Non-Steroidal Anti-Inflammatory Drugs Inhibit Bone Healing: a Review S

Review Article © Schattauer 2010 385

Non-steroidal anti-inflammatory drugs inhibit bone healing: A review S. Barry Washington State University, Department of Veterinary Clinical Sciences, Veterinary Teaching Hospital, Pullman, Wash- ington, USA

crine and autocrine activity, have since Keywords stems from prostaglandin inhibition and is been shown to regulate constitutive and in- Non-steroidal anti-inflammatory drugs, likely multifactorial. In human medicine ducible functions throughout the body, in- NSAID, bone healing NSAID are known to prevent heterotopic ossi- cluding bone healing (5–9). The mech- fication, however the clinical importance of anism of NSAID inhibition to bone healing Summary their effects on bone healing remains contro- is unknown, but is likely multifactorial. Re- The ability of non-steroidal anti-inflammatory versial. Although a small handful of reports searchers have suggested that NSAID affect drugs (NSAID) to inhibit bone healing has suggest that NSAID suppress bone healing in normal bone healing in multiple ways, with been established in experimental animal dogs and horses, there is little published infor- emphasis often (but not exclusively) placed models using mice, rats, and rabbits. The mation to direct veterinary practice in do- on processes related to the inflammatory mechanism of action is largely unknown but mestic species. stage. Deciphering the mechanism of NSAID inhibition requires an understanding of Correspondence to: Vet Comp Orthop Traumatol 2010; 23: 385–392 Sabrina Barry, DVM doi:10.3415/VCOT-10-01-0017 fracture healing. Fracture healing presents Washington State University Received: January 31, 2010 an exquisitely orchestrated series of coor- Department of Veterinary Clinical Sciences Accepted: June 23, 2010 dinated molecular and cellular events. Be- Veterinary Teaching Hospital Pre-published on: September 9, 2010 fore the molecular mechanisms of bone PO Box 647060 Pullman, WA 99164–7060 healing were appreciated, careful histologi- USA cal observations characterised the tem- Phone: +1 509 335 0711 poral stages of bone healing (10, 11). Fax: +1 509 335 3330 E-mail: [email protected] Direct (primary) fracture healing is characterised histologically by osteonal ‘cutting cones’ directly crossing the fracture gap, bypassing the intermediate stages of cartilaginous callus and endochondral Introduction directed by subsequent discoveries such as ossification (10–12). This type of healing is cyclooxygenase-2 (COX-2) and the bio- possible only with direct bone contact and The mechanism of action of non-steroidal logical mechanisms of bone healing. Today, rigid stabilisation. Osteonal activity in- anti-inflammatory drugs (NSAID) had NSAID are recognised as being effective increases near the injury – this phenomenon only recently been discovered when the hibitors of bone formation in experimental is referred to as ‘regional acceleratory phe- drugs’ ability to inhibit bone healing was models of fracture healing, spinal fusion, nomenon’ (RAP) and probably plays an first documented. In 1976, Ro and others bone ingrowth into implants, mechanically important role in direct fracture healing reported significantly reduced fracture induced bone remodelling, and hetero- (12). The mechanism of RAP is unknown, healing in rats treated with indomethacin topic ossification (3). This review focuses but the phenomenon may be mediated by (1). In the same year, a 64-year-old man on the effects of NSAID on bone healing the same signalling molecules as seen in with an ankle fracture/luxation was pur- after injury. other types of tissue repair (12). posefully treated with indomethacin to Indirect (secondary) fracture healing prevent callus formation in anticipation of can be classified into four histologically ob- surgery (2). The authors noted, “A dis- Mechanism of inhibition servable overlapping phases: inflamma- placed fibular fracture usually heals even tory, chondrogenic, osteogenic, and re- without external immobilisation. In the In 1971 Vane and others discovered that as- modelling (10, 13). The inflammatory present case, however, the fracture did not pirin, indomethacin, and salicylate exert phase is characterised by haematoma unite in spite of external immobilisation their potent antiphlogistic effects by information, local tissue hypoxia, and infor nine weeks.” hibiting COX, the rate-limiting enzyme in creased local production of inflammatory These initial reports led to a flurry of re- prostaglandin synthesis (4). Prostagland- cytokines and growth factors, including search which was periodically excited and ins, small signalling molecules with para- prostaglandins (6, 14). Mesenchymal stem

Vet Comp Orthop Traumatol 6/2010 386 S. Barry: Non-steroidal anti-inflammatory drugs inhibit bone healing

cells are recruited from intra- and extra- ing, are most prominent during endochon- wild type and COX-2-/- cell cultures to ap- osseous tissues to differentiate into chon- dral ossification in the osteogenic phase proximately the same degree. The authors drogenic and osteogenic cell types. In the (16). suggested that BMP-2 may be downstream chondrogenic phase, a cartilaginous callus Although the mechanism for NSAID- to prostaglandins during osteoblastogen- bridges the fracture fragments and intram- induced bone healing inhibition is un- esis and bone formation (21). embranous ossification occurs under the known, the predominant theory supposes Subsequent studies have investigated periosteum at the fracture edges. In the os- that NSAID, by inhibiting prostaglandin the relationship between COX-induced teogenic phase, the cartilaginous callus is synthesis, interfere with cell signalling dur- prostaglandin production and bone mor- replaced by woven bone via endochondral ing the inflammatory phase leading to an phogenetic proteins. A selective COX-2 in- ossification in a process reminiscent of uncoordinated healing response. Endogen- hibitor reduced BMP-2 expression in physeal growth. In the remodelling phase, ous prostaglandins E and F increase locally human mesenchymal stem cells; this effect young woven bone is replaced by lamellar in the first seven days after fracture, sug- was reversed with the addition of PGE2, bone to restore mechanical integrity. gesting a signalling function by prostag- suggesting COX-2 derived PGE2 induces The molecular mechanisms of bone landins in the early healing period (6–8). BMP-2 expression (24). Another study healing are complex and are the subjects of Attempts to simplify prostaglandin’s effects showed further support for this relation- current ongoing studies. Local and sys- on bone have been frustrated by the dif- ship, but indicated an opposite pathway in temic regulatory molecules including cyto- ficulty in simulating their naturally occur- that BMP-2 induced COX-2 expression in kines, growth and differentiation factors, ring expression and environment in vitro. cultured murine osteoblasts (25). In the hormones, and extracellular matrix mol- The small, quickly metabolised and locally same study, BMP-2-induced ectopic ossifi- ecules regulate processes such as cellular acting molecules can promote bone resorp- cation was inhibited in COX-2-/- mice, indi- migration, proliferation, differentiation, tion, bone formation, or both depending cating that BMP-2 effects on bone extracellular protein synthesis, and apopto- on the extracellular environment and the formation are in part due to COX-2 acti- sis (15). The most important local mol- intracellular chemical milieu (9). Exogen- vation. Finally, immunohistochemistry de- ecules currently recognised in fracture ous prostaglandin E administration in- tected a simultaneous rise in BMP-7 (also healing can be divided into three cat- creases regional acceleratory phenomenon known as osteogenic protein-1, or OP-1) egories: 1) pro-inflammatory cytokines, in- after rib fractures in dogs, increasing both and COX-2 at fracture calluses in mice. In cluding interleukin-1 (IL-1), interleukin-6 resorption and production in response to the same study, BMP-7-induced ectopic (IL-6), tumour necrosis factor-alpha injury (17, 18). Synthetic prostaglandin re- ossification was inhibited by diclofenac, an (TNF-α), platelet-derived growth factor ceptor agonists have an overall anabolic ef- NSAID (26). (PDGF), and others; 2) transforming fect and enhance bone healing in rats and Inhibition of angiogenesis has also been growth factor-beta superfamily members, dogs in experimental conditions (19, 20). suggested as a mechanism of NSAID effects which include bone morphogenetic pro- Inhibitory effects on mesenchymal stem on bone healing. Cyclooxygenase-2 meta- teins (BMP), transforming growth factor cell differentiation and activity have been bolic products (mainly prostaglandins) are beta isoforms (TGF-β), growth differenti- proposed as a mechanism for NSAID-in- stimulators of angiogenesis, which is essen- ation factors (GDF), activins, inhibins, and duced bone inhibition (21–23). In a no- tial for fracture healing (27, 28). Rofecoxib Müllerian inhibiting substance; and 3) an- table in vivo study involving an NSAID, di- inhibited osteotomy healing in mice fe- giogenic factors, including vascular en- clofenac profoundly reduced the number murs and significantly reduced blood flow dothelial growth factor (VEGF) and angio- of histologically visible osteoblasts at the within the femoral calluses (29). After ana- poietins (15, 16). Although there is con- injury site 10 days after bone drill defects lysing linear regression models, however, siderable overlap, these local molecules are were created in rat femora, suggesting in- the authors concluded that rofecoxib’s expressed in distinct temporal (and spatial) hibition of osteoblast differentiation or mi- negative effect on blood flow was indepen- patterns which can be loosely correlated to gration (23). Tibial fracture calluses in mice dent of rofecoxib’s negative effect on bone the histological stages of fracture healing lacking COX-2 expression (COX-2-/-) had healing. (14, 16). In general, inflammatory cyto- markedly fewer histologically visible os- kines are elevated in the inflammatory teoblasts and a higher incidence of fibrous phase and partially during secondary bone nonunion compared with wild-type mice In vivo evidence formation in the remodelling phase. Some (21). In a cell culture model by the same au- members of the TGF-β superfamily, such as thors, bone marrow cells from COX-2-/- Over thirty years of research have estab- bone morphogenetic protein-2 (BMP-2), mice failed to differentiate into osteoblasts lished that NSAID can slow fracture heal- are expressed at all stages of healing where- to the same degree as cells from wild-type ing in rodents. Using closed simple trans- as others are elevated at more specific mice when stimulated by osteogenic mol- verse long-bone fractures in rats and mice, times, such as during the cartilaginous ecules. The addition of prostaglandin E2 investigators have repeatedly documented phase (TGF-β) or during the osteogenic (PGE2) restored osteoblastogenesis to wild- bone healing inhibition as measured by phase (BMP-3, -8). Angiogenic factors, type levels. Bone morphogenetic protein-2 gross examination, nonunion rates, radio- though present throughout fracture heal- also induced osteoblastogenesis in both graphic scores, bone mineral density,

Vet Comp Orthop Traumatol 6/2010 © Schattauer 2010 S. Barry: Non-steroidal anti-inflammatory drugs inhibit bone healing 387

Table 1 Effect of non-steroidal anti-inflammatory drugs on rodent and rabbit models of bone healing.

Year, Author Animal Drug(s) Results 1976, Ro (1) Rats Indomethacin Decreased radiographical and histological healing, decreased mechanical strength of healed bone. 1979, Sudmann (57) Rabbits Indomethacin Inhibited Haversian remodelling in response to osteotomy. 1980, Allen (47) Rats Indomethacin, Both drugs caused dose-dependent decrease in fracture callus histological healing score. Aspirin 1980, Törnkvist (58) Rats Ibuprofen Decreased dry weight, ash weight, and organic matter weight of healing bones. 1982, Elves (59) Rats Indomethacin Decreased osteogenesis (strontium uptake) in response to drill defect. 1990, Keller (30) Rabbits Indomethacin No effect on Haversian remodelling close to a 2 mm drill defect, measured six weeks after operation. 1998, Reikeraas (31) Rats Indomethacin, Indomethacin weakened mechanical strength of healed fractures at 6 weeks; ketorolac had no Ketoralac effect at 6 weeks (3 days of treatment). 2002, Long (32) Rabbits Indomethacin, Indomethacin lowered spinal fusion rate; celecoxib had no effect. Celecoxib 2002, Simon (35) Rats, mice Indomethacin, Part I: Indomethacin delayed bone healing temporarily; rofecoxib and celecoxib had more Rofecoxib, permanent effects. Celecoxib Part II: Cox-2-/- mice had impaired bone healing. 2003, Beck (38) Rats Diclofenac Decreased bone density and mechanical strength of healing bone. 2003, Gerstenfeld Rats Ketorolac, Ketorolac decreased bone healing (histology and mechanical testing); parecoxib showed variable (34) Parecoxib results. 2003, Giordano (45) Rats Tenoxicam Decreased fracture healing (histomorphometry). 2003, Riew (46) Rabbits Indomethacin Decreased spinal fusion rate. 2004, Brown (36) Rats Indomethacin, Both drugs decreased histological healing scores; only indomethacin had effect on mechanical Celecoxib strength. 2005, Goodman (41) Rabbits Rofecoxib Decreased bony ingrowth into titanium chamber (histomorphometry). 2005, Endo (37) Rats Etodolac Time and duration-dependent reduction in fracture healing (radiographic scores and mechanical testing). 2005, Persson (60) Rats Indomethacin Decreased heterotopic bone formation, decreased mineralisation of traumatised femurs. 2006, Leonelli (42) Rats Rofecoxib, Rofecoxib decreased fracture healing (nonunion rate, radiographic score, histological score); Ibuprofen ibuprofen had little effect. 2006, Murnaghan Mice Rofecoxib Decreased fracture healing and blood flow across fracture callus. (29) 2007, Krischak (39) Rats Diclofenac Duration-dependent decrease in fracture healing (histomorphometry). 2007, Gerstenfeld (8) Rats Ketorolac, Duration-dependent decrease in fracture healing (nonunion rate, mechanical testing). Valdecoxib 2007, Simon (7) Rats Celecoxib Dose- and duration-dependent decrease in fracture healing (nonunion rate, radiographic score, and mechanical testing). 2008, Dimmen (40) Rats Parecoxib Transient reduction in bone mineral density during early healing phase (treatment for 7 days). 2009, Dimmen (43) Rats Indomethacin, Both drugs reduced bone mineral density and mechanical parameters of healing fractures. Parecoxib 2009, O'Connor (61) Rabbits Ibuprofen, Rofecoxib reduced mechanical strength more than ibuprofen. Rofecoxib 2010, Nyangoga (33) Rabbits Ketoprofen No effect on osteoconduction induced by β-tricalcium phosphate graft in femoral epiphyseal defects. 2010, Spiro (26) Mice Diclofenac Decreased fracture healing (micro-computed tomography, mechanical testing).

radiopharmaceutical uptake, histomor- dosages using allometric scaling calcu- particular time are studied (ǠTable 1). phometry, micro-computed tomography, lations. This research was contrasted by a Non-steroidal anti-inflammatory drugs and mechanical testing of fracture calluses study involving femoral fractures in male frequently used in veterinary medicine (ǠTable 1). The intensity and reliability of rats, where celecoxib inhibited fracture such as carprofen, meloxicam, deracoxib, this inhibition raises concerns about the healing as measured by radiographic scores firocoxib, and flunixin have not been clinical ramifications of prescribing and nonunion rate (35). Celecoxib failed to studied in vivo, with the exception of a re- NSAID to other animals in the period fol- produce significantly different mechanical cent abstract which evaluated carprofen in lowing a fracture or orthopaedic pro- testing scores than in untreated rats, how- dogs (44). cedure. Although conclusions are limited ever. Four years later, the same authors by experimental design and are sometimes showed that in female rats, celecoxib sig- conflicting, the current experimental data nificantly delayed femoral fracture healing NSAID may have their greatest would suggest the following points: when measured by mechanical testing (7). effect during the early phase of l All NSAID studied to date have the po- In their discussion, the authors pointed out bone healing tential to inhibit bone formation the discovery that intact male and female l NSAID may have their greatest effect Sprague-Dawley rats have markedly differ- Rats given etodolac for only the first week during the early phase of bone healing ent celecoxib metabolism, and that the after a femoral fracture had substantial de- l NSAID have a dose-dependent effect on plasma elimination half-life of celecoxib in creases in radiographic scores and callus bone healing female rats is nearly four times longer than strength measured three weeks after the frac- l NSAID often have duration-dependent that of male rats. ture (37). Their decreases were comparable and reversible effects on bone healing Cyclooxyenase-2 was discovered in 1991, to those of rats treated with etodolac for the l NSAID use before bone injury does not and the first coxibs – NSAID designed spe- entire three weeks of their healing period. In seem to adversely affect bone healing cifically to be COX-2 selective – were re- contrast, rats treated for only the last of three leased soon thereafter (5). Initial studies weeks had scores which were not signifi- comparing selective to non-selective COX cantly different from untreated rats. The All NSAID studied to date inhibitors argued that COX-2 selective seemingly permanent effects of short-term have the potential to inhibit NSAID did not have as profound an effect treatment with etodolac in this study bone formation against bone healing as indomethacin, a contrast with those of subsequent studies in traditional non-selective NSAID (32, 34, which the effects of short-term treatment From time to time, an article of research is 36). Subsequent genetic and in vivo studies with NSAID were reversible (7, 8, 39, 40). published which concludes that a particu- have provided strong evidence that inflam- However, in agreement with this study, other lar NSAID does not, like other NSAID, in- mation stimulated by the COX-2 isoenzyme studies also suggest that NSAID have their hibit bone formation (30–33). These is essential for fracture healing, and that greatest inhibitory effect in the early inflam- studies are often followed by subsequent COX-2 selective NSAID have an equal or matory phase of bone healing (7, 38, 45, 46). studies showing that with a different ex- greater negative effect on fracture healing in Simon and others included a ‘time delay’ perimental design, the same drug can in- rats than non-selective NSAID (7, 8, 21, group in which rats were allowed to heal un- hibit bone healing. For example, ketorolac 37–43). Thus, no single NSAID or group of treated for seven or 14 days after femoral (1 mg/kg/day) given for three days had no NSAID has been cleared as being ‘safe’, or fracture before receiving celecoxib for the effect on callus mechanical strength having no potential effect on bone healing. rest of the 28 day study period (7). Rats in the measured six weeks after fracture in rats Although adjusting the experimental group treated from seven to 28 days (21 day (31). However, ketorolac given at a higher design (increasing the dose or duration of treatment duration) had higher radio- dose (4 mg/kg/day) and for the entire time NSAID treatment) allows researchers to il- graphic scores and mechanical testing pa- course of the experiment reduced mechan- lustrate fracture healing inhibition for al- rameters at the study endpoint compared ical strength and decreased histological most any NSAID studied in rodent models, with rats treated from zero to 21 days (same scores of fracture calluses compared with it is important to acknowledge that even treatment duration). However, in statistical untreated rats (34). Similarly, the authors fractures in the NSAID-treated groups analysis, the treatment groups were com- of an experiment involving spinal fusion in often go on to heal. Additionally, it some- pared to untreated controls, and direct com- rabbits concluded that while indomethacin times takes sensitive outcome measures parisons between the time delay and other reduced spinal fusion in their model (vali- (such as histomorphometry) to detect a treatment groups were not performed. dating their study as one that can detect significant difference between groups. It is spinal fusion inhibition), celecoxib had no impossible to make conclusions about significant effect (32). In their discussion, clinical relevance in other species based on NSAID have a dose-dependent these authors admitted that the phar- these studies. effect on bone healing macokinetics of celecoxib in rabbits was Because most research in this field is unknown; the dose used in their study was geared towards human medicine, the most Few experiments include treatment groups extrapolated from recommended human promising NSAID for use in humans at a that differ in NSAID dose, however those

Vet Comp Orthop Traumatol 6/2010 © Schattauer 2010 S. Barry: Non-steroidal anti-inflammatory drugs inhibit bone healing 389 that do suggest that a higher dose is more tained significantly less bone and more car- in order to investigate whether NSAID detrimental to bone healing. A dose-de- tilage (39). Likewise, parecoxib given for would inhibit bone healing if discontinued pendent decrease in histological scores was seven days only transiently decreased bone at the time of bone trauma (7). This pre- seen in fracture calluses from rats given in- mineral density after tibial fractures in rats treatment with celecoxib had little to no ef- domethacin or aspirin (47). A dose-de- – bone mineral density returned to normal fect on fracture healing, suggesting that pa- pendent decrease in radiographic scores by three weeks (40). In a three-part study tients chronically taking NSAID, such as and mechanical properties was also ob- involving ketorolac and valdecoxib, the au- patients with osteoarthritis, would not served in fracture calluses from rats given thors not only noted a more detrimental ef- suffer diminished bone healing as a result celecoxib after femoral fractures (7). fect from both drugs when given for longer of the drug. Similar studies have not been durations – they also documented a drop in performed to investigate the effects of long- local prostaglandin E2 at the fracture callus er or more chronic durations of pre-treat- NSAID often have duration- during treatment followed by a rise in pros- ment. dependent and reversible effects taglandin E2 after the drugs were discon- on bone healing tinued (8). When measured seven days after NSAID discontinuation, local prostagland- Human clinical evidence Although some data suggest that effects of in E2 levels exceeded prostaglandin levels in early NSAID administration can persist untreated rats at the same time point after It is well-established that NSAID can pre- and be measured weeks after treatment has fracture, suggesting a rebound inflamma- vent heterotopic ossification in humans. ended, there is also evidence that NSAID- tory effect. However, controversy remains regarding induced inhibition is proportional to the the clinical effects of NSAID on fracture duration of treatment, and that bone can healing and spinal fusion. Only a handful of heal normally once NSAID administration NSAID use before bone injury does clinical studies exist evaluating the effect of is discontinued (8, 37–41, 43). Calluses not seem to adversely affect bone NSAID on fracture healing or spinal fusion from rats given diclofenac for seven days healing in humans (ǠTable 2). after osteotomy were not histologically dif- Heterotopic ossification, or ectopic ferent than calluses from rats given placebo Simon and others included in their study a bone formation, commonly occurs in the when evaluated at 21 days, whereas calluses group of 23 rats that were treated with ce- soft tissues around the hip joint following from rats treated for the entire 21 days con- lecoxib for five days before femoral fracture hip surgery in humans. It is estimated that

Table 2 Clinical research investigating non-steroidal anti-inflammatory drugs (NSAID) and bone healing in humans.

Year, Author Injury Drug(s) Results Prospective clinical trials 1988, Davis (52) Distal radial fractures Flurbiprofen No effect on healing in patients treated with flurbiprofen for 2 weeks. 1993, Adolphson (51) Distal radial fractures Piroxicam No effect on healing in postmenopausal women treated with piroxicam for 8 weeks. 2003, Burd (50) Long-bone fractures Indomethacin Increased nonunion rate in patients treated with indomethacin for 6 weeks. 2005, Reuben (62) Spinal fusion Celecoxib No effect on spinal fusion evaluated in 1 year in patients treated with celecoxib for 5 days after surgery. Article retracted by publisher due to fabricated data. Retrospective observational studies 1998, Deguchi (63) Spinal fusion Various NSAID Increased nonunion rate in patients taking NSAID. 1998, Glassman (64) Spinal fusion Ketorolac Increased nonunion rate by approximately 5 times in patients taking ketorolac. 2000, Giannoudis (65) Femoral fractures Various NSAID Increased nonunion rate in patients taking NSAID. 2005, Bhattacharyya (53) Humeral fractures Various NSAID Increased nonunion rate in patients taking NSAID or opioids 61-90 days after fracture. 2005, Reuben (66) Spinal fusion Celecoxib, No effect on spinal fusion evaluated 1 year after surgery in patients treated for 5 Rofecoxib, days with celecoxib, rofecoxib, or low-dose ketorolac; increased nonunion rate in Ketorolac patients treated with high-dose ketorolac. 2008, Pradhan (67) Spinal fusion Ketorolac No effect on spinal fusion evaluated 1 year after surgery in patients treated with ketorolac for 48 hours.

approximately one-third of humans the fracture site. Radiographic assessment Evidence in domestic undergoing total hip arthroplasty will de- of bone healing was limited to measure- velop some degree of heterotopic ossifica- ments of radial shortening and dorsal species tion, which can cause pain and reduced angulation of the radius over time. The pir- range-of-motion if moderate or severe oxicam group had a statistically greater de- Information about the effects of NSAID in (48). The etiopathogenesis is unknown, but gree of dorsal angulation than the placebo domestic species is lacking. One publication stimulation by trauma or surgery suggests group at four weeks but not at eight weeks. suggests an effect on bone remodelling in an inflammatory component. The ability of Of 42 patients, data from six fractures were horses and two studies address the effect on NSAID to inhibit heterotopic ossification omitted because excessive redislocation osteotomy healing in dogs (44, 54, 55). has been recognised since 1977, and a re- (defined by these authors as increased radi- An experimental study in horses found a cent meta-analysis concluded that medium al shortening and dorsal angulation) decreased mineral apposition rate (a to high doses of perioperative NSAID pro- necessitated re-reduction and surgical ex- measure of Haversian remodelling) in bone duce a substantial reduction in the inci- ternal fixation; all six of these patients were adjacent to tibial biopsy sites in horses dence of radiographic signs of heterotopic in the piroxicam group. The authors justi- treated with phenylbutazone for two weeks ossification (48, 49). Non-steroidal anti-in- fied the omission by pointing out that ex- (54). Although the healing biopsy sites con- flammatory drugs are commonly pre- ternal fixation would have had an effect on tained less mineralised tissue at 16 and 30 scribed to high-risk patients to prevent het- the bone mineral content proximal to the days in phenylbutazone-treated horses erotopic bone formation after hip surgery. fracture site. than in control horses, this difference was In a prospective, randomised, blinded Another prospective, double-blind, not significant. clinical study involving trauma patients with placebo-controlled study did not find that A prospective, randomised, placebo- acetabular fractures and concurrent long- flurbiprofen, given for two weeks at various controlled, cross-over study in dogs was bone fractures, 282 patients were allocated doses, had an effect on the healing of distal performed to evaluate the effects of phe- into three groups after open reduction and radial fractures (52). While bone area nylbutazone and indomethacin on the internal fixation of their acetabular fractures measurements in the hand (indices of os- postoperative course following experimen- (50). One group received indomethacin for teoporosis), hand and wrist function, and tal metacarpal osteotomy (55). Bone heal- six weeks as prophylaxis against heterotopic pain were measured in the short term, frac- ing was evaluated by subjective assessment ossification, while the other two groups re- ture healing was not assessed until one year of radiographs taken two, four, six, and ceived either no prophylaxis or a single dose after fracture. If a transient delay in bone eight weeks after surgery. No attempt to ob- of local radiation to the soft tissues around healing had occurred during flurbiprofen jectively score the radiographs was re- the hip joint. The patients receiving indome- treatment, this study would not have de- ported, and whether this assessment was thacin had a significantly higher nonunion tected it. However, this study is clinically blinded was also not reported. In dogs rate of their long-bone fractures (29%) helpful because it indicates that short-term treated with phenylbutazone the authors compared with patients who did not receive use of flurbiprofen after distal radial frac- reported, “…radiographs taken four and indomethacin (7%; p <0.004), strongly sug- ture does not result in nonunion or signifi- six weeks after surgery … revealed tenden- gesting that long-term therapy with in- cant malunion. cies in favour of placebo… After eight domethacin clinically inhibited long-bone Retrospective observational studies weeks, however, there were no noticeable fracture healing. have frequently found an association be- differences.” No noticeable radiographic In contrast, a prospective, randomised, tween NSAID use and increased nonunion differences in bone healing were seen in double-blind, placebo-controlled clinical rates after fracture or spinal fusion surgery dogs treated with indomethacin. Given the study did not find piroxicam to affect the in humans (ǠTable 2). Few studies make lack of objective assessment and the fact healing of distal radial fractures in post- an effort to identify confounding factors that these two NSAID are rarely used in menopausal women despite treatment for however, and their retrospective designs small animal medicine today, it is difficult eight weeks (51). Unfortunately, the out- make inferences about causation difficult. to make clinical decisions based on this come measures, data handling, and fund- For example, a large and frequently-cited study. ing source for this study lead the reader to retrospective study involving 9,995 patients Finally, a recent abstract described an question the conclusion with regard to with humeral shaft fractures reported a experimental study designed to evaluate bone healing. The stated objective of this strong association between NSAID use and the effect of carprofen on bone healing in study was to determine whether piroxicam nonunion (53). Because this association dogs (44). Healthy dogs received either car- changed the degree of osteopenia in the was only significant in the latest time phase profen (2.2 mg/kg PO q 12 hrs) or no treat- nearby radius and ulna after coaptation for studied (61–90 days after fracture) and be- ment for four months after a tibial osteot- fracture, rather than whether piroxicam af- cause the same pattern was observed with omy was stabilised with an intramedullary fected bone healing. The outcome opioid use, the authors concluded that pin. The NSAID group had delayed bone measures were geared towards this objec- NSAID and opioid use were markers for healing compared with the control group tive, and included bone mineral content patients in pain from unstable fractures, as detected by radiographical, biomech- analysis of the radius and ulna proximal to rather than the cause of nonunion. anical, and histological analysis.

Vet Comp Orthop Traumatol 6/2010 © Schattauer 2010 S. Barry: Non-steroidal anti-inflammatory drugs inhibit bone healing 391

chronic inflammatory disease, primary 8. Gerstenfeld LC, Al-Ghawas M, Alkhiary YM, et al. Discussion Selective and nonselective cyclooxygenase-2 in- musculoskeletal disease), drug adminis- hibitors and experimental fracture-healing. Re- While NSAID negatively affect bone heal- tration (corticosteroids, chemotherapeutic versibility of effects after short-term treatment. J ing, there is insufficient evidence to sup- agents, antibiotics, anticoagulants, bisp- Bone Joint Surg Am 2007; 89: 114–125. port withholding NSAID after a fracture or hosphanates, smoking), immobility, and 9. Blackwell KA, Raisz LG, Pilbeam CC. Prostagland- ins in bone: bad cop, good cop? Trends Endocrinol orthopaedic procedure in domestic others (56). The use of NSAID could have Metab 2010; 21: 294–301. species. Certainly, fractures and osteot- negative, additive, or synergistic effects on 10. McKibben B. The biology of fracture healing in long omies can heal in the face of NSAID ad- bone healing when combined with other bones. J Bone Joint Surg Br 1978; 60-B: 150–162. ministration, and NSAID are important for conditions or environmental factors. To 11. Einhorn TA. The cell and molecular biology of fracture healing. Clin Orthop Relat Res 1998; 355S: their potent anti-inflammatory and anal- date there is little to no information investi- S7-S21. gesic effects after bone injury. However, gating these combined effects or identify- 12. Frost HM. The regional acceleratory phenomenon: when the speed and effectiveness of bone ing patients at greatest risk for NSAID-in- A review. Henry Ford Hosp Med J 1983; 31: 3–9. healing is of particular concern, such as in duced delayed fracture healing. 13. Einhorn TA. The science of fracture healing. J Or- thop Traum 2005; 19(10 Suppl): S4-S6. delayed union, nonunion, tenuous ortho- 14. Cho TJ, Gerstenfeld LC, Einhorn TA. Differential paedic repairs, or in patients where delayed temporal expression of members of the transform- bone healing is expected, it is prudent to Conclusion ing growth factor ß superfamily during murine frac- consider the evidence when prescribing ture healing. J Bone Miner Res 2002; 17: 513–520. 15. Tsiridis E, Upadhyay N, Giannoudis P. Molecular NSAID. Research using rodent and rabbit models aspects of fracture healing: Which are the impor- In rodent models, COX-2 selective provides strong evidence that NSAID tant molecules? Injury 2007; 38(S1): S11-S25. NSAID do not seem to offer a benefit with negatively affect bone healing. Whether 16. Al-Aql ZS, Alagl AS, Graves DT, et al. Molecular regards to bone healing compared with this effect is clinically important is contro- mechanisms controlling bone formation during fracture healing and distraction osteogenesis. J nonselective NSAID. In humans, arguably versial in both human and veterinary medi- Dent Res 2008; 87: 107–118. the most convincing study showing a clini- cine. Prospective experimental and clinical 17. Shih MS, Norridin RW. Effects of prostaglandins on cally relevant effect on bone healing in- trials involving domestic species and vet- regional remodeling changes during tibial healing in beagles: a histomorphometric study. Calcif Tis- volved indomethacin, a nonselective erinary NSAID are needed to guide veter- sue Int 1986; 39: 191–197. NSAID (50). In dogs, carprofen, a COX-2 inarians as they decide which, when, at 18. Shih MS, Norridin RW. Effects of prostaglandin E2 selective NSAID, inhibited bone healing what dose, and for how long NSAID should on rib fracture healing in beagles: histomorpho- (44). The decision to use a COX-2 selective be prescribed after bone injury. metric study on periosteum adjacent to the fracture site. Am J Vet Res 1986; 47: 1561–1564. or nonselective NSAID should be made 19. Paralkar VM, Borovecki F, Cameron KO, et al. An based on other criteria, such as the poten- Acknowledgements EP2 receptor-selective prostaglandin E2 agonist in- tial for adverse effects, cost, and availability. duces bone healing. Proc Natl Acad Sci USA 2003; Experimental literature suggests that I gratefully acknowledge Dr. Steve Marti- 100: 6736–6740. 20. Tanaka M, Sakai A, Uchida S, et al. Prostaglandin E2 re- higher doses and longer durations of nez for his guidance and contribution to ceptor (EP4) selective agonist (ONO-4819.CD) accel- NSAID are more detrimental to bone heal- this manuscript. erates bone repair of femoral cortex after drill-hole in- ing. In addition, there is evidence to sup- jury associated with local upregulation of bone turn- port the notion that NSAID-induced in- over in mature rats. Bone 2004; 34: 940–948. 21. Zhang X, Schwarz EM, Young DA, et al. Cyclooxyge- hibition can be reversible once the NSAID nase-2 regulates mesenchymal cell differentiation is discontinued. Reports that illustrate References into the osteoblast lineage and is critically involved bone healing inhibition in humans and 1. Ro J, Sudmann E, Marton PF. Effect of indometha- in bone repair. J Clin Invest 2002; 109: 1405–1415. dogs involve NSAID administered for long cin on fracture healing in rats. Acta Orthop Scand 22. Kellinsalmi M, Parikka V, Risteli J, et al. Inhibition of 1976; 47: 588–599. cyclooxygenase-2 down-regulates osteoclast and durations (44, 50). Therefore, it makes osteoblast differentiation and favours adipocyte 2. Sudmann E, Hagen T. [Indomethacin-induced de- formation in vitro. Eur J Pharmacol 2007; 572: sense to discontinue NSAID use after a rea- layed fracture healing]. Arch Orthop Unfall-Chir sonable post-injury period if the NSAID is 1976; 85: 151–154. 102–110. 3. O’Connor JP, Lysz T. Celecoxib, NSAIDs and the 23. Krischak GD, Augat P, Blakytny R, et al. The non- no longer needed for analgesia. steroidal anti-inflammatory drug diclofenac re- Many other local and systemic factors skeleton. Drugs Today 2008; 44: 693–709. 4. Vane JR. Inhibition of prostaglandin synthesis as a duces appearance of osteoblasts in bone defect heal- can affect bone healing, making it difficult mechanism of action for aspirin-like drugs. Nat ing in rats. Arch Orthop Trauma Surg 2007; 127: 453–458. to isolate the independent effect of NSAID New Biol 1971; 231: 232–235. 24. Arikawa T, Omura K, Morita I. Regulation of bone 5. Clark TP. The clinical pharmacology of cyclooxyge- in a clinical setting. Local factors include morphogenetic protein-2 expression by endogen- type of fracture, fracture gap, fixation tech- nase-2-selective and dual inhibitors. Vet Clin Small Anim 2006; 36: 1061–1085. ous prostaglandin E2 in human mesenchymal stem nique (reduction, position, stability), the 6. Dekel S, Lenthall G, Francis MJO. Release of pros- cells. J Cell Physiol 2004; 200: 400–406. presence of infection or debris, and the de- taglandins from bone and muscle after tibial frac- 25. Chikazu D, Li X, Kawaguchi H, et al. Bone morpho- ture. J Bone Joint Surg 1981; 63-B: 185–189. genetic protein 2 induces cyclo-oxygenase 2 in os- gree of vascularisation. Systemic factors in- teoblasts via a Cbfa1 binding site: Role in effects of clude age and gender, metabolic and nutri- 7. Simon AM, O’Connor JP. Dose and time-dependent effects of cyclooxygenase-2 inhibition on frac- bone morphogenetic protein 2 in vitro and in vivo. tional status, concurrent disease (endocri- ture-healing. J Bone Joint Surg Am 2007; 89: J Bone Miner Res 2002; 17: 1430–1440. nopathies, neoplasia, immunodeficiency, 500–511.

26. Spiro AS, Beil FT, Baranowsky A, et al. BMP-7-in- 40. Dimmen S, Nordsletten L, Engebretsen L, et al. humeral shaft fractures. Arthritis Rheum 2005; 53: duced ectopic bone formation and fracture healing Negative effect of parecoxib on bone mineral dur- 364–367. is impaired by systemic NSAID application in ing fracture healing in rats. Acta Orthop 2008; 79: 54. Rohde C, Anderson DE, Bertone AL, et al. Effects of C57BL/6-mice. J Orthop Res 2010; 28: 785–791. 438–444. phenylbutazone on bone activity and formation in 27. Sahin M, Sahin E, Gümüslü S. Cyclooxygenase-2 in 41. Goodman SB, Ma T, Mitsunaga L, et al. Temporal horses. Am J Vet Res 2000; 61: 537–543. cancer and angiogenesis. Angiology 2009; 60: effects of a COX-2 selective NSAID on bone in- 55. Mbugua SW, Skoglund LA, Løkken P. Effects of 242–253. growth. J Biomed Mater Res 2005; 72: 279–287. phenylbutazone and indomethacin on the post-op- 28. Hausman MR, Schaffler MB, Majeska RJ. Preven- 42. Leonelli SM, Goldberg BA, Safanda J, et al. Effects of erative course following experimental orthopaedic tion of fracture healing in rats by an inhibitor of an- a cyclooxygenase-2 inhibitor (rofecoxib) on bone surgery in dogs. Acta Vet Scand 1989; 30: 27–35. giogenesis. Bone 2001; 29: 560–564. healing. Am J Orthop 2006; 35: 79–84. 56. Pountos I, Georgouli T, Blokhuis TJ, et.al. Phar- 29. Murnaghan M, Li G, Marsh DR. Nonsteroidal anti- 43. Dimmen S, Nordsletten L, Madsen JE. Parecoxib macological agents and impairment of fracture inflammatory drug-induced fracture nonunion: An and indomethacin delay early fracture healing: a healing: What is the evidence? Injury 2008; 39: inhibition of angiogenesis? J Bone Joint Surg Am study in rats. Clin Orthop Rel Res 2009; 467: 384–394. 2006; 88: 140–147. 1992–1999. 57. Sudmann E, Bang G. Indomethacin-induced in- 30. Keller J, Kjærsgaard-Andersen P, Bayer-Kristensen 44. Ochi H, Hara Y, Asou Y, et al. Effect of long-term ad- hibition of Haversian remodelling in rabbits. Acta I, et al. Indomethacin and bone trauma. Effects on ministration of non-steroidal anti-inflammatory Orthop Scand 1979; 50: 621–627. remodeling of rabbit bone. Acta Orthop Scand drugs on fracture healing in dogs. (Poster presenta- 58. Törnkvist H, Lindholm TS. Effect of ibuprofen on 1990; 61: 66–69. tion) ACVS Veterinary Symposium 2009 October mass and composition of fracture callus and bone. 31. Reikeraas O, Engebretsen L. Effects of ketorolac 8-10; Washington DC, USA. Scand J Rheumatol 1980; 9: 167–171. tromethamine and indomethacin on primary and 45. Giordano V, Giordano M, Knackfuss IG, et al. Effect 59. Elves MW, Bayley I, Roylance PJ. The effect of in- secondary bone healing. An experimental study in of tenoxicam on fracture healing in rat tibiae. Injury domethacin upon experimental fractures in the rat. rats. Arch Orthop Trauma Surg 1998; 118: 50–52. 2003; 34: 85–94. Acta Orthop Scand 1982; 53: 35–41. 32. Long J, Lewis S, Kuklo T, et al. The effect of cyclo- 46. Riew KD, Long J, Rhee J, et al. Time-dependent in- 60. Persson P, Sisask G, Nilsson O. Indomethacin in- oxygenase-2 inhibitors on spinal fusion. J Bone hibitory effects of indomethacin on spinal fusion. J hibits bone formation in inductive allografts but Joint Surg Am 2002; 84-A: 1763–1768. Bone Joint Surg Am 2003; 85: 632–634. not in autografts. Acta Orthop 2005; 76: 465–469. 33. Nyangoga H, Aguado E, Goyenvalle E, et al. A non- 47. Allen HL, Wase A, Bear WT. Indomethacin and as- 61. O’Connor JP, Capo JT, Tan V, et al. A comparison of steroidal anti-inflammatory drug (ketoprofen) pirin: effect of nonsteroidal anti-inflammatory the effects of ibuprofen and rofecoxib on rabbit does not delay β-TCP bone graft healing. Acta Bio- agents on the rate of fracture repair in the rat. Acta fibula osteotomy healing. Acta Orthop 2009; 80: mater 2010; 6: 3310-3317. Orthop Scand 1980; 51: 595–600. 597–605. 34. Gerstenfeld LC, Thiede M, Seibert K, et al. Differen- 48. Fransen M, Neal B. Non-steroidal anti-inflamma- 62. Rueben SS, Ekman EF. The effect of cyclooxyge- tial inhibition of fracture healing by non-selective tory drugs for preventing heterotopic bone nase-2 inhibition on analgesia and spinal fusion. J and cyclooxygenase-2 selective non-steroidal anti- formation after hip arthroplasty. Cochrane Data- Bone Joint Surg Am 2005; 87: 536–542. inflammatory drugs. J Orthop Res 2003; 21: base Syst Rev 2004; 3: CD001160. 63. Deguchi M, Rapoff AJ, Zdeblick TA. Posterolateral 670–675. 49. Almåsbakk K, Røysland P. Does indomethacin fusion for isthmic spondylolisthesis in adults: 35. Simon AM, Manigrasso MB, O’Connor JP. Cyclo- (IMC) prevent postoperative ectopic ossification in Analysis of fusion rate and clinical results. J Spinal oxygenase 2 function is essential for bone fracture total hip replacement? (Abstract) Acta Orthop Disord 1998; 11: 459–464. healing. J Bone Miner Res 2002; 17: 963–976. Scand 1977; 48: 556. 64. Glassman SD, Rose SM, Dimar JR. The effect of 36. Brown KM, Saunders MM, Kirsch T, et al. Effect of 50. Burd TA, Hughes MS, Anglen JO. Heterotopic ossi- postoperative nonsteroidal anti-inflammatory COX-2-specific inhibition on fracture-healing in fication prophylaxis with indomethacin increases drug administration on spinal fusion. Spine 1998; the rat femur. J Bone Joint Surg 2004; 86: 116–123. the risk of long bone nonunion. J Bone Joint Surg Br 23: 834–838. 37. Endo K, Sairyo K, Komatsubara S, et al. Cyclooxyge- 2003; 85-B: 700–705. 65. Giannoudis PV, MacDonald DA, Matthews SJ, et al. nase-2 inhibitor delays fracture healing in rats. Acta 51. Adolphson P, Abbaszadegan H, Jonsson U, et al. No Nonunion of the femoral diaphysis: The influence Orthop 2005; 76: 470–474. effects of piroxicam on osteopenia and recovery of reaming and non-steroidal anti-inflammatory 38. Beck A, Krischak G, Sorg I, et al. Influence of diclo- after Colles’ fracture. Arch Orthop Trauma Surg drugs. J Bone Joint Surg Br 2000; 82-B: 655–658. fenac (group of nonsteroidal anti-inflammatory 1993; 112: 127–130. 66. Reuben SS, Ablett D, Kaye R. High dose nonsteroi- drugs) on fracture healing. Arch Orthop Trauma 52. Davis TRC, Ackroyd CE. Non-steroidal anti-in- dal anti-inflammatory drugs compromise spinal Surg 2003; 123: 327–332. flammatory agents in the management of Colles’ fusion. Can J Anesth 2005; 52: 506–512. 39. Krischak G, Augat P, Sorg T, et al. Effects of diclofen- fractures. Br J Clin Pract 1988; 42: 184–189. 67. Pradhan BB, Tatsumi RL, Gallina J, et al. Ketorolac ac on periosteal callus maturation in osteotomy 53. Bhattacharyya T, Levin R, Vrahas MS, et al. Nonste- and spinal fusion: Does the perioperative use of ke- healing in an animal model. Arch Orthop Trauma roidal antiinflammatory drugs and nonunion of torolac really inhibit spinal fusion? Spine 2008; 33: Surg 2007; 127: 3–9. 2079–2082.