International Cartilage Repair Society

Transcription

1 OsteoArthritis and Cartilage (2007) 15, 266e272 ª 2006 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved. doi: /j.joca Measurement properties of the WOMAC LK 3.1 pain scale P. W. Stratford P.T., M.Sc.yz*, D. M. Kennedy P.T., M.Sc.xk, L. J. Woodhouse P.T., Ph.D.z{ and G. F. Spadoni P.T., M.Sc.x# y Department of Clinical Epidemiology and Biostatistics, School of Rehabilitation Science, McMaster University, Hamilton, ON, Canada z Division of Orthopaedic Surgery, Department of Surgery, Sunnybrook and Women s College, Health Sciences Centre, Toronto, ON, Canada x McMaster University, Hamilton, ON, Canada k Program Development for Hip and Knee Replacement, Holland Orthopaedic & Arthritic Centre of Sunnybrook Health Sciences Centre, Toronto, ON, Canada { School of Rehabilitation Science, McMaster University, Hamilton, ON, Canada # PRO Active Physiotherapy Clinic, Hamilton, ON, Canada Summary International Cartilage Repair Society Objective: The Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) is applied extensively to patients with osteoarthritis of the hip or knee. Previous work has challenged the validity of its physical function scale however an extensive evaluation of its pain scale has not been reported. Our purpose was to estimate internal consistency, factorial validity, testeretest reliability, and the standard error of measurement (SEM) of the WOMAC LK 3.1 pain scale. Method: Four hundred and seventy-four patients with osteoarthritis of the hip or knee awaiting arthroplasty were administered the WOMAC. Estimates of internal consistency (coefficient a), factorial validity (confirmatory factor analysis), and the SEM based on internal consistency (SEM IC ) were obtained. Testeretest reliability [Type 2,1 intraclass correlation coefficients (ICC)] and a corresponding SEM TRT were estimated on a subsample of 36 patients. Results: Our estimates were: internal consistency a ¼ 0.84; SEM IC ¼ 1.48; Type 2,1 ICC ¼ 0.77; SEM TRT ¼ Confirmatory factor analysis failed to support a single factor structure of the pain scale with uncorrelated error terms. Two comparable models provided excellent fit: (1) a model with correlated error terms between the walking and stairs items, and between night and sit items (c 2 ¼ 0.18, P ¼ 0.98); (2) a two factor model with walking and stairs items loading on one factor, night and sit items loading on a second factor, and the standing item loading on both factors (c 2 ¼ 0.18, P ¼ 0.98). Conclusion: Our examination of the factorial structure of the WOMAC pain scale failed to support a single factor and internal consistency analysis yielded a coefficient less than optimal for individual patient use. An alternate strategy to summing the five-item responses when considering individual patient application would be to interpret item responses separately or to sum only those items which display homogeneity. ª 2006 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved. Key words: Osteoarthritis, Outcome, Reliability, Validity, Knee, Hip. Introduction The prevalence of total hip and knee arthroplasty has increased substantially in the past decade, and with approximately 400,000 total knee replacements and 200,000 total hip replacements being performed in the United States each year, it is essential to have reliable and valid outcome measures to assess the effectiveness of this intervention 1. Many self-report health status measures were originally conceived to assess outcome in clinical trials 2e6. More recently, interest in applying health status measures in clinical practice to aid decision-making concerning individual *Address correspondence and reprint requests to: Professor P. W. Stratford, McMaster University, School of Rehabilitation Science, Institute of Applied Health Sciences, 1400 Main Street West, Room 430, Hamilton, ON, Canada L8S 1C7. Tel: x22523; Fax: ; stratfor@ mcmaster.ca Received 27 July 2006; revision accepted 2 September patients has emerged 7e10. When considering patients with osteoarthritis of the hip or knee and those subsequently progressing to total joint arthroplasty, the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) is frequently reported 3,4. This paper examines selected measurement properties of the WOMAC LK 3.1 pain scale and demonstrates how measurement error expressed in WOMAC pain scale points can be applied to assist clinical decision-making. At Outcome Measures in Rheumatoid Arthritis Clinical Trials III, pain, physical function, patient s global rating, and imaging tests were identified as core outcomes for patients with osteoarthritis 11. The WOMAC is a self-report functional status measure which professes to assess three health concepts, pain, stiffness, and physical functional 3,12. It has received extensive use over the past 20 years, however, recent work has consistently revealed the WOMAC lacks factorial validity 13e18. Specifically, the measure has been shown to have greater difficulty distinguishing between 266

2 Osteoarthritis and Cartilage Vol. 15, No the health concepts of pain and physical function than other assessment methods 13,14,19e21. Although one expects an association between the concepts of pain and physical function, this relationship appears inflated for the WOMAC and it is likely related to similar item phrasing and content on both scales 22. A consequence of this spuriously high association between pain and physical function was the WO- MAC s inability to detect deterioration in the functional status of patients immediately following hip or knee arthroplasty when performance measures and another self-report functional status measure detected significant deterioration in the ability of patients to move around 19. Although the validity of the physical function scale has been questioned, there is support for the WOMAC s ability to provide meaningful information concerning pain and change in pain 14,15,23e28. There are two scaling versions for WOMAC items: one applies a 100 mm visual analog scale (VAS) (VA3 series); the other uses a five-point Likert scale (LK3 series) 12. The WOMAC pain scale consists of five items: (1) walking on flat ground; (2) going up or down stairs; (3) at night while in bed; (4) sitting or lying; and (5) standing upright. Total WOMAC pain scale scores can vary from 0 to 500 when VA3 scaling is applied and from 0 to 20 when LK3 scaling is implemented. In many instances investigators have converted raw scores to a 100-point scale 9,10,29,30. The apparent rationale for this practice is that the conversion allows a comparison with other measures converted to the scale range. However, in the absence of common means and standard deviations this conversion does not achieve its intended goal. Moreover, from a resource perspective, this conversion takes time and requires a computational aid, both of which are likely to impede its application in clinical practice. For these reasons we believe it is of interest to present the measurement properties of the WOMAC LK is the same units as the original measurement. Our purpose was to estimate internal consistency, factorial validity, testeretest reliability, and the standard error of measurement (SEM) of the WOMAC LK 3.1 pain scale. A second goal was to illustrate how error estimates for the pain scale can be used to indicate the confidence in reported WOMAC LK 3.1 individual patient pain scores and change scores. Methods PATIENTS Our sample consisted of 474 patients with osteoarthritis of the hip or knee. All patients were awaiting primary total hip or knee arthroplasty. Two hundred and forty (51%) patients were female. Two hundred and fourteen patients (45%) had hip osteoarthritis of whom 95 were female. The sample s mean age (S.D.) and body mass index were 63.5 years (10.3) and 30.8 kg/m 2 (5.7), respectively. All patients contributed data to the estimation of internal consistency, factorial validity, and the SEM based on internal consistency (SEM IC ). Testeretest reliability and a SEM based on testeretest reliability (SEM TRT ) were calculated from a convenience subsample of 36 patients who were assessed on two occasions. Nineteen (52%) of these patients were female. Twenty patients (56%) had hip osteoarthritis of whom 10 were female. The testeretest reliability subsample s mean age and body mass index were 63 years (10.5) and 29.4 kg/m 2 (5.5), respectively. All patients provided informed consent and the project was approved by the institution s research and ethics committee. DESIGN A cross-sectional study design was applied to obtain estimates of internal consistency, factorial validity, and SEM IC. Data for the testeretest reliability sample were collected at two regularly scheduled clinic visits prior to surgery. The interval between assessments varied somewhat with the median interval being 106 days (first, third quartiles: 78, 133). Previous work has reported that similar patients were reasonably stable over a similar interval 31. MEASURE We administered the entire WOMAC LK 3.1, however, information from the pain scale only is of interest to this study. The pain scale consists of five items and all items are scored on a five-point scale (0e4) with higher scores representing greater levels of pain 32. Accordingly, pain scores can vary from 0 to 20. The five pain items are as follows: (1) walking on flat ground; (2) going up or down stairs; (3) at night while in bed; (4) sitting or lying; and (5) standing upright. ANALYSIS We calculated descriptive statistics and examined the distributions of scores for the entire sample and difference scores (Time 2 minus Time 1) for the testeretest reliability subsample. We calculated Pearson s correlation coefficients between WOMAC pain items. We calculated coefficient ap ffiffiffiffiffiffiffiffiffiffiffi as an index of internal consistency, and the SEM IC as s 1 a where s is equal to the standard deviation (S.D.) and a is coefficient alpha. Confirmatory factor analysis applying a maximum likelihood estimation method (AMOS 4.0, SmallWaters Corporation) was applied to assess the factorial structure of the pain scale 33. We conceptualized a one factor measurement model with uncorrelated error terms 34. The following indexes were applied to assess model fit: Comparative Fit Index (CFI); Relative Fit (RF); Tucker-Lewis Index (TLI); Root Mean-Square Error of Approximation (RMSEA); and the model fit chi-square and associated P-value 34. Although no single standard exists defining an acceptable fitting model, the following values are generally accepted; CFI, RF, and TLI values exceeding 0.95 indicate good fit; RMSEA values less than 0.05 represent good fit and values less than 0.08 indicate reasonable fit 34,35. We performed a randomized block analysis of variance and calculated patient, time, and residual error variance components. These variance components were used to calculate Shrout and Fleiss Type 2,1 intraclass correlation coefficients (ICC 2,1 ¼ patient variance/total p variance) 36. We also calculated the SEM ðsem TRT ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi residual varianceþ 37. Error estimates for score values were obtained by multiplying SEMs based on internal consistency and testeretest by 1.65, the Z-value associated with a two-sided 90% confidence interval. Minimal detectable change at a 90% confidence p level was calculated as follows: SEM TRT 1:65 ffiffi 2 38,39 p.the ffiffi 2 term acknowledges two measurements are being compared and the value 1.65 is the Z-value associated with a two-sided 90% confidence interval. The interpretation of minimal detectable change (MDC 90 ) is that 90% of stable patients will display random fluctuations equal to or less than this value when assessed on multiple occasions. Results Table I provides a summary of the WOMAC pain items descriptive statistics. Item means and standard deviations

3 268 P. W. Stratford et al.: WOMAC pain scale measurement properties Table I WOMAC pain items descriptive statistics Walk Stair Night Sit Stand Walk 2.1 (0.9)* Stair 0.70y 2.5 (0.9) Night (1.0) Sit (0.9) Stand (0.9) *Mean and S.D. ypearson s correlation coefficient. are reported on the main diagonal and inter-item correlations are presented in the other cells. Figure 1(a and b) display the distribution of WOMAC pain scores for the entire patient sample and reliability subsample, respectively. Table II displays summary statistics for the entire sample and testeretest reliability subsample. Also reported in Table II are estimates of SEM IC, SEM TRT, coefficient a, the Type 2,1 ICC, and MDC 90. The variance components used to calculate the Type 2,1 ICC were as follows: (1) patients 9.35; A Frequency B Frequency WOMAC Pain Score WOMAC Pain Difference Scores Fig. 1. (a) Distribution of WOMAC pain scores for entire sample (n ¼ 474). (b) Distribution of WOMAC pain difference scores for testeretest reliability sample (n ¼ 36) (2) time 0; and (3) error The zero variance associated with time term supports the premise that no systematic difference in pain intensity occurred between test and retest. The testeretest ICC was slightly less than the estimated values for coefficient a and SEM TRT was marginally greater than the estimates for SEM IC. Table III summarizes the fit statistics for the confirmatory factor analysis and Fig. 2 illustrates the one factor model including the factor loadings. Our hypothesized model of a single factor with uncorrelated error terms demonstrated a highly significant lack of fit. Specifically, modification indexes alerted us to significant correlations between at night while in bed and sitting or lying error terms and walking on flat ground and going up or down stairs error terms (modification indexes: nightesit 71.6, walkestair 29.8). To provide an insight into the factorial structure of the pain scale, we constructed two modified models. The correlated error model specified correlations between the night and sit error terms, and the walk and stair error terms [Fig. 3(a)]; the two factor model conceptualized walk and stair items loading on one factor, night and sit items loading on a second factor, and standing loading on both factors [Fig. 3(b)]. The fit statistics for both modified models were identical. Discussion Table II Summary statistics for WOMAC pain scores Entire sample (n ¼ 474) Testeretest sample (n ¼ 36) Time 1 score (mean, s*) 9.98 (3.66) 7.75 (3.66) Internal consistency measurements Coefficient a (95% CIz) 0.84 (0.82, 0.87) NAy SEM (95% CI) 1.48 (1.33, 1.57) NA Testeretest measurements Time 2 score (mean, s) NA 7.97 (3.32) Difference score (mean, s) NA 0.22 (2.39) Intraclass correlation NA 0.77 (0.59, 0.88) coefficient (95% CI) SEM (95% CI) NA 1.69 (1.37, 2.21) Minimal detectable change (90%) NA 3.94 *S.D., not applicable. ynot applicable. zconfidence interval. The rationale for summing multiple item responses into a single score is that the items are measuring a common trait or health concept. Factor analysis is a statistical technique which determines the extent to which responses Table III Confirmatory factor analysis fit statistics Initial model Correlated error model c 2 (df*, P) Two factor model 0.18 (3, 0.98) (5, <0.01) (3, 0.98) CFI 0.90 >0.99 >0.99 RF 0.79 >0.99 >0.99 TLI 0.80 >0.99 >0.99 RMSEAy 0.21 <0.01 <0.01 *Degrees of freedom. yroot mean-square error of approximation.

4 Osteoarthritis and Cartilage Vol. 15, No e_walk WALK A e_walk WALK e_stair STAIR.77 e_stair STAIR.69 e_night NIGHT.60 pain e_night NIGHT.57 pain e_sit SIT.74 e_sit SIT.80 e_stand STAND e_stand STAND Fig. 2. Initial confirmatory factor analysis model with factor loadings. B e_walk WALK.87 from multiple items can be represented by a smaller number of health concepts (factors); internal consistency examines the homogeneity of responses from a set of items. The WOMAC pain scale consists of five items conceived to assess a single health concept and a total score is obtained by summing the item responses. This approach would be reasonable provided the results from a factor analysis produced a single factor and the set of items displayed a high degree of internal consistency. We applied a confirmatory factor analysis and found that our hypothesized model of a single factor with uncorrelated error terms did not fit the data for the five WOMAC pain scale items. However, excellent fit was obtained from two modified models. One model specified correlations between the night and sit items error terms, and the walk and stair items error terms; the second model specified two factors with walk and stair items loading on one factor, night and sit items loading on a second factor, and the standing item loading on both factors. It is not a coincidence that both modified models produced the same fit statistics as the models essentially convey equivalent information concerning the relationship among pain items. We presented the two factor model to make explicit the impact of correlated error terms: something more than pain is influencing patients responses to WOMAC pain items. We cannot say with certainty what this isdpossible explanations include similarity of item content and yea-saying as the items with correlated error terms are adjacent to each other on the questionnairedhowever, it is important to note that factor analysis examines the covariance of responses among items and not the magnitude of responses. Accordingly, the fact that the night and sit items have lower pain scores should not be interpreted as causally influencing the association among items. Several studies have investigated the factorial structure of the WOMAC pain scale using exploratory factor analysis (principal component analysis). Faucher et al. commenting on 88 patients with osteoarthritis of the knee, reported the e_stair e_stand e_sit STAIR STAND SIT e_night NIGHT FACTOR 1 FACTOR 2 Fig. 3. (a) Modified confirmatory factor analysis model with correlated error terms and factor loadings. (b) Modified confirmatory factor analysis model with two factors and factor loadings. following factor loadings which accounted for 50% of the variance: (walk) 0.77; (stair) 0.61; (night) 0.63; (sit) 0.78; (stand) Guermazi et al. reporting on fewer than 50 patients (actual number not specified) with osteoarthritis of the knee, provided the following factor loadings which accounted for 53% of the variance: (walk) 0.73; (stair) 0.78; (night) 0.65; (sit) 0.71; (stand) Finally, Bellamy et al. also performed a principal component analysis on 17 patients with osteoarthritis of the knee, however, with this extremely small sample size the results cannot be interpreted with confidence 40. Although these factor loadings are comparable to those reported in our single factor model shown in Fig. 2, they were estimated using principal component analysis and cannot be compared directly. Guidelines for interpreting the magnitude of reliability coefficients for high stakes decisions concerning individuals.31.68

5 270 P. W. Stratford et al.: WOMAC pain scale measurement properties suggest a minimum value of 0.90 and ideally ,42. Our estimate of internal consistency was 0.84 and this is similar to published values which typically fall between 0.79 and ,27,28,43e45. Likewise, our estimate of testeretest reliability is reasonably consistent with reported values which are in the range of 0.77e0.86 for reassessment intervals of less than 22 days 27,28. The consequence of questionable factorial structure and less than optimal internal consistency is to increase the error and decrease the confidence associated with a reported patient s score. The SEM represents error in the same units as the original measurement and we have reported two SEMs. One SEM was based on internal consistency and the other on testeretest reliability. SEM IC is appropriate if one is interested in estimating the measurement error at an instant in time. As such it assumes that time stands still and the estimated score cannot be generalized beyond the instant in time at which a patient was assessed. Because a clinician is usually interested in generalizing a patient s scoredsay several days before the assessment or a day or two laterdbeyond the instant at which the assessment occurred, an error estimate based on SEM IC alone has little practical application. By contrast, random fluctuations in a stable patient s reported score over time are included in SEM TRT. Accordingly, SEM TRT is more appropriate when the intent is to generalize beyond the instant in time at which the test was performed. Although not intuitively obvious, SEM IC and SEM TRT share a common source of error variance which is influenced by the internal consistency. Accordingly, improving the internal consistency of a measure decreases both SEM IC and SEM TRT and increases one s confidence in a patient s reported score. Our estimates for SEM IC and SEM TRT were 1.48 and 1.69 WOMAC pain points, respectively. These values are consistent with estimates obtained from other investigations 31,45. For example, although Salaffi et al. did not report SEMs, it is possible to calculate from their report the values of 1.33 and 1.65 for SEM IC and SEM TRT, respectively, for the 20-point pain scale 28. Similarly, estimates of 1.42 and 1.97 for SEM IC and SEM TRT were calculated from Roorda et al. s study. Quintana et al. administered the Likert version of the WOMAC and converted their scores values out of These authors reported estimates of SEM IC of 7.71 and 8.24 (out 100 points) for patients receiving total hip and knee joint replacement, respectively 45. Converting their reported SEM IC values to the 20-point Likert scale, SEM IC values of 1.54 and 1.65 WOMAC pain points were obtained for the hip and knee samples, respectively. Kelly et al. in a study which examined the impact of waiting time on patients scheduled for total hip or knee arthroplasty administered the VAS version of the WOMAC 31. These authors concluded that the waiting time did not impact on the pain or physical function of patients as the mean changes in pain and function were 1 and 1.6 points, respectively, on a 100-point scale. Kelly et al. also presented the S.D. of the change scores and from these values we estimated SEM TRT. Converting Kelly s values, we obtained SEM TRT estimates of 1.88 and 2.16 for the hip and knee samples, respectively. In summary, SEMs obtained from the reports of other investigators fall within the 95% confidence interval for our estimated values. The second purpose of this paper was to illustrate how information from measurement studies reporting SEM TRT can be used to assist clinical decision-making concerning the confidence in a patient s score or change score. For the purpose of this illustration we will accept the estimated value for SEM TRT obtained from our study. One SEM is associated with a 68% confidence interval, however, in practice clinicians often desire greater certainty in their decisions and multiples of the SEMd1.65 for 90% and 1.96 for 95%dare frequently applied. To illustrate the application of the SEM, consider a patient who at the initial assessment reports a WOMAC pain score of 10/20, and at a subsequent assessment a score of 6/20. Applying an SEM TRT of 1.69 the following can be said: (1) the 90% confidence interval for this patient s initial pain score is from 7.2 to 12.8; (2) the 90% confidence interval for the follow-up score is from 3.2 to 8.8; and (3) 90% of truly unchanged patients will display random fluctuations of 3.9 WOMAC pain points when assessed on two occasions and we interpret the observed change of four points to represent a true change for this patient. There are several limitations associated with our work. First, our sample size for the testeretest component of this investigation was based on convenience and not a formal sample size calculation. The impact of this is a larger confidence interval width. For example, if one desired a lower one-sided 97.5% confidence interval width of 0.10, a sample size of 90 patients would be necessary assuming an obtained testeretest reliability of A second limitation is the extent to which our findings based on patients awaiting hip or knee arthroplasty are generalizable to all patients with osteoarthritis of the hip or knee on whom the WOMAC would be applied. A potential limitation for those interested in assessing patients with a smaller interval between assessments than that reported in our study is that the magnitude of measurement error may be related to the interval between assessments. Conceivably, a shorter interval between assessments may be associated with a smaller amount of measurement error however our estimated value was consistent with those calculated from the works of others who studied patients over a variety of retest intervals. In summary, our examination of the factorial structure of the WOMAC pain scale failed to support a single factor and internal consistency analysis yielded a coefficient less than optimal for individual patient use. Collectively, these findings question the wisdom of summing the five WOMAC pain scale item responses into a single score. An alternate strategy when considering individual patient application would be to interpret item responses separatelydpreferably with a minimum of seven response options per itemdor to sum only those items which display homogeneity. References 1. Kurtz S, Mowat F, Ong K, Chan N, Lau E, Halpern M. Prevalence of primary and revision total hip and knee arthroplasty in the United States from 1990 through J Bone Joint Surg Am 2005;87: 1487e Meenan RF, Anderson JJ, Kazis LE, Egger MJ, Altz- Smith M, Samuelson COJ, et al. Outcome assessment in clinical trials. Evidence for the sensitivity of health status measures. Arthritis Rheum 1984;27:1344e Bellamy N. An Evaluative Index for Clinical Trials. Hamilton: McMaster University 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. J Rheumatol 1988;15:1833e40.

6 Osteoarthritis and Cartilage Vol. 15, No Roland M, Fairbank J. The RolandeMorris Disability Questionnaire and the Oswestry Disability Questionnaire. Spine 2000;25:3115e Bellamy N, Campbell J, Haraoui B, Gerecz-Simon E, Buchbinder R, Hobby K, et al. Clinimetric properties of the AUSCAN Osteoarthritis Hand Index: an evaluation of reliability, validity and responsiveness. Osteoarthritis Cartilage 2002;10:863e9. 7. Angst F, Aeschlimann A, Stucki G. Smallest detectable and minimal clinically important differences of rehabilitation intervention with their implications for required sample sizes using WOMAC and SF-36 quality of life measurement instruments in patients with osteoarthritis of the lower extremities. Arthritis Rheum 2001; 45:384e Angst F, Aeschlimann A, Michel BA, Stucki G. Minimal clinically important rehabilitation effects in patients with osteoarthritis of the lower extremities. J Rheumatol 2002;29:131e8. 9. Quintana JM, Escobar A, Bilbao A, Arostegui I, Lafuente I, Vidaurreta I. Responsiveness and clinically important differences for the WOMAC and SF-36 after hip joint replacement. Osteoarthritis Cartilage 2005;13: 1076e Tubach F, Ravaud P, Baron G, Falissard B, Logeart I, Bellamy N, et al. Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: the minimal clinically important improvement. Ann Rheum Dis 2005;64:29e Bellamy N, Kirwan J, Boers M, Brooks P, Strand V, Tugwell P, et al. Recommendations for a core set of outcome measures for future phase III clinical trials in knee, hip, and hand osteoarthritis. Consensus development at OMERACT III. J Rheumatol 1997;24:799e Bellamy N. WOMAC Osteoarthritis Index User Guide IV. Queensland, Australia: University of Queensland Thumboo J, Chew LH, Soh CH. Validation of the Western Ontario and McMaster University osteoarthritis index in Asians with osteoarthritis in Singapore. Osteoarthritis Cartilage 2001;9:440e Faucher M, Poiraudeau S, Lefevre-Colau MM, Rannou F, Fermanian J, Revel M. Algo-functional assessment of knee osteoarthritis: comparison of the testeretest reliability and construct validity of the WOMAC and Lequesne indexes. Osteoarthritis Cartilage 2002;10:602e Faucher M, Poiraudeau S, Lefevre-Colau MM, Rannou F, Fermanian J, Revel M. Assessment of the testeretest reliability and construct validity of a modified WOMAC index in knee osteoarthritis. Joint Bone Spine 2004;71:121e Kennedy D, Stratford PW, Pagura SMC, Wessel J, Gollish JD, Woodhouse LJ. Exploring the factorial validity and clinical interpretability of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Physiother Can 2003;55:160e Guermazi M, Poiraudeau S, Yahia M, Mezganni M, Fermanian J, Habib Elleuch M, et al. Translation, adaptation and validation of the Western Ontario and McMaster Universities osteoarthritis index (WOMAC) for an Arab population: the Sfax modified WOMAC. Osteoarthritis Cartilage 2004;12:459e Wolfe F, Kong SX. Rasch analysis of the Western Ontario MacMaster questionnaire (WOMAC) in 2205 patients with osteoarthritis, rheumatoid arthritis, and fibromyalgia. Ann Rheum Dis 1999;58:563e Stratford PW, Kennedy DM, Hanna SE. Condition-specific Western Ontario McMaster Osteoarthritis Index was not superior to region-specific Lower Extremity Functional Scale at detecting change. J Clin Epidemiol 2004;57:1025e Terwee CB, van der Slikke RM, van Lummel RC, Benink RJ, Meijers WG, de Vet HC. Self-reported physical functioning was more influenced by pain than performance-based physical functioning in knee-osteoarthritis patients. J Clin Epidemiol 2006; 59:724e Ryser L, Wright BD, Aeschlimann A, Mariacher- Gehler S, Stucki G. A new look at the Western Ontario and McMaster Universities Osteoarthritis Index using Rasch analysis. Arthritis Care Res 1999;12: 331e Stratford PW, Kennedy DM. Does parallel item content on WOMAC s Pain and Function Subscales limits its ability to detect change in functional status? BMC Musculoskelet Disord 2004;5: Ostendorf M, van Stel HF, Buskens E, Schrijvers AJ, Marting LN, Verbout AJ, et al. Patient-reported outcome in total hip replacement. A comparison of five instruments of health status. J Bone Joint Surg Br 2004; 86:801e Tuzun EH, Eker L, Aytar A, Daskapan A, Bayramoglu M. Acceptability, reliability, validity and responsiveness of the Turkish version of WOMAC osteoarthritis index. Osteoarthritis Cartilage 2005;13: 28e Maly MR, Costigan PA, Olney SJ. Determinants of selfreport outcome measures in people with knee osteoarthritis. Arch Phys Med Rehabil 2006;87:96e McGregor AH, Rylands H, Owen A, Dore CJ, Hughes SPF. Does preoperative hip rehabilitation advice improve recovery and patient satisfaction? J Arthroplasty 2004;19:464e Roorda LD, Jones CA, Waltz M, Lankhorst GJ, Bouter LM, van der Eijken JW, et al. Satisfactory cross cultural equivalence of the Dutch WOMAC in patients with hip osteoarthritis waiting for arthroplasty. Ann Rheum Dis 2004;63:36e Salaffi F, Leardini G, Canesi B, Mannoni A, Fioravanti A, Caporali R, et al. Reliability and validity of the Western Ontario and McMaster Universities (WOMAC) osteoarthritis index in Italian patients with osteoarthritis of the knee. Osteoarthritis Cartilage 2003;11:551e Whitehouse SL, Lingard EA, Katz JN, Learmonth ID. Development and testing of a reduced WOMAC function scale. J Bone Joint Surg Br 2003;85:706e Lingard EA, Katz JN, Wright EA, Sledge CB. Predicting the outcome of total knee arthroplasty. J Bone Joint Surg Am 2004;86-A:2179e Kelly KD, Voaklander DC, Johnston DW, Newman SC, Suarez-Almazor ME. Change in pain and function while waiting for major joint arthroplasty. J Arthroplasty 2001;16:351e Bellamy N. Pain assessment in osteoarthritis: experience with the WOMAC osteoarthritis index. Semin Arthritis Rheum 1989;18:14e Streiner DL, Norman GR. Health Measurement Scales: A Practical Guide to Their Development and Use. 3rd edn. New York: Oxford University Press Byrne BM. Structural Equation Modeling with AMOS: Basic Concepts, Applications, and Programming. Mahwah, NJ: Lawrence Erlbaum Associates 2001.

7 272 P. W. Stratford et al.: WOMAC pain scale measurement properties 35. Schumacker RE, Lomax RG. A Beginner s Guide to Structural Equation Modeling. Mahwah, NJ: Lawrence Erlbaum Associates Shrout PE, Fleiss JL. Intraclass correlation: uses in assessing rater reliability. Psychol Bull 1979;86: 420e Stratford PW. Getting more from the literature: estimating the standard error of measurement from reliability studies. Physiother Can 2004;56:27e Beaton DE, Boers M, Wells GA. Many faces of the minimal clinically important difference (MCID): a literature review and directions for future research. Curr Opin Rheumatol 2002;14:109e Beaton DE, Bombardier C, Katz JN, Wright JG. A taxonomy for responsiveness. J Clin Epidemiol 2001; 54:1204e Bellamy N, Wells G, Campbell J. Relationship between severity and clinical importance of symptoms in osteoarthritis. Clin Rheumatol 1991;10:138e Nunnally JC. Psychometric Theory. Toronto: McGraw- Hill Kelley TL. Interpretation of Educational Measurements. Yonkers: World Books Stucki G, Sangha O, Stucki S, Michel BA, Tyndall A, Dick W, et al. Comparison of the WOMAC (Western Ontario and McMaster Universities) osteoarthritis index and a self-report format of the self-administered Lequesne-algofunctional index in patients with knee and hip osteoarthritis. Osteoarthritis Cartilage 1998;6: 79e Roos EM, Klassbo M, Lohmander LS. WOMAC osteoarthritis index. Reliability, validity, and responsiveness in patients with arthroscopically assessed osteoarthritis. Western Ontario and MacMaster Universities. Scand J Rheumatol 1999;28:210e Quintana JM, Escobar A, Arostegui I, Bilbao A, Azkarate J, Goenaga I, et al. Health-related quality of life and appropriateness of knee or hip joint replacement. Arch Intern Med 2006;166:220e Stratford PW, Spadoni GF. Sample size estimation for the comparison of competing measures reliability coefficients. Physiother Can 2003;55:225e9.