Article Text


Review of the safety and efficacy of long-term NSAID use in the treatment of canine osteoarthritis
  1. J. F. Innes, BVSc, PhD, CertVR, DSAS(orth), MRCVS1,
  2. J. Clayton1 and
  3. B. D. X. Lascelles, BVSc, PhD, DSAS(ST), DipECVS, DipACVS, MRCVS2
  1. 1 Small Animal Teaching Hospital and Musculoskeletal Research Group, Faculty of Veterinary Science, University of Liverpool, Leahurst Campus Neston, Wirral CH64 7TE
  2. 2 Comparative Pain Research Laboratory, College of Veterinary Medicine, North Carolina State University, Raleigh, 4700 Hillsborough Street, Raleigh, NC 27606, USA
  1. E-mail for correspondence: j.f.innes{at}


The published, peer-reviewed literature was systematically searched for information on the safety and efficacy of long-term (defined as 28 days or more of continuous therapy) NSAID use in the treatment of canine osteoarthritis. Online databases were reviewed in June 2008 and papers were selected based on their relevance. Fifteen papers were identified and evaluated. Six of seven papers indicated a benefit of long-term treatment over short-term treatment in terms of the reduction of clinical signs or lameness; one study showed no benefit. Fourteen papers evaluated safety with calculated experimental (adverse) event rates (EER) between 0 and 0.31, but there was no correlation between study length and EER (rs=-0.109, P=0.793). The balance of evidence for the efficacy of NSAIDs supports longer-term use of these agents for increased clinical effect. There is no indication in the literature that such an approach is associated with a reduction in safety, although robust data on the safety of long-term NSAID use are lacking in large numbers of dogs.

    Statistics from

    OSTEOARTHRITIS is a disease process for which clinical signs include pain, inflammation and lameness. It is associated with pathological changes in the tissues of the affected synovial joint. These pathological changes include a partial or total loss of articular cartilage. The disease is common, with reports estimating that up to 20 per cent of dogs over one year of age may develop the disease (Johnston 1997).

    There is no cure for osteoarthritis, although many different treatments are available in both human and veterinary medicine to aid the attempted management of the disease. These treatments vary greatly and may include dietary therapy, physiotherapy, ‘alternative’ therapies (for example, acupuncture) and surgery, with joint replacement surgery being employed in severe cases. One of the most common treatments is symptom-modifying pharmacotherapy, utilising NSAIDs or analgesics.

    NSAIDs are anti-inflammatory agents that aim to inhibit the cyclooxygenase (COX) enzyme, which catalyses the conversion of arachidonic acid to prostaglandins and thromboxane. The US Food and Drug Administration (FDA) has approved six NSAIDs for use in dogs: carprofen, meloxicam, tepoxalin, firocoxib, deracoxib and etodolac. The annual cost of the treatment of dogs with NSAIDs in the USA exceeds US$130 million a year and is growing by approximately 13 per cent a year (Schmit 2005).

    Recommendations for prescribing practices for NSAIDs in the management of canine osteoarthritis vary greatly, with some authors recommending intermittent ‘as needed’ therapy and others recommending continuous therapy. The potential benefits of continuous therapy include better control of pain, greater improvements in mobility and the potential slowing of the disease process through improved joint usage (for example, reduction in disuse-mediated muscle atrophy). The potential adverse effects of continuous therapy include tolerance over time to the drug, an increased incidence of adverse events associated with the use of the drug, as well as compliance issues.

    The central nervous system has plasticity, and the functional alteration in response to peripheral nociceptive input is called central sensitisation. Central sensitisation facilitates nociceptive throughput and amplifies signals, resulting in greater perceived pain. There is evidence that joint pain results in the development of central sensitisation (Menétrey and Besson 1982, Neugebauer and Schaible 1990). Furthermore, it has been demonstrated that COX enzymes play a role in central sensitisation (Samad and others 2001, Veiga and others 2004) and that COX inhibitors can inhibit this process (Veiga and others 2004). If the use of continuous NSAID therapy over time led to a reduction in central sensitisation, there should also be a concomitant progressive reduction in the pain perceived by the animal. There is a growing body of evidence that central sensitisation can actually drive the progression of disease in the periphery (joints), and that downward modulation of central sensitisation can result in decreased joint pathology (Sluka and others 1994, Fiorentino and others 2008).

    In addition, a direct effect of NSAIDs at the level of the joint may result in a reduction in disease progression (Pelletier and others 1999, 2004). One such mechanism is via the prevention of nitric oxideinduced cell death. Several studies have shown that osteoarthritic cartilage has a higher number of apoptotic chondrocytes than normal cartilage in animal models and human beings. The production of nitric oxide may represent an important component in the pathogenesis of osteoarthritis.

    Nitric oxide is produced in large quantities by chondrocytes following proinflammatory cytokine stimulation. Selective inhibition of COX-2 has been shown to significantly inhibit nitric oxide-induced cell death (Jovanovic and others 2002). Therefore, a number of lines of evidence suggest the potential theoretical benefits of continuous versus intermittent NSAID analgesic therapy for osteoarthritis.

    A possible, and potentially serious, effect of practising continuous NSAID therapy for the control of osteoarthritic pain is an increase in the incidence of adverse events. Gastrointestinal complications may occur in some individuals with the use of NSAIDs and it is most likely that it is the perception of this risk that restricts their long-term use. However, there are no accurate and controlled estimates for the incidence of adverse events with long-term NSAID use in dogs.

    To date, there are no studies that have evaluated the comparative efficacy of everyday versus intermittent NSAID therapy in dogs with osteoarthritis, and only one recent study has been conducted in human medicine (Luyten and others 2007). In that study, there were no significant differences between patients randomised to continuous or intermittent treatment, except for the intake of rescue analgesia for ‘flares’, which was less in the continuous treatment group.

    The aims of this review were to collate all the information on long-term (28 days or more) NSAID use and to evaluate the evidence for the safety and efficacy of long-term NSAID therapy for the treatment of osteoarthritis in dogs. Further aims were to evaluate the evidence for progressive decreases in pain, or progressive tolerance (lack of efficacy of the drug, which increases pain) over time; to evaluate the evidence for altered disease progression with long-term continuous use; and to evaluate the evidence for an increase (or decrease) in the incidence of adverse events with long-term use.

    Materials and methods

    Search design

    Five online databases were identified covering the veterinary literature in the English language. These were: Intute: Veterinary medicine; Journals@Ovid Full Text; Ovid MEDLINE; Veterinary Science Database; and ISI Web of Knowledge.

    The disease term identified was ‘osteoarthritis’. Three species terms were identified for the search within the veterinary databases, ‘dog’, ‘dogs’, and ‘canine’. The terms ‘NSAID’ and ‘non-steroidal’ were also used, as well as the generic names of a number of common NSAIDs, including FDA- or European Medicines Agency (EMEA)- approved drugs; these included carprofen, meloxicam, tepoxalin, firocoxib, deracoxib, etodolac and phenylbutazone. All terms were then combined using Boolean operators, firstly in an ‘AND’ combination and then in an ‘OR’ combination.

    Inclusion criteria

    Studies included in the review were those evaluating canine osteoarthritis affecting synovial joints of the appendicular skeleton only. Papers were only included if the NSAID therapy was continued for 28 days or more. All articles from the database searches were reviewed and included if they assessed either the safety or efficacy of prolonged NSAID use, or a combination of safety and efficacy. Efficacy was defined as any study that evaluated the functional or structure-modifying effects of NSAIDs on osteoarthritis in dogs.

    Evaluation criteria

    Evaluation criteria were based on the system for scientific data produced by the FDA (FDA 2009), with minor modifications (Tables 1, 2,3). This system is based on the Institute for Clinical Systems Improvement, as adapted by the American Dietetic Association (Myers and others 2001) and was designed to rate the strength of scientific evidence. The authors modified the system to apply to the veterinary application in that randomised, controlled intervention trials in canine experimental animal models of osteoarthritis were rated as type I studies.

    The papers were also categorised on the basis of whether a placebo control or positive comparator control was used. For evaluation of functional efficacy in veterinary studies, an objective primary outcome measure, for example, force platform, was graded more highly than a semi-objective or subjective outcome measure, for example, owner- or veterinarian-completed questionnaire. This decision was based on the fact that, at present, there is lack of strong validation and standardisation of owner- and veterinarian-based questionnaires. For structure-modification, histology of articular cartilage was rated higher than a surrogate outcome measure, for example, imaging or biomarkers. The success of the investigated therapy was based on the identified primary outcome only.

    Because of a likely placebo effect in therapeutic studies of osteoarthritis, it was decided that only studies comparing a treatment to a placebo would be classified as high quality. The placebo effect may have an impact on the perceived results. The rationale for rating placebo-controlled studies higher is that the presence of a treatment, independent of the treatment type, may correlate with an improvement in perceived clinical signs. Improvements attributed to the treatment type, rather than the presence of a treatment, can only be fully ascertained when comparing a drug with a placebo control.

    The study aimed to calculate estimates for ‘number needed to treat’ (NNT) (Laupacis and others 1988) to evaluate efficacy, and estimates for ‘number needed to harm’ (NNH) for safety.


    The comprehensive literature search of five different online databases in summer 2008 resulted in 31,106 hits. Of these, over 1500 papers were reviewed further. Many of these were duplicates or not relevant to the study. Those found that met the inclusion criteria were added to a spreadsheet. This resulted in 15 papers that finally met all of the inclusion criteria for the use of NSAIDs in dogs.

    Of these 15 papers, five studies looked solely at the use of carprofen, one solely at each of firocoxib, meloxicam and licofelone, and the other seven studies were comparator studies using a combination of two or more treatments, including carprofen (five studies), firocoxib (two studies), meloxicam (two studies), etodolac (three studies), buffered aspirin (two studies) and deracoxib (one study). Eight were rated as type I studies and seven as type III. Fourteen studies addressed func-tional safety in clinical cases of canine osteoarthritis and, of these, eight investigated functional efficacy. One study investigated the potential structure-modifying effects of a NSAID in experimentally induced osteoarthritis in dogs. Of the 15 studies, six were classed as high quality and the other nine fell into the moderate quality bracket.

    In total, 10 studies involving carprofen were reviewed (Table 4), of which three were classified as high quality and the rest as moderate quality. Five were of design type I and five were of design type III. Nine of the papers reported on the safety of carprofen. The use of carprofen was supported by the highest number of studies.

    The use of meloxicam was considered in four studies (Table 4). Three of these studies were design type I and one was of design type III. Three studies were of high quality and one was of moderate quality.

    Three studies looked at the use of etodolac (Table 4), of which two were of design type I and one was of design type III; all were judged to be of moderate quality. Two studies addressed efficacy and safety in dogs with osteoarthritis and one addressed safety only in healthy dogs.

    Three studies looked at the use of firocoxib (Table 4). Two were of design type I and one was of design type III. Two were classed as moderate quality and one as high quality. The numbers of animals included in the trials were sufficient to extrapolate to the target population.

    One study was included that investigated the safety of deracoxib (Table 4). It was rated as design type III, was of moderate quality and included a small study population.

    One study was included that investigated the safety and efficacy of licofelone (Table 4). It was a type I study that included an objective outcome measure (peak vertical force) with a placebo control, and was rated as high quality. The sample population was small and insufficient to extrapolate to the target population.

    Evidence of progressive decreases in pain with long-term NSAID use

    Seven studies were identified that included data on short-term (less than 28 days) efficacy of NSAIDs compared with long-term treatment. Of these, six reported results of a better outcome at 28 days or more, compared with before 28 days. This difference was statistically significant in two studies.

    In one type I study, Hanson and others (2006) showed improvement in clinical signs (lameness at walk, lameness at trot) in dogs treated with either firocoxib or etodolac between days 14 and 29 of treatment. However, whether this improvement was significant was not stated.

    Following 10 days of firocoxib treatment, a type III study (Ryan and others 2006) reported that veterinarians rated 88.2 per cent of dogs' conditions as improved (mildly to greatly improved) and owners rated 87.4 per cent of dogs as improved. However, at day 40, veterinarians rated 92.8 per cent of dogs as improved, and owners rated 90.8 per cent of animals as improved. These differences between day 10 and 40 were not statistically significant.

    Carprofen was the sole treatment in a type III study of dogs with osteoarthritis evaluated at entry and at 5, 30, 60, 90 and 120 days. Investigators rated the severity of the osteoarthritic condition and the dogs' clinical signs on a visual analogue scale (VAS); improvement between 0 and 120 days was significant (P<0.05) (Autefage and Gossellin 2007). A small (n=6) grade III study (Lipscomb and others 2002) rated dogs treated with carprofen at 0, 7 and 28 days. VAS scores were significantly lower at 28 days compared with entry, but data at seven days were not. Carprofen was also the sole treatment in a type III study in which 750 dogs completed treatment (Mansa and others 2007). In this study, the number of dogs veterinarians considered were treated successfully (responded positively to treatment) increased from 493 (65.7 per cent) on day 14 to 551 (73.5 per cent) on day 84. In the same period, the number of dogs becoming sound increased from 128 to 194. Most notably, a type I study (Moreau and others 2003) with a placebo control group (day 30 only) used an objective outcome measure (peak vertical force) at 0, 30 and 60 days in the evaluation of carprofen and meloxicam. Compared with baseline, results indicated a significant improvement at days 30 and 60 for meloxicam, and at day 60 only for carprofen.

    Conversely, a type I study (Pollmeier and others 2006) comparing carprofen and firocoxib showed no difference in veterinary evaluations at 30 days compared with 14 days for either drug. After 30 days, 92.5 per cent of the firocoxib-treated dogs and 92.4 per cent of the carprofen-treated dogs had improved in terms of their overall discontinuous ordinal scale scores, and after 14 days 93.4 per cent of the firocoxib-treated dogs and 92.4 per cent of the carprofen-treated dogs had improved.

    It was not possible to identify dichotomous primary outcome variable data to calculate NNT estimates for long-term NSAID treatment. For the one short-term study (14 days) with suitable data identified (Vasseur and others 1995), NNT for carprofen compared with placebo was 5 (95 per cent confidence interval [CI] 2 to 7).

    Potential structure-modifying effects of long-term use of NSAIDs

    One study with a 28-day treatment period investigated the potential structure-modifying effects of carprofen on experimentally induced osteoarthritis of the stifle joint (Pelletier and others 2000). It was rated as type I, of moderate quality, and involved a small number of animals; therefore, it could not extrapolated to the target population. The study showed that carprofen reduced histologically graded cartilage lesions in a Pond-Nuki model (Pond and Nuki 1973) of osteoarthritis after 28 days of treatment compared with placebo.

    Safety of long-term NSAID use compared with short-term use

    In the papers identified, a total of 1589 dogs had been entered into studies. However, there was inconsistency in the manner in which adverse events were reported. In addition, the lack of placebo control in the majority of studies limited the ability to calculate NNH estimates. The only study (Raekallio and others 2006) with placebo control for treatment of 28 days or more was a small study that contained an experimental (adverse) event rate (EER) of 0.31 and a control event rate (CER) of 0.33, producing a NNH estimate of -39 (equivalent to a NNT of 39). The odds ratio was not significant, and confidence intervals (CIs) for NNH were therefore not calculated. This result was compared with a study of short-term treatment with carprofen (Vasseur and others 1995), which reported an EER of 0.17 and a CER of 0.12 with a NNH of 21 (95 per cent CI 19 to 23). Only one adverse event, which was probably linked to treatment and considered serious, was reported in all the papers reviewed. This was a labrador with an episode of toxic idiosyncratic hepatitis; the dog was treated and survived. Overall, studies reported EERs between 0 and 0.31 (mean 0.11), but there was not a significant correlation between study length and EER (rs=-0.109, P=0.793).


    Canine osteoarthritis is a common diagnosis in small animal veterinary practice and there is a broad range of licensed and candidate therapies available for use. The efficacy and safety of these treatments varies, as well as the way in which they are used, including the duration of treatment. It is imperative, therefore, that a comparison study is undertaken not only to compare the efficacy of the treatments, but the safety as well, thus allowing a fully informed decision to be made by veterinarians regarding the correct usage of each drug, or drug class. Previous systematic reviews have addressed efficacy of different agents for treatment of canine osteoarthritis (Aragon and others 2007, Sanderson and others 2009). However, in this systematic review, the aim was to collate efficacy and safety information on the long-term use of NSAIDs in dogs for treatment of osteoarthritis. It should be noted that this review was of published peer-reviewed literature in the English language only. As such, a small number of studies may have been missed. Furthermore, it is likely that the published literature contains a bias towards studies with positive or supportive data for the agents being studied. The commercial funding of studies is also another likely source of bias in the literature (Chard and others 2000).

    Only 15 papers were identified, detailing a total of 1575 canine patients completing a course of treatment of 28 days or more. For the purpose of this study, long-term treatment was defined as 28 days or more of continuous therapy. Clearly, in a chronic disease such as osteoarthritis, therapy for clinical cases may extend way beyond 28 days, and this is a limitation of the present study and of the current peer-reviewed literature. Nevertheless, in some countries, and for some practitioners, therapy beyond 14 days may be unusual and the authors felt it relevant to explore any differences in efficacy and safety between short-term (14 days or fewer) and longer-term therapy.

    Evaluation of efficacy in published trials was hampered by the low number of studies with placebo control groups. While scientifically, this is a disappointment, many published studies are performed under guidelines from the regulatory authorities in the USA or Europe and they may insist on certain study design types (for example, positive control comparator studies). The authors would argue that ‘randomised, double-blind, placebo-controlled with rescue analgesia’ study designs would be more robust and would be ethically sound. Such designs are often used in human studies of osteoarthritis (Clegg and others 2006) and are considered ethical. Obviously, the use of validated outcome measures is important in such studies.

    Nevertheless, of the seven studies reporting efficacy at sampling points before or after 28 days, six reported results in favour of long-term therapy for additional efficacy over short-term therapy and the remaining study indicated no difference. Taken together, there is moderate evidence for recommending long-term use of NSAIDs, due to the additional beneficial treatment effects, for dogs with chronic osteoarthritis. However, this issue deserves further exploration. There was no information in the literature on whether this additional benefit was the result of peripheral or central changes in pain processing. Additionally, it is presently not known whether the additional benefit was indeed related to a reduction in pain, or in fact a progressive functional improvement due to increased muscle strength and range of motion as a result of the greater mobility resulting from initial pain relief.

    The review of safety was more challenging. Although 14 of 15 papers included data on safety, the majority were of moderate to low quality, again because of the lack of CER data. Thus the authors were unable to calculate reliable estimates of NNH; much larger numbers are required to provide such estimates with an acceptable degree of confi-dence. However, EER data did not indicate an increase in this parameter with duration of treatment. In addition, of the 1589 dogs entered into these studies, only one serious adverse event was identified that probably related to NSAID treatment; this dog was treated and survived. Thus it seems that the incidence of serious adverse events with NSAID use of 28 days or more is low. The authors suggest that adverse events might have more relation to an individual animal's inherent response to NSAIDs. Such a view suggests that adherence to a protocol of only short-term therapy does not make sense in the face of continued chronic pain from the underlying condition, in particular when there appears to be increased efficacy with longer-term therapy.

    Only one study considered the potential structure-modifying potential of NSAIDs in dogs. Because there is no clinically applicable and validated outcome measure for structure modification in dogs, a study on experimental models of osteoarthritis in dogs were included. This study (Pelletier and others 2000) reported a positive effect of carprofen in reducing histologically graded cartilage lesions; however, as is inevitable and appropriate for such experimental studies, the number of dogs involved was very low. In addition, an experimental model can never exactly replicate the clinical condition and therefore the results of such studies must be treated with caution. However, such studies are likely to provide the best evidence available for such effects in dogs for the foreseeable future.

    In summary, while the current literature has shortcomings in the objective evaluation of long-term NSAID use and, notwithstanding the limitations of this systematic review discussed above, the current evidence suggests that there is a clinical benefit of longer-term NSAID use for dogs with chronic osteoarthritis and that this is associated with a low risk of serious adverse events.


    J. C. was in receipt of a summer studentship stipend funded by Pfizer Animal Health, Paris, France.


      View Abstract

      Request permissions

      If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.