuron
injury
have excessive
activity
in these muscles during
mid to terminal
stance,5.?h
which
could
be greater
than the overactivity
commonly
present
in the quadriceps,
perhaps mitigating
the increased
external
knee extensor
torque.
Patients
with
lower
motor
neuron
injury,
on the other
hand,
may have less muscle
activity
than otherwise
and thus might
have less opportunity
for protection
against
the extensor
torque.
In conclusion,
the results
of the study demonstrate
that there
is considerable
range of knee extensor
torque
values
in patients
with knee recurvatum.
Although
overall,
patients
have higher
peak knee extensor
torque
values
compared
with control
subjects,
approximately
one quarter
of patient
limbs
have knee
extensor
torque
values
in the control
+ one standard
deviation
range.
The results
imply
that knowledge
of clinical
factors,
including
peak knee
hypcrextcnsion
angle,
can explain
only
about
half of the variance
in peak knee extensor
torque.
Although
this was a small sample
size for the number
of variables
analyzed,
a likely
reason for the poor relationship
between
clinical factors
and knee extensor
torque is that complex
interactions
about the trunk,
hip, knee, and ankle make it difficult
to predict
peak knee extensor
torque.
The findings
of this study support
routinely
using quantitative
gait analysis
to evaluate
peak extensor torque
in patients
with knee recurvatum
and to assess the
effects
of treatments
such as an AFO
aimed
at reducing
the
torque.
Acknowledgement:

MS. for his technical

statistical
assistance.

The authors acknowledge

Thomas A. Rihaudo.
assistance
and Richard
Goldstein.
PhD, for his

References
I

Hoguc RE, McCandless

S. Genu recurvalum:
auditory
biofeedback
treatment
for adult patients with stroke or head injuries. Arch Phys
Med Rchabil
1983;64:368-70.
2. Knutsson
E. Richards
C. Different
types of disturbed
motor control
in gait of hemiparetic
patients. Brain 1979; 102:405-30.
3. Simon SR. Deutsch
SD. Nuzzo RM. Mansour
MJ, Jackson
JL.
Koskinen
M, ct al. Genu recurvatum
in spastic cerebral palsy: report

Arch

Phys Mad

Rehabil

Vol77,

July

1996

650

4.
5.
6.

7.
8.
9.

IO.
I I.
12.
13.

14.
15.

16.

17.

Arch

BIOMECHANICS

OF KNEE

on findings by gait analysis.

J Bone Joint Surg Am 1978;60:88294.
Irwin CE. Genu recurvatum
following
poliomyelitis:
a controlled
method of operative
correction.
JAMA
1942; 120:277-80.
Perry J. Gait analysis:
normal and pathological
function.
Thorofare
(NJ): Slack, 1992.
Morris ME, Matyas TA, Bach TM, Goldie PA. Electrogoniometric
feedback
its effect on genu recmvatum
in stroke. Arch Phys Med
Rehabil 1992;73:1147-54.
Winter DA. Kinematic
and kinetic patterns in human gait: variability and compensating
effects. Hum Mov Sci 1984; 35-76.
Harris GF, Wertsch JJ. Procedures
for gait analysis. Arch Phys Med
Rehabil 1994;75:216-25.
Kadaba MP, Ramakrishnan
HK, Wootten
ME, Gainey J, Gorton
G, Cochran
GVB. Repeatability
of kinematic,
kinetic,
and electromyographic
data in normal adult gait. J Orthop
Res 1989;7:84960.
Bobath B. Evaluation
and management
of adult hemiplegia.
London: William
Heinemann
Medical
Books, 1978.
Knott M, Voss DE. Proprioceptive
neuromuscular
facilitation:
patterns and techniques.
New York: Harper & Row, 1968.
Koheil R, Mandel AR. Joint position
biofeedback
facilitation
of
physical
therapy in gait training. Am J Phys Med 1980;59:288-97.
Basaglia N, Mazzini N, Bold&i
P, Bacciglieri
P, Contenti E, Ferraresi G. Biofeedback
treatment
of gem-recurvatum
using an electrogoniometric
device with an acoustic
signal. Stand J Rehabil Med
1989;21:125-30.
Famcombe
PM. The Swedish Knee Cage: management
of the hyperextended
hemiplegic
knee. Physiotherapy
1980;66:33-4.
Isakov E, Mizrahi
IO, Susak Z. The control of genu recurvatum
by
combining
the Swedish Knee-Cage
and an ankle-foot
brace. Disabil
Rehabil 1992; 14: 187-91.
Rosenthal
RK, Deutsch
SD, Miller W, Schumann
W, Hall JE. A
fixed-ankle,
below-the-knee
orthosis
for the management
of genu
recurvatum
in spastic cerebral palsy. J Bone Joint Surg Am 1975;
57545-7.
Ferrigno
G, Pedotti A. ELITE:
a digital dedicated hardware
system
for movement
analysis
via real time TV processing.
IEEE Trans
Biomed Eng 1985;32:943-50.

Phys Med Rehabil

Vol77,

July

1996

RECURVATUM,

I8

I9

20

21

22

23

24

25

26

Kerrigan

Borghese
NA, Ferrigno
G. An algorithm
for 3-D automatic
movement detection
by means of standard
TV cameras.
IEEE Trans
Biomed Eng 1990;37:1221-5.
Pedotti A, Frigo C. Quantitative
analysis
of locomotion
for basic
research and clinical
applications.
Func Neurol Suppl 1992;7:4756.
Winter DA, Eng P. Energy generation
and absorption
at the ankle
and knee during fast, natural, and slow cadences.
Clin Orthop Rel
Res 1983; 175:147-54.
Huber PJ. The behavior
of maximum
likelihood
estimates
under
non-standard
conditions.
Proceedings
of the Fifth Berkeley
Symposium on Mathematical
Statistics
and Probability;
1965 June 21July 18. Berkeley,
CA: University
of California
Press, 1967; 1:22133.
White H. A heteroskedasticity
consistent
covariance
matrix estimator and a direct test for heteroskedasticity.
Econometrica
1980;48:
817-30.
Gonzalez
EG, Corcoran
PJ. Energy expenditure
during ambulation.
In: Downey
JA, Myers SJ, Gonzalez
EG, Lieberman
JS, editors.
The physiologic
basis of rehabilitation
medicine.
Ed. 2. Stoneham:
Butterworth-Heineman,
1994:413-46.
Corcoran
PJ, Jebsen RH, Brengelmann
GL, Simons BC. Effects of
plastic and metal leg braces on speed and energy cost of hemiparteci
ambulation.
Arch Phys Med Rehabil 1970;51:69-77.
Kenigan
DC, Viramontes
BE, Corcoran
PJ, LaRaia PJ. Measured
versus predicted
vertical displacement
of the sacrum during gait as
a tool to measure biomechanical
gait performance.
Am J Phys Med
Rehabil 1995;74:3-8.
Kenigan
DC, Gronley J, Perry J. Stiff-legged
gait in spastic paresis:
a study of quadriceps
and hamstrings
muscle activity.
Am J Phys
Med Rehabil I99 I ; 70:294-300.

Suppliers
a. Bioengineering
Technology
Systems,
Via Capecelatro,
66, 20148
Milan, Italy.
b. Advanced
Mechanical
Technology
Inc., AMTI,
I5 I California
Street,
Newton,
MA 02 158.
c. Epistat Services,
2011 Cap Rock Circle, Richardson,
TX 75080.
d. Stata Corporation,
702 University
Drive East, College Station, TX
77840.