Some key considerations regarding the conduct and interpretation of clinical trials for pharmacological therapies in OA patients have been reviewed by Buchanan and Kean. Although a well-conducted randomized control trial (RCT) ranks highly in the hierarchy of clinical evidence, care should be taken when attempting to generalise results of possibly unrepresentative study groups to the broader population. There are some common explanations for why a study sample may not adequately represent the broader patient base. Patients who agree to participate in such studies often have different characteristics from those who decline. Additionally, elderly patients with other severe illnesses in addition to OA are often excluded from trials.
Publication bias is also an issue that must be acknowledged with OA therapy RCTs. Trials with negative results are less likely to be published, thus resulting in an inflated view of a treatment's success and possibly a threat to the validity of meta-analyses. One must also be aware of the potentially variable OA diagnostic criteria across studies. Given the lack of a clear diagnostic test for OA, inclusion inconsistency between studies is common. This is likely acceptable in the context of interpreting an individual study (as long as internal consistency is maintained) but poses comparability concerns when looking at multiple studies. An additional consideration is the possible misattribution of pain to the condition of OA. This is most common in studies of hand OA (where carpal tunnel syndrome may be a strong contributor to the assessed pain), but is also a concern in studies of the knee and hip. It is rare that nerve conduction or EMG tests are done to exclude the possibility of confounding radiculopathy in enrolled subjects.[6,25] Another important issue to take into account is that quantitative analysis of improvement in RCTs deals with the average of treatment groups as a whole, rather than at the level of the individual patient.[6,26] Clinicians must bear this in mind when applying the findings of an RCT to the care of individuals.
Literature retrieval and analysis of trials
In preparing this review, the scope of our analysis was focused on ibuprofen therapy studies in patients with OA of the knee. This is the most common large joint that is affected by OA and is generally the simplest to quantify in terms of pain relief and functionality during clinical trials. The goal was to review and synthesize the findings of studies addressing the efficacy of ibuprofen in the management of knee OA.
The literature was searched thoroughly in duplicate to ensure the inclusion of all relevant studies. A search of PubMed was conducted using the following search terms: ‘osteoarthritis AND ibuprofen AND knee’ with the limits set to ‘RCTs OR clinical trials’ and ‘human subjects’. This yielded a list of 69 articles, which were then retrieved and their abstracts were screened. Studies were included if they followed the design of a double-blind RCT, they contained a placebo group for comparison, they involved patients with symptomatic OA of the knee, and they employed ibuprofen in at least one treatment group. Similar searches were conducted in the Medline and Embase databases (1950 to present). The reference lists of articles were scanned for the presence of any other relevant studies. Relevant articles were included in the final list, which consisted of 10 studies, to be analysed in full-text. Note: trials limited to the knee that fit our criteria were scarce and so exception was made for studies where the majority of subjects had knee OA (usually > 80% of subjects) but some had hip OA. Upon retrieval and full review, five of these studies were deemed to fit the given criteria and were included in the analysis. Two additional studies identified upon hand-searching of reference lists were included.
Efficacy in knee osteoarthritis
Table 2 lists the relevant randomized controlled trials and selected data examining the efficacy of ibuprofen in treating OA of the knee compared with placebo. The assessments were taken at baseline and then at the indicated follow-up times, and a quantification of improvement is presented for both ibuprofen and placebo groups. Various scales were used and different endpoints assessed, the most common being pain and physical function subscales of the Western Ontario and McMaster University Osteoarthritis Index (WOMAC). The pain and physical function subscales are abbreviated as WOMAC-PS and WOMAC-PFS, respectively, and patient global assessment of disease status is abbreviated as PGADS. A negative change from baseline indicates an improvement in a given endpoint. Dosages are shown in mg/day; in most studies, this dosage is divided equally over three administrations per day (t.i.d).
Table 2. Studies examining the efficacy of ibuprofen vs placebo in treatment of knee osteoarthritis
| || || ||12 weeks mean change from baseline (95% CI)|
|Knee or hip (>83% knee)||Puopolo et al. (2007) (n = 548)||Placebo||(All outcomes scored on a 100 mm visual analogue scale (VAS))|
|WOMAC-PS: −16.47 (−20.55, −12.40)|
|WOMAC-PFS: −13.56 (−17.59, −9.54)|
|PGADS: −17.85 (−22.41, −13.29)|
|Ibuprofen (2400 mg/day)||WOMAC-PS: −24.1 (−27.2, −20.99)|
|WOMAC-PFS: −20.09 (−23.87, −17.72)|
|PGADS: −25.97 (−29.39, −22.54)|
|Statistically significant difference (ibuprofen compared with placebo)?||P < 0.002 for all three outcome subscales.|
| || || ||12 weeks (mean change from baseline)|
|Knee or hip (primarily knee)||Wiesenhutter et al. (2005) (n = 528)||Placebo||Only the range of mean changes from baseline were provided for the WOMAC-PS, PFS, and PGADS instruments collectively:|
|Range from: −16.53 to −13.55|
|Changes specific to each subscale are not reported.|
|Ibuprofen||Range from: −26.53 to −22.97|
|Changes specific to each subscale are not reported.|
|Statistically significant difference?||P < 0.01 for each outcome measure|
| || || ||1 week (mean change from baseline)|
|Knee (non-prescription dose)||Schiff et al. (2004) (n = 298)||Placebo||Mean symptom score reduced by 20–25% from baseline (0–4 point categorical pain scale).|
|Symptoms measured: pain at rest, pain on passive motion, pain on weight bearing, stiffness after rest, day pain, night pain, 50-foot walk. Each was measured on a 0–4 point categorical pain scale, except the walk which was timed in seconds.|
|Ibuprofen (1200 mg/day)||Mean symptom score reduced 30–45% from baseline (0–4 point categorical pain scale).|
|Symptoms measured: pain at rest, pain on passive motion, pain on weight bearing, stiffness after rest, day pain, night pain, fifty-foot walk. Each was measured on a 0–4 point categorical pain scale, except the walk which was timed in seconds.|
|Statistically significant difference? (ibuprofen compared to placebo)||Pain at rest: P = 0.077 (not significant)|
|Pain on passive motion: P < 0.05|
|Pain on weight bearing: P < 0.01|
|Stiffness after rest: P < 0.01|
|Day pain: P < 0.01|
|Night pain: P = 0.193 (not significant)|
|50-Foot walk time: P < 0.05|
| || || ||6 weeks (mean change from baseline)|
|Knee or hip||Day et al. (2000) (n = 323)||Placebo||Outcomes scored on 100 mm visual analogue scale|
|Pain when walking on a flat surface (WOMAC question 1): −18.92 (−23.72 to −14.12)|
|WOMAC-PS: −11.89 (−15.98 to −7.80)|
|WOMAC-PFS: −8.76 (−12.72 to −4.79)|
|PGADS: −10.02 (−14.6 to −5.45)|
| || ||Ibuprofen (2400 mg/day)||Outcomes scored on 100 mm VAS|
|Pain when walking on a flat surface (WOMAC question 1): −33.55 (−36.26 to −30.84)|
|WOMAC-PS: −22.89 (−25.21 to −20.58)|
|WOMAC-PFS: −18.06 (−20.3 to −15.82)|
|PGADS: −25.28 (−27.87 to −22.69)|
| || ||Statistically significant difference? (Ibuprofen compared with placebo)||P ≤ 0.009 for all four outcome measures (ibuprofen versus placebo)|
| || || ||3 weeks after administration (change from baseline)|
|Hip or knee (36 subjects knee, 20 subjects hip)||Bliddal et al. (2000) (n = 56) (crossover trial)||Placebo||Pain Visual Analogue Scale (100 mm): 0 mm change from baseline; 95% CI (−3, 4)|
|Ibuprofen (1200 mg/day)||Pain Visual Analogue Scale (100 mm) Ibuprofen: −15 mm change from baseline; 95% CI (−23, −7.5)|
|Statistically significant difference? (Ibuprofen compared with placebo)|| |
| || || ||1 week after administration|
|Knee||Sacchetti et al (1978) (n = 24) (balanced incomplete block design- not a true randomised control trial)||Placebo||Pain scored on 0–4 scale (4 = very severe; 0 = no pain)|
|Daytime pain at rest (baseline): 2.44 ± 0.33|
|Daytime pain at rest (1 week): 1.81 ± 0.32|
|Night pain at rest (baseline): 2.56 ± 0.33|
|Night pain at rest (1 week): 1.94 ± 0.32|
|Ibuprofen (900 mg/day)||Pain scored on 0–4 scale (4 = very severe; 0 = no pain)|
|Daytime pain at rest (baseline): 2.75 ± 0.32|
|Daytime pain at rest (1 week): 1.25 ± 0.28|
|Night pain at rest (baseline): 3.00 ± 0.30|
|Night pain at rest (1 week): 1.62 ± 0.31|
|Statistically significant difference? (Ibuprofen compared with placebo)||P < 0.01 when ibuprofen group compared with placebo group daytime pain improvement.|
|P < 0.01 when ibuprofen group compared with placebo group night pain improvement.|
The studies vary significantly in design, measures and administration protocol. It is thus prudent to examine the various elements of the studies in combination with their tabulated results.
Puopolo et al. and Wiesenhutter et al. conducted large-scale multi-centre RCTs with three treatment arms (placebo, ibuprofen (2400 mg/day → 800 mg t.i.d) and etoricoxib), employing the highly validated WOMAC pain and physical function subscales, along with the PGADS. Both studies admitted subjects who had both clinical and radiographic evidence of OA, and fit within ARA functional class I-III over the previous six months. Puopolo et al. concluded that for each of these primary endpoints ibuprofen was superior to placebo, with statistical significance achieved (P < 0.002). In the ibuprofen group, 70.1% of subjects achieved at least minimal clinically important improvement on the WOMAC-PS compared with 55.1% in the placebo group. Clinically important improvement was defined as >15% improvement from baseline on the 100 mm scale used.[28,30] Subjects were assessed at two weeks and twelve weeks after treatment administration, and the results were only provided for the twelve-week point. Wiesenhutter et al., although less thorough in reporting results for each endpoint, showed similar results in support of ibuprofen's significant superiority over placebo (P < 0.01). The fact that ibuprofen showed significant efficacy across multiple validated endpoints (WOMAC subscales, PGADS), appears to provide strong evidence for its usefulness in managing OA.[28,29]
In the study by Puopolo et al. a greater proportion of the placebo group (18.9%) than the ibuprofen group (7%) dropped out due to lack of efficacy. A similar disproportionate drop-out rate due to lack of efficacy was seen when comparing the placebo and ibuprofen groups in the study by Wiesenhutter et al. (29.8% and 14%, respectively). This outcome-dependent drop-out rate raises flags for potential bias and skewed results in both studies. The placebo group efficacies reported in the studies may in fact be inflated due to the drop-out of subjects with poor efficacy.[28,29]
In a number of studies, paracetamol was provided as a rescue medication for unmanaged pain in subjects. The amount of rescue paracetamol required by each group over the course of the study would then be counted and analysed as a secondary endpoint in itself.[28–33] This could provide clues as to the efficacy of a particular treatment. In all studies where this was measured, rescue paracetamol consumption was greater in the placebo group than in the ibuprofen group; however, this difference was only statistically significant in two of the four studies that reported it.[28,33]
Numerous studies have noted a distinct trend whereby the majority of the improvement in all primary endpoints is seen within the first two weeks following administration of treatment.[28,29,32] This improved state is maintained over the remainder of the 12-week follow-ups, with only small further improvement. Thus, it may be fair to compare the results from six-week studies and twelve-week studies, because after the two-week mark there is minimal change and the drug effect seems to have reached its peak potential. However, in the study by Schiff et al. follow-up time was only one week, and as such it may not be appropriate to compare efficacy from that study with those of longer duration.
Schiff et al. assessed ibuprofen dosages of 1200 mg/day (400 mg t.i.d.) and dealt exclusively with subjects suffering from OA of the knee. Forgoing the use of validated standard scales, such as WOMAC, the study instead assessed seven symptoms on a categorical (0–4) scale at baseline and follow-up, noting the changes in each. The assessed symptoms include knee joint pain at rest, pain on passive motion, pain on weight bearing, severity of joint stiffness, pain severity the previous day, pain severity the previous night, and time taken to walk 50 feet. The authors noted ibuprofen was clinically effective in managing knee OA, with a 30–45% reduction in mean symptom score versus the placebo group's 20–25% score reduction. However, the authors did not define any threshold for clinical significance. The ibuprofen group achieved statistically significant (P < 0.05) improvement over the placebo group on five of the seven symptoms (pain at rest and night pain were exceptions). It is also notable that this trial used OTC dosages (1200 mg/day, i.e. 400 mg t.i.d.) of ibuprofen. The fundamental differences in the trial design and dosages make this study difficult to compare and integrate with the others previously discussed.
The crossover RCT by Bliddal et al. also yielded data that shows promise for the efficacy of 1200 mg/day (400 mg t.i.d) ibuprofen in the management of knee OA. Treatment with ibuprofen yielded a 15-mm decrease (improvement) from baseline on a 100-mm pain visual analogue scale. The statistical significance of ibuprofen's treatment effect compared with placebo was high (P < 0.0001). However, the study was fraught with questionable methodological elements. There were three treatments (ginger extract, ibuprofen and placebo) and subjects were randomized as to which treatment they would receive first. The RCT involved three consecutive treatment periods, each three weeks in duration. Although there was an initial washout period of one week before commencement of the study, there was no washout period after subjects finished one treatment period and were started on a different treatment. There is significant concern that carry-over effects from the previous treatment periods may skew findings in crossover RCTs (when there is no appropriate wash-out period). Eleven subjects were excluded secondarily from the study analysis, three of whom dropped out due to lack of effect; the study did not note to which group the excluded subjects belonged. This series of concerns should prompt careful consideration when interpreting and synthesizing this study's results with the others. Another crossover clinical trial by Sacchetti et al. was performed in 1978 using a balanced incomplete block design, and used a patient-scored (0–4) pain scale to assess daytime and night pain in women with knee OA. The study yielded statistically significant indications of improvement at one week in ibuprofen groups compared with placebo groups (P < 0.01) for both day and night pain. Other similar measured outcomes yielded congruent results. However, washout periods were not provided between treatment periods, and sufficient validation was not provided for the measurement scales used; these are important methodological notes that should indicate the need for caution in interpretation.
In addition to the patient-assessed endpoints in the clinical trials presented, there is recent physiological evidence supporting the effectiveness of ibuprofen in treating OA. NSAIDs have often been considered as symptom-modifying drugs with a debatable structural effect in the management of OA. The pathogenesis of OA has been highlighted earlier in the paper, with emphasis on progressive breakdown of the cartilage and synovium in the affected joint. Molecular markers have been identified whose urinary levels reflect turnover of cartilage and synovial tissue. These markers are c-telopeptide fragments of type II collagen (CTX-II) and glucosyl galactosyl pyridinoline (Glc-Gal-PYD), respectively.[35,36] It has been demonstrated that in subjects with knee OA accompanied by prominent inflammation, a group receiving placebo experienced increased levels of these CTX-II and Glc-Gal-PYD urinary markers over the course of 4–6 weeks. In a parallel group, subjects who were treated with 2400 mg/day of ibuprofen, levels of urinary CTX-II and Glc-Gal-PYD over the same period of time did not rise as much. These results indicate that ibuprofen may help prevent the high rate of cartilage and synovium catabolism characteristic of OA progression.
While this review has addressed the pharmacokinetics and clinical efficacy of only oral formulations of ibuprofen, another common method of administration is in the form of topical creams or gels, especially in treatment of knee OA. Trnavsky et al. and Rovensky et al. examined topical administration of 5% ibuprofen cream for treatment of knee OA in double-blind, placebo controlled RCTs and concluded that topical ibuprofen's efficacy is clinically and statistically relevant compared with placebo cream. These studies have been confirmed in a series of studies and evaluations by Underwood's group.[40–44] In essence, the conclusions from these large-scale evaluations in the general practice setting in southern England are that topical ibuprofen for knee OA, especially long term, has advantages over oral ibuprofen in that there are fewer side effects.