Home > Pharmacological management of...

PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

National Collaborating Centre for Chronic Conditions (UK). Osteoarthritis: National Clinical Guideline for Care and Management in Adults. London: Royal College of Physicians (UK); 2008. (NICE Clinical Guidelines, No. 59.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

7Pharmacological management of osteoarthritis

7.1. Oral analgesics

7.1.1. Clinical introduction

Appropriate pharmacological analgesia forms one of the key platforms for treating osteoarthritis when non-pharmacological therapy on its own is insufficient. The use of such analgesia may be aimed at different aspects of the patient’s pain, including night pain or exercise-associated pain. Oral analgesics, especially paracetamol, have been used for many years, with increasing use of opioid analgesics in recent years, partly fuelled by fears over the safety of NSAIDs. The exact mechanism of action of paracetamol is unclear, although it may work in part by inhibiting prostaglandin synthesis; its action seems to work via the central nervous system rather than through peripheral effects. Opioid analgesics work by action on endogenous opioid receptors in the central nervous system.

There is still surprisingly little data on how patients use these therapies, which may influence their efficacy (for example, intermittent usage only at times of increased pain versus regular daily dosing). There are also many assumptions made on the effectiveness of these therapies in osteoarthritis, based on concepts such as ‘analgesic ladders’ which are not well supported in osteoarthritis cohorts.

It should be noted that this section includes the use of tricyclic agents as analgesics in osteoarthritis. This refers to the concept of low-dose usage of these agents, rather than antidepressant doses; it has been suggested that such low-dose usage may result in significant antinociceptive effects. However, it is important to note that depression may be associated with any chronic painful condition such as osteoarthritis and may require treatment in its own right. Readers should refer to the NICE depression guidelines (National Institute for Health and Clinical Excellence 2007)

7.1.2. Methodological introduction: paracetamol versus NSAIDs including COX-2 inhibitors

We looked at studies on the efficacy and safety of paracetamol compared with oral NSAIDs or selective COX-2 inhibitors for symptomatic relief from pain in adults with osteoarthritis. We found one Cochrane meta-analysis (Towheed et al. 2006) of randomised controlled trials that addressed the topic. In addition, one RCT (Temple et al. 2006), four relevant N-of-1 trials (March et al. 1994; Nikles et al. 2005; Wegman et al. 2003; Yelland et al. 2007) and one cohort study (Fries and Bruce 2003) were identified. All studies were found to be methodologically sound and were included as evidence.

The meta-analysis included ten RCTs with comparisons between paracetamol and NSAIDs (celecoxib, diclofenac, ibuprofen, naproxen and rofecoxib). The analysis did not provide separate results for non-selective and COX-2 selective NSAIDs on pain outcomes, but did for gastrointestinal adverse events. Studies included in the analysis differed with respect to:

  • paracetamol dosage
  • site of disease
  • osteoarthritis diagnosis
  • trial design
  • funding sources
  • study site location.

To avoid double counting of participants receiving paracetamol, the analysis was stratified into three comparator groups involving paracetamol and:

  • ibuprofen 2400 mg, diclofenac, Arthrotec (diclofenac with misoprostol), celecoxib, naproxen (comparator 1)
  • ibuprofen 1200 mg, Arthrotec, rofecoxib 25 mg, naproxen (comparator 2)
  • ibuprofen 1200 mg, Arthrotec, rofecoxib 12.5 mg, naproxen (comparator 3).

The four N-of-1 trials reported on courses of paracetamol and NSAIDs given in random order to blinded participants acting as their own controls. There were high numbers of non-completers across all studies. One cohort study retrospectively examined the prevalence of serious gastrointestinal adverse events in participants taking paracetamol or ibuprofen.

The RCT (Temple et al. 2006) looked at paracetamol (4 g/day) versus naproxen (750 mg/day) in N=581 patients with knee or hip osteoarthritis in a 12-month or 6-month treatment phase.

7.1.3. Methodological introduction: paracetamol versus opioids, and paracetamol-opioid combinations

We looked at studies that investigated the efficacy and safety of i) paracetamol compared with opioids or opioid-paracetamol compounds, and ii) NSAIDs compared with opioid-paracetamol compounds to relieve pain in adult patients with osteoarthritis. One Cochrane systematic review and meta-analysis (Cepeda et al. 2006), six RCTs (Bianchi et al. 2003; Boureau et al. 1990; Irani 1980; Kjaersgaard et al. 1990; McIntyre et al. 1981; Parr et al. 1989) and one prospective cohort study (Mitchell 1984) were found on paracetamol versus opioids, paracetamol versus paracetamol-opioids, NSAIDs versus paracetamol-opioids and opioids versus NSAIDs. The cohort study had a mixed arthritis population, did not stratify the study findings in terms of diagnostic category, and had multiple methodological limitations: it was excluded.

The Cochrane meta-analysis only included one RCT comparing the opioid tramadol (up to 300 mg/day) with the NSAID diclofenac (up to 150 mg/day) for 28 days of treatment in N=108 patients with hip or knee osteoarthritis. The RCT was assessed for quality and found to be methodologically sound.

The included RCTs addressing individual questions were as follows:

Studies differed with respect to the anatomical site of osteoarthritis, and treatment regimens (doses and treatment length). All studies included as evidence had methodological issues, including:

  • small sample sizes
  • inadequate blinding
  • no washout period for previous analgesic medication
  • ITT analysis was rarely performed.

7.1.4. Methodological introduction: opioids

We looked at studies that investigated the efficacy and safety of low-dose opioids with or without paracetamol compared with higher-strength opioids with respect to symptoms, function and quality of life in adults with osteoarthritis. Two systematic reviews and meta-analyses (Bjordal et al. 2007; Cepeda et al. 2006) and four RCTs (Andrews et al. 1976; Bird et al. 1995; Gana 2006; Jensen and Ginsberg 1994) were found that addressed the question. One RCT (Andrews 1976) was excluded due to methodological limitations.

The Cochrane systematic review (Cepeda et al. 2006) included three RCTs (N=467 patients) comparing tramadol (opioid) with placebo and two RCTs (N=615 patients) comparing tramadol vs paracetamol vs placebo and tramadol vs diclofenac.

Opioid versus placebo

The three RCTs included in the MA were similar in terms of trial design (parallel-group studies), blinding (double blind) and study quality. However, trials varied in terms of:

  • osteoarthritis site (two RCTs knee, one RCT hip or knee)
  • treatment regimen – dose of tramadol one RCT 200 mg/day, two RCTs up to 400 mg/day)
  • trial size and length.

Opioid–paracetamol combinations versus placebo

The two RCTs included in the MA were similar in terms of trial design (parallel-group studies), blinding (double blind) and study quality. However, trials varied in terms of:

  • trial size and length
  • dose of tramadol 37.5 mg/day, paracetamol 325 mg/day (increased to 4 or 8 tablets/day further into the trial).

The second systematic review (Bjordal et al. 2007) included 63 RCTs (of which N=6 RCTs compared opioids with placebo, N=1057 patients) and assessed the outcome of pain. Trials were similar in terms of osteoarthritis site (knee osteoarthritis) and study quality. However, trials varied in terms of:

  • trial size and length
  • treatment – type of opioid used (N=2 RCTs tramadol, N=2 RCTs oxymorphone, N=1 RCT oxycodone, N=1 RCT codeine, N=1 RCT morphine sulphate).

Note: the Bjordal MA (Bjordal et al. 2007) includes two RCTs that were also included in the Cepeda MA (Cepeda et al. 2006). However, both MAs included a number of different additional studies and thus both MAs were included as evidence.

The three included RCTs were methodologically sound and assessed patients with knee and/or hip osteoarthritis. The first RCT (Bird et al. 1995) was a cross-over study and compared low dose tramadol with pentazocine in N=40 patients for a 2-week treatment period. The second RCT (Jensen and Ginsberg 1994) used parallel group design and compared dextropropoxyphene with high dose tramadol in N=264 patients for a 2-week treatment period. The third RCT (Gana et al. 2006) compared tramadol (at increasing doses 100, 200, 300 and 400 mg/day) with placebo for a 12-week treatment period.

The cross-over study (Bird et al. 1995) did not include a wash-out period between treatment periods. However, in an attempt to reduce the influence of any carry-over effects, the final 7 days of each treatment period were used to compare the treatments. This study also had a high withdrawal rate (48%), but was otherwise fairly well conducted. The parallel group study (Jensen and Ginsberg 1994) was methodologically sound.

7.1.5. Methodological introduction: paracetamol versus placebo

We looked at studies that investigated the efficacy and safety of paracetamol compared with placebo with respect to symptoms, function, and quality of life in adults with osteoarthritis. We found one Cochrane systematic review and meta-analysis (Towheed et al. 2006) and two RCTs (Altman et al. 2007; Herrero-Beaumont et al. 2007) on paracetamol versus placebo.

The Cochrane meta-analysis assessed the RCTs for quality and pooled together all data for the outcomes of symptoms, function and AEs. However, the outcomes of quality of life and GI AEs were not reported. The results for these outcomes have been taken from the individual RCTs included in the systematic review. No relevant RCTs, cohort or case-control studies were found.

Outcomes in the RCTs of the MA were analysed by a number of different assessment tools, using either categorical or quantitative data. For continuous outcome data, the MA has used SMD (standardised mean difference) to pool across RCTs. For dichotomous outcome data, the MA has calculated RR.

The meta-analysis included seven RCTs (with N=2491 participants) that focused on comparisons between paracetamol and placebo. Studies included in the analysis differed with respect to:

  • paracetamol dosage (five RCTs 1000 mg daily, two RCTs 4000 mg daily)
  • site of disease (five RCTs knee two RCTs knee or hip)
  • osteoarthritis diagnosis (five RCTs radiological one RCT clinical and radiological one RCT Lequesne criteria)
  • trial length and design (four RCTs were parallel group design, three RCTs cross-over design)
  • funding sources (three RCTs had involvement of a pharmaceutical company).

The two RCTs (Altman et al. 2007; Herrero-Beaumont et al. 2007) not included in the systematic review were parallel studies that focused on the outcomes of symptoms, function and AEs. The first RCT (Altman et al. 2007) was methodologically sound (randomised and double-blind) and compared paracetamol extended release (ER) (3900 mg/day) versus paracetamol ER (1950 mg/day) versus placebo in N=483 patients with knee or hip osteoarthritis in a 12-week treatment phase. The second RCT (Herrero-Beaumont et al. 2007) was methodologically sound (randomised and double-blind) and compared paracetamol ER (3000 mg/day) versus placebo or glucosamine sulfate in N=325 patients with knee osteoarthritis in a 6-months treatment phase. The results for the glucosamine arm are not presented here.

7.1.6. Methodological introduction: tricyclics, SSRIs and SNRIs

We looked for studies that investigated the efficacy and safety of tricyclics/SSRI/SNRI drugs compared with placebo with respect to symptoms, function, and quality of life in adults with osteoarthritis. One RCT (Scott 1969) was found that on the outcomes of symptoms and function. No relevant cohort or case-control studies were found.

The RCT (Scott 1969) (N=24) used a cross-over design and involved a mixed population of osteoarthritis (N=7), RA (N=14) or ankylosing spondylitis (N=1) patients who were randomised to treatment with the tricyclic antidepressant imipramine or placebo. Results for osteoarthritis patients only are reported here. The study length was 6 weeks (3 weeks for each treatment). The results for each patient were reported separately and therefore the osteoarthritis data have been extracted. The anatomical site of osteoarthritis was not mentioned and AEs were not reported for the separate osteoarthritis subgroup. Overall, the study was fairly well conducted (although it did not include a wash-out period between treatments) and is therefore included as evidence.

7.1.11. Evidence statements: tricyclics, SSRIs and SNRIs

Symptoms: pain

1 RCT (Scott 1969) (N=7) found that when imipramine was given as the first treatment, the pain severity score (measured change from baseline) improved when measured after imipramine treatment (−0.8) but stayed the same when measured after placebo. (1+)

The same RCT (Scott 1969) (N=7) found that when placebo was given as the first treatment, the pain score stayed the same when measured after imipramine treatment and after placebo. (1+)

Symptoms: function

One RCT (Scott 1969) (N=7) found that when imipramine was given as the first treatment, function score and grip strength (measured change from baseline) improved when measured after imipramine treatment (−0.4 and +19 mmHg respectively) but stayed the same when measured after placebo. (1+)

The same RCT (Scott 1969) (N=7) found that when placebo was given as the first treatment, function score stayed the same when measured after imipramine treatment and after placebo. However, grip strength increased after treatment with imipramine and after placebo, the increase being greater after imipramine (+42.5 and +12.5 mmHg respectively). (1+)

Global assessment

One RCT (Scott 1969) (N=7) found that when imipramine was given as the first treatment, most of the patients and physicians preferred imipramine to placebo (three out of four patients for both). (1+)

The same RCT (Scott 1969) (N=7) found that when placebo was given as the first treatment, no patients preferred imipramine to placebo. (1+)

7.1.12. From evidence to recommendations

There is a good amount of evidence from RCTs on the efficacy of paracetamol in knee osteoarthritis, with less evidence supporting its use in osteoarthritis of other sites. The long-term safety data on paracetamol from observational studies are reassuring. The GDG noted that patients commonly use infrequent dosing of paracetamol which may lead to reduced efficacy. There are limited data on the efficacy of paracetamol used in combination with other pharmacological therapies, and most such data are drawn from studies where paracetamol is used as ‘escape’ analgesia.

The evidence supporting the use of opioid analgesia in osteoarthritis is poor, and it must be noted there are virtually no good studies using these agents in peripheral joint osteoarthritis patients. There is little evidence to suggest that dose escalation of these agents is effective. There are also few data comparing different opioid formulations or routes of administration. Toxicity remains a concern with opioid use, especially in the elderly. Constipation, nausea, itchiness, drowsiness and confusion remain important side effects to be considered.

There is no good evidence to support the use of low dose tricyclic agents for osteoarthritis pain. However, consideration of sleep and mood disturbance is part of the assessment of the osteoarthritis patient and appropriate pharmacological therapy may be warranted. The reader is also referred to the NICE depression guideline (National Institute for Health and Clinical Excellence 2007).

RECOMMENDATIONS

R21.

Healthcare professionals should consider offering paracetamol for pain relief in addition to core treatment (see Fig 3.2); regular dosing may be required. Paracetamol and/or topical NSAIDs should be considered ahead of oral NSAIDs, COX-2 inhibitors or opioids.

R22.

If paracetamol or topical NSAIDs are insufficient for pain relief for people with osteoarthritis, then the addition of opioid analgesics should be considered. Risks and benefits should be considered, particularly in elderly people.

7.2. Topical treatments

7.2.1. Clinical introduction

Topical NSAIDs, capsaicin and rubefacients and are widely used to treat osteoarthritis.

After topical application, therapeutic levels of NSAIDs can be demonstrated in synovial fluid, muscles and fasciae. They may have their pharmacological effects on both intra-and extra-articular structures (Dominkus et al. 1996; Lin et al. 2004; Rolf et al. 1999). It is assumed that their mechanism of action is similar to that of oral NSAIDs. Topical NSAIDs produce a maximal plasma NSAID concentration of only 15% that achieved following oral administration of a similar dose (Dominkus et al. 1996; Heyneman et al. 2000). Thus, it would be expected that topical NSAIDs would have far fewer systemic side effects than oral NSAIDs. Even if their pain relieving effect is less than that of oral NSAIDs, they may be an attractive option for the treatment of osteoarthritis because they will produce fewer NSAID-related adverse effects.

It is possible that the act of rubbing and expectation of benefit may also contribute to any therapeutic effect from topical preparations (Arcury et al. 1999; Vaile and Davis 1998). This may partially account for the continued popularity of rubefacients. Rubefacients produce counterirritation of the skin that may have some pain relieving effect in musculoskeletal disorders.

Capsaicin is derived from chilli peppers. As well as a counter-irritant effect it depletes neuro-transmitters in sensory terminals reducing the transmission of painful stimuli. There may be a delay of some days for the effects of topical capsaicin to be evident, perhaps due to this progressive neurotransmitter depletion.

7.2.2. Methodological introduction

We looked for studies that investigated the efficacy and safety of topical agents (NSAIDs/capsaicin/rubefacients) compared with oral NSAIDs or placebo with respect to symptoms, function and quality of life in adults with osteoarthritis. Two systematic reviews and meta-analyses (Lin et al. 2004; Towheed 2006) were found on topical NSAIDs and 10 additional RCTs (Algozzine et al. 1982; Altman et al. 1994; Deal et al. 1991; McCleane 2000; Niethard et al. 2005; Rothacker et al. 1994, 1998; Schnitzer et al. 1994; Shackel et al. 1997; Trnavsky et al. 2004) on topical NSAIDs, capsaicin and rubefacients.

Both of the meta-analyses assessed the RCTs for quality and pooled together all data for the outcomes of symptoms, function and AEs. However, the outcome of quality of life was not reported. No QoL data were reported by the individual trials in the Towheed MA (Towheed 2006). However, QoL was reported in the individual RCTs included in the Lin MA (Lin et al. 2004). Results for quality of life have therefore been taken from the individual RCTs included in this systematic review.

Topical NSAIDs

Two SRs/MAs (Towheed 2006b; Lin et al. 2004) and two RCTs (Niethard et al. 2005, Trnavsky et al. 2004) were found on topical NSAIDs.

The first MA (Lin et al. 2004) included 13 RCTs (with N=1983 participants) that focused on comparisons between topical NSAIDs versus placebo or oral NSAIDs in patients with osteoarthritis. All RCTs were randomised and double-blind. Studies included in the analysis differed with respect to:

  • osteoarthritis site (eight RCTs knee osteoarthritis; three RCTs hand osteoarthritis; one RCT hip, knee and hand osteoarthritis; one RCT hip and knee osteoarthritis)
  • type of topical NSAID used
  • type of oral NSAID used
  • treatment regimen
  • trial design (two RCTs cross-over; 11 RCTs parallel group studies), size and length.

The second MA (Towheed 2006b) included four RCTs (with N=1412 participants) that focused on comparisons between topical diclofenac in DMSO carrier versus placebo or oral diclofenac in patients with knee osteoarthritis. All RCTs were randomised, double-blind parallel group studies. Studies included in the analysis differed with respect to:

  • treatment regimen (three RCTs vs placebo, 50 drops 4 times daily; one RCT vs oral diclofenac, 50 drops 3 times daily)
  • trial size and length.

The two RCTs not included in the systematic review focused on the outcomes of symptoms, function and quality of life in patients with knee osteoarthritis. They were both parallel group studies and were methodologically sound (randomised, double-blind, ITT analysis). However, they differed in terms of study intervention, sample size and study duration.

Topical capsaicin

Four RCTs were found on topical capsaicin versus placebo (given 4 times daily) and focused on the outcomes of symptoms, function and quality of life in patients with osteoarthritis. All trials were parallel group studies and were methodologically sound.

However, they differed in terms of osteoarthritis site, sample size and study duration. One RCT (Altman et al. 1994) looked at 113 patients with knee, ankle, elbow, wrist and shoulder osteoarthritis and treatment lasted for 12 weeks. The second RCT (Deal et al. 1991) looked at 70 patients with knee osteoarthritis and treatment lasted for 4 weeks. The third RCT (McCleane 2000) looked at 200 patients with knee, hip, shoulder and hand osteoarthritis and treatment lasted for 6 weeks. The fourth RCT (Schnitzer et al. 1994) looked at 59 patients with hand osteoarthritis and treatment lasted for 9 weeks.

Topical rubefacients

Four RCTs were found that focused on topical rubefacients versus placebo and focused on the outcomes of symptoms, function and quality of life in patients with osteoarthritis. All trials were methodologically sound (randomised and double-blind; two RCTs also included ITT analysis) (Rothacker et al. 1994; Shackel et al. 1997).

However, they differed in terms of: osteoarthritis site, trial design, sample size, study duration and study intervention. One RCT (Algozzine et al. 1982) compared trolamine salicylate to placebo in 26 patients with knee osteoarthritis and treatment lasted for 7 days. The second RCT (Rothacker et al. 1994) compared trolamine salicylate to placebo in 50 patients with hand osteoarthritis and where treatment was a single application. The third RCT (Rothacker et al. 1998) compared trolamine salicylate to placebo in 86 patients with hand osteoarthritis and where treatment was a single application. The fourth RCT (Shackel et al. 1997) compared copper salicylate to placebo in 116 patients with knee and/or hip osteoarthritis and treatment lasted for 4 weeks. Two of the RCTs were parallel group studies (Rothacker et al. 1998, Shackel et al. 1997) and the other two RCTs (Algozzine et al. 1982; Rothacker et al. 1994) used a cross-over design, both of which included a wash-out period between cross-over treatments.

7.2.6. Health economic evidence

We looked at studies that conducted economic evaluations involving topical NSAIDs, capsaicin or rubefacients. Three papers, two from the UK and one from Australia, relevant to this question were found and included as evidence. After the re-run search one further study was included.

Two UK papers from the early 1990s conducted cost-minimisation analyses rather than full cost effectiveness or cost–utility analysis.

One UK paper compares oral ibuprofen (1200 mg/day) to topical Traxam and oral Arthrotec (diclofenac 50 mg/misoprostol 200 mg one tablet twice daily) (Peacock and Rapier 1993). The study considers the drug cost of each treatment as well as the cost of ulcers caused by the treatment using a simple economic model. It does not include other GI adverse events or CV adverse events. Including these would make the oral NSAID appear more expensive. Ulcer incidence rates are estimated based on findings in the literature, and some simple sensitivity analysis is undertaken around this. In conducting a cost-minimisation analysis the authors have implicitly assumed equal efficacy of the treatments, which may not be appropriate. The duration considered in the study is one month.

Another UK study considers oral ibuprofen (1200 mg/day) and topical piroxicam gel (1g three times daily) (McKell and Stewart 1994). The cost per patient of each treatment is calculated using a decision tree which includes ulcers and dyspepsia as adverse events. CV adverse events are not included. Adverse event rates are estimated using data in the published literature. Importantly, the efficacy of the treatments is assumed to be equal and hence only costs are considered. The duration of the study is three months.

The Australian study considers a number of different treatments for osteoarthritis, one of which is topical capsaicin compared with placebo (Segal et al. 2004). The paper is generally well conducted. Data regarding the effectiveness of capsaicin are taken from the literature (Altman et al. 1994; Deal et al. 1991). The transfer to utility (TTU) technique was used to transform the pain improvement data available in trials into a quality adjusted life year (QALY) gain. The paper assumes that capsaicin does not increase the risk of adverse events over the levels experienced by the general population, and so the only costs included in the study are the specific drugs cost. The study takes a 1-year time period and calculates the incremental cost–effectiveness ratio (ICER) of topical capsaicin compared with placebo.

It is of note that a study protocol for a trial assessing the costs and benefits associated with treating patients with chronic knee pain with topical or ibuprofen was published in November 2005.

One UK study which is yet to be published investigates oral ibuprofen compared with topical ibuprofen in 585 patients with knee pain. The study had an RCT arm and a patient preference arm, and includes 12-month and 24-month data.

7.2.7. Health economic evidence statements

Oral ibuprofen vs topical Traxam or topical piroxicam and Arthrotec

Table 7.53Cost (1993 £) of treating 1000 patients for 1 month

Ibuprofen (1200 mg/day)TraxamArthrotec
41,4087,31917,924

Table 7.54Cost (1991–1992 £) per patient for 3 months

Ibuprofen (1200 mg/day)Piroxicam (1g tid)
89.1254.57

The tables above show the results of the two studies from the UK (McKell and Stewart 1994; Peacock and Rapier 1993). They offer evidence that treatment with topical NSAIDs is likely to be cheaper than treatment with oral NSAIDs. However, it must be noted that the studies are incomplete with regards to the adverse events included (neither include CV adverse events, and not all GI adverse events are included). Including these adverse events would result in topical NSAIDs leading to a higher cost saving compared with oral NSAIDs, providing topical NSAIDs result in fewer of these events than oral NSAIDs. Also the results of the studies are of limited use with regards to cost effectiveness since a health outcome is not included. Equal efficacy is assumed, but if oral NSAIDs are in actuality more effective, then there remains a possibility that they could be considered cost effective despite being more expensive.

In summary, evidence suggests that treatment with topical NSAIDs will result in lower costs than treatment with oral NSAIDs due to the higher prevalence of adverse events with oral NSAIDs. The cost effectiveness of oral NSAIDs depends on their clinical efficacy compared with topical NSAIDs. If oral NSAIDs are of equal efficacy compared with topical NSAIDs it is likely that topical NSAIDs would be cost effective.

Topical capsaicin vs placebo

The table above shows the cost effectiveness of a number of drugs as calculated by the Australian study (Segal et al. 2004). NSAIDs, COX-2 inhibitors, and glucosamine sulfate are included to allow some comparison of cost effectiveness between the drugs, although each is only compared with placebo in the analysis, rather than to each other. Where the cost–effectiveness ratio is said to range ‘to infinity’ this is because the benefits of the drug are not assured.

These results suggest that topical capsaicin brings more QALY gain than NSAIDs or COX-2 inhibitors compared with placebo, while resulting in lower total costs than COX-2 inhibitors (although the total costs are higher than for NSAIDs). Therefore capsaicin appears dominant compared with COX-2 inhibitors. The incremental cost-effectiveness ratio between NSAIDs and topical capsaicin [(236–174)/(0.053–0.043) = $6,200] suggests that topical capsaicin is likely to be cost effective compared with NSAIDs. However, the incremental cost-effectiveness ratio between topical capsaicin and glucosamine sulfate only shows borderline cost effectiveness (236–180)/(0.053–0.052) = $56,000 per QALY. Because the cost of topical capsaicin is relatively low and QALY gains are accrued, the incremental cost-effectiveness ratio of $4,453 stated in Table 7.55 suggests the treatment is cost effective compared with placebo.

Table 7.55. Segal’s estimates of cost effectiveness.

Table 7.55

Segal’s estimates of cost effectiveness.

Some care has to be taken with these results because of the relative lack of studies which show the benefits of capsaicin and glucosamine sulfate. The transfer to utility approach for calculating QALY gains has also been questioned in the literature. The study is also from an Australian perspective which may not be transferable to a UK setting.

It is of interest that in the UK 45 g of topical capsaicin costs £15.04. If this size tube was sufficient for one month of treatment the UK yearly cost of treatment with topical capsaicin would be £180.48 (taking into account only drug costs). Some sources suggest this size tube would in fact not be sufficient for one month of treatment (www.pharmac.govt.nz/pdf/0804.pdf). This is significantly more expensive than the $236 cost stated by the Australian study, which equates to £95.57, but does assume that the patient uses the treatment continuously for one year. Using this cost, the incremental cost-effectiveness ratio of topical capsaicin compared with placebo would be (180.48/0.053) £3405 per QALY which is still relatively low.

However, in comparison to other drugs, topical capsaicin appears likely to be closer to the cost of COX-2 inhibitors, and significantly more expensive than some NSAIDs in a UK setting. In the UK celecoxib costs £21.55 per 60-cap 100 mg pack, suggesting a yearly drug cost of approximately £21.55 × 12 = £259 (BNF 51). Diclofenac sodium costs £1.52 for an 84-tab pack of 25 mg, suggesting a yearly drug cost of approximately £1.52 × 12 = £18.24, (BNF 51) although these estimates do not include the adverse event costs of these drugs.

Given this, it is difficult to make reliable recommendations regarding topical capsaicin compared with COX-2 inhibitors and NSAIDs based on this Australian data.

In summary, evidence from an Australian study suggests that topical capsaicin is cost effective compared with placebo, since it brings QALY gains at relatively low cost.

Topical ibuprofen vs oral ibuprofen

The study finds that the effectiveness of the two treatments is not statistically significantly different, but that oral ibuprofen appears slightly better. Oral ibuprofen is generally a more expensive treatment option, due to more gastroprotective drugs and cardiovascular drugs being prescribed alongside it. Overall, oral ibuprofen is generally found to be cost effective compared with topical ibuprofen. However, the authors note that the study considered a population at low risk of adverse events and the prevalence of adverse events in the study was lower than expected. Given the risks known to be associated with taking oral NSAIDs, it may be that in a higher risk population oral NSAIDs would not be cost effective.

In summary:

  • in a population at low risk of adverse events, oral ibuprofen is likely to be a cost-effective treatment compared with topical ibuprofen
  • treatment with topical ibuprofen is likely to be cheaper than treatment with oral ibuprofen.

7.2.8. From evidence to recommendations

A number of studies, mainly of knee osteoarthritis, have shown short-term (less than four weeks) benefits from topical NSAID gels, creams and ointments when compared with placebo. There are no data on their long-term effectiveness when compared with placebo. There are limited studies comparing other topical gels, creams and ointments with oral NSAIDs. One study with three month follow-up found topical diclofenac in dimethyl sulfoxide to be equivalent to oral diclofenac for knee osteoarthritis over three months.

The data from RCTs have demonstrated a reduction in non-serious adverse effects when compared with oral NSAIDs, although topical preparations may produce local skin irritation. The RCT data do not allow a conclusive judgement on whether using topical NSAIDs reduces the incidence of serious NSAID-related adverse effects. However, it seems logical that there would be a reduced risk given the total dose of NSAIDs from topical application to one joint area is much less than when used orally. Thus, since there are some data supporting the effectiveness of topical NSAIDs, they are likely to be preferred to using oral NSAIDs as early treatment for osteoarthritis, particularly for patients who do not have widespread painful osteoarthritis. However, there are no data comparing topical NSAIDs to paracetamol or on the comparative risk and benefits from the long-term use of oral or topical NSAIDs.

Topical NSAIDs are relatively costly but are cost effective given that they prevent or delay use of oral NSAIDs with their associated serious adverse events. Most of the clinial evidence is for the preparation of diclofenac in DMSO (Pennsaid), but overall there is little evidence and the group did not find sufficient justification to single out this brand in the recommendations. At the time of writing, Pennsaid was not the cheapest alternative in this class.

There are limited data showing some positive effects from topical capsaicin, with short-term follow-up. Although the evidence is limited to knee osteoarthritis, the GDG were aware of widespread use in hand osteoarthritis as part of self-management and felt that the data on efficacy at the knee could reasonably be extrapolated to the hand. No serious toxicity associated with capsaicin use has been reported in the peer-reviewed literature. The evidence base, however, does not support the use of rubefacients.

Topical treatments are used in self-management, which helps change health behaviour positively. Often, people with osteoarthritis will use the topical treatment on top of daily paracetamol and exercise to cope with flare-ups. This is in line with the evidence, which shows short-term benefit. As a safe pharmaceutical option, topical NSAIDs were regarded by the GDG as one of the second-line options for symptom relief after the core treatments. They have therefore been placed on an equal footing with paracetamol.

RECOMMENDATIONS

R23.

Healthcare professionals should consider offering topical NSAIDs for pain relief in addition to core treatment (see Fig 3.2) for people with knee or hand osteoarthritis. Topical NSAIDs and/or paracetamol should be considered ahead of oral NSAIDs, COX-2 inhibitors or opioids.

R24.

Topical capsaicin should be considered as an adjunct to core treatment for knee or hand osteoarthritis; rubefacients are not recommended for the treatment of osteoarthritis.

7.3. NSAIDs and highly selective COX-2 inhibitors

7.3.1. Clinical introduction

Non-steroidal anti-inflammatory drugs (NSAIDs) have been available for many years and are thought to work by reducing the production of pro-inflammatory and pain-related prostaglandins. The discovery of different cyclooxygenase (COX) enzymes with different physiological actions brought with it the concept that differential blockade of COX-1 (important in normal regulation of the gastro-intestinal (GI) mucosa) and COX-2 (up-regulated at sites of inflammation among other functions and thought responsible for pro-inflammatory mediator production) may provide effective analgesic/anti-inflammatory actions without the common GI complications of traditional NSAIDs. These GI complications are well known to clinicians and include a spectrum of problems from dyspepsia and ulcers to life-threatening perforations, ulcers and bleeds. However the blocking of COX-2 always carried the potential for a pro-thrombotic effect, by changing the balance of pro- and anti-thrombotic mediators.

The first novel agents to be classed COX-2 selective were celecoxib and rofecoxib, although existing agents were also recognised for their high COX-2/COX-1 inhibitory ratios (etodolac, meloxicam). Of these agents, rofecoxib in particular demonstrated the expected outcomes, in that initial studies demonstrated reduced serious GI problems compared with traditional NSAIDs. Importantly, there was no evidence to suggest that any of these agents would differ from traditional NSAIDs with respect to efficacy. However, the initial, pivotal study also demonstrated increased pro-thrombotic cardiovascular problems (an increase in myocardial infarctions). This brought a spotlight to bear on the cardiovascular safety of all such agents, but also on traditional NSAIDs which had varying degrees of COX-2 selectivity. This remains a complex field because of issues including that:

  • long-term toxicity must be assessed from longitudinal, observational databases with their inherent problems, including lack of thorough assessment of an individual’s cardiovascular risk factors
  • more detailed trial data are only available on newer agents
  • drug dose in studies do not reflect usual prescribed doses or patient use.

As a result of further scrutiny, there seems less reason to use the terms ‘traditional NSAIDs’ and ‘COX-2’ selective agents. It would appear that it may be more useful to return to the generic term NSAIDs with a concomitant awareness of the differing degrees of COX-2 selectivity and different (though not always consequent) side-effect profiles.

7.3.2. Methodological introduction

Three questions were posed in the literature searches for this section of the guideline.

  • In adults with osteoarthritis, what are the benefits and harms of COX-2 inhibitors compared with i) non-selective NSAIDs or ii) placebo in respect to symptoms, function and quality of life?
  • In adults with osteoarthritis, what are the relative benefits and harms of i) selective COX-2 inhibitors versus nonselective NSAIDs plus GI protective agents and ii) selective COX-2 inhibitors plus GI protective agents versus nonselective NSAIDs plus GI protective agents?
  • In adults with osteoarthritis taking aspirin, what are the relative benefits and harms of selective COX-2 inhibitors versus nonselective NSAIDs versus each of these combined with GI protective agents?

We looked firstly at studies that focused on investigating the effects of COX-2 inhibitors compared with non-selective NSAIDs or placebo for the outcomes of symptoms, function, quality of life, and adverse events (AEs) where the latter where reported. Due to the high number of studies in this area only randomised double-blinded controlled trials were considered for inclusion as evidence for all osteoarthritis sites. However, for knee osteoarthritis studies, only double-blinded RCTs with N400 participants and with a duration of longer than 4 weeks were considered for inclusion.

For the second question, we found two studies (Chan et al. 2007; Scheiman et al. 2006) that investigated the effects of esomeprazole versus placebo in adults with osteoarthritis or RA receiving concomitant COX-2 inhibitors or non-selective NSAIDs. Although these studies included a mixed osteoarthritis/RA population, it was decided to include them, since they were the only studies reporting on the results of well-designed RCTs on this topic. One other RCT (Lai 2005) was found but excluded from the evidence since it was an open-label study and thus did not fulfil the inclusion criteria.

Finally, two studies (Schnitzer et al. 2004; Singh et al. 2006) selected for the first question also included data on adverse gastro-intestinal events in adults with osteoarthritis taking low-dose aspirin. They were therefore relevant to the third question, which focuses on the relative benefits and harms of COX-2 inhibitors and non-selective NSAIDs in adults with osteoarthritis receiving concomitant low-dose aspirin.

The relevant data are reported under the adverse events section of the evidence statements. No other studies were identified that addressed the third question.

7.3.3. Evidence statements: COX-2 inhibitors vs placebo and NSAIDs

Summary

Symptoms: pain

Overall, the studies found that both COX-2 inhibitors were superior to placebo in terms of reducing pain over treatment periods ranging from six weeks to six months. The majority of the data reported here are for outcomes on the VAS and the WOMAC. The limited data on direct comparisons of COX-2 inhibitors and non-selective NSAIDs for this outcome suggested these two drug classes were equivalent. Only a small number of studies reported significant differences when comparing COX-2 inhibitors with NSAIDs:

  • knee: two studies reported in favour of celecoxib compared with naproxen (N=1061) (Kivitz 2001); (N=1608) (Zhao et al. 1999)
  • knee and hip: one study reported in favour of naproxen compared with etodolac (N=76) (Chikanza and Clarke 1994)
  • mixed sites: one study reported in favour of diclofenac compared with meloxicam (N=10,051) (Hawkey 1998).
Knee osteoarthritis

Fifteen RCTs (Bensen et al. 1999; Fleischmann et al. 2006; Gottesdiener et al. 2002; Kivitz et al. 2001; Lehmann 2005; Lund 1998; McKenna 2001; Sheldon 2005; Smugar et al. 2006; Tannenbaum et al. 2004; Williams et al. 2000, 2001; Zhao et al. 1999) focusing on knee osteoarthritis were identified. Two studies (Suarez-Otero 2002; Williams and Osie 1989) were excluded due to multiple methodological limitations, including absence of reported washout period prior to baseline assessment. All other studies were included as evidence.

The studies below reported significant reductions in pain for the following COX-2 inhibitors compared with placebo for treatment periods ranging from 3 to 13 weeks:

The studies below reported on outcomes for the following drug interventions for treatment periods ranging from 12 to 14 weeks:

  • celecoxib 100 mg resulted in significant reductions in pain compared with naproxen 2000 mg in WOMAC pain (p<0.001). Celecoxib 200 and 400 mg and naproxen 1000 mg (NS) (N=1061) (Kivitz et al. 2001)
  • celecoxib 100 mg and 200 mg had significant reductions in pain scores (WOMAC) compared with naproxen 1000 mg (% change from baseline celecoxib 100 mg −29.5, celecoxib 200 mg −25.2 versus naproxen −21.8) (N=1003) (Zhao et al. 1999)
  • celecoxib 100 mg and diclofenac 50 mg (NS) (N=600) (McKenna 2001)
  • etoricoxib 20 to 360 mg and diclofenac 150 mg (NS) (N=617) (Gottesdiener et al. 2002)
  • etoricoxib 30 mg and celecoxib 200 mg (NS) (N=599 and 608) (Bingham et al. 2007).
Hip osteoarthritis
  • etodolac 100 to 400 mg versus placebo joint tenderness on pressure, all measures of weight bearing pain (standing, walking, retiring/arising, standing from chair), and night pain for participants receiving (all p0.05) at 12 weeks (N=36) (Sanda et al. 1983)
  • melixocam 15 mg and piroxicam 20 mg (NS) (N=285) (Linden et al. 1996)
  • celecoxib 100 mg and dexibuprofen 400 mg (NS) (N=148) (Hawel 2003).
Hand osteoarthritis

In favour of lumiracoxib 200 and 400 mg compared with placebo (VAS and AUSCAN) at 4 weeks (N=594) (Grifka 2004).

Foot osteoarthritis

Etodolac 800 mg and naproxen 1000 mg at 5 weeks (NS) (N=60) (Jennings and Alfieri 1997).

Knee and hip osteoarthritis

Eleven RCTs (Chikanza and Clarke 1994; Hosie et al. 1996; Hosie 1997; Leung and Malmstrom 2002; Perpignano 1994; Pincus 2004; Rogind 1997; Sowers and White 2005; Wiesenhutter and Boice 2005; Yocum 2000; Zacher 2003) focusing on knee and hip osteoarthritis were identified.

The studies below compared the following COX-2 inhibitors with placebo, all reporting significant reductions in pain in favour of the active drug treatment(s) for treatment period’s ranging from 6 to 12 weeks:

The studies below reported on outcomes for the active drug comparisons for treatment periods ranging from 6 weeks to 6 months:

Mixed sites of osteoarthritis

Three RCTs (Dequeker 1998; Hawkey 1998; Hawkey and Svoboda 2004) included populations of adults with knee, hip, hand or spinal osteoarthritis, while two other RCTs (Schnitzer 2004; Singh 2006) included populations of adults with knee, hip or hand osteoarthritis.

The studies below reported on outcomes for the following active drug comparisons over treatment periods ranging from 28 days to 52 weeks:

  • diclofenac 100 mg showed a statistically significant reduction in pain on active movement (VAS) compared with meloxicam 7.5 mg at 28 days (mean difference 2.29, 95%CI 1.38 to 3.20). For the outcome of pain at rest (VAS) (NS) (N=10051) (Hawkey 1998)
  • lumiracoxib 400 mg, naproxen 1000 mg and ibuprofen 2400 mg (NS) (N=18,325) (Schnitzer et al. 2004)
  • celecoxib 200 or 400 mg compared with naproxen 1000 mg and diclofenac 100 mg (NS) (N=13274) (Singh et al. 2006)
  • lumiracoxib 200 or 400 mg, celecoxib 200 mg and ibuprofen 2400 mg (NS) (N=1042) (Hawkey and Svoboda 2004)
  • meloxicam 7.5 mg and piroxicam 20 mg (NS) (N= 9286) (Dequeker 1998).

Summary

Symptoms: stiffness

Overall, the studies found that both COX-2 inhibitors were superior to placebo in terms of reducing pain over treatment periods ranging from 15 days to six months. The majority of data reported here are for outcomes on the WOMAC and VAS. The limited data available indicated that COX-2 inhibitors and non-selective NSAIDs were comparable with regard to the outcome of stiffness reduction. Only a small number of studies reported a significant difference when comparing COX-2 inhibitors with NSAIDs:

Knee osteoarthritis

Twelve RCTs (Bensen et al. 1999; Fleischmann et al. 2006; Gottesdiener et al. 2002; Kivitz et al. 2001; Lehmann 2005; McKenna 2001; Sheldon 2005; Smugar et al. 2006; Tannenbaum et al. 2004; Williams et al. 2000, 2001; Zhao et al. 1999) focusing on knee osteoarthritis were identified.

The studies below all reported significant improvements in stiffness for the COX-2 inhibitors compared with placebo for treatment periods ranging from 6 to 13 weeks:

The studies below reported outcomes for the following active drug comparisons in WOMAC stiffness for treatment periods ranging from 6 to 14 weeks:

  • celecoxib 100 mg had statistically significant reductions in stiffness scores (WOMAC) compared with naproxen 1000 mg (% change from baseline celecoxib 100 mg −25.5 versus naproxen −22.0) (Zhao et al. 1999)
  • celecoxib 100 mg showed significantly reductions in stiffness scores (WOMAC) compared with naproxen (p<0.001). Celecoxib 200 and 400 mg and naproxen 1000 mg on this outcome (NS) (N=1061) (Kivitz et al. 2001)
  • etoricoxib 20 to 360 mg and diclofenac 150 mg (NS) (N=617) (Gottesdiener et al. 2002)
  • celecoxib 100 mg and diclofenac 50 mg (N=600) (NS) (McKenna 2001).
Hip osteoarthritis

One RCT found that use of etodolac 100 to 400 mg resulted in significant reductions in morning stiffness compared with placebo at 12 weeks (N=36) (Sanda et al. 1983).

Celecoxib 100 mg and dexibuprofen 400 mg (NS) (N=148) (Hawel 2003).

Hand osteoarthritis

One RCT found that at 4 weeks lumiracoxib 200 mg and lumiracoxib 400 mg groups both had statistically significant improvements in pain scores (VAS, AUSCAN) compared with placebo (N=594) (Grifka 2004).

Knee and hip osteoarthritis

Nine RCTs (Chikanza and Clarke 1994; Hosie et al. 1996; Hosie 1997; Leung and Malmstrom 2002; Rogind 1997; Sowers and White 2005; Wiesenhutter and Boice 2005; Yocum 2000; Zacher 2003) focusing on knee and hip osteoarthritis were identified.

The studies below reported a significant difference in favour of the following COX-2 inhibitors compared with placebo for treatment period of 12 weeks:

Out of the studies comparing two active drug comparisons, only one reported a significant reduction in stiffness (WOMAC p=0.02), favouring celecoxib 200 mg versus naproxen 500 mg in participants with hypertension and diabetes after 12 weeks (N=404) (Sowers and White 2005).

The remaining studies reported no statistical differences for the active drug comparisons for treatment period’s ranging from 6 weeks to 6 months:

Summary: general function/global efficacy measures

Overall, it was found that both COX-2 were superior to placebo in terms of improving patient’s and physician’s assessments of disease and overall function scores. The data on direct comparisons of COX-2 inhibitors and non-selective NSAIDs indicate these two drug classes had similar effects for these outcomes. Outcomes were assessed using a number of measures including the Patients’ and Physicians’ Global Assessments and WOMAC. The treatment period’s ranged from 15 days to 52 weeks. Only a small number of studies reported a significant difference on comparisons between two active drug interventions:

Knee: one RCT found in favour of celecoxib compared with naproxen (N=1003) (Bensen et al. 1999) and one found in favour of naproxen compared with celecoxib (N=1061) (Kivitz et al. 2001).

Knee osteoarthritis

Fourteen RCTs (Bensen et al. 1999; Bingham et al. 2007; Fleischmann et al. 2006; Gottesdiener et al. 2002; Kivitz et al. 2001; Lehmann 2005; Lund 1998; McKenna 2001; Sheldon 2005; Smugar et al. 2006; Tannenbaum et al. 2004; Williams et al. 2000, 2001) focusing on knee osteoarthritis were identified.

The studies below reported in favour of the COX-2 inhibitors in comparison with placebo for treatment periods ranging from 3 to 13 weeks:

The studies below reported on outcomes for comparisons between two or more drug interventions for treatment period’s ranging from 12 to 14 weeks:

  • celecoxib 100 mg had a significant improvement in osteoarthritis Severity Index compared with naproxen (p0.05) (N=1003) (Bensen et al. 1999)
  • naproxen 1000 mg had significantly greater improvements compared with celecoxib 100 mg and 400 mg (p0.05) on the outcome of Patient’s Global Assessment, with NS differences between naproxen and doses of celecoxib for all other measures (NS) (N=1061) (Kivitz et al. 2001)
  • lumiracoxib 100 to 400 mg and celecoxib 200 mg (NS) (N=1608) (Fleischmann et al. 2006); (N=1684) (Lehmann 2005); (N=1551) (Sheldon 2005)
  • etoricoxib 20 to 360 mg and diclofenac 150 mg (NS) (N=617) (Gottesdiener 2002); etoricoxib 30 mg and celecoxib 200 mg (NS) (N=599) (Bingham et al. 2007); (N=608) (Bingham et al. 2007)
  • celecoxib 100 mg and diclofenac 50 mg (NS) (N=600) (McKenna 2001).
Hip osteoarthritis

Etodolac 100 to 400 mg resulted in significant improvements on global efficacy measures compared with a placebo group in adults with hip osteoarthritis at 12 weeks (N=36) (Sanda et al. 1983). Two other RCTs found non-significant differences between COX-2 inhibitors and non-selective NSAIDs on global efficacy measures, namely meloxicam and piroxicam (N=285) (Linden et al. 1996) and celecoxib 100 mg and dexibuprofen 400 mg (N=148) (Hawel 2003).

Hand osteoarthritis

One RCT found that at 4 weeks lumiracoxib 200 mg and lumiracocib 400 mg groups both had statistically significant improvements in Patient’s and Physician’s Global Assessments of Disease and patient’s functional status (AUSCAN total score) compared with placebo (N=594) (Grifka 2004).

Knee and hip osteoarthritis

Nine RCTs (Hosie et al. 1996; Hosie 1997; Leung and Malmstrom 2002; Perpignano 1994; Rogind 1997; Sowers and White 2005; Wiesenhutter and Boice 2005; Yocum 2000; Zacher 2003) were identified that focused on knee and hip osteoarthritis.

The studies below reported significant improvements on measures of global efficacy and function scores in favour of the following COX-2 inhibitors compared with placebo for a treatment period of 12 weeks:

The following studies reported on outcomes for comparisons between the active drug comparisons over treatment period’s ranging from 6 weeks to 6 months:

Mixed sites of osteoarthritis

Three RCTs (Dequeker 1998; Hawkey et al. 1998; Hawkey and Svoboda 2004) included populations of adults with knee, hip, hand or spinal osteoarthritis, while two other RCTs (Schnitzer et al. 2004; Singh et al. 2006) included populations of adults with knee, hip or hand osteoarthritis. The treatment period’s ranged from 28 days to 52 weeks:

  • diclofenac 100 mg showed statistically significant improvements in measures of global efficacy and function outcomes compared with meloxicam 7.5 mg at 28 days. However, these differences did not appear to be clinically significant (NS) (N=10,051) (Hawkey 1998)
  • lumiracoxib 400 mg, naproxen 1000 mg and ibuprofen 2400 mg (NS) (N=18325) (Schnitzer et al. 2004)
  • lumiracoxib 200 and 400 mg, celecoxib 200 mg and ibuprofen 2400 mg (NS) (N=1042) (Hawkey et al. 2004)
  • celecoxib 200 and 400 mg and naproxen 1000 mg and diclofenac 100 mg (NS) (N=13274) (Singh et al. 2006)
  • meloxicam 7.5 mg and piroxicam 20 mg (NS) (N= 9286) (Dequeker 1998).

Summary: physical function

Overall, both COX-2 inhibitors were superior to placebo in terms of improving physical function. In general, data are presented for outcomes on the WOMAC. The treatment period’s ranged from 6 to 14 weeks. The limited data on direct comparisons of COX-2 inhibitors and non-selective NSAIDs for this outcome suggested these two drug classes may be comparable for this outcome. Only two studies reported a significant difference between active drug interventions in the knee, in favour of celecoxib compared with naproxen (N=1003) (Zhao et al. 1999); (N=1061) (Kivitz et al. 2001).

Knee osteoarthritis

Eleven RCTs (Bensen et al. 1999; Bingham et al. 2007; Fleischmann et al. 2006; Gottesdiener et al. 2002; Kivitz et al. 2001; Lehmann 2005; McKenna 2001; Smugar et al. 2006; Williams et al. 2000, 2001; Zhao 1999) focussed on knee osteoarthritis The studies below reported in favour of the following COX-2 inhibitors in comparison to placebo for treatment period’s ranging from 6 to 12 weeks:

The studies below reported on outcomes for comparisons between for the following active drug comparisons for treatment period’s ranging from 12 to 14 weeks:

  • celecoxib 100 mg had statistically significant improvements in physical function scores (WOMAC) compared with naproxen (% change from baseline celecoxib 100 mg −26.8 versus naproxen −21.3) (N=1003) (Zhao et al. 1999)
  • celecoxib 100 mg showed significantly greater improvement in WOMAC physical function compared with naproxen (p<0.001). There was NS difference between other celecoxib dose groups and naproxen on this outcome (N=1061) (Kivitz et al. 2001)
  • etoricoxib 20 to 360 mg and diclofenac 150 mg (NS) (N=617) (Gottesdiener et al. 2002); etoricoxib 30 mg and celecoxib 200 mg (NS) (N=599) (Bingham et al. 2007); (N=608) (Bingham 2007)
  • celecoxib 100 mg and diclofenac 50 mg (NS) ((N=600) McKenna 2001).
Hip osteoarthritis

Celecoxib 100 mg and dexibuprofen 400 mg (NS) (N=148) (Hawel 2003)

Knee and hip osteoarthritis

Five RCTs (Leung and Malmstrom 2002; Sowers and White 2005; Wiesenhutter and Boice 2005; Yocum 2000; Zacher 2003) were identified focusing on hip and knee osteoarthritis.

The studies below reported in favour of the following COX-2 inhibitors compared with placebo on WOMAC for a treatment period of 12 weeks:

The following studies reported outcomes for comparisons between the drug interventions for treatments period’s of 6 to 12 weeks:

Physical examination findings

Hip osteoarthritis

In favour of etodolac 100 to 400 mg compared with placebo on the outcomes of ROM hip adduction, ROM external rotation, and ROM internal rotation (all p0.05) at 12 weeks. Outcomes of ROM hip abduction, walking time, and climbing stairs (NS) (N=36) (Sanda et al. 1983).

Celecoxib 100 mg and dexibuprofen 400 mg (NS) (N=148) (Hawel 2003).

Foot osteoarthritis

In favour of etodolac 800 mg compared with naproxen 1000 mg at 5 weeks on walking up steps (p=0.03). Outcomes of walking down stairs, chores, running errands, and walking on a flat surface (NS) (N=60) (Jennings and Alfieri 1997).

Hip osteoarthritis

Celecoxib 100 mg and dexibuprofen 400 mg (NS) (Hawel 2003) (N=148).

Knee and hip osteoarthritis

Two RCTs found NS differences between COX-2 inhibitors and non-selective NSAIDs, meloxicam 15 mg and piroxicam 20 mg (N=455) (Hosie 1997) and meloxicam 7.5 mg and diclofenac sodium 100 mg (N=336) (Hosie et al. 1996) in terms of quality of life outcomes at six month follow-up in adults with hip or knee osteoarthritis.

Gastro-intestinal adverse events

Knee osteoarthritis

Fourteen RCTs (Bensen et al. 1999; Curtis et al. 2005; Fleischmann et al. 2006; Gottesdiener et al. 2002; Kivitz et al. 2001; Lehmann 2005; Lund 1998; McKenna 2001; Sheldon 2005; Smugar et al. 2006; Tannenbaum et al. 2004; Williams et al. 2000, 2001; Zhao et al. 1999) focussed on knee osteoarthritis. Statistical significance testing of differences between treatment groups was not done. COX-2 inhibitors generally had higher percentages of GI AEs compared with placebo, but lower percentages compared with non-selective NSAIDs.

Two RCTs found that celecoxib 200 mg was significantly better than placebo (N=1521) (Smugar et al. 2006); (N=1082) (Smugar et al. 2006) (N=1082) in terms of:

  • discontinuation due to lack of efficacy over 6 weeks (end of study)
  • use of rescue analgesia over 6 weeks (end of study)
  • number of patients with SAEs.

Two RCTs found that there was NS difference between celecoxib 200 mg and placebo (N=1521) (Smugar et al. 2006); (N=1082) in terms of:

  • number of patients with drug-related AEs
  • number of patients with GI AEs
  • number of patients with 1 or more clinical AE.

For the number of withdrawals due to AEs there was no significant difference for etoricoxib 30 mg and placebo (N=599) or celecoxib 200 mg and placebo (N=599) (Bingham et al. 2007); etoricoxib 30 mg and celecoxib 200 mg (NS) (N=599) (Bingham et al. 2007).

One study reported that etoricoxib 30 mg and celecoxib 200 mg were significantly better than placebo for withdrawals due to AEs (N=608) (Bingham et al. 2007) (N=608).

Hip osteoarthritis

Three RCTs focusing on hip osteoarthritis (Hawel 2003, Linden et al. 1996, Sanda et al. 1983) reported on the percentages of GI AEs for COX-2 inhibitors versus non-selective NSAIDs and placebo. Statistical significance testing of differences between treatment groups was not done. COX-2 inhibitors had higher percentages of GI AEs compared with placebo, but lower percentages compared with non-selective NSAIDs.

Hand osteoarthritis

One RCT (Grifka JK 2004) (N=594) reported percentages of GI AEs for COX-2 inhibitors versus placebo. Statistical significance testing of differences between treatment groups was not done. COX-2 inhibitors had higher percentages of GI AEs compared with placebo.

Knee and hip osteoarthritis

Nine RCTs (Hosie et al. 1996; Hosie 1997; Leung and Malmstrom 2002; Perpignano 1994; Pincus 2004; Rogind 1997; Wiesenhutter and Boice 2005; Yocum 2000; Zacher 2003) reported on the percentages of GI AEs for COX-2 inhibitors versus non-selective NSAIDs and placebo. Statistical significance testing of differences between treatment groups was not done for most studies. COX-2 inhibitors generally had higher percentages of GI AEs compared with placebo, but lower percentages compared with non-selective NSAIDs.

Mixed

Three RCTs (Dequeker 1998; Hawkey 1998; Hawkey and Svoboda 2004) included populations of adults with knee, hip, hand or spinal osteoarthritis, while two other RCTs (Schnitzer et al. 2004, Singh et al. 2006) included populations of adults with knee, hip or hand osteoarthritis. These studies found that generally COX-2 inhibitors were associated with fewer GI AEs than non-selective NSAIDs. In people not taking low-dose aspirin, COX-2 inhibitors were associated with fewer GI AEs than non-selective NSAIDs in one study, but not in another. However, there was no difference between the two drug classes in terms of the incidence of GI AEs for people taking low-dose aspirin.

Cardiovascular adverse events

Knee osteoarthritis

Four RCTs (Curtis et al. 2005; Fleischmann et al. 2006; Lund 1998; Sheldon 2005) focusing on knee osteoarthritis reported percentages of different cardiovascular AEs in the table below. Statistical significance testing of differences between treatment groups was not done. There was no visible trend in the direction of the results across the studies:

Hip osteoarthritis

One RCT focusing on hip osteoarthritis (Hawel 2003) reported on the percentages of cardiovascular complaints for COX-2 inhibitors versus non-selective NSAIDs. Statistical significance testing of differences between treatment groups was not done. COX-2 inhibitors had higher percentages of CV AEs in this study compared with non-selective NSAIDs:

Hand osteoarthritis

One RCT (Grifka 2004) (N=594) reported percentages of cardiovascular AEs for COX-2 inhibitors versus placebo. Statistical significance testing of differences between treatment groups was not done. COX-2 inhibitors had lower percentages of CV AEs in this study compared with placebo:

Knee and hip osteoarthritis

Four RCTs (Leung and Malmstrom 2002, Rogind 1997, Wiesenhutter and Boice 2005, Zacher 2003) reported percentages for CV AEs for COX-2 inhibitors versus non-selective NSAIDs and placebo. Statistical significance testing of differences between treatment groups was not done. COX-2 inhibitors had lower percentages of CV AEs in most of these studies compared with non-selective NSAIDs:

Mixed sites of osteoarthritis

One RCT (Hawkey 1998) included populations of adults with knee, hip, hand or spinal osteoarthritis and reported percentages of cardiac failure events without statistical significance testing. Two other RCTs (Schnitzer et al. 2004, Singh et al. 2006) included populations of adults with knee, hip or hand osteoarthritis. One study (Singh et al. 2006) found NS difference between COX-2 inhibitors and non-selective NSAIDs on the rate of myocardial infarction, but found that non-selective NSAIDs had a higher rate of cardiac failure episodes compared with COX-2 inhibitors. A second study (Schnitzer 2004) with a 52-week treatment and follow-up period found that COX-2 inhibitors and non-selective NSAIDs had similar incidences of cardiovascular AEs in adults with osteoarthritis, regardless of concurrent use or non-use of low dose aspirin.

Renal and hepatic adverse events

Knee osteoarthritis

Four knee osteoarthritis studies reported data on renal AEs. One study (Bensen et al. 1999) found that participants receiving celecoxib had a slightly higher percentage of peripheral edema and hypertension than participants on naproxen or placebo, and had similar percentages of participants with abnormal liver function for each study drug. A second study (McKenna 2001) found that participants receiving diclofenac had significant changes in renal values in comparison with placebo, with celecoxib having lower percentage increases in these values than diclofenac, with most being equivalent to placebo. A third study (Tannenbaum et al. 2004) found that participants receiving celecoxib had slightly higher percentage increases in liver function values than lumiracoxib. The fourth study (Williams et al. 2000) found NS difference between celecoxib and placebo in terms of abnormal renal values.

Knee and hip osteoarthritis

Three studies including participants with knee and/or hip osteoarthritis reported data on renal AEs. One study (Perpignano 1994) reported a significant increase in urea values from baseline in the tenoxicam group, whereas there was NS increase in these levels in the etodolac group. There were NS differences between etodolac and tenoxicam in terms of abnormal changes in any of the other renal values reported. A second study (Rogind 1997) found NS differences between etodolac and piroxicam for renal values reported. The third study (Zacher 2003) found that participants receiving etoricoxib had slightly lower percentages of peripheral edema and hypertension compared with those receiving diclofenac. A lower percentage of participants in the etoricoxib group had abnormal increases in liver values compare to the diclofenac group.

Mixed sites of osteoarthritis

Three studies (Dequeker 1998; Hawkey 1998; Schnitzer et al. 2004) included adults with osteoarthritis in different sites (knee, hip, hand, spine). Two studies (Dequeker 1998; Hawkey 1998) found a significantly lower percentage of abnormalities in a number of renal values for COX-2 inhibitors versus non-selective NSAIDs. The other study (Schnitzer 2004) reported no significant difference between the two drug classes in terms of the percentages of major renal events and serious liver abnormalities found. However, this same study found that significantly more participants taking lumiracoxib had abnormal increases in transaminase levels compared with participants taking NSAIDs.

7.3.4. Evidence statements: co-prescription of a proton pump inhibitor

All evidence statements in section 7.3.4 are level 1++.

Adverse events

One study (Scheiman et al. 2006) reported on two identically designed RCTs (VENUS N=844; PLUTO N=585) that investigated the effect of esomeprazole 20 mg or 40 mg versus placebo in adults with osteoarthritis or RA currently using either COX-2 inhibitors or non-selective NSAIDs over a period of 26 weeks. Outcomes reported included the occurrence of gastric and duodenal ulcers and upper GI AEs. Esomeprazole reduced the occurrence of both types of ulcer and upper GI AEs over a 6-month period in participants receiving either COX-2 inhibitors or non-selective NSAIDs in comparison to users of these anti-inflammatory drugs who received placebo instead of a PPI.

Table 7.56Incidence of adverse events with PPI

StudyUlcer typePlaceboEsomeprazole 20 mgEsomeprazole 40 mg
VENUSGastric34/267 (12.7%)12/267 (4.5%)10/271 (3.7%)
Duodenal10/267 (3.7%)0/267 (0.0%)0/271 (0.0%)
GU + DU2/267 (0.7%)0/267 (0.0%)1/271 (0.4%)
PLUTOGastric19/185 (10.3%)7/192 (3.6%)6/196 (3.1%)
Duodenal1/185 (0.5%)1/192 (0.5%)2/196 (1.0%)
GU + DU0/185 (0%)1/192 (0.5%)0/196 (0.0%)

Occurrence of GI ulcers in participants receiving NSAIDs or COX-2 inhibitors

In a stratified pooled analysis of the two studies, significantly fewer participants on esomeprazole compared with placebo developed ulcers when taking a non-selective NSAID or a COX-2 inhibitor after 6 months of treatment.

For participants receiving non-selective NSAIDs, 17.1% (95% CI 12.6 to 21.6) of those on placebo developed ulcers compared with 6.8% (95% CI 3.9 to 9.7, p<0.001) of those who received esomeprazole 20 mg and 4.8% (95% CI 2.3 to 7.2, p<0.001) who received esomeprazole 40 mg.

For participants receiving COX-2 inhibitors, 16.5% (95% CI 9.7 to 23.4) of those on placebo developed ulcers over 6 months compared with 0.9% (95% CI 0 to 2.6, p<0.001) of those who received esomeprazole 20 mg and 4.1% (95% CI 0.6 to 7.6, p=0.002) of those who received esomeprazole 40 mg.

Significant reductions in ulcers occurred for COX-2 inhibitor users taking either dose of esomeprazole in each study versus COX-2 inhibitor users taking placebo (p<0.05). For non-selective NSAID users, esomeprazole significantly reduced ulcer occurrence in the VENUS study (p<0.001) but not in the PLUTO study versus NSAIDs users taking placebo.

GI ulcer incidence in low-dose aspirin users

In participants taking concomitant low-dose aspirin, the ulcer incidence at 6 months was similar to that of the whole study population for all treatment groups (placebo: 12.2%, esomeprazole 20 mg: 4.7%, esomeprazole 40 mg: 4.2%).

Serious GI AEs

Overall, there were more serious GI AEs in participants on placebo (12/452 2.7%) than in participants receiving esomeprazole (9/926 1.0%) across the two studies.

7.3.5. Health economic evidence

We looked at studies that focused on economically evaluating nonsteroidal anti-inflammatory drugs (NSAIDs) and COX-2 treatments, GI protective agents, or placebo for the treatment of adults with osteoarthritis. Sixty-one studies (16 through cross-referencing) were identified through the literature search as possible cost-effectiveness analyses in this area. On closer inspection 56 of these studies were excluded for:

  • not directly answering the question
  • not including sufficient cost data to be considered a true economic analyses
  • involving a study population of less than 30 people
  • not including cardiovascular adverse events in the analysis.

Five papers were found to be methodologically sound and were included as health economics evidence. However, none of the papers were UK-based and of an acceptable standard to satisfy the GDG as suitable evidence from which to make recommendations. For this reason this area was outlined as important for additional economic modelling. Due to this what follows is simply a brief review of the included studies.

One Canadian study (Maetzel et al. 2003) conducts a detailed cost–utility analysis assessing rofecoxib and celecoxib compared with non-selective NSAIDs. The model involved a Markov model with a decision tree within each health state. Myocardial infarction (MI) was included as a cardiovascular (CV) adverse event, but no other CV adverse events were included. The model had a 5-year duration, but was limited in that once one MI had occurred a patient could not suffer any further CV events. Direct health care costs (in 1999 Canadian $) were calculated and QALYs were estimated using utility values obtained by a standard gamble technique from a survey of 60 randomly selected individuals. The patient population was people with OA or rheumatoid arthritis (RA) who were not prescribed aspirin. The study assumed equal effectiveness of the drugs and only considered differences in adverse events.

The study results were as follows:

  • for average-risk patients, the cost per additional QALY of treating patients with rofecoxib rather than naproxen was $455,071
  • for average-risk patients, the cost per additional QALY of treating patients with diclofenac rather than ibuprofen was $248,160, and celecoxib was dominated by diclofenac
  • for high-risk patients, treatment with rofecoxib dominated treatment with naproxen + PPI; the cost per additional QALY of treating patients with rofecoxib + PPI compared with rofecoxib on its own was $567,820
  • for high-risk patients, treatment with celecoxib dominated treatment with ibuprofen + PPI. The cost per additional QALY of treating patients with diclofenac + PPI compared with celecoxib was $518,339. Treating patients with celecoxib + PPI was dominated by treating patients with diclofenac + PPI.

Hence the study concluded that treatment with COX-2 inhibitors is cost effective in high risk patient groups with OA and RA, but not in average risk groups.

A US study considered the cost effectiveness of COX-2 inhibitors compared with non-selective NSAIDs for people with arthritis from the Veterans Health Administration perspective (Schaefer 2005). Two patient groups were considered – those of any age who had a history of perforation, ulcer or bleed (PUB); and those aged 65 years or older, regardless of their PUB history. Both of these groups are regarded as being at ‘high risk’ of a gastrointestinal (GI) event. CV events included were MI and chronic heart failure (CHF). Costs are in 2001 US$ and QALY weights were obtained from the literature. The time period modelled was one year, but a scenario was also included where the costs for MI were calculated for a 10-year period. The study assumed equal effectiveness of the drugs and only considered differences in adverse events.

The results of the study were as follows:

  • the cost per additional QALY for celecoxib compared with non-selective NSAIDs was $28,214 for the PUB history analysis, rofecoxib was dominated by celecoxib and non-selective NSAIDs
  • the cost per additional QALY for celecoxib compared with non-selective NSAIDs was $42,036 in the elderly patient analysis, again rofecoxib was dominated by both celecoxib and non-selective NSAIDs
  • sensitivity analysis showed that with a threshold cost per QALY value of $50,000, there was an 88% probability that celecoxib would be cost effective in the elderly population, and a 94% probability that it would be cost effective in the PUB history population.

Another US study (Spiegel et al. 2003) conducted a cost-utility analysis comparing COX-2 inhibitors to nonselective NSAIDs. The patient population was 60-year-old patients with OA or RA who were not taking aspirin and who required long-term NSAID therapy for moderate to severe arthritis pain. A lifetime duration was adopted. CV events were included in sensitivity analysis. Patients with a history of ulcer complications were included in sensitivity analysis. A third party payer perspective was adopted for costs (estimated in 2002 US$) and utility values validated by previous investigators were used to allow QALYs to be calculated. The study assumed equal effectiveness of the drugs and only considered differences in adverse events.

The results of the study were as follows:

  • the cost per additional QALY of treating patients with a COX-2 inhibitor (celecoxib or rofecoxib) rather than naproxen was $395,324
  • the cost per additional QALY of treating patients with a COX-2 inhibitor rather than naproxen assuming a high-risk cohort was $55,803.

A UK study conducted a cost-minimisation analysis based on patients aged 18 or over with acute osteoarthritis of the hip, knee, hand or vertebral spine, taking an NHS perspective (Tavakoli 2003). The treatments considered were meloxicam, diclofenac, and piroxicam, and all resource use associated with GI and non-GI adverse events were included as costs, calculated in 1998 £s. However, the duration of the model was only 4 weeks, giving little time for costs to be accrued.

The results of the study were as follows:

  • cost per patient was least for meloxicam (£30), followed by piroxicam (£35) and diclofenac (£51).

An Australian study conducted a cost-utility analysis on a number of different interventions for OA (Segal et al. 2004). One of these analyses involved comparing diclofenac and naproxen with celecoxib. Efficacy was included in the analysis by allocating QALY gains due to pain relief. PUBs and CHF were included as adverse events. Health service costs were considered and are calculated in 2000–2001 Aus$, and QALYs were calculated using the transfer to utility (TTU) technique. The drugs were compared with placebo. The analysis is based on 12 months of treatment. A significant problem with the study is that QALY scores for non-fatal AEs are not incorporated into the modelling, meaning that only fatal AEs are reflected in the results.

The results of the study were as follows:

  • the best estimate of cost per additional QALY of treating patients with naproxen rather than placebo (paracetamol) was $7,900 per additional QALY, incorporating a 5% discount rate
  • the best estimate of cost per additional QALY of treating patients with diclofenac rather than placebo (paracetamol) was $40,800 per additional QALY, incorporating a 5% discount rate
  • the best estimate of cost per additional QALY of treating patients with celecoxib rather than placebo (paracetamol) was $32,930 per additional QALY, incorporating a 5% discount rate
  • the study does not directly compare non-selective NSAIDs to COX-2 inhibitors, but the results suggest that net utility gains are similar for the two types of drugs, while non-selective NSAIDs result in lower costs.

7.3.6. Health economic modelling

We conducted a cost-effectiveness analysis, comparing paracetamol, standard NSAIDs and COX-2 inhibitors at doses relevant to clinical practice for which there were robust clinical trial data sufficient to draw reliable conclusions: paracetamol 3000 mg, diclofenac 100 mg, naproxen 750 mg, ibuprofen 1200 mg, celecoxib 200 mg, and etoricoxib 60 mg. We also tested the cost effectiveness of adding omeprazole, a proton pump inhibitor, to each of these NSAIDs/COX-2 inhibitors. It should be noted that we did not consider the cost-effectiveness of other NSAIDs, meloxicam or etodolac, due to a lack of suitable data.

The analysis was based on an assumption that the NSAIDs and COX-2 inhibitors are equally effective at controlling OA symptoms, but that they differ in terms of GI and CV risks. The adverse event risks were taken from three key studies: MEDAL, CLASS and TARGET. As the doses of both standard NSAIDs and COX-2 inhibitors were very high in these trials, we adjusted the observed rates to estimate the impact of more commonly-used and licensed doses. The effectiveness of NSAIDs/COX-2 inhibitors and paracetamol at controlling OA symptoms was estimated from a meta-analysis of RCTs. Given these assumptions, lower doses of a drug will always be more cost effective than a higher dose of the same drug. In practice, though, some individuals may require higher doses than we have assumed in order to achieve an adequate therapeutic response.

One clear result of our analysis is that it is cost effective to add a generic PPI to standard NSAIDs and COX-2 inhibitors. We did not test the relative cost effectiveness of other gastroprotective agents, because of the superior effectiveness evidence for PPIs, and the currently very low cost of omeprazole at this dose.

Given our assumptions and current drug costs, celecoxib 200 mg is the most cost effective of the included NSAIDs/COX-2 inhibitors. This result was not sensitive to the assumed duration of treatment (from 3 months to 2 years), or to the baseline risk of GI events in the population (55 years vs 65 years). It was also relatively insensitive to the baseline risk of CV events. In patients who cannot tolerate celecoxib, etoricoxib 30 mg would be a cost-effective alternative. The relative cost effectiveness of these two options in this context depends primarily on their cost.

However, it is important to note substantial uncertainties over the relative rates of adverse events associated with the COX-2 inhibitors estimated from the MEDAL, TARGET and CLASS studies. In particular, the estimated risk of stroke for celecoxib from CLASS was surprisingly low. If this is an underestimate, then etoricoxib 30 mg could be more cost effective than celecoxib 200 mg. The full data submitted to the American Food and Drug Administration were used for the economic model.

Observational data imply a less attractive cost–effectiveness ratio for celecoxib (around £30,000 per QALY), though this estimate may be biased by its use in selected higher-risk patients in clinical practice. There were no observational data for the other COX-2 inhibitors.

For patients who cannot, or do not wish to, take a COX-2 inhibitor, the relative cost effectiveness of paracetamol and standard NSAIDs depends on their individual risk profile, as well as the dose required to achieve an adequate therapeutic response:

  • with low GI and CV risk (patients aged under 65 with no risk factors), standard NSAIDs with a PPI do appear to be relatively cost effective in comparison with paracetamol or no intervention
  • for patients with raised GI or CV risk (aged over 65 or with risk factors), standard NSAIDs are not a cost-effective alternative to paracetamol; in our model, the risks of these treatments outweighed the benefits of improved control of OA symptoms, as well as incurring additional costs for the health service.

The model provides cost-effectiveness estimates at a population level, including for NSAIDs in people with increased GI risk. Clearly, for many of these people NSAIDs will be contraindicated and thus the average cost effectiveness in those who remain eligible will be better than the estimate given here. The relative cost effectiveness of particular NSAIDs and COX-2 inhibitors will vary depending on individual patients’ GI and CV risk factors.

The model assesses which of the drugs is most suitable as the first choice for treatment. In instances where the drug is not tolerated or gives inadequate relief, and a different drug from this class is sought as the second choice, treatment needs to be carefully tailored to the individual and it is not possible to provide useful recommendations in a national clinical guideline for this.

The relative costs of the standard NSAIDs employed in this model (diclofenac 100 mg, naproxen 750 mg and ibuprofen 1200 mg) prescribed concurrently with a PPI are similar, and uncertainties over the relative incidence of adverse events with these drugs make it difficult to draw clear conclusions about their comparative cost effectiveness.

The doses and costs considered in the model are shown in Appendix D, available online at www.rcplondon.ac.uk/pubs/brochure.aspx?e=242. Because the incremental cost effectiveness ratios are affected by dose and individual risk factors, the Guideline Development Group felt it would be unwise to single out specific drugs and doses within these classes, except for etoricoxib 60 mg, which was consistently dominated (more expensive and has overall lower gain in QALYs than comparator drugs) in the model results. Readers should be alert to changes in available drug doses and costs after this guideline is published.

7.3.7. From evidence to recommendations

A large amount of clinical trial evidence supports the efficacy of both traditional NSAIDs and COX-2 selective agents in reducing the pain and stiffness of osteoarthritis with the majority of studies reflecting short-term usage and involving knee or hip joint osteoarthritis. There is no strong evidence to suggest a consistent benefit over paracetamol, although some patients may obtain greater symptom relief from NSAIDs. There are again no data to suggest benefits above opioids, but there is a lack of well-designed comparator studies.

All NSAIDs, irrespective of COX-1 and COX-2 selectivity are associated with significant morbidity and mortality due to adverse effects on the GI, renal and cardiovascular system. It should be noted again that clinical trials recruit patients without the serious comorbidities that would be present in routine clinical practice and that supra-normal doses of newer agents are commonly used in clinical trials in order to demonstrate safety.

GI toxicity

There are some data to support that certain COX-2 selective agents reduce the incidence of serious GI adverse events (such as perforations, ulcers and bleeds) when compared with less selective agents, while the evidence for other agents has been more controversial. Dyspepsia, one of the most common reasons for discontinuation, remains a problem with all NSAIDs irrespective of COX-2 selectivity.

Liver toxicity

At the time of writing, lumiracoxib has been withdrawn from the UK market, following concerns about liver toxicity associated with high doses. The model therefore represents the current situation regarding liver toxicity. The GDG were mindful that further safety data will emerge in the lifetime of this guideline; prescribers should be aware of the Summaries of Product Characteristics.

Cardiovascular toxicity

All NSAIDs have the propensity to cause fluid retention and to aggravate hypertension, although for certain agents this effect appears to be larger (etoricoxib). Increasingly a pro-thrombotic risk (including myocardial infarction and stroke) has been identified with COX-2 selective agents in long-term studies, and there does seem to be some evidence for a dose effect. These observational studies also demonstrate an increased cardiovascular risk from older agents such as diclofenac, which has high COX-2 selectivity. It is possible that naproxen does not increase pro-thrombotic risk. All NSAIDs may antagonise the cardio-protective effects of aspirin.

Summary

All potential adverse effects must be put in perspective of patient need and individual risk, including the influence of the patient’s age on their GI risk. Best estimates of toxicity data, along with the uncertainty in these values, are detailed in Appendix D, available online at www.rcplondon.ac.uk/pubs/brochure.aspx?e=242. The recommendations mention assessment and monitoring of risk factors, but are unable to specify these because of the rapidly emerging evidence base in this area. Prescribers will be informed by the regularly updated Summaries of Product Characteristics.

There is likely to be a continuing role for NSAIDs/COX-2 inhibitors in the management of some patients with OA. Allowing for the inevitable differences in individual patient response, in general the choice between NSAIDs and COX-2 inhibitors is influenced by their separate side-effect profiles, which tend to favour COX-2 inhibitors, and cost, which tends to favour NSAIDs. Extensive sensitivity analyses showed that these are the two factors which most strongly influence the results of the economic model.

Given that costs are constantly changing and that new data on adverse events will become available, the GDG deemed it unwise to suggest a particular ranking of individual drugs. Indeed, there is no clear distinction between the two sub-classes. Meloxicam and etodolac were not included in the model because of a lack of comparable trial data, and other NSAIDs were excluded because of the rarity of use in the UK, according to the Prescription Pricing Authority (see Appendix D for details, available online at www.rcplondon.ac.uk/pubs/brochure.aspx?e=242). It is beyond the role of a clinical guideline to attempt to categorise meloxicam or etodolac into one of the two sub-classes. However, it is worth noting that each of the drugs in this section varies in its COX-1/COX-2 selectivity.

There was a consistent difference between etoricoxib 60 mg and the other drugs in the model, and therefore in line with the original aim of the economic model, advice is given against the use of etoricoxib 60 mg as the first choice for treatment.

The GDG also noted that the incidence of potentially serious upper GI problems can be reduced by the use of PPIs, and the potential benefit of coprescription of PPIs was an important element of the cost-effectiveness analysis. In fact, the analysis found that it was always more cost effective to coprescribe a PPI than not to do so. The primary paper discussed was the Scheiman paper (Scheiman 2006). The Lai paper was excluded as it was an open-label trial and the Chan paper (Chan 2007) had several limitations: i) a population following hospitalisation for upper GI bleeding (which was not what we were looking at for the model); and ii) it had a zero event rate in the PPI arm of the trial. This meant that we were unable to calculate a relative risk, which is required for the model. Hence the Chan paper corroborates the effectiveness of adding a PPI to a COX-2, but has not been used for the sensitivity analysis. The GDG have attempted to balance all these factors in the following recommendations.

Although NSAIDs and COX-2 inhibitors may be regarded as a single drug class of NSAIDs, these recommendations continue to use the two terms for clarity, and because of the differences in side-effect profile. The recommendations in this section are derived from extensive health economic modelling, which included December 2007 NHS drug tariff costs. This guideline replaces the osteoarthritis aspects only of NICE technology appraisal guidance 27 (National Institute for Health and Clinical Excellence 2001). The guideline recommendations are based on up-to-date evidence on efficacy and adverse events, current costs and an expanded health-economic analysis of cost effectiveness. This has led to an increased role for COX-2 inhibitors, an increased awareness of all potential adverse events (gastrointestinal, liver and cardio-renal) and a recommendation to coprescribe a proton pump inhibitor (PPI).

RECOMMENDATIONS
R25.

Where paracetamol or topical NSAIDs are ineffective for pain relief for people with osteoarthritis, then substitution with an oral NSAID/COX-2 inhibitor should be considered.

R26.

Where paracetamol or topical NSAIDs provide insufficient pain relief for people with osteoarthritis, then the addition of an oral NSAID/COX-2 inhibitor to paracetamol should be considered.

R27.

Oral NSAIDs/COX-2 inhibitors should be used at the lowest effective dose for the shortest possible period of time.

R28.

When offering treatment with an oral NSAID/COX-2 inhibitor, the first choice should be either a standard NSAID or a COX-2 inhibitor (other than etoricoxib 60 mg). In either case, these should be coprescribed with a proton pump inhibitor (PPI), choosing the one with the lowest acquisition cost.

R29.

All oral NSAIDs/COX-2 inhibitors have analgesic effects of a similar magnitude but vary in their potential GI, liver and cardio-renal toxicity and therefore when choosing the agent and dose, healthcare professionals should take into account individual patient risk factors, including age. When prescribing these drugs, consideration should be given to appropriate assessment and/or ongoing monitoring of these risk factors.

R30.

If a person with osteoarthritis needs to take low dose aspirin, healthcare professionals should consider other analgesics before substituting or adding an NSAID or COX-2 inhibitor (with a PPI) if pain relief is ineffective or insufficient.

7.4. Intra-articular injections

7.4.1. Clinical introduction

This section of the guideline is concerned with those therapies that require use of an intra-articular injection, including corticosteroid and hyaluronan injections. It should be noted that all such therapies should be delivered by appropriately trained individuals, and that even experienced clinicians can miss intra-articular placement of such injections, especially in small joints.

Corticosteroids

Corticosteroid injections are used to deliver high doses of synthetic corticosteroids to a specific joint, while minimising systemic side effects. Corticosteroids have marked anti-inflammatory effects, and it is assumed that their analgesic action in osteoarthritis is in some way related to their anti-inflammatory properties. Certainly intra-articular corticosteroids can reduce the volume of synovitis of osteoarthritis (Ostergaard et al. 1996). However, the relationship between osteoarthritis synovitis and pain is less clear. It is recognised that clinical examination is not sensitive in detecting inflammation (synovial hypertrophy or effusions) when compared with imaging methods such as ultrasonography or MRI (D’Agostino et al. 2005), so clinical prediction of response to a corticosteroid injection is unreliable. The presence of an effusion is not in itself an indication for corticosteroid injection, unless there is significant restriction of function associated with the swelling. Rather, the indication should be based on severity of pain and disability.

Hyaluronans

Endogenous hyaluronan (HA, previously known as hyaluronic acid) is a large, linear glycosaminoglycan and is a major non-structural component of both the synovial and cartilage extracellular matrix. It is also found in synovial fluid and is produced by the lining layer cells of the joint. Hyaluronan is removed from the joint via the lymphatic circulation and degraded by hepatic endothelial cells. Its key functions in the joint are to confer viscoelasticity, lubrication and help maintain tissue hydration and protein homeostasis by preventing large fluid movements and by acting as an osmotic buffer.

Synthetic HA was isolated from roosters’ comb and umbilical cord tissue and developed for clinical use in ophthalmic surgery and arthritis in the 1960s. The beneficial effects in ophthalmic surgery were followed by the use of HA in osteoarthritis: the rationale was to replace the properties lost by reduced HA production and quality as occurs in osteoarthritis joints, a concept known as viscosupplementation. Commercial preparations of HA have the same structure as endogenous HA although cross-linked HA molecules (known as hylans) were later engineered by linking HA molecules in order to obtain greater elasto-viscosity and intra-articular dwell-time.

However, the mechanism by which HA exerts its therapeutic effect, if any, is not certain, and evidence for restoration of rheological properties is lacking. Given the relatively short intra-articular residency (hours), any hypothesis for its mechanism of action must account for the sometime reported long-duration of clinical efficacy (months).

7.4.2. Methodological introduction: corticosteroids

We looked for studies that investigated the efficacy and safety of intra-articular injection of corticosteroid compared with placebo with respect to symptoms, function and quality of life in adults with osteoarthritis. One Cochrane systematic review and meta-analysis on knee osteoarthritis patients (Bellamy et al. 2006) and three additional RCTs on osteoarthritis of the hip (Flanagan et al. 1988; Qvistgaard et al. 2006) or thumb (Meenagh et al. 2004) were found. No relevant cohort or case-control studies were found.

The meta-analysis assessed the RCTs for quality and pooled together all data for the outcomes of symptoms, function and AEs. However, the outcome of quality of life was not reported. The results for quality of life have therefore been taken from the individual RCTs included in the systematic review.

The meta-analysis included 12 RCTs (with N=653 participants) with comparisons between intra-articular corticosteroids and intra-articular placebo injections in patients with knee osteoarthritis. Studies included in the analysis differed with respect to:

  • type of corticosteroid used (one RCT prednisolone acetate; four RCTs triamcinolone hexacetonide; one RCT methylprednisolone; three RCTs hydrocortisone solution; two RCTs triamcinolone acetonide; one RCT cortivazol; one RCT methylprednisolone acetate)
  • treatment regimens
  • trial design, size and length.

Tests for heterogeneity were performed for any pooled results, but no evidence of heterogeneity was found between studies that were combined. Unless otherwise stated, all evidence statements are derived from data presented in the systematic review and meta-analysis.

The three additional RCTs focused on the outcomes of symptoms, function and quality of life. The three included RCTs were similar in terms of osteoarthritis diagnosis (radiologically). However, they differed with respect to osteoarthritis site, corticosteroid agent, and sample size.

7.4.3. Methodological introduction: hyaluronans

We looked for studies that investigated the efficacy and safety of intra-articular injection of hyaluronic acid/hyaluronans compared with placebo or steroid injection with respect to symptoms, function and quality of life in adults with osteoarthritis. One Cochrane systematic review and meta-analysis on knee osteoarthritis patients (Bellamy et al. 2006) and four additional RCTs on hip/knee/other osteoarthritis sites (Fuchs et al. 2006; Petrella et al. 2006; Qvistgaard et al. 2006; Stahl 2005) were found. No relevant cohort or case-control studies were found.

The meta-analysis assessed the RCTs for quality and pooled all data for the outcomes of symptoms, function and AEs. However, the outcome of quality of life was not reported. The results for quality of life have therefore been taken from the individual RCTs included in the systematic review.

The meta-analysis included 40 RCTs (with N=5257 participants) on comparisons between intra-articular hyaluronic acid/hyaluronans and intra-articular placebo injections in patients with knee osteoarthritis. Studies included in the analysis differed with respect to:

  • type of HA used (seven RCTs Artz; one RCT BioHy; one RCT Durolane; 14 RCTs Hyalgan; nine RCTs Synvisc; one RCT Suvenyl; five RCTs Orthovisc; one RCT Replasyn; one RCT Suplasyn)
  • treatment regimens
  • trial design, size and length
  • mode of HA production (includes bacterial and animal sources).

Additionally, the meta-analysis included nine RCTs (with N=755 participants) on comparisons between intra-articular hyaluronic acid/hyaluronans and intra-articular corticosteroid injections. Studies included in the analysis differed with respect to:

  • type of HA used (five RCTs Hyalgan; two RCTs Synvisc; two RCTs Orthovisc)
  • type of corticosteroid used (five RCTs methylprednisolone acetate; two RCTs triamcinolone hexacetonide; one RCT betamethasone; one RCT betamethasone sodium phosphate-betamethasone acetate)
  • treatment regimens
  • trial size and length
  • mode of HA production (includes bacterial and animal sources).

Results in the Cochrane meta-analysis were presented by product and also pooled by drug-class. This was because products differed in their molecular weight, concentration, treatment schedule and mode of production, and therefore the results for the individual products, rather than pooled class data, are presented here. The chi-squared test for heterogeneity was performed on any pooled results, and evidence of significant heterogeneity (p<0.10) was found between pooled studies for many of the outcomes.

Unless otherwise stated, all evidence statements are derived from data presented in the systematic review and meta-analysis.

  • The four additional RCTs focused on the outcomes of symptoms, function and quality of life. However, one of these (Stahl et al. 2005) was excluded as evidence due to multiple methodological limitations. The trials were similar in terms of osteoarthritis diagnosis (radiologically).
  • However, they differed with respect to osteoarthritis site, study intervention, sample size and study duration. One RCT (Qvistgaard et al. 2006) had a 3-week treatment phase and a follow-up at 3 months; the second RCT (Petrella and Petrella 2006) had a 2-month treatment phase and a follow-up at 12 weeks, and the third RCT (Fuchs et al. 2006) had a 3-week treatment phase and a follow-up at 26 weeks.

7.4.4. Evidence statements: Intra-articular (IA) corticosteroids vs placebo

Knee

Overall, the evidence appraised by the Cochrane review suggests a short-term benefit (up to 1 week) in terms of pain reduction and patient global assessment after IA injections with corticosteroids in the knee. Beyond this period of time there were non-significant differences between IA corticosteroids and IA placebo as reported by most of the studies identified.

There was evidence of pain reduction between two weeks to three weeks but a lack of evidence for efficacy in functional improvement.

No significant differences between corticosteroids and placebo were reported at any time point by studies evaluating the following outcomes in patients with knee OA:

  • functional improvement (eg walking distance, range of motion)
  • stiffness
  • safety
  • study withdrawals.

7.4.5. Evidence statements: Intra-articular (IA) hyaluronans vs placebo

Knee

Overall, the evidence suggests that hyaluronans and hylan derivatives seem to be superior to placebo in terms of efficacy* and quality of life outcomes in patients with OA in the knee at different post-injection periods but especially at the 5- to 13-week post injection period.

No major safety issues were identified relating to these agents when compared with placebo but a definitive conclusion is precluded due to sample-size restrictions.

It should be noted that alongside the by drug class (pooled) results, outcomes are presented by therapeutic agent due to the differential efficacy effects for different products on different variables at different timepoints** found by the Cochrane review (Bellamy 2006).

7.4.6. Evidence statements: Intraarticular (IA) therapy: Hyaluronans vs Corticosteroid

Knee

Data from the Cochrane meta-analysis (Bellamy et al. 2006b) suggest that IA therapy with hyaluronans may have a more prolonged effect than IA with corticosteroids.

Hip and hand

No significant differences between hyaluronans and corticosteroids were reported at any time point by the two studies evaluating efficacy and function outcomes in patients with OA in the hip (Qvistgaard 2006) and the hand (Fuchs 2006).

Table 7.72. Pain in knee OA

Table 7.73. Function in knee OA

Table 7.74. Adverse events in knee OA

Table 7.75. Withdrawals in knee OA

7.4.7. Health economic evidence

We looked at studies that conducted economic evaluations involving corticosteroids or hyaluronans versus placebo or compared with each other. Four papers: one French, one Canadian, one Taiwanese and one from the USA, and all evaluating hyaluronans, are included here as a summary of the health economic evidence currently available. However, due to some methodological limitations the use of these papers is limited and evidence statements cannot be made from them.

The French study (Kahan et al. 2003) compared treatment including hyaluronan with ‘conventional treatment’ for patients with knee OA. The authors carried out a cost-effectiveness analysis using the Lequesne index, WOMAC and the SF-12. The duration of the analysis was 9 months, and the hyaluronan treatment arm involved three intrarticular injections spaced one week apart. No information is given on what the conventional treatment arm involved so unfortunately it is difficult to interpret the results precisely.

An additional problem with the study is that the full costs of hyaluronan were not included, since its cost was based on the 65% reimbursement rate instituted in France in September 2000. All other costs were reported in 1998 euros.

Table 7.76Results of Kohan’s economic evaluation

Conventional treatmentHyaluronan treatment (Synvisc)p value
Total cost (societal)€829.40€829.10
Total cost (medical only)€777.90€785.30
Total cost (medical only, including total Hyaluronan cost€777.90€848.14
Reduction in:
 Lequesne index−1.6−3.60.0001
WOMAC (total)−8.1−19.80.0001
WOMAC (pain)−12.2−24.60.0001
WOMAC (stiffness)−7.7−20.70.0001
WOMAC (function)−7.0−18.40.0001
 Pain on walking (VAS)−24.4−37.40.0001
Improvement in quality of life (SF12):
 Physical2.35.50.0001
 Mental1.62.9Not stated

The results taken from the French paper are shown in the table above (Kahan et al. 2003). They appear to show hyaluronan treatment with Synvisc dominating (less expensive and more effective) conventional treatment. If only medical costs are considered, Synvisc does not appear dominant, but given it was found to be significantly more effective than conventional treatment for the outcome measures considered for a small cost increment, then Synvisc would appear to be cost effective.

Including the full hyaluronan costs, the cost difference per patient becomes more substantial and the cost effectiveness of the treatment becomes more uncertain. For example, the cost per one point improvement in the SF12 Physical Quality of Life scale over a 9-month period is € 21.95 when comparing hyaluronan treatment to conventional treatment.

The lack of detail included regarding what treatment ‘conventional treatment’ involves, and the fact that the total cost of the hyaluronan treatment was not included due to the reimbursement regime in France means that this study cannot be used to make evidence statements.

The Canadian study (Torrance et al. 2002) conducted a cost–utility analysis to compare appropriate care with hyaluronan treatment to appropriate care without hyaluronan treatment for patients with knee OA. Appropriate care is described as the preferred management strategy of specialists, rheumatologists or orthopaedic surgeons, encouraged to follow treatment guidelines published by the American College of Rheumatology and instructed to treat conservatively. Appropriate care did include corticosteroid injections for some patients. The clinical data were taken from the clinical trial run alongside the economic analysis, which was a 1-year prospective, randomised, open-label, parallel design trial (Raynauld 2002). The unblinded nature of the trial may bias the results.

The base-case analysis took the societal perspective, but a healthcare payer perspective was considered in sensitivity analysis and this had little effect on costs differences between treatment groups, and so affected results only minimally. All costs were reported in Canadian dollars at 1999 prices.

Table 7.77Costs from Torrance’s economic evaluation

Appropriate careAppropriate care + HyaluronanDifference
Total societal cost (Canadian $)$1,414.58$2,124.71$710.13
Total healthcare payer cost (Candian $)Not statedNot stated$705

Table 7.78Results of Torrance’s economic evaluation

Incremental costQALY gainCost per QALY gain
Base case$7050.071$9,930
SA on outcomes
High$7050.117$6,026
Low$7050.025$28,200
SA on costs
High$1,0080.071$14,197
Low$4020.071$5,662

The results of the Canadian study are shown in the tables above. The study provided little information on what appropriate care involves. The cost data provided show that some patients in both treatment arms received corticosteroid injections, with the cost per patient of this treatment being higher for patients in the appropriate care arm. However, the cost per patient ($18.45) is still very low and the difference in ‘medication’ costs (for example, NSAIDs) per patient between the groups is far larger ($237.32 vs $305.10). Hence it is not clear that this analysis allows a comparison of hyaluronan treatment and corticosteroid treatment as there are no data that suggest that patients substitute one treatment for another depending on which group they are in, particularly as some patients in the hyaluronan arm clearly receive both treatments.

Hyaluronan can also not be compared with placebo using this study as a placebo was not used in the study. The fact that the study was not blinded may also weaken any evidence statements made, and so again, given these problems no evidence statements are made.

The study conducted in Taiwan was well conducted, but is of limited use for the questions being addressed here. This is because the study compared hyaluronan with celecoxib and naproxen in a cost-effectiveness analysis (Yen et al. 2004). Being set in Taiwan makes the study of limited use for a UK guideline, and also CV adverse events were not included, which severely harms the credibility of the naproxen and celecoxib analysis. Given that a placebo arm was not included, this harms the interpretation of the hyaluronan treatment arm as the treatments it is compared with may not be credible. All costs were reported in US dollars at 2002 prices. Table 7.79 shows the results of the study.

Table 7.79. Results of Yen’s economic evaluation.

Table 7.79

Results of Yen’s economic evaluation.

These results meant that hyaluronan was not considered cost effective compared with celecoxib and naproxen in Taiwan. If CV effects were included, celecoxib and naproxen would appear slightly worse, making hyaluronan appear better in comparison. However, due to the setting of the study and the comparators used, no useful evidence statements can be made from this study.

The US study (Waddell et al. 2001) builds a pharmacoeconomic model to calculate the cost effects of including hyaluronan treatment in a standard osteoarthritis treatment pathway for patients with mild, moderate, and severe OA. This treatment path does include steroid injections for some patients. The study is not a formal cost-effectiveness or cost–utility analysis as it does not include a measure of health gain attributable to the treatment. Instead, it assumes that hyaluronan treatment reduces the need for other treatments, and delays total knee replacement operations, and analyses the difference this makes to costs from a health-insurer perspective.

The chief problem with this study is that the duration used is not appropriate for the modelling assumptions. It is assumed that hyaluronan treatment delays total knee replacement (a key cost driver) by approximately one year. The authors analyse a 3-year time period and then include no future costs. Given that a certain proportion of the theoretical cohort has a total knee replacement each year, and that it is assumed that the hyaluronan treatment delays such treatment by a year, it is obvious that the hyaluronan treatment group will incur much less costs in a fixed time period. However, these costs would simply be accrued the following year in reality, unless the treatment actually prevents total knee replacement surgery, which is not what the paper assumes. For this reason the results of the paper are very misleading, and so again no evidence statements can be made.

7.4.8. From evidence to recommendations

Corticosteroids

Generally the research evidence demonstrates that intra-articular corticosteroid injections provide short-term (1–4 weeks) reduction in osteoarthritis pain, although effects on function appear less marked. The effects have been best demonstrated for knee osteoarthritis, although there are some data for efficacy in hip and hand osteoarthritis. The GDG noted that these injections are widely used in many osteoarthritis sites. There is no clear message from this evidence on whether any particular corticosteroid preparation is more effective than another, or on which dose of a given preparation is most effective. In clinical practice, the short-term pain relief may settle flares of pain and also allow time for patients to begin other interventions such as joint-related muscle strengthening.

The risks associated with intra-articular corticosteroid injection are generally small. A small percentage of patients may experience a transient increase in pain following injection. Subcutaneous deposition of steroid may lead to local fat atrophy and cosmetic defect. Care should always be taken when injecting small joints (such as finger joints) to avoid traumatising local nerves. There is a very small risk of infection. The question of steroid-arthropathy, that is, whether intra-articular steroids may increase cartilage loss, remains controversial and is currently based on animal model and retrospective human studies. Nevertheless, caution should be applied if injecting an individual joint on multiple occasions and other osteoarthritis therapies should be optimised.

Hyaluronans

The research evidence on the efficacy of HAs is often difficult to interpret because of confounders including:

  • different molecular weights of HA
  • different injection schedules (ranging from once weekly to a series of five injections)
  • poor trial design despite large numbers of studies, for example lack of intention-to-treat analyses, limitations in blinding.

On balance, the evidence seems to suggest a benefit for reducing pain up to 3 months after a series of three to five injections, although the effect size is generally small. Given this, and the cost of the therapies together with increased clinician visits required for injections, hyaluronan injections were assessed in the cost-consequence analysis (see Appendix C for details, available online at www.rcplondon.ac.uk/pubs/brochure.aspx?e=242). The Cochrane review (Bellamy 2006b) regarded pooled estimates across different products as potentially misleading, and also warned about pooled estimates because of different study designs. Also, meta-analysis was only possible for two of the WOMAC sub-scales, ruling out the use of the transfer to utility technique. With this in mind, and given the effect that different injection schedules have on cost estimates, the cost-consequence analysis looked at three products individually, using estimates from individual trials in each case. This allows a more thorough sensitivity analysis across different hyaluronan products. In all cases, the cost-effectiveness estimate is outside the realms of affordability to the NHS, and in one case is dominated by placebo. Sensitivity analyses on the individual estimates give a consistent message: that the efficacy would have to be three to five times higher than the estimates from the trials before reaching the standard threshold for cost effectiveness to the NHS.

Clinical trials do not suggest sub-groups of osteoarthritis patients may have greater benefit from HA therapies thereby improving cost effectiveness. A research recommendation is therefore made in section 10 to this effect.

The toxicity of intra-articular HA appears small. A small percentage of patients may experience a transient increase in pain following injection, and some get a frank flare of arthritis with marked effusion. As with any injection procedure there is a very small risk of infection.

RECOMMENDATIONS
R31.

Intra-articular corticosteroid injections should be considered as an adjunct to core treatment for the relief of moderate to severe pain in people with osteoarthritis.

R32.

Intra-articular hyaluronan injections are not recommended for the treatent of osteoarthritis. See Chapter 9 for the associated recommendation for future research.

Footnotes

*

Efficacy was assessed in terms of pain relief, function improvement and patient global assessment.

**

There was a high variability observed across the agents evaluated (different molecular weight, concentration, treatment schedules, and mode of production) and the heterogeneity of the RCTs included.

Copyright © 2008, Royal College of Physicians of London.

All rights reserved. No part of this publication may be reproduced in any form (including photocopying or storing it in any medium by electronic means and whether or not transiently or incidentally to some other use of this publication) without the written permission of the copyright owner. Applications for the copyright owner’s written permission to reproduce any part of this publication should be addressed to the publisher.

Cover of Osteoarthritis
Osteoarthritis: National Clinical Guideline for Care and Management in Adults.
NICE Clinical Guidelines, No. 59.
National Collaborating Centre for Chronic Conditions (UK).

NICE (National Institute for Health and Care Excellence)

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...