In clinical investigations of osteoarthritis it has been recommended that efficacy should be measured by pain, function, and global measures evaluated by the patient.^1,2 According to OARSI and OMERACT guidelines, several instruments may be considered within each domain provided they have adequate reliability, validity, and responsiveness.² For example, pain could be measured by a global measure of pain, the WOMAC (Western Ontario and McMaster Universities) pain subscale or as an average of the four questions of the Lequesne functional severity index that focus on pain (pain at night, pain in the standing position, pain during walking, and pain while switching from a sitting to a standing position).^1,3,4 Function in osteoarthritis trials is most commonly measured by WOMAC or the Lequesne functional severity index.^3,4 In addition, a generic instrument such as the 36 item short-form health survey (SF-36) is often used to measure health related quality of life in osteoarthritis trials.⁵

Osteoarthritis is a heterogeneous disorder.⁶ Observing an effect of a treatment in one major joint does not necessarily mean that it will be equally effective in another joint. Guidelines for clinical trials recommend that patients with knee osteoarthritis should be evaluated separately from patients with hip osteoarthritis.⁷ However, this increases the costs for clinical trials. It would be possible to decrease the time and cost for trials if patients with either knee or hip osteoarthritis could be combined in the same analysis. However, this would require that the response characteristics to treatment of these two joints were the same.

The present study represents an exploratory ad hoc analysis of data from a large multinational randomised clinical trial including both knee and hip osteoarthritis, comparing the gastrointestinal safety and efficacy of the COX inhibiting nitric oxide donator (CINOD) AZD3582 with naproxen.⁸ The primary end point of that study was the six week incidence of endoscopic gastroduodenal ulcers. The purpose of the present study was to evaluate the relative improvement in hip and knee osteoarthritis during treatment with naproxen, as measured by the disease specific instrument WOMAC and the generic measure SF-36. For this, the effect size was separately determined for hip and knee to see whether or not the knee and hip data could be combined for trial analysis.

METHODS

RESULTS

Naproxen 500 mg twice daily was given to 403 patients (280 with knee osteoarthritis and 123 with hip osteoarthritis), and placebo to 108 patients (75 knee, 33 hip). The treatment groups were well balanced at baseline for demographic characteristics (table 1⇓).

Further, there were no differences in baseline WOMAC scores between treatment groups or between those with hip or knee osteoarthritis (table 2⇓). The baseline scores for all SF-36 domains were similar between treatment groups and between those with knee or hip osteoarthritis, as were the proportions of subjects taking rescue medication for osteoarthritis related pain during the wash out period (data not shown). Compliance of ⩾70% of the prescribed drug intake for the knee was 96.6% for naproxen and 96.2% for placebo, and for the hip, 96.9% for naproxen and 91.9% for placebo. Rescue medication was taken by 88% of knee patients in the naproxen group and 92% in the placebo group, and by 90% of the hip patients in the naproxen group and 92% in the placebo group.

As shown in table 2⇑, for naproxen treatment improvement at week 6 was between 4 and 7 mm greater for the knee than for the hip for all three WOMAC subscales (pain, Δ = 4.7 mm (p = 0.03); stiffness, Δ = 6.6 mm (p = 0.004); function, Δ = 4.8 mm (p = 0.016)). Placebo response showed the same trend, but with non-significant statistical differences between knee and hip (pain, Δ = 2.9 mm; stiffness, Δ = 0.1 mm; function, Δ = 2.3 mm). The effect size was about 0.8 for the knee and between 0.5 and 0.6 for the hip for the naproxen treated group.

For the SF-36 domains bodily pain and role–physical, knee patients treated with naproxen improved by 4.6 units (p = 0.033) and 10.3 units (p = 0.014) more than hip patients, respectively, with no statistically significant difference for placebo treatment between knee and hip patients.

The patients taking naproxen rated the treatment as good or very good in 75% of the cases for the knee and 66% for the hip. For placebo, the numbers were 50% for the knee and 35% for the hip. The investigators rated the naproxen treatment as good or very good for the knee in 73% of the cases, and 64% for the hip. For placebo, the numbers were 47% for the knee and 38% for the hip.

DISCUSSION

To our knowledge, this is the first within-trial comparison of the improvement of knee and hip osteoarthritis during treatment with naproxen. As the mean of the baseline scores were lower for the knee group than for the hip group, the greater improvement found in the knee patients during treatment with naproxen is unlikely to be caused by a regression to the mean. Is this a true difference in naproxen efficacy between hip and knee, or does the WOMAC questionnaire have a higher responsiveness for knee osteoarthritis compared with hip osteoarthritis? The fact that a higher proportion of both patients and investigators rated the overall treatment as good or very good for the knee than for the hip supports the first explanation. The subscales of SF-36 most relevant to musculoskeletal problems and most comparable to the content of WOMAC are physical functioning, role–physical, and bodily pain. Seeing the same pattern of improvement in two of these three subscales as in WOMAC implies the difference between knee and hip being a treatment effect and not a psychometric phenomenon with regard to the WOMAC instrument.

An effect size of 0.5, which was found for the hip in the present study, is considered a moderate change, while an effect size of 0.8 as found for the knee is considered large.^11–,¹³ However, the placebo response was not subtracted from these effect size values.

The mean knee–hip differences in the WOMAC subscales pain (4.7 mm), stiffness (6.6 mm), and physical functioning (4.8 mm) were less than the proposed minimal perceptible clinical improvement (MPCI) of 10 mm for WOMAC.¹⁴ However, in that study the baseline values were considerably higher than in the present study, at 65.0 mm VAS v 44.2 mm for pain (knee, naproxen; table 2⇑), 65.9 v 48.9 mm for stiffness, and 63.9 v 47.1 mm for physical functioning, and it has been shown that higher baseline scores require larger raw changes to represent a clinically important difference.¹⁵ Further, to be included in that study, the patients had to have at least a 15 mm increase in the pain walking score after the washout, and a washout score of ⩾40 mm.¹⁴ Neither of this was required in the present study.

The cited studies deal with the concept of how a patient perceives a change during the course of a treatment.^14,15 This might not be the same as a perceived difference between treatments. Therefore, it is difficult on the basis of those studies to draw any firm conclusions as to whether the difference between hip and knee in the present study is clinically relevant or not, but using the 11-point numerical rating scale, it was concluded that a pain reduction of approximately 30% represents a clinically important difference.¹⁵ In the present study the reduction in WOMAC knee pain for naproxen was 38% (16.6/44.2) and in WOMAC hip pain 26% (12.3/47.3). Using the suggested cut off point of 30% would imply that the reduction in pain was clinically important for the knee but not for the hip.

The results of the present study strongly influence trial power and number of patients needed per treatment arm in clinical trials. Based on the effect sizes for pain, 108 subjects with hip osteoarthritis compared with only 54 subjects with knee osteoarthritis would need to be included in a clinical trial to determine a significant difference against baseline with 80% power. The findings support the recommendation that trials concerning efficacy of treatment for osteoarthritis of the knee and hip should be stratified with respect to target joint or evaluated separately,⁷ and they warrant further investigation concerning the clinical relevance for the individual patient.