WOMAC

WOMAC in Osteoarthritis Trials and Economic Evaluation

Trials of treatments for hip and knee osteoarthritis often record the WOMAC as their patient-reported outcome, and an economic evaluation built on such a trial needs a way to turn WOMAC scores into QALYs. Wailoo, Hernández Alava and Escobar Martinez tackled that problem with 7,072 observations from Spanish patients and found that the stiffness subscale has limited relationship to EQ-5D, whereas functional disability and pain are strong predictors. The three subscales therefore carry different weight once the index enters a cost-utility analysis, even though they sit side by side on the questionnaire. The name refers to the osteoarthritis index developed by Bellamy and colleagues at the University of Western Ontario, which exists in several versions that differ in their response format.

Development inside a trial of two anti-inflammatory drugs

Bellamy, Buchanan, Goldsmith, Campbell and Stitt reported the validation of the WOMAC in the Journal of Rheumatology in 1988. They described it as a new multidimensional, self-administered health status instrument for patients with osteoarthritis of the hip or knee, and they validated it within a double-blind randomised controlled parallel trial of two nonsteroidal anti-inflammatory drugs. Their abstract reports that the pain, stiffness and physical function subscales fulfil conventional criteria for face, content and construct validity, reliability, responsiveness and relative efficiency.

The setting matters for interpretation. The WOMAC was designed as an evaluative instrument for detecting change in drug trials, with content limited to the hip or knee. Bellamy's own review of twenty years of experience with the index, published in 2002, noted that the same development work led to AUSCAN, a separate index for hand osteoarthritis, so hand osteoarthritis has its own instrument.

Items, subscales and response formats

Wailoo and colleagues describe the WOMAC as 24 questions, with responses to each given on a Likert scale of none (0), mild (1), moderate (2), severe (3) or extreme (4) problems. The questions form three subscales. Pain has 5 questions with a range of 0 to 20, stiffness has 2 questions with a range of 0 to 8, and functional impairment has 17 questions with a range of 0 to 68.

$$S_j = \sum_{i \in j} x_i, \qquad x_i \in \{0, 1, 2, 3, 4\}$$

where $S_j$ is the score for subscale $j$ (pain, stiffness or physical function), $x_i$ is the response to question $i$ and the sum runs over the questions in that subscale. Higher scores mean worse symptoms or more limitation. Adding the three subscale maxima gives 20 + 8 + 68 = 96 for the whole index.

A visual analogue version also exists. Kersten, White and Tennant tested the VAS version of WOMAC 3.0 in 221 patients with chronic hip or knee pain awaiting joint replacement. Factor and Rasch analysis supported unidimensional pain and physical functioning scales that could be transformed to interval scaling, and the stiffness scale fitted the Rasch model. Scores from the two formats sit on different scales, so a pooled analysis has to keep them apart or convert them explicitly.

Mapping WOMAC subscales to EQ-5D utilities

The WOMAC has no preference weights, so its scores cannot serve as a health state utility value. When an osteoarthritis trial collected the WOMAC without a preference-based measure, a mapping function can predict EQ-5D values from the subscale scores, and the general design choices are discussed under mapping study.

A WOMAC result can enter an economic evaluation in two ways. As the effectiveness measure of a cost-effectiveness analysis, the WOMAC gives a cost per point of improvement on a subscale, which is simple to compute but cannot be judged against a threshold set per QALY or compared with results in other diseases. Through a mapping to a preference-based measure, the same scores support a cost-utility analysis and a comparison with other uses of the same budget, at the price of the prediction error that every mapping carries.

Wailoo, Hernández Alava and Escobar Martinez published such a mapping in 2014. They used 7,072 observations from a large registry of Spanish patients with knee or hip osteoarthritis who completed both the WOMAC and the EQ-5D, and applied the UK EQ-5D tariff. Their mixture model was built on distributions bespoke to the EQ-5D UK value set, and a five-class model fitted the observed data closely at all levels of disease severity. The observed data and the estimated model were very similar for the mean (0.542 and 0.542), the median (0.620 and 0.636) and the percentage of observations at full health (15 and 14.8). The authors concluded that EQ-5D can be reliably estimated from WOMAC subscale scores without any systematic bias.

Two features of this mapping shape its use. The predictions are EQ-5D-3L values on the UK tariff, so an analysis that needs a different EQ-5D version or value set needs a different mapping. The weak link between stiffness and EQ-5D also means that a treatment whose main effect is on stiffness will gain few mapped QALYs, One interpretation, not tested in the paper, is that the EQ-5D descriptive system has no dimension for stiffness, so the weak link need not mean that patients place little value on it.

Minimal clinically important improvement in the WOMAC

Tubach and colleagues estimated the minimal clinically important improvement (MCII) for WOMAC function in a prospective multicentre four-week cohort study of 1,362 outpatients with knee or hip osteoarthritis. The MCII for absolute change, with relative change in brackets, was −9.1 (−26.0%) for knee osteoarthritis and −7.9 (−21.1%) for hip osteoarthritis. The MCII was affected by the initial severity of symptoms but not by age, disease duration or sex.

The relative figures are easier to carry into a model. An absolute threshold depends on the scale on which the function subscale is reported, while a percentage change from baseline does not change when the subscale is rescaled by a constant factor. A responder definition built on the minimal clinically important difference should still state the joint, the baseline severity and the version of the WOMAC it assumes.

Illustrative worked example with one knee osteoarthritis patient

All numbers in this example are illustrative. A patient with knee osteoarthritis completes the Likert version before and after a twelve-week treatment.

1. Baseline subscale scores. The five pain answers are 3, 3, 2, 2 and 2, giving 3 + 3 + 2 + 2 + 2 = 12 out of 20. The two stiffness answers are 3 and 2, giving 5 out of 8. The seventeen function answers sum to 40 out of 68. The total is 12 + 5 + 40 = 57.

2. Follow-up subscale scores. Pain answers of 2, 1, 1, 1 and 1 give 6, stiffness answers of 2 and 1 give 3, and the function answers sum to 28. The total is 6 + 3 + 28 = 37.

3. Relative change in function.

$$\frac{F_1-F_0}{F_0} = \frac{28-40}{40} = -0.30$$

where $F_0$ is the baseline function score and $F_1$ is the follow-up score. The 30% improvement is larger than the 26.0% relative MCII that Tubach and colleagues reported for knee osteoarthritis, so the patient would count as improved under that threshold.

4. Mapped utility and cost per QALY. Suppose a mapping predicts EQ-5D utilities of 0.50 at baseline and 0.62 at follow-up, and the gain lasts one year. The QALY gain is 0.62 − 0.50 = 0.12. With an incremental cost of GBP 2,400,

$$\frac{\Delta C}{\Delta Q} = \frac{2400}{0.12} = 20000$$

where $\Delta C$ is the incremental cost in GBP and $\Delta Q$ is the QALY gain, so the treatment costs GBP 20,000 per QALY gained in this illustration.

The pain score halved while stiffness fell by 2 points, yet only the pain and function changes would move the mapped utility much, given the weak relationship Wailoo and colleagues found for stiffness. A trial report that highlights the WOMAC total therefore says less about the QALY gain than the subscale scores do.

How the WOMAC differs from the Oxford scores and the HAQ-DI

The WOMAC is a patient-reported outcome measure specific to hip and knee osteoarthritis. The Oxford Hip Score and the Oxford Knee Score are each confined to one joint, while a single WOMAC form is used for hip or knee osteoarthritis and keeps pain, stiffness and function apart.

The HAQ-DI was first validated in rheumatoid arthritis and measures whole-body physical function across eight categories of functioning in one index scored 0 to 3. The WOMAC asks about a target joint and reports pain separately, which suits trials of treatments aimed at that joint. Neither is preference-based, so both reach QALYs only through a mapping, and Hernández Alava and Wailoo are among the authors of published mixture-model mappings for each.

Limitations and common misreadings

Several mistakes recur when WOMAC results feed an economic model. Most arise from treating the index as one score on a fixed scale.

• Reading the total as a utility. A WOMAC total of 0 means no reported symptoms in the target joint, not full health, and a higher total means worse health, the opposite of a utility.
• Pooling different versions. Likert and visual analogue versions produce scores on different scales, so results from the two need to be kept apart or converted explicitly before they are combined.
• Treating raw scores as interval measures. Kersten and colleagues found responsiveness estimates lower once VAS data were transformed to interval scaling than with the raw ordinal data, so responsiveness judged on the raw scale may be overstated.
• Weighting the subscales equally. Stiffness has only two questions and, in the Wailoo mapping, little relationship to EQ-5D, so a change in the total can overstate or understate the change in mapped utility.
• Using one responder threshold for every patient. Tubach and colleagues found the MCII depends on baseline severity and differs between knee and hip osteoarthritis.
• Transferring a mapping without checking its population. The Wailoo mapping comes from Spanish patients scored with the UK EQ-5D-3L tariff, and its predictions should be tested for the population and EQ-5D version of the target analysis.

Sources

• Bellamy N. WOMAC: a 20-year experiential review of a patient-centered self-reported health status questionnaire. Journal of Rheumatology. 2002;29(12):2473-2476. https://www.jrheum.org/content/29/12/2473
• Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. Journal of Rheumatology. 1988;15(12):1833-1840. https://pubmed.ncbi.nlm.nih.gov/3068365/
• Kersten P, White PJ, Tennant A. The visual analogue WOMAC 3.0 scale: internal validity and responsiveness of the VAS version. BMC Musculoskeletal Disorders. 2010;11:80. doi:10.1186/1471-2474-11-80
• Tubach F, Ravaud P, Baron G, et al. Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: the minimal clinically important improvement. Annals of the Rheumatic Diseases. 2005;64(1):29-33. doi:10.1136/ard.2004.022905
• Wailoo A, Hernández Alava M, Escobar Martinez A. Modelling the relationship between the WOMAC osteoarthritis index and EQ-5D. Health and Quality of Life Outcomes. 2014;12:37. doi:10.1186/1477-7525-12-37