Article Text


NSAIDs, including coxibs, probably do more harm than good, and paracetamol is ineffective for hip OA
  1. J M Bjordal1,
  2. A E Ljunggren1,
  3. A Klovning1,
  4. L Slørdal1
  1. 1Department of Public Health and Primary Health Care, University of Bergen, 5018 Bergen, Norway
  1. Correspondence to:
    Dr J M Bjordal
  1. M Doherty2,
  2. W Zhang2
  1. 2University of Nottingham, UK

Statistics from

New EULAR treatment guidelines in hip osteoarthritis (OA) were recently reviewed.1 The recommendations are allegedly based on top level scientific (grade Ia) data, and advocate the use of paracetamol (acetaminophen) and non-steroidal anti-inflammatory drugs (NSAIDs), including selective cyclo-oxygenase 2 inhibitors (coxibs), in OA of the hip. We interpret the underlying data differently.

The review1 states that owing to a lack of studies evaluating paracetamol in hip OA alone, recommendations rely on studies that also include OA conditions in other joints. A review of four placebo controlled trials demonstrated an effect size (ES) of 0.21 (95% confidence interval (CI) 0.02 to 0.41).2 Two small (n = 50 and n = 60) trials could not be pooled for lack of data. Thus, ES calculations are from two trials, where the smallest reported no significant effect, leaving us with a final sample of 387 patients. The EULAR review does not mention that a recent large (n = 779) knee OA trial reported a non-significant effect of paracetamol corresponding to a minuscule 0.8 mm on a visual analogue pain scale (VAS).3 Together, available data clearly demonstrate a lack of effect of paracetamol in large joint OA.

Concomitant with the withdrawal of rofecoxib (Vioxx), and an increasing awareness that other coxibs may also induce serious cardiovascular toxicity, EULAR recommends the use of coxibs as second line treatment after paracetamol for patients at risk for gastrointestinal adverse effects.1 However, no references to randomised controlled trials (RCTs) or systematic reviews on efficacy are given. To our knowledge only two RCTs have evaluated the use of coxibs in hip OA alone. A trial funded by Pfizer with valdecoxib found a difference over placebo of only 1.6 from a baseline value of 10.8 on the WOMAC subscale of pain,4 which corresponds to a modest 7.9 mm on a VAS. Similar results were found for celecoxib in another Pfizer-funded trial.5 These results are well below the threshold of 15 mm on a VAS, which defines minimal clinically important differences.6 On the other hand, pivotal questions about the safety of coxibs remain unresolved. In our view, this calls for caution and suggests that the use of these drugs should be suspended until comprehensive safety data for all coxibs are available.

The EULAR review states that NSAIDs are effective, but that adverse effects may counter their benefit.1 This conclusion is based on one systematic Cochrane review,7 which allegedly included 14 placebo controlled trials. The ES for pain relief was 0.69 (95% CI 0.12 to 1.26). However, the cited Cochrane review does not include a single study published after 1994. It is not based on 14 placebo controlled trials, but on 14 placebo controlled comparisons. The number of placebo controlled trials included was three (with 9, 146, and 104 patients, respectively) of 2–8 weeks’ duration. The given ES of 0.69 was calculated from the results of a single 8 week trial (n = 146),8 which reported a mean difference over placebo corresponding to 10 mm on a VAS.

To our knowledge, no further hip-specific RCTs with NSAIDs have been published except the two coxib trials mentioned above, where groups of patients treated with naproxen were included. The efficacy of naproxen in these trials was not significantly different from that of the coxibs. A recent analysis of 23 trials on the efficacy of NSAIDs (including coxibs) in knee OA demonstrated a small effect of treatment on pain corresponding to 10.1 mm on a VAS,9 which is too modest to be of any clinical significance in the average patient. No long term trials (>13 weeks) have demonstrated beneficial effects of NSAIDs over placebo in hip or knee OA. Although all NSAIDs may induce serious adverse effects, relevant safety issues cannot be dealt with by short term studies. In our view, it is highly likely that long term use of NSAIDs, including coxibs, does more harm than good in patients with OA of the hip.

We support the EULAR guidelines initiative as a positive contribution to evidence based care of patients with arthritis. However, we believe that some of the recommendations are erroneous and misleading.

In our view, the available information unequivocally demonstrates that standard pharmacological interventions with paracetamol and NSAIDs, including coxibs, cannot be recommended for the osteoarthritic patient.


Authors’ reply

We thank Dr Bjordal for his critical comments on paracetamol and NSAIDs (both conventional and coxibs), which were included in our recent evidence based recommendations for hip osteoarthritis (OA).1 We agree that direct research evidence for paracetamol in the treatment of hip OA is insufficient as already described in the paper.

Unlike other guidelines, the EULAR evidence based recommendations are generated according to evidence from both clinical expertise and research. We first used a Delphi consensus approach to generate the key clinical propositions for the management of hip OA. We then went on to search systematically for research evidence for these propositions. Paracetamol had been proposed through the Delphi approach, but the research evidence to support its use in hip OA is lacking.

The results were reported to the committee and the strength of recommendation was then determined based both on clinical expertise and research evidence (that is, on all forms of “evidence” in accord with evidence based principles). It is a group decision-making exercise with multidimensional trade-off procedure, taking into account efficacy, side effects, cost effectiveness, logistics of delivery, and patient acceptance and tolerability, not just a simple systematic review of research evidence for efficacy from randomised controlled trials (RCTs) for paracetamol. As a result, paracetamol was still recommended given the evidence from clinical experience (efficacy, safety, availability, etc) and indirect research evidence from knee OA. Not surprisingly, because of this weakness, the strength of recommendation for paracetamol is quite low (table 8 in Zhang et al1).

We know of the large scale RCT in knee OA,2 but this was reported after the EULAR recommendations were submitted. The caveats of that trial and its possible influence on the use of paracetamol have been discussed by us elsewhere.3 Clearly, such data will be included for consideration when we next update the EULAR recommendations.

We appreciate the comments made in relation to NSAIDs, including coxibs, in particular the effect size taken from the Cochrane review,4 which was calculated from one trial as you rightly point out. Two other trials funded by Pfizer were not included for assessment because of the strategies of the EULAR Taskforce—evidence will only be included for assessment when a higher grade of evidence is not available. This may be arbitrary but reflects a consensus for the level of evidence, where systematic review is graded as Ia, and allows us to consider not only quality but also the contribution of any single study in the context of the entire literature pool.

As to what should be considered a clinically meaningful improvement, we used effect size as a general outcome in order to cross-compare the treatment programmes. It is suggested that 0.2 is mild, 0.5 is moderate, and over 0.8 is a large clinical effect. Although the threshold of 15% pain reduction may be useful for single programmes, it is less useful for the guideline development, where cross-comparisons between programmes are normally required.


View Abstract

Request permissions

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