Animal Model of Fracture Repair
All animal experiments were carried out in accordance with the United
Kingdom Animals (Scientific Procedures) Act (1986) at the Queen's University
Belfast Biomedical Research facility. Four-month-old male Carworth Farms
Lane-Petter mice (Laboratory Services, Queen's University of Belfast, Belfast,
Northern Ireland, United Kingdom) underwent a standard open femoral osteotomy
as previously
described26.
General anesthesia was induced with use of a gaseous mixture of 3% isoflurane
in a 50:50 mixture of N2O2:O2 at 2 L/min in a
sealed chamber. Once anesthetized, the animals were transferred to the
operating table, with the same mixture of gases delivered by way of a Hunt
mask. The animals were given a bolus dose of 1 mL of 0.9% sodium chloride
containing 0.1 µg of buprenorphine to maintain fluid homeostasis and for
postoperative analgesia. Through an anterolateral approach, the left femoral
shaft was exposed and clamped in a specially designed jig. Four holes, each
angled at 7° to the perpendicular, were drilled through both cortices with
use of a 0.55-mm drill bit. Fixator pins (0.51 mm) were inserted into the
holes, and a metal fixator crossbar was loosely placed over the pins and held
at 10 mm from the bone surface while a low-energy transverse osteotomy was
made between the two central drill holes. The geometry of the drilling jig and
crossbar ensured that the distance between the central two pinholes in the
crossbar was 0.7-mm less th an the distance between the two pinholes in the
femur; thus, sliding the crossbar down the pins and toward the bone produced
compression of the fracture gap at a standard bone-to-bar distance of 4 mm.
The tops of the pins were then cut and secured with use of a small drop of
quick-drying glue, and the wound was closed. To enable the assessment of
sequential radiographic and vascular changes across the fracture gap, the
external fixator bar was modified from a four-hole to a seven-hole construct.
Two perpendicular holes were created to allow connection with an x-ray jig,
and a central conduit was made to enable passage of an optical fiber.
In total, 235 animals (103 control animals and 132 NSAID-treated animals)
underwent surgery to induce a fracture. To further validate the model for
sequential assessment of radiographic changes and blood flow at the fracture
site, five additional animals underwent sham surgery. These animals had a
routine surgical approach with the application of an external fixator and a
laser Doppler probe; however, no actual fracture was induced. It was
hypothesized that as there was no fracture to heal, these animals would have
no changes in either radiographic or Doppler readings. A radiographic
assessment of healing was carried out for all 240 animals, histologic analysis
was obtained for fifty-four, biomechanical testing was performed on fifty-two,
and laser Doppler blood-flow measurements were acquired for sixty.
Delivery of Treatment
The human dose of rofecoxib is 12.5 to 25 mg per day for osteoarthritis and
25 to 50 mg per day for acute pain. We used an equivalent murine dose to the
human dosing regimen of 25 mg per day. The murine equivalent dosage (5.1 mg/kg
per day) was calculated with use of the approved dose calculator of the United
States Food and Drug
Administration27.
Once daily, with use of an oropharyngeal feeding tube, the animals received a
0.2 mL oral solution that contained either 0.25 mg of rofecoxib in a 0.5%
methylcellulose solution (the NSAID group) or 0.5% methylcellulose solution
only (the control and sham surgical groups).
Outcome Measures
Radiographic Assessment
Following surgery, each of the animals had a digital radiograph made of the
fractured limb (at days 4, 8, 16, 24, and 32). With the animals under light
general anesthesia, the x-ray jig was placed through the external fixator bar,
thereby controlling for rotation in all planes, and the animal was placed on
its side in the x-ray chamber. Digital radiographs were quantitatively
analyzed with use of a freely available software program (UTHSCSA ImageTool;
Department of Dental Diagnostic Science, University of Texas Health Science
Center at San Antonio, San Antonio, Texas
[]).
The pixel density of the fracture gap was compared with two adjacent areas of
uninjured bone. As callus was laid down at the fracture gap during repair, the
pixel density of the fracture gap increased with respect to that of the
uninjured adjacent bone. Results were expressed as "relative
bone-mineral content." Intraobserver analysis with use of this method
was associated with a low error rate with a single-measure intraclass
correlation coefficient of 0.9769 (p < 0.0001).
For all animals, the density of the fracture gap was assessed until the
animals were killed or until it was noted that the fracture had >50%
displacement or that the fixator had displaced or broken. If this occurred,
the animals were killed and removed from further study.
Histologic Analysis
After each animal was killed (fifty-four animals total, days 8, 16, 24, and
32), the hind limb was disarticulated and amputated below the knee. Specimens
were coded and then fixed in 4% paraformaldehyde for twenty-four hours before
undergoing decalcification in 7% formic acid. Following decalcification, the
external fixators were carefully removed, taking care not to damage the
fracture morphology. The specimens were then embedded in paraffin blocks, from
which 7 µm sections were cut, mounted on slides, and stained with
hematoxylin and eosin. Slides were analyzed under a light microscope with use
of qualitative as well as semiquantitative scoring systems. Each slide
received a score that was based on the type of tissue that was observed at the
fracture gap: a score of 1 indicated the presence of mostly fibrous tissue; 2,
mostly cartilage; 3, both cartilage and bone; 4, little cartilage remaining
and mostly bone; and 5, remodeled bone. Specimens were also analyzed with use
of an image analysis software program (Scion Image; Scion Corporation,
Frederick, Maryland) to calculate areas of bone, new callus, cartilage, or
fibrous tissue. All results were calibrated to the femoral diameter in order
to standardize for variances in animal size. Slides were examined on two
occasions at two months apart with the examiner blinded to time of harvest and
to whether or not the animal had received drug treatment. Examination revealed
a high intraobserver correlation (intra-class correlation coefficient 0.8461;
p < 0.000). To further validate the model, radiographic and histologic
results were correlated, and they demonstrated positive relationships at both
day 24 (r = 0.849, p = 0.016) and 32 (r = 0.626, p = 0.013).
Biomechanical Testing
After each animal was killed (fifty-two animals total, days 24 and 32), the
hind limbs were disarticulated and amputated below the knee. Sharp dissection
was used to dissect off all soft tissues, taking care not to disturb the
callus. The specimens were stored in moist gauze for transport to the testing
facility. All femora were analyzed with use of three-point bending (NEXYGEN MT
data-analysis software; Lloyd Instruments, United Kingdom) within six hours
after harvesting. The femur was placed on custom-made struts, 9 mm apart, with
a 100-N superior load cell delivered through a superior strut directed to the
mid-diaphyseal region at a rate of 3 mm/min. Data were collected concerning
peak load to failure, and stiffness was calculated from load-displacement
curves. Data were also collected from the contralateral intact limb of each
animal, and the results were expressed as a percentage of this limb to control
for variations in animal size.
Laser Doppler Flowmetry
To measure blood flow at the fracture gap in real time, the principles of
laser Doppler flowmetry were utilized in sixty animals. We used a laser
Doppler flowmeter (Oxylab ML192 Dual Flow meter; Oxford Optronics, Oxford,
United Kingdom), which comprised a light source and receiver, optical cables,
and data-analysis software. This unit utilized a 1-MW Class-1 laser with a
780-nm wavelength and a sampling rate of 10 to 22 Hz. A 0.5-mm chronically
implantable optical fiber was passed through the central conduit of the
external fixator and secured adjacent to the fracture gap. The tip of this
cable became surrounded by the callus during repair. The remaining cable was
secured inside a snap-fit cylindrical container that surrounded the external
fixator, and was removed only at testing intervals (days 4, 8, 16, 24, and
32). To minimize any extraneous influences of atmospheric conditions and
movement artifact during testing, the animals were placed on a heating mat
(37° C) and kept under light general anesthesia, the overhead lights were
switched off, and after a three-minute settling period, the lowest value of
flow was measured over a one-minute testing period. Following testing, the
optical cable was disconnected from the light source and replaced in the
protective cylindrical container.
Statistical Analysis
All data were transferred to statistical spreadsheets and analyzed with use
of a commercially available statistical program (Statistical Package for the
Social Sciences [SPSS] software, version 11; SPSS, Chicago, Illinois).
Whenever possible, data were analyzed with use of parametric measures
(independent-samples t test). However, because of the small size of some
groups and the nonconformity of some results to the normal distribution, it
was also necessary to use a nonparametric mode of analysis (Mann-Whitney U
test). Differences due to treatment and healing time were considered
significant at p < 0.05 for all tests. Correlations were analyzed with use
of the Spearman rank-correlation test, and backwards linear regression was
then performed. Data were expressed on box and scatter plots where
appropriate.
Etiological Variables
Animal age (and standard deviation) ranged from ten to twenty weeks (mean
14.1 ± 2.98 weeks). It was noted that the NSAID animals were, on the
average, two weeks older than the control animals (p = 0.014). Animal weight
ranged from 35.6 to 56.0 g (mean 45.28 g ± 4.28), and there was an
average non-significant weight change of -0.42 g during the investigation.
Radiographic Assessment
Radiographic assessment revealed that the sham group had a small drop-off
in bone-mineral content at the fracture gap during the period of the
experiment, but the magnitude of this change was small when compared with the
change in the two treatment groups (change in bone mineral content at the
fracture gap from day 0 to day 32: sham group, 0.35%; control group, 29.8%;
NSAID group, 24.5%). There were no statistical differences between the two
groups (control and NSAID) immediately after surgery, and both groups
demonstrated a stepwise increase in bone density during the experiment with a
gradual widening in the range of results as time progressed. There was a trend
toward significant differences between the control and the NSAID groups at day
32, with the control group exhibiting greater bone-mineral content at the
fracture gap (p = 0.051, Fig.
1).
Histologic Analysis
Results of the qualitative histologic analysis are illustrated in
Figure 2 and demonstrate a
progressive increase in scores in both groups. Control animals scored
significantly better than the NSAID animals from day 8 to day 24 and through
day 32 (p = 0.016 and 0.003, respectively) and between day 16 and day 32 (p =
0.013). There were no significant differences for the NSAID-treated animals
across these time intervals. There was a significant difference between the
treatment groups at day 32 (p = 0.004), with the control animals having more
advanced stages of repair. Semiquantitative analysis suggested that control
animals tended to exhibit more callus at day 8 (p = 0.050) and at day 32 (p =
0.070). NSAID-treated animals had a significantly greater area of cartilage at
day 8 (p = 0.014) and more fibrous tissue at day 32 (p = 0.024).
Biomechanical Testing
Control animals achieved a greater median value for both peak load to
failure and stiffness across time compared with NSAID-treated animals. These
differences, however, were not significant. A significantly larger proportion
of animals from the NSAID group failed to survive until the predesignated
harvesting date (58% of the NSAID group compared with 28% of the control
group; p = 0.001). This was primarily due to fixator failure and fracture
displacement. If all animals that were entered into this limb of the study
were examined on an "intention-to-treat" basis and assigned a
value of 1% if removed from the study due to fixation and/or fracture
displacement, then the control animals did exhibit a significantly greater
biomechanical integrity for peak load and stiffness at day 32 (p = 0.028 for
both).
Laser Doppler Flowmetry
Results for control, NSAID, and sham blood flows are illustrated in
Figure 3. The sham group had a
small but significant rise in flow from day 0 at both days 8 and 24 (each p =
0.043). Both the control and the NSAID groups exhibited a greater magnitude of
change. There were no significant differences between the two groups at day 0.
Thereafter, flows in both groups rose, peaking at days 8 to 16 before falling
back to resting levels at day 32. Control animals exhibited a greater median
flow at all time intervals, which was significant at days 4 (p = 0.050), 16 (p
= 0.032), and 24 (p = 0.007).
Correlations and Regressions
There were significant relationships between higher blood flow and improved
fracture repair at various time intervals; higher flows at day 4 were
associated with improved healing radiographically at day 24 (p = 0.020, r =
0.657), and higher flows at day 16 were associated with improved healing
radiographically also at day 16 (p = 0.045, r = 0.645). A greater change in
blood flow from day 0 was also associated with better repair histologically at
day 32 (p = 0.002, r = 0.724).
To investigate if the relationships between alterations in flow and
fracture repair were causal, linear regression models were then completed.
These results suggested that, although alterations in blood flow had a
significant association with fracture repair, this effect was independent of
the effect from NSAID treatment.
The animal model used in this study was both reliable and valid as an
assessment of fracture-healing. The results from the intraobserver correlation
analysis for both radiographic and histologic analysis illustrated that the
mechanisms of deriving data and the subsequent analysis of those data were
highly repeatable. Outcome measures also correlated across the variables as
reflected by positive correlations between radiographic and histologic
outcomes both at days 24 and 32. The validity of the radiographic and
blood-flow outcomes that were used in demonstrating real changes across the
fracture gap was reflected by the differences that would be expected, and
these were noted when comparing the differences between control and sham
fracture animals. The small drop in bone-mineral content noted in the sham
group was likely due to some degree of osteopenia secondary to decreased
weight-bearing and to a stress-shielding effect of the external fixator. The
small changes noted in blood-flow measurements were likely to be secondary to
increased flow due to soft-tissue repair or callus formation at the tips of
the external fixator pins.
Results from the fracture outcome analysis confirm what would be expected
in a fracture-repair model in an animal of this size. A stepwise progression
in healing was identified histologically: a maximum area of cartilage
formation was seen at day 8, progression to bone formation was seen at day 24,
and a decline in healing was seen at day 32, secondary to remodeling.
Radiographic analysis showed a gradual increase in density throughout the
investigation. These time scales of healing corresponded to those of other
authors who used similar models of repair: Connolly et al. reported maximum
bone formation and osseous union histologically at day 24 in the same model as
in this study26,
whereas Zhang et al., using as a model a murine femoral fracture treated with
an intramedullary nail, noted abundant callus formation at day 14
histologically and evidence of union radiographically at day
217.
When the histologic characteristics of the animals treated with rofecoxib
were examined, NSAIDs had no effect on cortical bone resorption. They did not
produce as much new bone as their control counterparts did, but they exhibited
a greater area of cartilage. We had expected to see a rise in cartilage
formation in the control animals during endochondral ossification, but it is
possible that the earliest time interval (day 8) was too late to identify such
a change. Analysis illustrated that the rofecoxib-treated animals attained
poorer radiographic as well as biomechanical profiles of repair and had a
greater propensity toward failure of fracture fixation. This is consistent
with a similar response to NSAIDs as reported in the literature. Connolly
illustrated that meloxicam inhibited fracture repair; treated animals had less
bone and more cartilage both at days 16 and 24 and had less biomechanical
integrity at day
1628. Zhang et al.,
in a COX knockout model, demonstrated significantly less bone formation
alongside decreased mesenchymal cell differentiation at day
167. Simon et al.
demonstrated that while NSAID treatment did not prevent the formation of
callus as seen on radiographs, few animals achieved radiographic union and
treated animals attained very poor biomechanical integrity with less callus
apparent on histologic
analysis6. In a
similar manner to the present study, Simon et al. experienced a high
pin-slippage rate across all NSAID treated animals, which was particularly
evident in the animals that had been treated with rofecoxib.
There are several methods of assessing changes in bone vasculature that
have been described in the literature. The most frequently used assessments
are based on vessel counting of histologic samples, with use of either
traditional staining techniques or immunolocalization of the vessels. These
methods are based on anatomy and do not make any reference to the functional
compatibility of the vessels, and each measure requires killing the animals
for assessment. In addition, these methods do not account for vessel size, as
a larger arteriole establishes a similar "count" as a smaller,
newly formed and friable capillary, although capacity for the volume of
nutrient and cellular delivery is vastly different. In our study, we employed
a functional outcome measure of flow that would enable sequential assessment
in the in vivo environment.
Following fracture, the animals in the present study exhibited a rise in
blood flow until day 16, at which time the blood flow returned to resting
levels at day 32. Both control and NSAID animals followed a similar pattern,
and no differences were identified between the two groups on the day of
surgery, before any rofecoxib had been prescribed. Thereafter, the magnitude
of change exhibited by the control animals was significantly higher (days 4,
16, and 24), and that group displayed a higher median blood flow at all time
points. These differences between the groups were noted despite the fact that
approximately 50% more NSAID animals than control animals were entered into
the trial for the Doppler study. This was secondary to greater numbers of
NSAID-treated animals being removed from investigation after they had
development of fracture displacement or a failure of the fracture fixation. It
may be that, while some NSAID-treated animals were able to achieve healing of
the fracture, the added insult of carrying the laser Doppler and Doppler
protection container was too great a burden and therefore compromised the
healing process in others. The flow results in this study correspond well to
those reported in the literature with regard to angiogenesis and
fracture-healing. Pufe et al. and Uchida et al. suggested that the maximal
release of the cytokines involved in signaling angiogenic growth occurs at
approximately day 5 following injury in murine models of
repair29,30.
In human studies, maximum flow to a tibial corticotomy (identified with use of
Doppler ultrasound) has been identified as occurring between fourteen and
twenty-eight days following
injury31. The
increase in flow in our study, peaking at day 16, would correspond to the
principle first predicted in an early work by Trueta and Morgan that
angiogenesis directly precedes
osteogenesis32.
This was subsequently illustrated by Winet with use of intravital microscopy
in bone-chamber
models33.
The molecular basis behind this effect is likely to be by an action on
vascular endothelial growth factor (VEGF). NSAIDs inhibit circulating levels
of tumor necrosis factor alpha
(TNF-a)28, a
cytokine that has been noted to be activated early in the inflammatory cascade
and that is both a trigger for the release of pro-angiogenic cytokines from
polymorphonuclear
cells34 and also a
pre-activator of cells to a "VEGF-responsive"
status35. VEGF has
been illustrated to be of importance in enhancing both leukocyte adhesion and
rolling, thus enabling recruitment by chemotaxis and diapedesis of
inflammatory cells from the circulation to a site of
injury36. These
polymorphonuclear cells are known to secrete the cytokines required in the
early stages of repair, with the degranulation process of these cells being
induced by
TNF-a37, an
effect illustrated to be negated by the addition of the COX-2 specific
inhibitors meloxicam and
piroxicam38.
The results found in this study suggest that while NSAIDs do exert an
inhibitory action on vascularity, the effect on fracture-healing is
independent of this reduction in blood flow. Although TNF-a has
importance as an upstream regulator of vasculogenic proliferation, it has also
been shown to influence osteogenic development as well. Inhibition of the
above mechanisms involved in cell adhesion by NSAIDs may decrease the
attraction and transport, and thus the accumulation, of circulating
mesenchymal stem cells to the site of
injury39.
Gerstenfeld et
al.40,41
demonstrated that TNF knockout mice have a slower recruitment of osteoblasts
and an almost complete lack of intramembranous ossification. It has also been
noted that inhibiting the release of early inflammatory cytokines can result
in enlargement of the soft callus and a delay in chondrocyte hypertrophy.
Specifically, TNF-a is important in the regulation of the chondrocytic
population, and a decrease in circulating levels of this cytokine may lead to
desensitization of cells with a resultant loss of an apoptotic response by the
hypertrophic
chondrocytes42.
This finding of an increased mass of soft callous and cartilage with slower
progression to bone formation has been noted in the present study at day 8 and
also by Connolly et
al.26, who
postulated that this alteration in the tissue volume and content was secondary
to an inhibition of mesenchymal cell differentiation or an inhibition of
chondrocytic apoptosis. Zhang et
al.7 proposed a
similar conclusion following work that utilized a COX knockout murine tibial
fracture model and demonstrated a delay in mesenchymal differentiation in
COX-2 knockout but not COX-1 knockout animals, with a resultant increase in
fibrous nonunion.
Although the newer COX-2 inhibitors are marketed as having a lower
side-effect profile than traditional NSAIDs, particularly with regard to
gastrointestinal effects, this study has illustrated that they continue to
exert an inhibitory action on fracture repair. We therefore propose that it
might be appropriate to withhold NSAIDs when treating patients following
osseous injury. Withholding NSAID treatment has particular importance in the
treatment of fractures that are associated with a delay in healing and that
often are accompanied by a reduction in blood flow to the fracture site. This
caution should include fractures to bones that, by their very nature, have a
tenuous blood supply (scaphoid, talus, subcapital neck of femur, and
diaphysis), fractures and/or injuries that compromise the vascular supply
itself (fractures associated with vascular damage, and crush and high-energy
injuries), and fractures in patients with concomitant conditions that may
already predispose to delayed osseous repair (smoking, diabetes, anemia, and
malnutrition). ?