Nonsteroidal anti-inflammatory drugs are some of the most widely used
analgesics for the treatment of acute postoperative and posttraumatic
pain1,2.
They also have been extensively used to alleviate chronic pain and discomfort
associated with
osteoarthritis3. The
primary mode of action of this class of compounds is the inhibition of
cyclooxygenase (COX) activity leading to a reduction in prostaglandin
production and its effects on both inflammation and
pain4-6.
The COX enzyme family is encoded by two genes, COX-l and COX-2. Each of these
genes encodes several isotypic variants that arise by differential
splicing7,8.
The first generation of drugs developed, the nonsteroidal anti-inflammatory
drugs, inhibit the activities of both COX enzymes; however, their use has been
associated with prolonged bleeding time and gastric mucosal irritation,
sometimes leading to ulceration. Subsequent research demonstrated that COX-1
was constitutively expressed in many tissues and that two of its primary
functions are the protection of gastric mucosa and the regulation of platelet
aggregation9,10.
The COX-2 enzyme was shown to be less broadly expressed and is induced by a
wide variety of factors, including mechanical stimulation, physical trauma,
and numerous inflammatory
mediators11-13.
Newer generations of drugs (coxibs) that selectively inhibit COX-2 were
subsequently developed. These compounds were shown to have both selective
anti-inflammatory and analgesic effects while appearing to have lesser effects
on gastric mucosa and blood
coagulation10,14.
Considerable evidence has accumulated that conventional nonspecific
nonsteroidal anti-inflammatory drugs, such as ibuprofen, ketorolac, and
indomethacin, have an inhibitory effect on fracture-healing as well as other
forms of postoperative bone
repair15-24.
In recent studies comparing nonselective nonsteroidal anti-inflammatory drugs
and COX-2-selective drugs in rats, impaired fracture-healing was observed,
with a greater effect seen in association with the COX-2-selective
drugs25. Moreover,
in studies of mice that were homozygous for a null mutation in the COX-2 gene,
induced prostaglandin production was shown to be essential for optimal
mesenchymal cell differentiation during skeletal
repair26. The
histology of the fractures in the COX-2-null mice showed a persistence of
undifferentiated mesenchyme and a marked reduction in osteoblastogenesis,
resulting in a high rate of nonunion. In a number of retrospective human
studies examining the effects of nonsteroidal anti-inflammatory drugs in
various clinical settings of bone-healing, inhibitory effects have been
reported20,27.
However, to our knowledge, there have been no randomized, controlled,
prospective trials in
patients28. It is
interesting to note that recent studies on the use of selective
prostaglandin-receptor agonists have shown that stimulation of the EP2 or EP4
prostaglandin E2 (PGE2) receptor specifically promotes
bone-healing29.
Both animal and retrospective human studies have raised concerns about the
use of these compounds as analgesics even for short periods after skeletal
injury or surgery. However, there is limited information regarding the
reversibility of these effects on skeletal healing or the relationship between
the inhibition of bone repair and the influence of drug dose on the production
of PGE2 in skeletal tissues. To further elucidate the effects of
COX-2 inhibition on bone repair, we tested the hypothesis that the impairment
of fracture-healing produced by the administration of either a selective or
nonselective inhibitor of COX-2 is reversible and is related to the recovery
of PGE2 levels in the fracture callus.
Materials
Double trocar pointed stainless steel 0.045 × 9-in (0.11 ×
22.9-cm) Kirschner wires (Howmedica Osteonics, Allendale, New Jersey) were
used for intramedullary fixation. Ketorolac (tromethamine salt) was obtained
from Cayman Chemical, Ann Arbor, Michigan (catalog number 70690). Valdecoxib
(Bextra) was produced as a sodium salt and was provided by Pfizer, Groton,
Connecticut.
General Experimental Design
Experiment 1: Effect of Duration and Discontinuation of Drug
Treatment on Fracture-Healing
This experiment involved three treatment groups: the first group (the
control group) was treated with vehicle only (0.5% methylcellulose) and no
drug, the second group was treated with a nonspecific nonsteroidal
anti-inflammatory drug (ketorolac; 4 mg/kg), and the third group was treated
with a COX-2-specific inhibitor (valdecoxib; 5 mg/kg). Higher drug dosages
were selected on the basis of the ED50 (median effective dose) of these two
drugs6. Treatments
lasted for seven or twenty-one days, and animals were killed and calluses were
tested at twenty-one and thirty-five days. In each of these six experimental
groups, seven animals were assigned to histomorphometric analysis, twenty-five
were assigned to bio-mechanical testing at twenty-one days, twenty-three were
assigned to biomechanical testing at thirty-five days, and seven were assigned
to PGE2 analysis.
Experiment 2: Effect of Dose and Withdrawal of Drug Treatment on
Local PGE2 Levels in the Fracture Calluses
The first study examined PGE2 levels in normal, untreated
calluses across the whole experimental time-course of healing. The second
study addressed the effect of daily treatment with nonselective and selective
COX-2 inhibitors on PGE2 levels in the callus and also addressed
whether PGE2 levels rebound in the tissue after drug withdrawal. In
this study, ketorolac (4 mg/kg) and valdecoxib (5 mg/kg) were administered on
a daily basis for seven days beginning at twenty-four hours after fracture.
Calluses were then harvested on either the seventh day after the fracture
(that is, after seven days of treatment) or the fourteenth day after the
fracture (that is, after seven days of treatment and seven days without
treatment), and PGE2 levels in the calluses were measured. In a
third study, the dose response of PGE2 levels in the fracture
callus was measured. In this study, fractures were created and, six days after
the fracture, a single treatment of three different doses of the two drugs was
administered. The doses of ketorolac were 4 mg/kg (high dose), 1.4 mg/kg
(intermediate dose), or 0.4 mg/kg (low dose), and the doses of valdecoxib were
5.0 mg/kg (high dose), 1.5 mg/kg (intermediate dose), or 0.5 mg/kg (low dose).
Six hours after the drugs were administered, the animals were killed and the
calluses were harvested for PGE2 analysis. Seven animals were
assigned to each time-point at which PGE2 was assayed.
All sample sizes for individual experimental groups were determined with
power statistics calculations, based on a coefficient of variation of 25% in
the types of data that are collected, and after accepting a and ß
errors of 5%.
Surgical Procedure, Animal Care, Tissue Harvest and Storage
Animal research was conducted in conformity with all federal and United
States Department of Agriculture guidelines, under an Institutional Animal
Care and Use Committee-approved protocol. Male Sprague-Dawley rats (Harlan
Bio-products for Science, Indianapolis, Indiana) with a mean weight (and
standard deviation) of 449 ± 39 g and an age of about seven to nine
months were used for all experiments. Animals were individually housed at
22°C with free access to food and water on a twelve-hour light and dark
cycle. Animals were allowed at least forty-eight hours to acclimate after
shipment before the performance of any procedure. Closed simple transverse
fractures were generated as described by Bonnarens and
Einhorn30. The
exclusion criteria included inappropriate pin placement; postoperative pin
slippage, defined as loss of fracture fixation (with the pin backing out
beyond the site of the fracture); postoperative infection; and a fracture that
was situated too far from the mid-diaphyseal region of the femur. Of the total
number of rats enrolled, only nineteen were excluded: twelve in the
seven-day-treatment group (five for pin slippage, one for inappropriate pin
placement, and six for inappropriate fracture placement) and seven in the
twenty-one-day-treatment group (two for pin slippage, four for inappropriate
fracture placement, and one for postoperative infection). Radiographic
assessment of the fracture was performed immediately after the fracture was
created. Intramedullary pins were removed after the animal was killed, and a
terminal radiograph was made immediately after pin removal. Once the terminal
radiograph was made, whole femora that were designated for mechanical testing
were wrapped in saline solution-soaked gauze and were immediately frozen at
—20°C, samples that were designated for histological assessment were
placed in ice-cold fixative, and samples that were designated for
PGE2 analysis were snap-frozen by immersion in liquid nitrogen and
stored at —80°C.
Drug Preparation and Administration
Drugs were supplied in powdered form and were suspended in delivery vehicle
containing 0.5% methylcellulose (weight per volume) (catalog number M-O262;
Sigma Chemical, St. Louis, Missouri) and 0.025% Tween 20 (volume per volume)
(catalog number P-1379; Sigma) in sterile deionized water. Drug suspensions
were resterilized by filtration through a 0.45-µM filter and stored at
4°C. Drug solutions and vehicle were delivered once daily in a volume of
0.4 mL by means of oral gavage. Drug administration was commenced at
twenty-four hours after the creation of the fracture and was continued daily
for the specified times for each study group. As specified for each
experiment, each operative group was then retrieved at defined times after
creation of the fracture.
Biomechanical Analysis
Bones were subjected to biomechanical testing to failure with use of a
servo-actuated rapid-load torsion-testing device at a rate of 10 N/mm/s.
Specimens were thawed just prior to testing, and the proximal and distal ends
were potted in 2-cm2 aluminum blocks that were filled with a lead
alloy (Cerrobend; Cerro Medical Products, Bellefonte, Pennsylvania) that melts
at low temperature (158°F [70°C]). The bones were then positioned in
the aluminum blocks when the metal alloy was still melted such that 1 cm of
bone was exposed (0.5 cm proximal and 0.5 cm distal to the center of the
fracture site). The applied moment and angular deformation of the femora were
measured and plotted. Values were obtained for shear modulus (stiffness) and
maximum torque (torsional strength).
A nonunion was considered to be present if, after the pin was removed and
muscle attachments were removed by means of dissection, the bone was in two
pieces. A nonunion was also considered to be present if, after the pin was
removed, the fracture site demonstrated gross motion on physical manipulation
and could not vertically support the weight of the aluminum potting blocks.
This latter group of bones all showed values of <50 N/mm when tested, and
these results were excluded from the testable results.
Histological Analysis
After the animals were killed, the femora were removed and fixed for three
days in 4% paraformaldehyde and phosphate-buffered saline solution at 4°C.
After fixation, the bones were rinsed in sterile phosphate-buffered saline
solution, decalcification was carried out with use of 14% EDTA for two to four
weeks at 4°C, and the specimens were stored in sterile phosphate-buffered
saline solution at 4°C until being embedded in paraffin for sectioning.
After embedding, all specimens were sagittally sectioned at a 7-µm
thickness on a Leica Microsystems Polycut microtome (Wetzler, Germany).
Histological analysis was carried out on midsagittal sections of five stained
specimens from at least five animals per data point. Each specimen was
examined at multiple magnifications (×40 to ×600) to ensure proper
identification of cell types.
Sections were stained with safranin O-fast green as previously described to
discriminate mature cartilage from bone and noncartilage connective
tissues31. Each
section was photographed at 10.25× magnification with a light microscope
(Olympus BX51; Olympus, Center Valley, Pennsylvania) that was attached to a
digital camera, and the image was downloaded into an Image-Pro Plus program
(version 4.1.0.0 for Windows; Media Cybernetics, Silver Spring, Maryland). An
area of interest was created by loading a uniform box (5 × 7.9 mm) onto
the photograph and centering the callus within the box. The bone was then
outlined within the uniform area of interest, excluding any muscle, soft
tissue, or periosteum. With use of a color-match program, the total area of
cartilage (red) and bone (green) was initially identified and quantified with
use of a filter range of 573.9 to 5.74 × 103 µm.
Subsequently, color-matched areas defining specific tissues were individually
assessed and final areas were hand-traced. Mean values were calculated for
specimens sampled from each bone and then were used to create group means,
standard deviations, and standard errors with respect to both time after the
fracture and animal
group31.
Assessments of PGE2 Levels in the Callus
Seven animals were used for PGE2 determinations for each
time-point and drug-dose group. On the day of tissue retrieval, the femur was
rapidly removed from each rat and a uniform area of callus and underlying bone
was cut with use of a dental burr such that tissue was excised both 5 mm
proximal and distal to the fracture site. For samples that were isolated at
twenty-four hours and three days after the fracture, a small margin of muscle
tissue was also retrieved along with the underlying bone in order to retain
the hematoma and the initial cell populations that had been recruited to the
injury site. As soon as the tissue was retrieved, it was immediately frozen in
liquid nitrogen and stored until assay at —80°C. In order to isolate
PGE, the callus and bone tissues were powdered under liquid nitrogen with use
of a SPEX Tissue Mill 6750 (SPEX CentriPrep, Metuchen, New Jersey).
PGE2 was extracted in methanol from 1 mg of the tissue powder from
each callus. The organic methanol phase was appropriately diluted and then
analyzed with use of a direct PGE2 EIA assay (GE Healthcare
Bio-Sciences, Piscataway, New Jersey) according to the manufacturer's
directions.
Statistical Analysis
The effects of treatment on torque and shear modulus on Days 21 and 35 were
evaluated with analysis of variance and multivariate analysis of variance
(general linear models) (SAS, release 8.01; SAS Institute, Cary, North
Carolina). The models were run separately for the seven and twenty-one-day
treatment protocols. The main effects in the models were treatment group and
time. An interaction term for treatment relative to time (days) was included
in the models. When the interaction was significant, it indicated that the
direction and/or magnitude of change in the outcome measurement was not the
same in all treatment groups. When any of the treatment effects (the day
effect or the interaction term) was significant at p < 0.05, pairwise
contrasts of means of interest were performed to determine which treatments
and time-points differed. The distribution of moduli was skewed, so it was
transformed with the log function for the analysis of variances. Assessments
of statistical effect on nonunion rate were determined with use of a Fisher
exact test at p < 0.05.
Effects of Duration of Nonspecific and COX-2-Specific Enzyme
Treatment
All of the experimental groups showed excellent callus formation with
slightly larger-appearing calluses at twenty-one days as compared with
thirty-five days. While the short course of treatment (seven days) produced a
higher percentage of nonunions when assessed at twenty-one days (16% in the
ketorolac group, 21.7% in the valdecoxib group, and 8.3% in the control
group), these differences were not significant
(Table I). In contrast, the
long course of treatment (twenty-one days) produced a higher percentage of
nonunions in the valdecoxib group than in either the ketorolac group or the
control group when assessed at twenty-one days (13% in the ketorolac group,
36% in the valdecoxib group, and 4% in the control group; p = 0.05). However,
by thirty-five days, there was no significant difference in the rate of
nonunion among the treatment groups.
Torque-to-failure testing was used to assess the overall effect of COX-2
enzyme inhibition on the biomechanical properties of the fracture calluses in
bones that demonstrated sufficient healing to undergo testing. These results
showed that neither the nonspecific nonsteroidal anti-inflammatory drug,
ketorolac, nor the COX-2-specific drug, valdecoxib, affected the overall
biomechanical strength of the healing calluses when administered for the short
duration of time (seven days). However, when assessed at thirty-five days,
there was a significant decrease in the strength of the calluses in the
animals that had been treated with valdecoxib for twenty-one days as compared
with controls (p = 0.03; Table
II). Interestingly, significant reductions in the shear modulus
were seen in the valdecoxib group as compared with the control group after
both the short course of treatment (p = 0.051) and the long course of
treatment (p = 0.01) when assessed at thirty-five days
(Table III). The diminished
levels in shear modulus in the valdecoxib-treated groups suggest that there
may be longer-term effects on this mechanical property than those that affect
torsional strength. The effects of the three treatments (control, ketorolac,
and valdecoxib) on torque and shear modulus at Days 21 and 35 were evaluated
with analysis of variance and multivariate analysis of variance (general
linear models). The results of the multivariate analysis that examined the
data for interactions seen within groups across time are summarized in Tables
II and
III.
The second analysis that was carried out examined the seven and
twenty-one-day treatment groups and tested for main effects between both
treatment group and time. These data showed no significant differences in
torque for time, treatment, or time-by-treatment effects. A small interaction
effect was seen for shear modulus (p = 0.076), which was indicative of a
trend.
Effect of Nonsteroidal Anti-Inflammatory Drug or Coxib Treatment on
PGE2 Levels Within the Fracture Callus
The highest levels of PGE2 within the callus tissues were
reached twenty-four hours after the fracture, after which time the levels
steadily declined. By thirty-five days after the fracture, the levels had
fully returned to baseline values (Fig. 1,
A). Both ketorolac and valdecoxib produced a reduction in
PGE2 levels within the callus tissues after seven days of
treatment. However, when the drug was withdrawn after seven days of treatment
and the tissues were examined at Day 14, the levels had recovered. An
interesting aspect of these results is that, after cessation of treatment,
calluses from both treatment groups showed significantly higher values for
PGE2 than were seen in the control tissues at fourteen days (p =
0.0157) (Fig. 1,
B).
In a separate experiment involving another set of animals, the levels of
PGE2 in fracture calluses after a single six-hour period of
treatment with various doses of either the nonspecific or the COX-2-specific
nonsteroidal anti-inflammatory drug were examined
(Fig. 2). It is interesting to
note that, for the nonspecific nonsteroidal anti-inflammatory drug, ketorolac,
all three doses were about twice as inhibitory to the local PGE2
levels in the calluses as were the three doses of the COX-2-specific drug,
valdecoxib (p = 0.0001). Statistical analysis of the various treatment groups
showed that, within a group (ketorolac or valecoxib), all three doses of a
given drug were significantly different from the control for that group (p =
0.0001). Comparisons between the groups showed that, for all three doses of a
given drug, the PGE2 levels in the animals that had been treated
with the nonspecific nonsteroidal anti-inflammatory drug were also different
from those in the animals that had been treated with the COX-2-specific
nonsteroidal anti-inflammatory drug.
Finally, when the single six-hour treatment was compared with the levels
after daily treatment for seven days, the inhibitory levels were seen to be
comparable (ketorolac, 4.0 mg/kg; 2.5 pg/mg callus tissue daily treatment
seven days versus 1.8 pg/mg callus tissue six-hour treatment) and (valdecoxib,
5 mg/kg; 10 pg/mg callus tissue daily treatment seven days versus 12 pg/mg
callus tissue twenty-four-hour treatment). A comparison of these data shows
that, although these two studies were performed at different times and with
different groups of animals, the response to both drugs was relatively
constant and daily dosing achieved a steady-state effect comparable with
single six-hour dosing over a prolonged period.
Histological Analysis
Five to seven fracture callus specimens from each of the six experimental
groups were examined, and the ratios of cartilage to bone were determined.
Greater cartilage contents were noted in the short-term drug-treatment groups.
No other differences were seen (Figs. 3-A
and 3-B). However, measurement of the cartilage percentage showed
that short-term treatment, when assessed at twenty-one days, was associated
with a significant increase in cartilage content in callus in the animals that
had been treated with valdecoxib (Table
IV). More detailed tissue analysis at higher magnifications showed
obvious qualitative differences in both tissue and cellular morphologies
(Fig. 3-C). This assessment had
several striking features. The first was the greater number of trabeculae with
areas of mineralized cartilage that were observed in calluses from animals in
both the short and long-term valdecoxib treatment groups. This is seen as the
central red-staining regions of unremodeled calcified cartilage in the
trabeculae of these tissues that were largely absent in the control specimens.
The second feature was the diminished cellularity of the marrow elements and
the paucity of lining cells observed on the trabecular surfaces within the
calluses from animals that had been treated with valdecoxib for twenty-one
days. It is interesting to note the intermediate appearance of the marrow
element in the specimens from the seven-day-treatment group.
Fracture-healing following a traumatic or surgical injury is initiated in
response to regulatory factors associated with inflammation and the innate
immune
response3,32,33.
Descriptive studies examining the expression of various peptide-signaling
molecules, inflammatory cytokines, and other biochemical mediators of
inflammation and repair have shown that pros-taglandins are critical to this
process and are induced within the first three days after a
fracture23,25,26,32,34.
Indeed, the crucial role that prostaglandins play in the bone-repair process
has been demonstrated in studies examining the effects of COX-2 inhibition on
fracture-healing22,25,26,35.
In animal studies in which a nonselective nonsteroidal anti-inflammatory drug
was compared with a COX-2 selective drug in terms of the inhibitory effects on
fracture-healing, impaired fracture-healing was shown to be greater when the
COX-2 selective drug was
used25. Moreover,
studies of mice that were homozygous for a null mutation in the COX-2 gene
demonstrated that induced pros-taglandin production was essential for optimal
mesenchymal cell differentiation during skeletal
repair25,26.
In a number of retrospective human studies examining the effects of
nonsteroidal anti-inflammatory drugs and coxibs in various clinical settings
of bone repair, inhibitory effects were
observed20,21.
However, a more recent study of patients who underwent spinal fusion showed no
effect of short-term coxib use on the rate of non-union when fusion was
assessed one year after the
operation36.
The goals of the present study were to determine if COX-2 inhibition
produced by either nonsteroidal anti-inflammatory drug treatment or coxib
treatment is reversible and if the restoration of prostaglandin synthesis
would rescue bone repair. The results confirmed that inhibition of COX-2 is
associated with an initial delay in healing and that this delay is dependent
on both the dose of COX-2 inhibitor and the duration of treatment but that the
healing process is restored when inhibition is removed. The data also showed
that the inhibitory effects of the coxib (valdecoxib) were greater than those
of the nonsteroidal anti-inflammatory drug (ketorolac) when used in doses
equivalent to the ED50 and comparable with those used in clinical
settings.
These results showed that neither ketorolac nor valdecoxib, when
administered for seven days, resulted in a significant rate of nonunion when
compared with the rate in control animals that had been treated with vehicle
only when healing was assessed at either twenty-one or thirty-five days after
a fracture. However, when valdecoxib was administered for twenty-one days,
inhibition of healing was demonstrated at that time-point. This effect was
shown to be reversible as no effect on the nonunion rate was noted in
fractures that were assessed at thirty-five days. The results of the
biomechanical testing experiments showed that neither ketorolac nor valdecoxib
affected the torque strength of the healing calluses when administered for the
short duration of time (seven days). However, when testing was performed at
thirty-five days, there was a significant decrease in the biomechanical
strength of the calluses in the animals that had been treated with valdecoxib
for twenty-one days. Interestingly, significantly diminished tissue
stiffnesses were seen in the valdecoxib-treated animals in both the short and
long-duration treatment groups when assessed at thirty-five days. It is
important to note that these results are based on the data obtained from bones
that had sufficient mechanical integrity to be biomechanically testable. Taken
together, we interpret these findings to suggest that inhibition of COX-2 is
associated with impaired fracture-healing and that removal of drug inhibition
is associated with gradual recovery.
Previous histological studies from our laboratory showed a persistence of
cartilage in fracture callus in association with COX-2
inhibition22. In
the current study, we confirmed these qualitative findings and demonstrated
that treatment for seven days, when assessed at twenty-one days, was
associated with a greater amount of persistent cartilage in the valdecoxib
group. However, neither short-term treatment with ketorolac nor long-term
treatment with either drug produced this effect. It is possible that because
normal endochondral fracture-healing is associated with a transition from
cartilage to bone beginning at about twenty-one days in the rat, the healing
process is more sensitive to this effect at the early time-point. It is also
possible that once the drug is removed, there is a compensatory response that
leads to more cartilage formation. This is consistent with the general rebound
in PGE2 levels in the callus tissues after COX-2 inhibition was
removed and was clearly reflected by the greater strength and stiffness in the
calluses that were tested at twenty-one days. Hence, both biomechanical and
histological analysis showed that selective COX-2 inhibition produced
observable biological effects on early events in fracture callus tissues and
that these effects were gradually reversed after discontinuation of the
inhibition.
In a report on fracture-healing in transgenic animals lacking COX-2, data
consistent with inhibition equally affecting bone and cartilage formation
throughout healing suggested that COX-2 also functions in the context of
coupled
remodeling26. Our
data suggest that COX-2 inhibition primarily affects the actions of
prostaglandins on the events of early fracture-healing but that later events,
particularly those involving callus remodeling, are also affected. Other
studies have shown a loss of lining cells and osteogenic recruitment in the
marrow space in the absence of COX-2
enzyme26, and the
histological sections of this study were consistent with those previous
findings. The histological findings of the present study were also consistent
with the biomechanical findings that are reported here.
One of the most intriguing findings of the present study is that the coxib
only partially inhibited PGE2 levels in the calluses relative to
the nonspecific nonsteroidal anti-inflammatory drug. While it generally has
been assumed that PGE2 is the major product of COX-2 activity and
is the primary mediator of the anabolic activities in skeletal
cells6, the
inhibition of prostaglandin production in the callus does not fully explain
the observed biological activities affecting fracture-healing. Indeed, the
general profile of inhibition of fracture-healing in studies on transgenic
mice that were null for COX-2 is similar to that seen for the coxib-treated
animals in the present study and provides strong supporting evidence that the
pharmacological activities of the coxibs are related to their specificity for
COX-2. One study, however, showed that selective COX-2 inhibitors were only
about a third as effective as nonsteroidal anti-inflammatory drugs in
inhibiting PGE2 within tissues from local sites of adjuvant-induced
arthritis37. That
report further demonstrated that nonselective and COX-2-selective drugs have
very different biological effects on the regulation of local expression of
COX-1 and COX-2 enzymes (the coxib downregulated COX-2, whereas nonselective
inhibitors did not) as well as differential effects on local expression of
IL-6 and TNF-a mRNA. These observations lead to questions regarding how
the partial inhibition of COX-2 affects prostaglandin synthesis and how this
inhibition compares with the more complete inhibition produced by nonspecific
nonsteroidal anti-inflammatory drugs with regard to their effects on various
biological activities. Such differences may reside in the different systemic
pharmacokinetics, such as those that affect platelet activity and the effects
on coagulation6. The
recent focus on the cardiac toxicity of coxibs has led to a number of reports
showing that different pharmacological agents have differential inhibitory
effects on COX-1 and COX-2 through selective alterations in the production of
other prostaglandin metabolites such as prostacyclin (PGI) and thromboxane A2
(TXA2)38.
Selective COX-2 inhibitors appear to depress PGI2 but not
COX-1-derived
TXA239,40.
In the case of cardiac toxicity, these two prostaglandins have broadly
differing and counterbalancing biological activities, with PGI2
having vasodilatory and platelet inhibitory effects and with TXA2
having inverse
activities41.
Uncoupling of the balance of these effects by a selective inhibitor of COX-2
apparently leads to detrimental effects on coagulation control. It may
therefore be speculated that a similar type of uncoupled inhibition of COX
enzymes may influence the process of bone repair through effects on the local
ratios of different prostaglandins.
Data from this and other preclinical studies suggest that nonsteroidal
anti-inflammatory drugs and coxibs are associated with varying degrees of
inhibition of the processes of bone repair. However, in the absence of
randomized, controlled trials effectively powered to determine the effects of
nonsteroidal anti-inflammatory drugs or coxibs in patients with skeletal
injuries, it is difficult for surgeons to make decisions regarding the care of
patients. On the basis of what is known of the inhibitory effects of
nonsteroidal anti-inflammatory drugs or coxibs on bone repair in animal
systems, it is reasonable for surgeons to exercise caution regarding their
clinical use in the early postoperative period. The data presented here, while
confirming that COX-2 inhibition impairs fracture-healing, suggest that the
effects are reversible and that the discontinuation of drug treatment may
result in a restoration of healing. ?