Journal of Biomechanics 33 (2000) 415}425

The in#uence of mechanical stimulus on the pattern of tissue

di!erentiation in a long bone fracture * an FEM study
T.N. Gardner!,*, T. Stoll!, L. Marks", S. Mishra!, M. Knothe Tate#
!Oxford Orthopaedic Engineering Centre, Nuzeld Orthopaedic Centre, University of Oxford, Windmill Road, Headington, Oxford, OX3 7LD, UK
"New Technology Engineering Ltd, 14 Meadow View, Long Crendon, Bucks HP 18 9EQ, UK
#Institute of Biomedical Engineering and Medical Informatics University and Federal Institute of Technology, Zurich, Switzerland
Accepted 13 October 1999

Abstract

2D, coronal plane, "nite elements models (FEMs) were developed from orthogonal radiographs of a diaphyseal tibial fracture and
its reparative tissue at four di!erent time points during healing. Each callus was separated into regions of common tissue histology by
computerised radiographic analysis. Starting point values of tissue material properties from the literature were re"ned by the model to
simulate exactly the mechanical behaviour of the subject's callus and bone during loading. This was achieved by matching measured
inter-fragmentary displacements with calculated inter-fragmentary forces. Stress and strain distributions in the callus and bone were
calculated from peak inter-fragmentary displacements measured during natural walking activity, and were correlated with the
subsequently observed pattern of tissue di!erentiation and maturation of the callus. The growth and sti!ening of the external callus
progressively reduced the inter-fragmentary gap strain. Partial maturation of the gap tissue was apparent only one week before "xator
removal. Principal stresses in the callus were compared with &yield stresses' in corresponding tissue from the literature. This indicated
the presence of stress concentrations medial and lateral to the fracture gap, which probably caused tissue damage during normal
activity levels. Tissue damage may also have precipitated partial structural failure of the callus, both of which were believed to have
delayed healing during the middle third of the "xation period. Had the "xation device provided greater inter-fragmentary support
during early healing, this may have prevented callus failure and the consequent delay in healing. A further bene"t of this would have
been the reduction of the initially high intra-gap tissue strains to a magnitude more conducive to earlier maturation of the bridging
tissue that united the bone. ( 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Fracture; Finite Element; Displacement; Callus; Healing

1. Introduction (Hall, 1968; Yamagishi and Yoshimura, 1955), whereas

The in#uence of mechanical stimulus upon the healing
of bone fractures has been demonstrated in many diverse
studies. Experimental tests on model fractures in animals
(Yamagishi and Yoshimura, 1955; Cheal et al., 1991),
have examined morphological responses to applied
strain. These have suggested a strong correlation be-
tween the nature and magnitude of ambient stress or
strain in reparative tissue (callus), and its subsequent
histological development (McKibbin, 1978; Brighton,
1984). It is believed that an environment of mechanical
compression initiates the growth of cartilaginous tissue
* Corresponding author. Tel: #1-44-1865-227451; fax: #1-44-
1865-742348.
E-mail address: trevor.gardner@ooec.ox.ac.uk (T.N. Gardner)
combined cyclical tensile and shear stresses may produce
"brous collagenous tissue (Cheal et al., 1991).
Clinical evidence of the in#uence of mechanical stimu-
lus upon healing (Kenwright and Goodship, 1989; White
et al., 1977) suggests that periodical cyclical strain enhan-
ces healing when it is externally applied through cyclical
inter-fragmentary displacement soon after injury. Fur-
ther evidence suggests a temporal correlation between
naturally occurring inter-fragmentary displacement and
the growth and increasing stability of the external callus
(Gardner et al., 1997). In vitro studies on osteoblast-like
bone cells support the theory that mechanical strains are
transduced into cellular and molecular responses (Jones
et al., 1991; Jones and Bingmann, 1991; Pan and Einhorn,
1992).
Models for examining the correlation between mech-
anical stimulus and the pattern of tissue di!erentiation

0021-9290/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 1 - 9 2 9 0 ( 9 9 ) 0 0 1 8 9 - X
416 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425
have so far proved to be of limited use, because of the
di$culty of simulating accurately the mechanical envi-
ronment under which fractures normally heal. The prob-
lem for modelling is exacerbated since the environment of
one fracture may be substantially di!erent from another.
This is because of variability in the structure and ge-
ometry of fracture surfaces in bone, and in the fracture
callus that may surround the bone. Additionally, fracture
callus is heterogeneous and its material properties will
vary temporally; as will the shape and volume of the
callus. Further variability that adds to the complexity
of modelling is brought about by the type of "xation
device selected to stabilise the fracture. Di!erent types
and con"gurations of devices provide variable inter-frag-
mentary support (Gardner and Evans, 1992; Gardner et
al., 1996a) and inter-fragmentary loading not only varies
with support but varies according to an individuals'
weight, mobility, sensitivity to discomfort and functional
requirements. In summary, little commonality between
fractures exists, and therefore generalised models may be
uncharacteristic of individual fractures and will not simu-
late genuinely representative conditions.
Modelling studies reported in the literature adopt
a simpli"ed fracture geometry or apply arbitrary forces
or displacements, often with arbitrary material proper-
ties. Cheal et al. (1991) applied tensile displacement longi-
tudinally across 1 mm transverse sheep osteotomies and
used "nite element models (FEM) to calculate tissue
stresses and strains arising from bone-gap extensions
distributed linearly across the bone width from 10 to
100%. They were able to demonstrate that the morpho-
logy of fracture healing is dependent on local strain
magnitude, which in their case was a consequence of the
particular modelling conditions imposed. They observed
that callus tissue proliferated in regions of high strain, the
di!erentiation of soft tissue into bone occurred sooner in
regions of low strain, and bone resorption was found to
occur in the fragment ends where stresses were highest.
DiGioia et al. (1986) also modelled intra-gap stresses and
strains arising from displacements applied to sheep os-
teotomies. Severe strain gradients occurred over the
thickness of the gap, and radially between the endosteal
and periosteal surfaces at the gap. Predictably, strains
were heavily dependent on the extent of displacement
relative to gap size and were therefore a consequence of
the particular modelling conditions imposed. Neither
study sought to simulate natural conditions of loading to
observe corresponding histological e!ects on tissue. Beig-
ler and Hart (1992) modelled idealised fractures of ribs
and radii and assigned values of material properties
taken from the literature. They illustrated the in#uence of
callus geometry and material composition on the frac-
ture's response to loading, and to stress and strain distri-
bution in the callus. Although Meroi and Natal (1989)
used MRI data to model a femoral fracture supported by
a bi-lateral external "xator, arbitrary values were as-
signed for the material properties of the fracture and the
applied loading. Therefore, the correlation between the
support provided by the "xator and the physical re-
sponse of the callus could not be unequivocally estab-
lished.
Di!erent approaches have been made by Carter et al.
(1988) and Blenman et al. (1989) who examined the cor-
relation between the spatial distribution of &osteogenic
index' in the callus, and the pattern of bone formation
during endochodral ossi"cation. The osteogenic index
de"ned a combined stress condition, comprising relative
proportions of shear (deviatoric) and &hydrostatic'
(dilatational) stress. Again, fracture geometry was simpli-
"ed and the material properties were arbitrarily selected,
as were the forces. Therefore, as the load and displace-
ment environment of the model was not necessarily in
accordance with the geometry and properties selected for
the callus, such correlations between stress distributions
and patterns of bone formation may not necessarily be
found in real fractures. Further di!erent approaches have
been made such as retrieval studies, which have exam-
ined the interface at unicondylar knee components his-
tologically and by FEM (Giori et al., 1995). These have
been useful in showing that cartilaginous extracellular
matrix may be produced by cyclical dilatational stresses
of 0.7 MPa, and frequent distortional strains of 10% may
stimulate "brous extracellular matrix production. Mi-
cromotion devices implanted into the condyles of dogs
demonstrated by FE analysis the potential of mechanical
stimuli to regulate tissue di!erentiation (Prendergast et
al., 1997). Interdisciplinary research, involving experi-
mental fractures in animals, cell culture studies and FEM
studies have indicated that intramembranous bone
formation occurs when dilatational stresses and strains
are below 0.15 MPa and 5%, and endochondral ossi"ca-
tion occurs when stress exceeds this but strain remains
below 15% (Claes et al., 1998).
The new approach adopted by the present study was
to model accurately the natural conditions of loading
under which a real fracture healed, and to observe the
corresponding growth and maturation of the fracture
callus. A combination of direct measurement and math-
ematical modelling was used to develop computerised
(FEM) of the fracture at four temporal points during its
healing. Material properties of the callus tissue were
calculated by iterative analysis at each time point. Longi-
tudinal stresses within the tissue, arising from peak dis-
placements measured during walking activity, were
calculated and correlated with the pattern of tissue di!er-
entiation and the geometry of the callus assessed from
radiographs and from the derived tissue properties. Max-
imum (principal) stresses were calculated and compared
with failure stresses reported in the literature for corre-
sponding bone tissue to examine the in#uence upon the
callus of stress concentrations likely to cause tissue dam-
age (Perren and Cordey, 1977).
 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425
2. Method
The male subject of 20 yr of age had sustained a closed,
non-comminuted, oblique, diaphyseal fracture of the left
tibia, which was stabilised by an Ortho"x unilateral
external "xator with three 6 mm bone screws inserted in
the antero-medial cortex either side of the fracture. This
was the subject's "rst long bone fracture, he received no
additional therapy (such as hormonal) during treatment,
and he was not a habitual smoker or drinker. The follow-
ing measurements were taken at 4, 8, 12 and 16 weeks
post-fracture to provide data for developing 2D, plane
stress, FEM at these four temporal points during healing.
2.1. Displacements and forces at the fracture
3D inter-fragmentary displacement was measured dur-
ing two-legged stance and during walking at the subject's
routine gait frequency of 0.5 Hz. This was done by an
instrumented spatial linkage (Gardner et al., 1996b) at-
tached to the inner pair of bone screws across the frac-
ture. The instrument used six magnetic "eld Hall e!ect
devices to measure displacement in six degrees of free-
dom. Displacements at the linkage were translated trig-
onometrically to monitor 3D motion of the distal bone
fragment in relation to the proximal at the centre of the
fracture. A test simulation using a vernier measurement
jig established that displacements measured by the in-
strument are accurate to within$0.025 mm and$0.025
(Gardner et al., 1996b). Tibial forces and moments at the
fracture during two-legged stance were calculated from
measured ground reaction forces, using a correlation
between the two which was developed from a lower limb
biomechanics model (Lu et al., 1997b,c). This is a 3D
mathematical model of the lower limb, with the knee as
a parallel spatial mechanism, the hip as a ball and socket
joint, and the ankle as a two-revolute joint. 36 muscles in
the lower limb are included, and the knee extensor mech-
anism is characterised as a patellofemoral joint model
with a point patella. A solution procedure is used to
calculate forces transmitted by the muscles, ligaments
and bones (Collins, 1995). Validation tests of the bi-
omechanics model have been carried out on two tumour
patients with instrumented proximal femoral prostheses
(Lu et al., 1995,1996,1997a,b), and calculation of the axial
forces in the femora agreed well with the measured forces
during standing.
2.2. Fracture geometry and callus histology
Antero-posterior radiographs of the fractured bone
and callus were scanned and digitised to provide 2D
geometry of the computerised fracture models that repre-
sented the four time points. Intervals of colour scale
range were displayed sequentially to establish the bound-
aries between regions of common image intensity. As
417
spatial variation in image intensity correlates with spatial
variation in X-ray absorption arising from di!erences in
the tissue properties of density and atomic number (De-
nby and Heaton, 1987), it was concluded that regions of
common image intensity possessed common tissue prop-
erties. Since regional patterns (Fig. 1(i) * peripheral,
central and adjacent callus) conformed with those identi-
"ed by Sarmiento and Latta (1995), it was concluded that
regional tissue histology in these four models also con-
formed to their descriptions. Peripheral callus is de-
scribed as the dense well-orientated "brous layer that
progresses to new bone relatively quickly and is bounded
externally by the new periosteum. Central callus which
bridges the proximal and distal fragments is initially
formed from soft, "brous and cartilaginous tissue that is
"nally replaced by hard calci"ed tissue during endochon-
dral ossi"cation. Adjacent callus is attached to the intact
cortical bone at the periosteum and endosteum in a loose
trabecular structure of relatively mature tissue that forms
rapidly and converts from soft to hard tissue. Maturation
of the tissue in each region was evident from progressive
increases in the regional properties of the callus that were
derived by the model, and was interpreted histologically
in the light of the regional histology and maturation of
tissue de"ned by Sarmiento and Latta (1995).
2.3. FE modelling
Computerised 2D geometric models of the coronal
plane were provided for FE analysis by mapping the
internal and external boundaries of the bone and the
three regions of callus at each of the four temporal points.
Cosmos/M (Structural Research and Analysis Corpora-
tion, Los Angeles, CA, USA) performed the FE analysis,
and between 4517 and 5026 elements were used with
non-uniform mesh density which was speci"ed for each
region depending on complexity of shape. Triangular
elements were used since square or quadrilateral ele-
ments were not able to "t the awkward shapes. As the
average element in the callus was small, discontinuity at
element borders could generally be neglected and the
system considered compatible. Element thickness was
found by calculating the thickness of an equivalent rec-
tangular transverse section through the model fracture.
The equivalent section had the same second moment of
area as the transverse section at mid-height through the
real fracture that was reconstructed from orthogonal
radiographs. Measured inter-fragmentary displacements
were applied to the centre of the model fracture at the end
of a rigid beam element that was connected to the upper
bone pin insertion point (Fig. 1(ii)). Reaction forces and
moments at the "xed lower insertion point, arising from
these displacements, were calculated by plane stress, FE
analysis consistent with a material of linear elastic, iso-
tropic, monophasic behaviour. This was considered to be
valid since DiGioia et al. (1986) showed that accurate
418 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425
Fig. 1. (i) External and inter-regional boundaries of the tibial fracture
callus in the coronal plane at the four temporal points in healing. The
three callus regions of common tissue histology are central (c), adjacent
(a) and peripheral (p). The peripheral callus, medial (left) to the fracture,
is discontinuous at week 12, probably as a result of the additional
angulation and translation in fragment alignment that took place
between weeks 8 and 12. Reprinted with permission from Gardner et al.
(1998). (ii) The external "xator was "xed to the antero-medial cortical
surface of the fractured tibia (a). Boundary conditions of the model (b).
The distal fragment was &"xed' distally at the lower pin group insertion
point. Displacement was applied to the proximal fragment at the distal
end of a rigid bar, which was "xed at the upper pin group insertion
point.
results could be obtained for strains of less than 10%. In
the present study, models for weeks 8, 12 and 16 exhib-
ited strains of approximately 10% or below, while at
4 weeks only localised regions of the central callus ex-
ceeded this. The FEMs of the present study have been
presented brie#y by Gardner et al. (1998), in a discussion
of the modelling technique.
2.4. Analysis
Initial starting point values for regional material prop-
erties (Table 1) were designated from the literature for
each of the four FEMs (Perren and Cordey, 1980; Duck,
1990; DiGioia 1986; Davy and Connolly, 1982; Huber-
Betzer et al., 1990). Reaction forces and moments
calculated by each FEM, arising from displacements
measured during two-legged stance, were compared with
the corresponding forces and moments calculated using
the lower limb biomechanics model. Disparities between
the two were gradually diminished by iteratively re"ning
the elastic moduli of each region by the same factor until
good agreement was achieved by the FEMs. The re"ned
material properties were assigned to each region at each
time point, and the corresponding peak displacements
measured during walking were then applied to the four
displacement-driven FEMs to calculate stress (longitudi-
nal and principal) on the callus and bone. Comparisons
were made between the greatest principal stresses occur-
ring in each region and the yield and ultimate stresses
extrapolated from Reilly and Burstein (1975) and Davy
and Connolly (1982) for tissue of corresponding material
properties and histology (Table 2). Values for the central
callus are omitted since radiographs indicated the pres-
ence of soft immature tissue up to week 16, which re-
duced its susceptibility to yield.
3. Results
3.1. The pattern of callus growth and tissue diwerentiation
during healing
Fig. 1(i) shows the external and inter-regional bound-
aries of the oblique fracture and callus at the four
time points in healing. These are coronal plane models,
with the lateral surface on the right. Almost the full
extent of the central and adjacent callus (endosteally
and periosteally) was evident at week 4. Whereas the
soft "brous layer of peripheral callus, which appeared
initially at the periosteal surface of the fragments, was
not fully formed at the periphery until week 8. At week 12
it was no longer intact medial to the fracture, probably
due to partial "xation failure between weeks 8 and 12.
Failure was indicated by sequential radiographs that
showed a further 23 of angulation between bone
fragments, and a further 1 mm of relative transverse
 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425 419

Table 1
Tissue histology and Young's moduli of the three regions of fracture callus, at 4, 8, 12 and 16 weeks post-fracture

Stage CENTRAL Modulus (MPa) ADJACENT Modulus (MPa) PERIPHERAL Modulus (MPa)
(weeks) tissue type tissue type tissue type
Initial "nal initial "nal initial "nal

4 Haematoma, granulated 0.05 0.19 Soft connective tissue 1.0 3.8 Soft "brocartilage tissue 20 76
tissue with invading
vasculature
8 Fibrous perichondrial 20 28 Woven bone, 25% 500 700 Dense "brous tissue, 2000 2860
tissue maturation 45% maturation
12 Fibrous perichondrial 30 30.6 Woven bone, 25% 900 765 Dense "brous tissue, 3600 3060
tissue maturation 45% maturation
16 Fibrous cartilage tissue, 30 75 Woven bone, 60% 2000 5000 Bone, 100% maturation 8000 20 000
10% maturation maturation

Table 2
Yield stresses and ultimate stresses extrapolated from the literature (C"compression, T"tension)

Week Adjacent (MPa) Peripheral (MPa)

Yield Ultimate Yield Ultimate

C T C T C T C T

8 8.6 4.5 8.9 5.9 32.5 18.2 33.2 22.8

12 9.4 4.9 9.7 6.5 36.0 20.2 36.6 25.1
16 57.4 33 58.3 40.4 196 121 198 140.6

translation. Subsequently, there was an increase in the
size of the adjacent and central callus at this location,
and also in the adjacent callus lateral of the fracture at
week 16.
3.2. Tissue histology and stiwness
Table 1 gives "nal values for the elastic properties of
tissue at each of the four time points (previously sum-
marised in a review paper by Kenwright and Gardner,
1998); in each case these were derived from no more than
four re"nement iterations. At week 4, re"ned values of
elastic moduli (0.19, 3.8, 76 MPa) were four times the
initial starting point estimates provided from the litera-
ture (0.05, 1.0, 20 MPa). Since they were several orders of
magnitude less than the modulus of cortical bone (18.4
GPa), tissues were not sti! enough to contribute appreci-
ably to tibial load bearing. Values suggest that the central
callus had changed little, progressing from haematoma
and granulation tissue at week 4 to slightly sti!ened
tissue at week 16 indicating the commencement of matu-
ration. Adjacent callus appeared to have rapidly produc-
ed sti! tissue by week 8, for which the delineated regional
histology indicated partially calci"ed woven bone. Peri-
pheral callus properties at week 8 and regional histology
indicated partially calci"ed "brous tissue, able to support
compressive load (as will be seen in the next section).
Between weeks 8 and 12, the material properties of tissue
ceased to increase in all three regions. Although a small
change in shape and volume of callus occurred medial of
the fracture, increases in volume and progress in sti!en-
ing were otherwise inhibited during the period fragment
angulation and translation occurred. From week 12 to
16, progressive sti!ening enabled the adjacent and peri-
pheral callus to contribute increasingly to compressive
load bearing.
3.3. Longitudinal stresses and the pattern of changing load
transfer through the fracture
Re"ned values of the material properties (Table 1) were
assigned to each region at each time point, and the
corresponding peak displacements measured during
walking (Table 3) were applied to each of the four dis-
placement-driven FEMs to calculate tissue stresses.
Fig. 2 shows contours of maximum stress longitudinal to
the bone, in line with the predominant force. These "g-
ures illustrate load transfer routes through the bone and
callus by unbroken paths of elevated compressive stress.
At week 4, all three regions of the callus de"ned in
Fig. 1(i) exhibited low stresses; this was because tissue
moduli at this point were too low to enable the callus to
contribute signi"cantly to tibial load bearing. Load was
supported by direct load transfer arising from fragment
420 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425
Table 3
Displacements measured during walking. (y (longitudinal), x (trans-
verse), Z (rotational in x}y plane)
Week x (mm) y (mm) Z (radians)
4 0.283 !0.377 !0.00329
8 !0.015 !0.159 0.00201
12 !0.092 !0.260 0.00169
16 !0.129 !0.111 !0.00215
end-bearing contact; this was indicated by compressive
stress paths through the fracture fragments and stress
concentrations within their abutting ends. By week 8,
load-bearing paths were predominantly diverted through
the external callus as a result of its increasing sti!ness.
The external callus (outer adjacent callus and sections of
the peripheral callus) exhibited compressive stresses of up
to 10 MPa, while the compressive stress concentrations
at the fragment ends reduced to below 2 MPa. Tensile
stresses of up to 1 MPa predominated in the bulk of the
adjacent callus at the endosteum and close to the lateral
external cortical surface of the bone. No signi"cant
change in the longitudinal stress patterns occurred be-
tween week 8 and 12, but by week 16 the whole of the
adjacent callus (endosteally and periosteally) was in com-
pression and supported weight bearing. Stresses in the
central callus became tensile at week 8 and remained low
throughout healing until the stabilising device was re-
moved at week 17.
3.4. Maximum stress levels * the inyuence of yield
(tissue damage and callus failure)
Fig. 3 shows maximum principal stresses at each tem-
poral point in healing. Data inspection showed that lo-
calised peak values increased from!0.06 to!0.33 MPa
(compression) in the central callus, from !0.11 to 11.3
MPa (compression to tension) in the adjacent callus and
5.2 to 8.2 MPa (tension) in the peripheral callus. Local-
ised compressive stress concentrations were con"ned to
cortical bone at the abutting fragment ends, to the central
callus between opposing bone fragments, and to the
central callus near the external cortex. Stresses in the
callus were below yield at weeks 4 and 16, but exceeded
yield at weeks 8 and 12. At week 8, yield stress (4.5 MPa)
and ultimate stress (5.9 MPa) were exceeded in tension as
a result of relatively large tissue distortion within the
adjacent callus close to the medial end of the central
callus. At week 12, yield stress (4.9 MPa) and ultimate
stress (6.5 MPa) were exceeded medially, and also
laterally, in the adjacent callus at the cortical surface of
the proximal fragment. Locations of yield medial to the
fracture gap at weeks 8 and 12 corresponded with small
additions to the callus volume observed here at weeks 12
and 16. Similarly, callus growth was observed at weeks 12
and 16 subsequent to yield stresses lateral to the fracture
(seen at week 12).
4. Discussion
4.1. Stages of callus maturation (Brighton, 1984) and
mechanical function
It can be seen from longitudinal stress paths at week
4 that during the initial period of the &soft callus stage'
load was largely transferred longitudinally between frac-
ture fragments across the inter-fragmentary gap. Gar-
dner et al. (1998), who previously reported peak strain in
the intra-gap tissue of these four models, showed that
strain was as great as 70% at week 4. During the
early part of the &hard callus stage' (commencing before
week 8), compressive load was increasingly diverted
through stress paths in the external callus at the periph-
ery as a result of sti!ening of the tissue. By the middle of
the &hard callus stage' (at week 12), load was no longer
transferred across the gap, but was diverted wholly
through the external callus. By week 16, the external
callus had provided su$cient inter fragmentary stability
to reduce strains substantially in the intra-gap connective
tissue to below 2%, the level reported by Gardner et al.
(1998). This reduction probably initiated the "nal intra-
gap bone repair process which ultimately united the
fracture, since a low-strain environment is more conduc-
ive to the di!erentiation and maturation of hard tissue
(Cheal et al., 1991). This process was evident for the "rst
time at week 16 from the presence of sti!ening intra-gap
tissue towards the end of the &hard tissue stage', as one
would expect endochondral ossi"cation to commence
replacing the soft "brous and cartilage tissue with hard
calci"ed tissue (Sarmiento and Latta, 1995). Although
within one week of this the bone was considered to be
&clinically healed' (at week 17) and the "xator was re-
moved, the "nal &remodelling stage' was probably de-
layed by a post-"xation extension of the &hard tissue
stage'. Observations on similar fractures with the same
"xation (Gardner et al., 1997) have shown that further
callus growth occurs peripherally in response to the addi-
tional mechanical stimulus that is provided by the in-
creased angular inter-fragmentary motion arising from
"xator removal.
4.2. The inyuence of tissue damage upon healing
The FE models revealed the in#uence of tissue damage
due to high stresses and callus failure upon the pattern of
tissue di!erentiation and maturation. Damage to the
external callus tissue probably occurred in locations of
stress concentrations that exceeded tissue failure stress.
These occurred medial and lateral to the fracture gap at
weeks 8 and 12, where subsequently there was a change
 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425 421

Fig. 2. Contours of peak stresses longitudinal to the bone seen during walking at the four temporal points in healing. At week 4 (a), stress
concentrations are evident within the ends of the cortical bone fragments, due to end bearing as the fracture gap compresses during walking. At week
8 (b) and week 12 (c), tensile stresses predominated in the adjacent callus, and the outer areas of the external callus were under greater compressive
stress than were the abutting fragments. This indicated that tibial load was being diverted through the callus rather than across the fracture.

in callus shape and an increase in size between weeks
8 and 16. The angular and transverse inter-fragmentary
translation, which indicated callus failure between week
8 and 12, may have been precipitated by this damage and
by partial failure of the "xator frame. As the sti!ness of
the bulk of the adjacent and peripheral callus remained
constant between week 8 and 12, the drive towards the
bone-bridging stage would appear to have been inhibited
during a callus damage and repair phase. Damage,
repair and stress redistribution that dissipated the stress
422 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425

Fig. 3. Contours of maximum principal stresses are shown. At week 8 (b) and week 12 (c), stress concentrations in the adjacent callus close to the
medial end of the gap (and the lateral end at week 12 only) exceeded tensile yield stress measured in similar tissue. Unrecoverable displacement here
would have damaged the tissue, possibly weakening the fracture callus. This weakening resulted in further bone fragment angulation, apparently
stimulating the subsequent growth of callus external to these locations.

concentrations was "nally achieved between week 12 and external callus (evident at week 16) subsequent to the
16. The loss of callus stability arising from the suspected damage repair stage. Thus the drive towards the bridging
callus failure between week 8 and 12 probably provided stage at the inter-fragmentary gap appears to have been
the mechanical stimulus for the increase in volume of the restored by week 16. A parallel may be drawn between
 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425
this stimulus and that which initiated the original callus
proliferation in response to the loss in stability caused by
the fracturing of the bone.
This interpretation of healing pattern is further sub-
stantiated by the maximum strains in the central callus
reported by Gardner et al. (1998). Strain almost doubled
(from 6 to 11%) between week 8 and 12 because of the
reducing stability of the callus, but subsequently as stab-
ility was recovered strain reduced (to below 2%) by week
16. Whether stress concentrations caused tissue damage
and this inhibited healing between week 8 and 12, or
whether they precipitated the mechanical failure of the
callus which was responsible for inhibited healing, is
unclear. However, it is extremely likely that high stress
concentrations were either directly or indirectly respon-
sible for a delay in healing.
4.3. Tissue properties
Material properties of the callus tissues were compared
with corresponding values obtained from the literature.
These were estimated from Markel et al. (1990) by nor-
malising their canine callus indentation sti!nesses with
those of cortical bone, and by assuming that these ratios
apply also to the relative values of elastic moduli in
human callus and bone. Elastic moduli of the present
study compared well with those derived from Markel et
al. for &adjacent' and &peripheral' callus at corresponding
time points, allowing for some di!erence in callus re-
gionalisation, but the &central' callus was substantially
di!erent in the latter stages (weeks 12 and 16). For
example, at week 12 the elastic moduli of the central
callus region (30.6 MPa * Table 1) was substantially
lower than the modulus of &gap' tissues (2700 MPa) at the
corresponding point (8 weeks) derived from Markel et al.
To examine the in#uence of this di!erence upon the
FEM solution, the model was re-analysed at 12 weeks
using moduli produced from Markel et al. It was found
that the new solution, produced largely the same pattern
and magnitudes of longitudinal stress in the &adjacent'
and &peripheral' callus as the original solution. It also
provided similar concentrations of principal stress that
exceeded yield stress in the &adjacent' callus medial and
lateral to the &central' region. It was therefore concluded
that observations based on the original solution remain
valid since the model is not sensitive to variations in the
sti!ness of central callus, probably because the large
volume of relatively sti! external callus dominates tibial
load bearing. This is advantageous, as it is not possible to
re"ne the ratio between the values of elastic moduli in
each region of tissue, since it arises from the relationship
between the starting point values extrapolated from the
literature and maintained during the re"nement process.
However, it is worth noting that the ratio between the
radiographic image intensities (30 : 10 : 1) of each region
of callus (peripheral : adjacent : central) at week 12 was
423
more consistent with the ratio between the initial starting
point values of regional elastic moduli than with Mar-
kel's moduli. If it is assumed that X-ray absorption is
proportional to tissue density, and elastic modulus is
approximately proportional to density, then regional
moduli predicted by relative intensities are 6000, 2000
and 200 MPa, respectively, which are not too dissimilar
to the re"ned values produced by the model of 3600, 900
and 36 MPa.
4.4. Fracture treatment
The pattern of healing suggests that the primary mech-
anical function of the callus was to reduce routine intra-
gap strains to a level at which mature bone could directly
bridge the inter-fragmentary gap to unite the bone. Two
questions arise from this observation. Firstly, if the ex-
ternal "xator had reduced maximum intra-gap strains
closer to the required level prior to the &soft callus stage',
would this have produced a smaller volume of support-
ing callus that would have taken less time to form?
Secondly, how would this have in#uenced the inhibitive
in#uence on healing caused by stresses above yield, the
risk of callus fracture, and the corresponding delays to
healing? Since mechanical stimulus in#uences callus de-
velopment that modulates the original mechanical stimu-
lus, the in#uence upon healing of this approach can only
be clari"ed by similarly modelling fractures that are
managed in this way. However, parallels provided by
other treatment methods may suggest a resulting pattern
of healing that is more characteristic of #exible, rather
than rigid, plate "xation (Cheal et al., 1991).
4.5. Limitations of the study
Although the study examines a single subject for whom
conclusions are directly relevant, it should be possible to
extrapolate the mechanical stimuli and tissue responses
detected to fracture healing mechanisms in general. Be-
fore doing so, technical di$culties in modelling proced-
ure need to be appraised in the light of conclusions drawn
from the models.
The solution of an FEM that simulates real geometry
and behaviour is often sensitive to errors in de"ning the
model shape and its material properties. As the material
properties were derived in this study rather than esti-
mated, and as Sarmiento and Latta (1995) showed that it
is reasonable to assume that tissue is largely homogene-
ous within the regions of callus that they de"ned,
errors in this study due to modelling were con"ned to
the process of locating these regional boundaries.
These boundaries were at times partially obscured, be-
cause radiographic absorption was in#uenced by the
variability in transmission path lengths through callus
tissue and bone in addition to the spatial variation
in material properties that was under examination.
424 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425
However, variations in radiographic image intensity due
to tissue material properties were generally more abrupt
than were variations due to transmission path lengths;
this kept accuracy of boundary location to within
$1 mm, despite the presence at these boundaries of
tissues of transitional histology. The image analysis tech-
nique also compensated for di!erences in magni"cation
and exposure between radiographs.
A direct approach was taken to model the mechanical
behaviour of the callus tissue, consistent with a linear
elastic, isotropic, monophasic material. Therefore, appro-
priate starting point values for material properties were
available from the literature, and these were re"ned by
the models to reproduce the actual load versus displace-
ment behaviour of the subject's callus and bone at each of
the time points. This would not have been possible if
a porohyperelastic approach (Simon, 1990) or a bi-
phasic, mixture, penalty approach (Spilker et al., 1990)
had been adopted, because this would have necessitated
the arbitrary selection of the values of at least "ve
variables to de"ne callus and bone behaviour. Addition-
ally it is probable that the bulk, elastic, monophasic
behaviour of the callus modelled in this study, is a valid
simulation of the rate of loading that occurs during
walking. Since inter-fragmentary displacement was
found to be instantly recovered during the unloaded
sequence of walking, physiological loading rates were
apparently too rapid for signi"cant relative motion to
have occurred between solid and #uid phases as a result
of a &drained' poroelastic response (Zienkiewicz and
Bettess, 1982). This leaves only a complex porohyperelas-
tic undrained response as a reasonable alternative, for
which a process of re"nement of the material properties is
not possible.
Model analysis to examine tissue properties in the
two-legged stance test did not account for the unknown
contributions to fracture stability made by the unilateral
external "xator frame and by the fractured "bula. This is
because a previous study on this type of fracture and
"xation had shown that at 4 weeks post-fracture less than
10% of the predominantly axial tibial load during two-
legged stance was supported by the frame; the balance of
load being transferred directly across the fracture be-
cause of inter-fragmentary gap closure (Gardner and
Evans, 1992). Thereafter, the "xator contribution to load
bearing was further reduced because of the sti!ening of
the external callus. Conversely, the axial tibial load car-
ried by the healing "bula would have increased progress-
ively up to 7% of tibial load by week 16 (Goh et al., 1992).
Overall therefore, the error arising from omitting the
"xator and "bula from the model was estimated to be less
than 10% of the tibial load at each of the four temporal
points during healing.
Finally, as the lower limb biomechanics model re-
vealed that forces at the fracture were predominantly
compressive during two-legged stance and at peak load
during the stance phase of walking, angular, transverse
and torsional contributions to load bearing in the model
were considered to be negligible. Taking all the above
considerations into account, overall accuracy of the
model is di$cult to quantify but the procedures adopted
should ensure that results are valid and meaningful. Cer-
tainly the patterns of stress and strain distribution are
accurate. Also, if absolute values are within 20%, the
conclusions drawn from results remain valid in regard to
the general identi"cation of tissue, the spatial distribu-
tion of stresses and strains and the location of stress
concentrations that exceed yield. Therefore, it has been
possible for the present study to determine the probable
mechanical function of each region of the callus at each of
the four stages in healing by modelling the actual
load}displacement behaviour of the callus and bone. It
has also been possible to correlate the timing of callus
growth and maturation with the stimulus provided by
the original loss of stability caused by the bone fracture,
the concentrations of stress and strain and the resulting
tissue damage and failure of the callus.
Acknowledgements
The authors wish to thank Professor Kenwright and
Professor Simpson of the Nu$eld Department of Or-
thopaedic Surgery, University of Oxford, for clinical ad-
vice.
References
Beigler, F.B., Hart, T.H,. 1992 Finite element modelling of long bone
fracture healing. In: Middleton, J. (Ed.), Proceedings of the Interna-
tional Symposium on Computer Methods in Biomechanics and
Bioengineering (Ed.), Middleton J., University of South Wales,
Swansea, pp. 30}39.
Blenman, P.R., Carter, D.R., Beaupre, G.S., 1989. Role of mechanical
loading in the progressive ossi"cation of a fracture callus. Journal of
Orthopaedic Research 7, 398}407.
Brighton, C.T., 1984. Principles of fracture healing. In: Murray, J. A.
(Ed.). Instructional Course Lectures. Vol. 33, C. V. Mosby Co., St.
Louis. pp. 60}82.
Carter, D.R., Blenman, P.R., Beaupre, G.S., 1988. Correlations between
mechanical stress history and tissue di!erentiation in initial fracture
healing. Journal of Orthopaedics Research 6, 736}748.
Cheal, E.J., Mansmann, K.A., DiGioia III, A.M., Hayes, W.C., Perren,
S.M., 1991. Role of interfragmentary strain in fracture healing: ovine
model of a healing osteotomy. Journal of Orthopaedics Research
9 (1), 131}142.
Claes, L.E., Heigele, C.A., Neidlinger-Wilke, C., Kaspar, D., Seidl, W.,
Margevicius, K.J., Augut, P., 1998. E!ects of mechanical factors on
the fracture healing process. Clinical Orthopaedics and Related
Research 355, 132}147.
Collins, J.J., 1995. The redundant nature of locomotion optimization
laws. Journal of Biomechanics 28 (3), 251}267.
Davy, D.T., Connolly, J.F., 1982. The biomechanical behaviour of
healing canine radii and ribs. Journal of Biomechanics 15 (4),
235}247.
 T.N. Gardner et al. / Journal of Biomechanics 33 (2000) 415}425
Denby, P.P., Heaton, B., 1987. Physics for Radiologists. Blackwells
Scienti"c Publications, Oxford, UK.
DiGioia III, A.M., Cheal, E.J., Hayes, W.C., 1986. Three-dimensional
strain "elds in a uniform osteotomy gap. Journal of Biomechanical
Engineering 108, 273}280.
Duck, F.A., 1990. Physical Properties of Tissues. Academic Press,
London, UK.
Gardner, T.N., Evans, M., 1992. Relative sti!ness, transverse displace-
ment and dynamisation in comparable external "xators. Clinical
Biomechanical 7, 231}239.
Gardner, T.N., Evans, M., Kenwright, J., 1996a. The in#uence of ex-
ternal "xators on fracture motion during simulated walking. Jour-
nal of Medical Engineering and Physics 18 (4), 305}313.
Gardner, T.N., Evans, M., Kyberd, P. J., 1996b. An instrumented
spatial linkage for measuring the three dimensional displacement of
one point relative to another. Journal of Biomedical Engineering
American Society of Mechanical Engineering 118, 586}594.
Gardner, T.N., Hardy, J.R.W., Evans, M., Kenwright, J., 1997. Tem-
poral changes in dynamic inter fragmentary motion and callus
formation in fractures. Journal of Biomechanics 30 (4), 315}321.
Gardner, T.N., Stoll, T., Marks, L., Knothe-Tate, M., 1998. Mathemat-
ical modelling of stress and strain in bone fracture repair tissue. In:
Middleton, J., Pandy, G.N. (Eds.), Computer Methods in Bi-
omechanics and Biomedical Engineering, Vol. 2. Gordon and
Breach Science Publishers, London, pp. 247}254.
Giori, N.J., Ryd, L., Carter, D.R., 1995. Mechanical in#uences on tissue
di!erentiation at bone}cement interfaces. Journal of Arthroplasty
10 (4), 513}522.
Goh, J.C.H., Lee, E.H., Ang, E.J., Bayon, P., Pho, R.W.H., 1992. Bi-
omechanical study on the load-bearing characteristics of the "bula
and the e!ects of "bula resection. Clinical Orthopaedics and Re-
lated Research 279, 223}228.
Hall, B.K., 1968. In vitro studies on the mechanical evocation of
adventitious cartilage in the chick. Journal of Experimental Zool-
ogy 168, 283}306.
Huber-Betzer, H., Brown, T.D., Mattheck, C., 1990. Some e!ects of global
joint morphology on local stress aberations near imprecisely reduced
intra-articular fractures. Journal of Biomechanics 23 (8), 811}822.
Jones, D.B., Bingmann, D., 1991. How do osteoblasts respond to
mechanical stimulations. Cells and Materials 1 (4), 329}340.
Jones, D. B., Nolta, H., Scholubbers, J-G., Turner, E., Veltel, D. 1991.
Biochemical signal transduction of mechanical strain in osteoblast-
like cells. Biomaterials 12, 101}110.
Kenwright, J., Gardner, T.N., 1998. Mechanical in#uences on tibial
fracture healing. Clinical Orthopaedics and Related Research 355S,
179}190.
Kenwright, J., Goodship, A.E., 1989. Controlled mechanical stimula-
tion in the treatment of tibial fractures. Clinical Orthopaedics Re-
lated Research 241, 36}47.
Lu, T.W., Taylor, S., Gill, H.S., Walker, P.S., Ling, R.S.M., O'Connor,
J.J., 1995. Gait analysis of patients with instrumented massive
proximal femoral prostheses. In: Proceedings of the British Ortho-
paedics Research Society; Journal of Bone and Joint. Surgery
77B(Suppl. III), 321.
425
Lu, T.W., O'Connor, J.J., Taylor, S., Walker, P.S. 1996. Interpretation
of isometric tests on patients with instrumented massive prosthesis.
In: Proceedings of the European Society of Biomechanics, August,
269 pp.
Lu, T.W., O'Connor, J.J., Taylor, S.J.G., Walker, P.S.; 1997a. Compari-
son of femoral forces during double and single leg stance: applica-
tion to knee replacement. In Proceedings of European Orthopaedic
Research Society, April 1997(1).
Lu, T.W., Taylor, S.J.G., O'Connor, J.J., Walker, P.S., 1997c. In#uence
of muscle activity on the forces in the femur: an in vivo study.
Journal of Biomechanics 30 (11/12), 1101}1106.
Lu, T.W., O'Connor, J.J., Taylor, S.J.G., Walker, P.S., 1997b. Valida-
tion of a lower limb model with in vivo femoral forces telemetered
from two subjects. Journal of Biomechanics 31, 63}69.
Markel, M.D., Wilkenheiser, M.A., Chao, E.Y.S., 1990. A study
of fracture callus material properties: relationship to the torsional
strength of bone. Journal of Orthopaedics Research 8 (6),
843}850.
McKibbin, B., 1978. The biology of fracture healing in long bones.
Journal of Bone and Joint Surgery 60B, 150}162.
Meroi, E.A., Natal, A.N., 1989. A numerical approach to the bi-
omechanical analysis of bone fracture healing. Journal of Biomedi-
cal Engineering 11, 390}397.
Pan, W. T., Einhorn, T. A., 1992. The biochemistry of fracture healing.
Current Orthopaedics. Longman Group UK Ltd, pp. 207}213.
Perren, S.M., Cordey, J., 1977. Die Gewebsdi!erenzierung in der Frac-
turheilung. Monatsschrift f. Unfallheilkunde 80, 161}164.
Perren, S. M., Cordey, J., 1980. Current Concepts of Internal Fixation
of Fractures. Springer Berlin.
Prendergast, P.J., Huiskes, R., Soballe, K., 1997. Biophysical stimuli on
cells during tissue di!erentiation at implant interfaces. Journal of
Biomechanics 30 (6), 539}548.
Reilly, D.T., Burstein, A.H., 1975. The elastic and ultimate properties of
compact bone tissue. Journal of Biomechanics 8, 393}405.
Sarmiento, A., Latta, l. l., 1995. Functional Fracture Bracing. Tibia,
Humerus and Ulna. Springer, Berlin.
Simon, B. R., 1990. Poroelastic "nite element models for soft tissue
structures. In: Hukins, D. (Ed.), Connective Tissue Matrix, Part
2 , MacMillan, London, pp. 66}90.
Spilker, R.L., Suh, J.K., Vermilyea, M.E., Maxian, T.A., 1990. Alternate
hybrid, mixed and penalty "nite formulations for the biphasic
model of soft hydrated tissues. In: Ratcli!e, A., Woo, S.L-Y. (Eds.),
Biomechanics of Diarthrodial Joints. Springer, New York, pp.
400}435.
White III, A.A., Panjabi, M.M., Southwick, W.O., 1977. E!ects of
compression and cyclical loading on fracture healing * a quantita-
tive biomechanical study. Journal of Biomechanics 10, 233}239.
Yamagishi, M., Yoshimura, Y., 1955. The biomechanics of fracture
healing. Journal of Bone and Joint. Surgery 37A, 1035}1068.
Zienkiewicz, O.C., Bettess, P., 1982. Soils and other saturated media
under transient, dynamic conditions; general formulations and the
validity of various simplifying assumptions. In: Pandy, G.N., Zien-
kiewicz, O.C. (Eds.), Soil Mechanics * Transient and Cyclic Loads.
Wiley, UK, pp. 1}15.