Gait and Posture 12 (2000) 156 – 161

www.elsevier.com/locate/gaitpost

Leg length discrepancy — an experimental study of compensatory

changes in three dimensions using gait analysis
M. Walsh *, P. Connolly, A. Jenkinson, T. O’Brien
Gait Laboratory, Central Remedial Clinic, Clontarf, Dublin 3, Ireland

Received 30 January 2000; received in revised form 16 May 2000; accepted 3 June 2000

Abstract

Patients with leg length discrepancy (LLD) develop compensatory mechanisms, which result in kinematic alterations in the
lower limbs and pelvis. We investigated these compensatory mechanisms. Seven normal subjects underwent three-dimensional
(3-D) gait analysis using a CODA MPX 30® analyser. The subjects were fitted with raises of varying heights to one lower limb
and then analysed in static and dynamic modes. Pelvic obliquity was the most common mechanism used to compensate for small
degrees of leg length discrepancy up to 2.2 cm, particularly in static standing. With larger degrees of discrepancy, the subjects
developed flexion of the knee of the longer leg. Compensatory mechanisms during walking were more complex and included a
combination of kinematic changes at the pelvis, knee and ankle. We developed mathematical models for the compensatory
mechanisms at all the three levels. We conclude that 3-D gait analysis is a useful tool in the assessment of patients with functional
and structural leg length discrepancies. © 2000 Elsevier Science B.V. All rights reserved.

Keywords: Leg length discrepancy; Compensatory changes; 3-D gait analysis

1. Introduction
It has been suggested that lower limb inequality
greater than 2 cm affects at least one in every 1000
people [1,2] and the causes varied [3,4]. Differences in
leg length are divided conveniently into two categories,
structural and functional [5,6]. In structural leg length
discrepancy (LLD), there can be an alteration in length
of any of the lower limb segments. Functional leg
length discrepancy may be caused by a rigid or dynamic
contracture of a lower limb joint [5 – 7]. These deformi-
ties may be rigid or dynamic. When the deformity is
dynamic, as often occurs in patients with cerebral palsy
[8], it will not be possible to detect its effect by clinical
examination alone. Three-dimensional (3-D) gait analy-
sis can assess how a patient adapts to a structural or
functional LLD.
Patients with a LLD may accommodate for this by
changing the kinematic pattern of the lower limb joints,
both in static standing and dynamically during walking
[3,5,9–12]. The commonly used techniques are tilting
* Corresponding author. Tel.: + 353-31-8057431.
the pelvis in the coronal plane, flexing the hip and knee
on the long side and plantarflexing the ankle on the
short side in the sagittal plane. The extent to which
these techniques are used and how they accommodate
for a LLD have not been clearly defined.
In a recent study, we demonstrated that the unaf-
fected long limb in patients who have hemiplegic cere-
bral palsy had kinematic changes in the sagittal plane
which we attributed to a compensation for their LLD
[13]. Failure to fully extend the hemiplegic knee exacer-
bated the structural shortening of that limb while the
presence of an equinus contracture functionally length-
ened the involved limb. Such compensations in the
normal and involved limbs can equalise lower limb
length in both functional and structural leg length
discrepancy in patients with hemiplegic cerebral palsy.
It can, however, be argued that some of the kinematic
abnormalities in these patients may be related to the
underlying diagnosis of cerebral palsy. Similarly, most
studies, which have been performed on subjects with
established LLD are complicated by the existence of
other primary pathology [14–16].
The purpose of this study was to examine in 3-D the
main compensatory mechanisms of a normal popula-

0966-6362/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 9 6 6 - 6 3 6 2 ( 0 0 ) 0 0 0 6 7 - 9
 M. Walsh et al. / Gait and Posture 12 (2000) 156–161
tion in response to an artificially imposed LLD of
varying height. This would allow us to define the
normal compensatory mechanisms adopted by a nor-
mal population to an acutely imposed LLD. We also
set out to model mathematically the effect that these
compensatory mechanisms have on functional leg
length.
2. Subjects and methods
Seven normal subjects underwent a full clinical exam-
ination and clinical estimation of leg lengths. No signifi-
cant LLD i.e. greater than 0.5 cm was identified in any
subject. A graduated series of LLD ranging from 1 to 5
cm was created by applying a range of raises of differ-
ent heights to the sole of one foot using pelite. After the
raise had been applied, each subject underwent gait
analysis using the CODA MPX 30® system. After the
raise had been applied, the subjects were allowed to
acclimatise to the imposed discrepancy. Once they were
comfortable, each was asked to stand in front of the
CODA system. A static analysis of their relaxed stand-
ing posture was obtained (the Static Test). At the end
of the static test, each subject was asked to equalise
their lower limbs by going into equinus on the short leg
when wearing a 5-cm raise. This was performed to
check the validity of the mathematical model developed
for the ankle. Gait analysis was performed while the
subject walked at their comfortable walking velocity
and this velocity was standardised for each subject
using a metronome (the Dynamic Test). We analysed
the kinematic parameters for each lower limb joint for
discrepancies from 0 to 5 cm using an Excel spread-
sheet. A total of 60 individual parameters from the
long- and short-limbs were analysed, which included
joint positions at important events during the gait cycle
such as the instant of initial contact of both long- and
short-legs and mid to late swing of the long leg.
2.1. Mathematical calculations
2.1.1. Pel!ic accommodation for LLD
We defined the hip joint centres from the gait analy-
sis and designated pelvic width as X, the distance
between the centres and the angle of pelvic obliquity as
!. By trigonometry, pelvic accommodation can be cal-
culated as X sin !. Therefore, the ability to accommo-
date an LLD is governed by the following formula:
Pelvic accommodation for LLD
= sin (pelvic obliquity) pelvic width.
2.1.2. Knee kinematics to accommodate the LLD
In the static standing test, we were able to model
157
mathematically the effect that knee flexion had on
accommodating for, or producing, an LLD. If the
segment length of the femur is designated as a, the
segment length of the tibia as b, and the resultant
functional length of the lower limb as c, the following
formula allows a calculation of the effect of knee
flexion in altering the functional length of the lower
limb:
c 2 =a 2 + b 2 − 2ab cos(180°−knee flexion)
=a 2 + b 2 + 2ab cos(knee flexion)
We tested the validity of the mathematical models for
the pelvis and knee by applying the above formulae to
the mean values of pelvic obliquity and knee flexion
achieved during the static standing tests.
2.1.3. Alteration in ankle kinematics to accommodate
for an LLD
The functional length of the foot was designated as Y
and the angle of ankle plantarflexion was designated as
". By trigonometry, the functional height gain due to
the equinus can be calculated as Y sin ". Therefore, the
ability for an equinus to increase the functional length
of a leg is governed by the formula:
Increase in length
= (length of foot) sin(angle of plantarflexion)
We tested the validity of the mathematical model by
asking each subject to plantarflex their short side ankle
while wearing a 5-cm raise.
3. Results
3.1. Kinematic changes in stance and swing
The kinematic changes resulting from the induced
LLD from 1 to 5 cm are summarised in Tables 1–3. In
the static standing tests, subjects used two mechanisms
to compensate for LLD. These were pelvic obliquity
(tilted upwards on the long side) and knee flexion of the
long leg. The most significant changes in pelvic obliq-
uity occurred in accommodating between 2 and 3 cm of
discrepancy while the most significant changes in knee
flexion occur beyond 2 cm (Table 1).
In the dynamic tests, the subjects tended to use a
combination of separate mechanisms at various points
of the gait cycle, which lengthened functionally the
short leg and shortened the long leg. The kinematic
changes which occurred in both legs during the stance
phase of walking are shown in Table 2. The pelvis
displayed a gradual increase in obliquity (tilted up on
the long side) but never reached the values achieved in
the static test. During walking, the long side hip and
knee became more flexed while the long side ankle
158 M. Walsh et al. / Gait and Posture 12 (2000) 156–161

became more dorsiflexed. The ankle gradually reached
peak dorsiflexion later in the stance phase. Some sub-
jects displayed increased ankle pronation on the long
side as another method of compensating for the longer
limb but this was not consistent. On the short side, the
main changes during stance were an increase in knee
extension and ankle plantarflexion. As in the static
tests, the more significant changes in knee kinematics
occurred with greater degrees of LLD. However,
changes in ankle kinematics were seen even with small
discrepancies.
The main kinematic changes which occurred during
swing of the long leg are outlined in Table 3. On the
long limb, the hip and knee gradually became more
flexed to aid in limb clearance and the ankle on the
short side became less dorsiflexed (the short leg was in
stance).
3.2. Mathematical models
3.2.1. Pel!ic obliquity to accommodate for an LLD
In the static and dynamic tests, the pelvis accommo-

Table 1
Kinematic changes in response to applied raises during static standing testsa

Kinematic variables Height of raises applied (cm)

0 1 2 2.5 3 5

Pelvic obliquity — long limb −0.5 (1.6)b 3.2 (1.6) 4.8 (1.7) 5.6 (2.3) 6.1 (2.5) 5.1 (3.8)
Knee flexion — long limb 6.9 (3.5) 8.1 (3.9) 13.8 (8.5) 16.7 (9.9) 18 (11.3) 33.5 (4.2)

a
Mean value in degrees (S.D. in brackets).
b
A minus value indicates that the pelvis is tilted down.

Table 2
Kinematic changes during stance in response to applied raises during dynamic walking testsa

Kinematic variables Height of raises applied (cm)

0 1 2 2.5 3 5

Pelvic obliquity — long limb (at the instant 0.3 (2.2) 1.3 (1.9) 2.5 (1.4) 3.4 (1.6) 2.7 (1.7) 2.7 (1.7)
of ground contact by the short limb)
Hip flexion — long limb (at instant of 32.2 (3.6) 33.5 (3.2) 35.0 (3) 35.1 (3.1) 36.0 (3.8) 39.7 (4.5)
ground contact by the long limb)
Knee flexion — long limb (at the instant of 13.6 (5.4) 16.9 (4.5) 15.4 (3.6) 14.5 (5.6) 15.9 (2.7) 20.8 (4.5)
ground contact by the short limb)
Knee flexion — short limb (at the instant of 49.8 (11.3) 48.2 (8.7) 48 (7.1) 48.3 (7.4) 48.4 (7) 43.7 (7)
load acceptance by long limb)
Ankle flexionb — long limb (at the instant of 5.3 (3) 9.6 (3) 11.2 (2.2) 14.6 (3.6) 15.1 (5.2) 14.7 (5)
ground contact by the short limb)
Ankle flexionb — short limb (at the instant of −17.8 (6.9) −21.5 (5) −21.9 (5.8) −21.3 (6.3) −21.6 (4.9) −24.3 (9.7)
load acceptance by long limb)

a
Mean value in degrees (S.D. in brackets).
b
A positive value indicates dorsiflexion while a negative value indicates plantarflexion.

Table 3
Kinematic changes during swing in response to applied raises during dynamic walking testsa

Kinematic variables Height of raises applied (cm)

0 1 2 2.5 3 5

Hip flexion — long limb (at the instant of mid to late swing 31.9(3.6) 33.4(3) 36.1(2.4) 36.9(3.5) 36.6(3.4) 40.2(4.2)
of long limb)
Knee flexion — long limb (at the instant of mid to late 27.4(3) 27.7(3.4) 32.0(2.4) 31.3(2.1) 29.7(4.2) 35.2(4.6)
swing of long limb)
Ankle flexionb — short limb (at the instant of mid to late 9.3(3.4) 7.9(4) 6.5(4) 6.0(4.2) 6.7(2.8) 2.6(7.4)
swing of long limb)

a
Mean value in degrees (S.D. in brackets).
b
A positive value indicates dorsiflexion.
 M. Walsh et al. / Gait and Posture 12 (2000) 156–161
Fig. 1. Pelvic obliquity — tilting in the coronal plane.
dated for an induced LLD by tilting only in the coronal
plane i.e. by developing pelvic obliquity. The pelvic
correction for an LLD tilted the pelvis away from the
long leg (Fig. 1). In static standing, pelvic obliquity was
the most significant compensatory mechanism for dis-
crepancies between 2 and 3 cm. On average, the maxi-
mum pelvic obliquity developed by the subjects was
6.1°.
The amount of correction allowed by pelvic obliquity
was found to be independent of the actual leg lengths
or proportion of discrepancy in relation to the actual
leg lengths. For example, the subject with the shortest
stature was found to accommodate an LLD of 2 cm
just as adequately as the tallest subject (Fig. 2). We
found the distance between the hip joint centres (Fig.
3), which was calculated from the CODA MPX 30®
system, was the determining factor in accommodating
for an LLD. The wider the pelvis the greater the ability
to accommodate an LLD.
In our subjects, the average maximum pelvic obliq-
uity was 6.1° and the average distance between the hip
joint centres was 20.3 cm. Using the formula
Pelvic accommodation for LLD
= sin(pelvic obliquity)pelvic width
the average subject was able to accommodate for an
LLD of 2.2 cm by tilting the pelvis in the coronal plane.
Similarly, a fixed pelvic obliquity of 6.1° would be
expected to result in a functional LLD of 2.2 cm.
3.2.2. Alteration in knee kinematics to accommodate
for an LLD
At the knee, the most significant changes occurred in
the sagittal plane. In the static test, the long leg flexion
became significant when the LLD was greater than 2
cm. Similarly in the dynamic test, the long side knee
steadily became more flexed during stance especially
during double support and during swing to clear the
limb. The short side knee showed a significant and
steady increase in mid to late stance extension as the
Fig. 2. Diagram demonstrating how short stature accommodates for
an LLD as efficiently as a taller stature does.
159
Fig. 3. Mathematical model for the effect of pelvic obliquity.
discrepancy increased. There were no significant
changes in the coronal or transverse planes.
For example, a femoral length a of 45 cm and a tibial
length b of 40 cm, a knee flexion angle of 30° shortens
the functional length of the lower limb by 2.9 cm. From
the formula (Fig. 4)
c 2 = a 2 + b 2 −2ab cos(180°− knee flexion)
= a 2 + b 2 +2ab cos(knee flexion)
it can be seen that bending a knee by a further 30° will
have a much greater effect than bending from the fully
extended position. Similarly, straightening a fixed flex-
ion deformity of 30° at the knee will result in a func-
tional lengthening of 2.9 cm.
We tested the validity of the mathematical models for
the pelvis and knee by applying the above formulae to
the mean values of pelvic obliquity and knee flexion
achieved during the static standing tests (Table 4). The
combined predictions of height changes accommodated
by pelvic obliquity and knee flexion closely match the
applied LLD.
3.2.3. Alteration in ankle kinematics to accommodate
for an LLD
At the short leg ankle, there was a progressive ten-
dency towards equinus as LLD increased. The func-
tional effect of this was to lengthen the short leg. We
found that the amount of lengthening was determined
by the functional length of the foot (the distance from
the ankle joint centre to the head of the first
metatarso–phalangeal joint) and the degree of
plantarflexion (Fig. 5).
Increase in length
=(length of foot) sin(angle of plantarflexion)
For example, an equinus deformity of 20° with a
functional foot length of 15 cm would result in a
functional lengthening of the leg of 5.1 cm. Similarly,
correcting an equinus deformity of 20° will result in a
functional shortening of 5.1 cm. The maximum func-
tional increase due to an equinus deformity is equal to
the length of the foot and occurs when the angle of
plantarflexion equals 90°. The validity of the mathemat-
ical model is confirmed in the results in Table 5. It must
160 M. Walsh et al. / Gait and Posture 12 (2000) 156–161

Table 4
Comparison of height accommodation by pelvic obliquity and knee flexion (estimated from mathematical models) with height of actual raises
applied (n =7)a

Applied raise (cm) Pelvic obliquity Estimated accommodation by Knee flexion Estimated accommodation by Sum of
pelvic obliquity knee flexion accommodations

0 −0.5 −0.2 6.9 0.15 −0.05

1 3.2 1.1 8.1 0.2 1.3
2 4.4 1.6 13.8 0.6 2.2
2.5 5.6 2 16.7 0.9 2.9
3 6.1 2.2 18 1 3.2
5 5.1 1.8 33.5 3.6 5.4

a
Mean angles in degrees from static tests are used. Estimated heights in centimetres.

be noted, however, that this model is only valid when
the knee is straight. In many clinical circumstances, an
equinus deformity is accompanied by a knee flexion
contracture and the resultant effect on limb length is
more complex to model mathematically.
4. Discussion
In this study, we have defined how a number of
separate mechanisms, such as pelvic obliquity and
changes in the sagittal plane kinematics of the hip, knee
and ankle, are used by normal individuals to compen-
sate for an artificially induced LLD. These compensa-
tory mechanisms occur in both static standing and
walking, although there are important differences in
how the legs and pelvis adapt in these different circum-
stances. In our normal subjects, we observed that no
significant compensatory mechanisms to deal with LLD
occurred in the transverse plane.
The most common mechanism for dealing with mi-
nor degrees of limb length discrepancy was the induc-
tion of pelvic tilting in the coronal plane, pelvic
obliquity. In the normal individuals studied, the aver-
age maximum pelvic obliquity was 6.1°, which totally
accommodated for a LLD of 2.2 cm. The anatomical
structure of the pelvis and lumbar spine obviously
accommodates, easily and comfortably, for such small
leg length discrepancies. Pelvic obliquity appears to be
the common pathway for dealing with small degrees of
LLD up to about 2 cm. Our findings correlate well with
the clinical impression that patients with minor degrees
of LLD up to 2 cm tolerate the discrepancy well
without treatment [3,9]. This may be explained by the
fact that pelvic obliquity accommodates for such dis-
crepancies while larger discrepancies require the devel-
opment of significant knee flexion. While pelvic
obliquity does not require extra energy expenditure, the
maintenance of knee flexion puts considerable extra
pressure on the knee extensor mechanism.
With LLD which are greater than 2 cm, the coronal
plane mechanism is necessarily augmented by changes
in the sagittal plane, which occur at the hip and knee in
standing. These kinematic changes are accompanied by
alterations of sagittal plane kinematics of the ankles
during walking. The combination of these changes have
the effect of shortening the functional length of the long
limb both in the stance and swing phases while length-
ening the shorter limb during the stance phase. These
changes are best illustrated by the kinematic alterations

Fig. 4. Mathematical model for the effect of knee flexion. Fig. 5. Mathematical model for the effect of an equinus deformity.
 M. Walsh et al. / Gait and Posture 12 (2000) 156–161
Table 5
Estimated height accommodation by chosen equinus angles to accommodate for 5-cm raise using mathematical model of equinus (n= 7)a
Subject ID Equinus angle (°)
FD 16.7
JR 16.7
MMG 19
MW 15.4
PCO 16.8
PCA 17
LMG 20
a
Formula, sin(equinus angle)×functional foot length.
at the ankle. The long leg ankle demonstrates increased
dorsiflexion at terminal stance and a delay in the timing
of this peak, while the short leg ankle produces early heel
rise and greater degrees of plantar flexion during stance.
Many of the compensatory mechanisms which we
have identified in our normal population are seen in
pathological populations with LLD [15,16]. In cerebral
palsy, both structural and functional LLD is frequently
encountered especially in patients with hemiplegia and
asymmetric diplegia [13]. The great difficulty in this
population is distinguishing between primary pathology
and secondary compensations. This has even greater
implications when surgical procedures are being under-
taken. For example, a fixed equinus contracture in a
patient with hemiplegia and structural shortening may in
fact be providing the extra functional length to the limb,
which allows balanced walking. Similarly, knee flexion in
a patient with asymmetric diplegia may be a mechanism
for compensating for a shorter contralateral leg and not
necessarily due to a primary pathology affecting the
hamstrings.
Our mathematical models, for the calculation of the
effect of pelvic obliquity, knee flexion and ankle equinus
provide a scientific basis for planning of intervention in
patients with functional and structural LLD. Currently,
the most commonly used methods of assessing LLD and
its management involve static testing only. This applies
to both radiological and clinical assessment. It may be
more important to assess LLD and the response to leg
equalisation during dynamic walking rather than static
standing. This is particularly true in cerebral palsy where
many patients have dynamic spasticity. 3-D gait analysis
provides us with the best means of objectively assessing
the dynamic problem of LLD and the response to leg
equalisation in these patients.
161
Foot length (cm) Calculated height increase (cm)
15.2 4.4
16 4.6
15.8 5.1
16.7 4.4
15.9 4.6
16.1 4.7
15 5.1
References
[1] Guichet JM, Spivak JM, Trouilloud P, Grammont PM. Lower
limb length discrepancy-an epidemiologic study. Clin Orthop
1991;272:235 – 41.
[2] Rush WA, Steiner HA. A study of lower extremity length
inequality. Am J Roentgenol 1946;56:616 – 23.
[3] Menelaus MB. The Management of Limb Inequality. London:
Churchill Livingstone, 1991.
[4] Green WT, Anderson M. The problem of unequal leg length.
Pediatr Clin North Am 1955;2:1137 – 55.
[5] Baylis WJ, Rzonca EC. Functional and structural limb length
discrepancies: evaluation and treatment. Clin Pediatr Med Surg
1988;5:509 – 20.
[6] Danbert RJ. Clinical assessment and treatment of leg length
inequalities. J Manip Phys Ther 1988;11:290 – 5.
[7] Amstutz HC, Sakai DN. Equalization of leg length. Clin Orthop
1978;136:2 – 6.
[8] Gage JR. Gait Analysis in Cerebral Palsy. Oxford: Blackwell
Scientic Publications, 1991.
[9] Kaufman KR, Miller LS, Sutherland DH. Gait asymmetry in
patients with limb-length inequality. J Pediatr Orthop
1996;16:144 – 50.
[10] Liu XC, Fabry G, Molenaers G, Lammens J, Moens P. Kine-
matic and kinetic asymmetry in patients with leg length dis-
crepancy. J Pediatr Orthop 1998;18:187 – 9.
[11] Delacerad FG, Wikoff OD. Effect of lower extremity asymmetry
on the kinematics of gait. J Orthop Sport Phys Ther 1982;3:105–
7.
[12] D’Amico JC, Dinowitz HD, Polchaninoff M. Limb length dis-
crepancy. An electrodynographic analysis. J Am Podiat Assoc
1985;75:639 – 43.
[13] Allen P, Jenkinson A, Stephens M, O’Brien T. Abnormalities in
the uninvolved lower limb in children with spastic hemiplegia.
The effect of actual and functional leg length discrepancy. J
Pediatr Orthop, 2000;20:88 – 92.
[14] Bhave A, Paley D, Hazenberg J. Improvement in gait parameters
after legnthening for the treatment of limb length discrepancy. J
Bone Jt Surg 1999;81A:529 – 34.
[15] Abraham D, Dimon JH. Leg length discrepancy in total hip
arthroplasty. Orthop Clin North Am 1992;23:201 – 9.
[16] Gofton JP, Trueman GE. Studies in osteoarthritis of hip. Part II.
Osteoarthritis of hip and leg length disparity. Can Med Assoc J
1971;104:791 – 9.