Volume 11, Issue 4 pp. 229-237
Research Article
Full Access

Different Wording of the Patient Global Visual Analogue Scale (PG-VAS) Affects Rheumatoid Arthritis Patients’ Scoring and the Overall Disease Activity Score (DAS28): A Cross-Sectional Study

Tracy French

Tracy French

Rheumatology Department, Bristol Royal Infirmary, Bristol, UK

Search for more papers by this author
Sarah Hewlett

Sarah Hewlett

University of West of England, Bristol, and Academic Rheumatology, Bristol Royal Infirmary, Bristol, UK

Search for more papers by this author
John Kirwan

John Kirwan

University of Bristol and Division of Medicine, Bristol Royal Infirmary, Bristol, UK

Search for more papers by this author
Tessa Sanderson

Corresponding Author

Tessa Sanderson

University of West of England, Bristol, and Academic Rheumatology, Bristol Royal Infirmary, Bristol, UK

Correspondence: Tessa Sanderson, Academic Rheumatology, Bristol Royal Infirmary, Lower Maudlin Street, Bristol, BS2 8HW, UK. Tel: +44 (0)117 342 2901; Fax: +44 (0)117 342 3841

Email: Tessa2.Sanderson@uwe.ac.uk

Search for more papers by this author
First published: 04 February 2013
Citations: 33

Abstract

Objective

The Disease Activity Score in 28 joints (DAS28) is a key measure in clinical practice and clinical trials. There are at least five different versions of the ‘Patient Global’ Visual Analogue Scale (PG-VAS) being used in the DAS28. The developers suggested that the PG-VAS can be an assessment of global health or disease activity, but did not specify the wording of the question. There is no consensus on what the PG-VAS is intended to capture, and the different words and phrases have not been evaluated. The aim of this study was to test if phrasing affects PG-VAS scores and hence yields different results for the DAS28.

Methods

Fifty patients with rheumatoid arthritis taking biologic agents in a rheumatology outpatient department completed a self-administered questionnaire containing five versions of the 100 mm PG-VAS.

Results

All PG-VAS versions correlated strongly with each other (rho = 0.67–0.87, p < 0.0001). However, individual scores for each PG-VAS, when compared with the comparator on a Bland–Altman chart had wide limits of agreement—the largest being −42 mm to +45 mm. The five overall DAS28 scores were calculated for each patient using the five different PG-VAS. The largest difference in DAS28 scores was 0.63.

Conclusion

Different phrasing of the PG-VAS gives different DAS28 results. As the DAS28 is a key outcome measure, such differences have the potential to influence clinical decisions relating to eligibility for biologic agents and evaluation of new therapies. We urgently need to decide on the concept to be measured and the phrasing required to capture this. The PG-VAS phrasing should then be standardized and validated. Copyright © 2013 John Wiley & Sons, Ltd. Copyright © 2013 John Wiley & Sons, Ltd.

Introduction

Rheumatoid arthritis (RA) is controlled using disease-modifying anti-rheumatic drugs (DMARDs) such as methotrexate and glucocorticoids and/or biologic therapies, depending on the severity of disease. Assessing the efficacy of these drugs in clinical practice is critical to ensure that erosive damage to joints is prevented and to maintain quality of life for patients. The ultimate goal is to achieve remission. The cost of biologic therapies is far higher than for traditional DMARDs so there is the extra consideration of financial implications when assessing response to biologics. The British Society for Rheumatology Biologics Register (2009) and the European League Against Rheumatism (EULAR) (Smolen et al., 2010) recommend that the Disease Activity Score in 28 joints (DAS28) is used in clinical practice to evaluate the response to therapies by assessing clinical disease activity, and it is commonly used in clinical trials and treat-to-target strategies. The DAS28 is a composite index of four process variables: 28-tender joint count, 28-swollen joint count, a measure of inflammatory markers [either C-reactive protein (CRP) or erythrocyte sedimentation rate] and a ‘Patient Global’ Visual Analogue Scale (PG-VAS) (van der Heijde et al., 1992). The DAS was developed based on the clinical judgement of six rheumatologists during a prospective study, aiming to standardize the comparison of treatment efficacy across trials (van der Heijde et al., 1992). However, there was no patient involvement in developing the wording of the PG-VAS and there is no ‘gold standard’ for the phrasing of the PG-VAS. Publications on the development of the DAS do not specify the actual phrasing of the stem question for the PG-VAS, or the timeframe (Prevoo et al., 1995; van der Heijde, 1998; van der Heijde et al., 1992, 1993; van Gestel et al., 1996).

The DAS was validated using an unspecified ‘general health’ PG-VAS (van der Heijde et al., 1993), but the DAS website (DAS Group, 2001) suggests that a ‘disease activity’ PG-VAS can also be used, although this has not been validated. Several authors have variously reported that the PG-VAS (unspecified phrasing) is presumed to capture ‘disease activity’, ‘general health’ and ‘arthritis impact’ (Harrison et al., 2009; Kalyoncu et al., 2009; Smedstad et al., 1997). It is unlikely that one question can assess all these aspects of RA. The literature shows a disparity in the phrasing of the PG-VAS (Meenan et al., 1980; Scott et al., 1993; van Gestel and van Riel, 1997; van Gestel et al., 1998). Porter and colleagues (2011) have highlighted this issue and suggested that a standardized approach should be used when assessing the PG-VAS, although this was not based on data. For the assessment of remission in RA, the American College of Rheumatology (ACR)/EULAR have suggested that the patient global assessment uses the following phrasing: ‘Considering all of the ways your arthritis has affected you, how do you feel your arthritis is today?’ (Anchors: very well–very poor) (Felson et al., 2011). This version of the question is very important for future clinical practice, but may affect remission rates because it could capture non-inflammatory issues such as low back pain and functional limitations (Masri et al., 2012). The PG-VAS is also an important component of the RA core set (Felson et al., 1995), indices such as the Clinical Disease Activity Index (CDAI) (Aletaha et al., 2005) and may be used independently by clinicians to assess patient status; whenever it is used, the same limitations apply. However, we focus here on the DAS28 because of its role to determine eligibility for biologic agents.

There is no clarity about the concept that the DAS PG-VAS should capture, or how it should be phrased in order to do so. The DAS28 algorithm uses a weighting of 0.014 for the PG-VAS. Given this small weighting, it is possible that differences in phrasing would result in little difference to the overall DAS28 score but, alternatively, the differences in scoring could be significant. The current study therefore aimed to test whether i) different phrasing affects how patients score the PG-VAS, and ii) if any differences in scores were large enough to alter their subsequent calculated DAS28 score.

Methods

Patients

Consecutive adult patients with RA attending nurse-led biologic clinics in one centre over a two-month period were invited to participate. The questionnaires could be completed at home or within the clinic, but no professional assistance was given during completion to ensure that no bias occurred. Patients were excluded if they could not complete the questionnaire independently or if they were not fluent in English. Written consent was obtained and ethics approval was granted by the Frenchay Research Ethics Committee (10/HO107/13).

Outcome measures

Five published or commonly used versions of the PG-VAS were tested (Table 1). Four were taken from the literature (see Table 1) and one from a rheumatology nurse pilot survey of current clinical practice (French, 2010, unpublished), but which is not found in the literature. All of the PG-VASs were presented in the same format, with a horizontal 100 mm scale without marks and with the same timeframe (previous seven days). Version 5 [Arthritis Impact Measurement Scales (AIMS) phrasing] (Meenan et al., 1980) is used locally and does not specify a timeframe; it was decided not to change this, because it would alter clinical practice for all patients attending the centre. One of the PG-VAS versions that had originally been a statement was presented as a question to aid comprehension and provide consistency.

Table 1. Versions of the PG-VAS used in the study
PG-VAS version Stem Instructions Anchors Source and changes
1. Feeling How do you feel concerning your arthritis over the last week? Please place a mark on the line below Very well to Extremely bad (van Gestel et al., 1998: 1846) Timeframe added
2. Disease activity How active has your disease been this week? Please place a mark on the line below Not active at all to Extremely active (Scott et al., 1993: 27) Statement turned into a question
3. Well-being How has your overall well-being been this week? Please indicate on the scale below Best imaginable health state to Worst imaginable health state (British Society for Rheumatology, 2009) Timeframe added
4. Best/worst If 0 is the best you have ever been and 100 is the worst you have ever been, where do you think you have been over the last week? Please place a mark on the line below Best you have ever been to Worst you have ever been (French, 2010), unpublished
5. AIMS Considering all the ways your arthritis affects you… …mark on the line below how well you are doing. Very well to Very badly (Meenan et al., 1980)
  • AIMS, arthritis impact measurement scales; PG-VAS, ‘patient global’ visual analogue scale

PG-VAS versions 1–4 were embedded in a self-administered questionnaire, separated by a larger set of standard questions and other VAS measures (detailed below) which served as ‘distracters’. The PG-VAS versions 1–4 were differently ordered in four questionnaire packs, and these were given out in sequence, so that any order effect bias could be detected. Version 5, the locally used phrasing of the PG-VAS used as a comparator, was collected as part of routine assessment during the consultation. Swollen and tender joint counts, CRP and the Health Assessment Questionnaire (HAQ) Disability Index (Fries et al., 1980) were collected, along with 100 mm VASs for fatigue (‘Please place a mark on the line to show your level of fatigue over the last 7 days’; anchors: ‘No fatigue’ and ‘Totally exhausted’), emotional well-being (‘Considering all the ways your arthritis affects you, please mark on the line below to show your sense of emotional well-being today’; anchors: ‘Little emotional well-being’ and ‘Great emotional well-being’) and pain (‘How much pain have you had over the last 24 hours?’; anchors: ‘No pain’ and ‘Severe pain’).

Statistical methods

A sample size of 50 provides 90% power to detect a 0.1 unit effect on the DAS28 (based on a 7 mm difference in the PG-VAS score equalling a 0.1 change in DAS28, using the DAS28 algorithm with four variables) (DAS Group, 2001). Prior to any analysis of the data, a one-way (repeated measure) analysis of variance test was conducted to identify any order effect there may have been from the four differently ordered questionnaires. Parametric or non-parametric analyses were conducted as appropriate to the data distribution. The shared variance (r2) between all pairs of PG-VAS was calculated and Bland–Altman plots were produced to illustrate the limits of agreement between individual scores for the comparator PG-VAS version 5 and each of the other PG-VASs. Correlation coefficients were also calculated to determine the strength and direction of the relationship between the five DAS28 scores, calculated with the five versions of PG-VAS. Each PG-VAS was also correlated with the other clinical variables collected, and with a three-variable DAS28 (swollen joint count, tender joint count and CRP, omitting the patient global question) to assess their relationship with these disease activity measures.

Results

Of 54 sequential patients approached, 53 were recruited and 50 completed the study. The sample characteristics are presented in Table 2, and show a wide range in disease activity (DAS28 1.79–7.69).

Table 2. Sample characteristics
Variable Number (%) or mean (SD)
Gender, female 39 (78%)
Ethnicity, White British 46 (92%)
Age (years) 58.2 (13.9)
Disease duration (years) 16.4 (9.2)
HAQ-DI (n = 49) 1.65 (0.9)
DAS28 (PG-VAS version 5), taken on day recruited 4.31 (1.5)
  • DAS28, disease activity score in 28 joints; HAQ-DI, Health Assessment Questionnaire - Disability Index; PG-VAS, ‘patient global’ visual analogue scale; SD, standard deviation

The repeated measures one-way analysis of variance gave a Wilks’ λ of 0.984 (p = 0.870), showing that there was no order effect of versions 1–4 of the PG-VAS on patient scores. Scores for PG-VAS versions 1–4 were normally distributed (0.073–0.112 Kolmogorov–Smirnov, p > 0.05, indicating normality) but the PG-VAS version 5 was non-normally distributed (0.15 Kolmogorov–Smirnov), so non-parametric tests were used in the analysis. However, the DAS28 scores calculated from the PG-VAS scores in the second part of the analysis were normally distributed so they were analysed using parametric tests. The inter-quartile ranges for the five PG-VASs are presented in Table 3, illustrating that the mean scores were very similar, but that all PG-VASs had a wide distribution of scores.

Table 3. Descriptive statistics for the five versions of the PG-VAS (N = 50)
PG-VAS version Mean (SD) Median Inter-quartile range
1. Feeling 43.22 (27.01) 41.00 19, 65
2. Disease activity 41.36 (28.41) 39.50 17, 65
3. Well-being 41.40 (24.61) 41.00 24, 62
4. Best–worst 42.00 (25.40) 39.50 21, 61
5. AIMS 43.06 (24.29) 39.00 26, 62
  • AIMS, arthritis impact measurement scales; PG-VAS, ‘patient global’ visual analogue scale

All PG-VAS scores correlated strongly and positively (rho = 0.67–0.87, p < 0.0001) (Table 4), and the coefficient can be interpreted as large if r ≥ 0.50 (Bland and Altman, 1986). Shared variance was calculated, showing that the ‘Disease Activity’ PG-VAS (version 2) had the least shared variance with the three general health/well-being PG-VASs (versions 3, 4 and 5). The ‘Feeling’ and ‘Best/Worst’ PG-VASs (versions 1 and 4) showed the highest amount of shared variance. Correlations between the PG-VAS and other clinical measures are also shown in Table 4. The ‘Feeling’ PG-VAS correlated strongly with most other measures: Pain VAS (rho = 0.81), disease activity [DAS28 with three variables (i.e. tender, swollen joint count and CRP)] (rho = 0.74), Emotional Well-being PG-VAS (rho = 0.69) and HAQ (rho = 0.62), all p < 0.0001 (Table 5). The ‘Disease Activity’ PG-VAS correlated least well compared with the other versions: Disease activity (DAS28 with three variables tender, swollen joint count and CRP) (rho = 0.64, p < 0.0001), HAQ-DI (rho = 0.45, p < 0.001) and Fatigue VAS (rho = 0.48, p < 0.0001). The effect of differences in PG-VAS scores on the overall DAS28 was assessed by calculating the five DAS28 scores for each patient using the different PG-VAS scores. These correlated strongly and positively—all better than 0.98 (96% variance, p < 0.001).

Table 4. Correlations (rho) between the five versions of the PG-VAS and the five versions of the DAS calculated from them, and shared variance (%), and other measures (all p < 0.0001, except where indicated) (N = 50)
PG-VAS scores Other measures
2. Disease activity 3. Well-being 4. Best/worst 5. AIMS Fatigue VAS (N = 50) Emotional well-being VAS (N = 49) HAQ (N = 49) Pain VAS (N = 49) DAS28 (CRP) Swollen joint count Tender joint count CRP
1. Feeling 0.70 (49.0) 0.76 (57.8) 0.87 (75.7) 0.82 (67.2) 0.69 0.69 0.62 0.81 0.75 0.69 0.62 0.37*
2. Disease activity 0.67 (44.9) 0.74 (54.8) 0.67 (44.9) 0.48 0.67 0.45 0.71 0.69 0.70 0.65 0.00
3. Well-being 0.79 (62.4) 0.74 (54.8) 0.81 0.66 0.49 0.63 0.71 0.64 0.63 0.21
4. Best/worst 0.70 (49.0) 0.67 0.67 0.51 0.76 0.76 0.72 0.66 0.22
5. AIMS 0.64 0.65 0.54 0.80 0.80 0.74 0.71 0.34**
  • All p < 0.0001 except:
  • * p < 0.05
  • ** p < 0.01
  • Not significant
  • Calculated without the PG-VAS (i.e. on three clinical variables only)
  • AIMS, arthritis impact measurement scales; CRP, C-reactive protein; DAS28, disease activity score in 28 joints; HAQ, health assessment questionnaire; VAS, visual analogue scale
Table 5. The 95% limits of agreement (mm) in Bland-Altman plots for the AIMS PG-VAS (version 5) compared to the PG-VAS versions 1–4, and for the AIMS DAS28 (version 5) compared to DAS versions 1–4
AIMS PG-VAS AIMS DAS
1. Feeling −30.34 to +30.02 −0.43 to +0.42
2. Disease activity −41.86 to +45.26 −0.59 to +0.63
3. Well-being −32.27 to +35.59 −0.45 to +0.50
4. Best/Worst −35.39 to +37.51 −0.50 to +0.53
  • AIMS, arthritis impact measurement scales; DAS, disease activity score; PG-VAS, ‘patient global’ visual analogue scale

The amount of agreement between individual participant scores for versions 1–4 of the PG-VAS and the comparator [‘AIMS’ PG-VAS (version 5)] and for the DAS28 versions 1–4 and the comparator [‘AIMS’ DAS (version 5)] are shown in Table 5. There were wide 95% limits of agreement, the largest being between the ‘Disease Activity’ PG-VAS (version 2) and the ‘AIMS’ PG-VAS (version 5) with a range of −41.86 mm to +45.26 mm from a maximum possible range of −100 mm to +100 mm. For the DAS28 values, the smallest limits of agreement were between the ‘Feeling’ PG-VAS (version 1) and the ‘AIMS’ PG-VAS (version 5) (−0.43 to +0.42), and the largest between the ‘Disease Activity’ PG-VAS (version 2) and the ‘AIMS’ PG-VAS (version 5) (−0.59 to +0.63).

Discussion

The effect of phrasing on PG-VAS scores

The current study showed that different words or phrasing of the Patient Global VAS change the scoring of the PG-VAS and that this changes the value of the subsequently calculated DAS score. The strong and positive correlations found between the different versions of the PG-VAS are to be expected as they are measuring a similar global variable which is broadly determined by the overall state of the patient's arthritis. However, as Bland and Altman (1986) demonstrate, it is the ‘limits of agreement’ which better describe the measurement accuracy of two measures of the same construct. There were wide 95% limits of agreement between the PG-VAS scores, indicating a clear lack of agreement between the different versions of the PG-VAS. This is the case for other measures of disease activity when testing reliability (e.g. Uhlig et al., 2009). In the Bland–Altman analysis, 50% or more patients had a difference between PG-VAS versions 1–4 and the comparator ‘AIMS’ PG-VAS (version 5). This strongly suggests that the different words and phrases do capture different information.

The PG-VAS has a weighting of 0.014 within the DAS28 calculation, so that the maximum difference that the PG-VAS can make to the overall DAS28 is 1.4 (shown by holding the other variables constant and using first a PG-VAS score of 0 mm, then of 100 mm). The different scores for the PG-VAS versions, when translated into the DAS28, showed differences in the DAS28 scores. The largest DAS28 difference reported here was 0.63, which is easily large enough to affect the assessment of suitability for, and response to, biologic agents and DMARDs. Treatment with anti-tumour necrosis factor (TNF) therapy should only be continued in the National Health Service if there is an adequate response (an improvement in DAS28 of 1.2 points or more) at six months following initiation of therapy (National Institute for Health and Clinical Excellence, 2007). However, physicians interpret DAS scores, and the differences shown in the current study may not influence their interpretations and subsequent clinical decisions in other circumstances. Nevertheless, tight control strategies (Bakker et al., 2012) imply the use of outcome measures such as the DAS.

Different phrasings of the PG-VAS correlate differently with specific clinical variables. The ‘Disease Activity’ PG-VAS had the least amount of shared variance with three other PG-VAS and had the widest 95% limits of agreement on the Bland–Altman plot. It also correlated least well with other measures (DAS28 with three variables, HAQ-DI and Fatigue VAS), whereas the ‘Feeling’ PG-VAS correlated well with these. Therefore, based on these data it would be recommended that the phrasing ‘How do you feel concerning your arthritis over the last week?’ (with anchors: ‘Very well’ and ‘Extremely bad’) be used for the PG-VAS for the patient component of the DAS28. That is, if the VAS is to capture the global patient experience, this phrasing is most effective. While this version of this PG-VAS has similarities to that proposed by the ACR/EULAR remission group (Felson et al., 2011), we recommend that the ‘Feeling’ PG-VAS (version 1) phrasing is used as it is based on data, whereas the ACR/EULAR remission group did not report the decision-making process that led to their proposed version. Contrary to Masri and colleagues’ (2012) contention that the PG-VAS is problematic because it captures non-inflammatory factors, we argue that this broader measurement is favourable.

Since the current study was undertaken, Dougados and colleagues (2011) have published a comparison of two PG-VASs: a general health status PG-VAS (‘In general, how would you rate your health over the last 2–3 weeks?’) and an RA disease activity PG-VAS (‘Please estimate your disease activity over the last 48 hours’). This was carried out in a larger sample of 108 patients in a clinical trial of etanercept, from screening to 12 weeks after commencing therapy. They used the intra-class correlation coefficient to compare the results for the two techniques and found similar scores, but they concluded that the DAS28 was not significantly affected by the different techniques used to assess the PG-VAS. There are, however, differences between the studies which may limit direct comparison of the results. First, the sample was in a clinical trial rather than being ‘real life’ patients. Second, Dougados and colleagues (2011) did not specify where the phrasing of the two PG-VASs they tested originated. In spite of their results, they recommend that standardization of the phrasing is preferable, to reduce heterogeneity in scoring of the DAS28, and recommend that further analysis of this issue is needed.

Fries and Ramey (1997) compared a quality of life (EuroQoL) vertical feeling thermometer with a horizontal VAS using the AIMS phrasing (PG-VAS version 5) and concluded that the two scales produced similar results. Their study was much larger (n = 663), with 43 patients being retested six months later. Their correlations were r = −0.68 (p < 0.0001) with strong retest scores. They concluded that the small differences found were due to the differences in phrasing and presentation of the two scales. However, they did not examine their data to look at variance or agreement between the two scales. Further analysis of individual scores (with a Bland–Altman plot), rather than groups of scores (with Pearson's correlation), may have shown a lack of agreement between the two scales, as was the case in the present study.

Harrison et al. (2009) compared responses to a global health VAS, presented both as a 10 cm horizontal VAS with no incremental markers, and a vertical 20 cm VAS with 1 cm markers. They concluded that different presentation of scales, order effect and incremental markers on VAS affect scoring. The present study addressed these possibilities for bias and showed that phrasing alone affects scoring.

Wolfe and Michaud (2009) studied over 20,000 RA patients to assess whether treatment success as determined by outcome measures (including the DAS28) were also defined as successful by patients in terms of health satisfaction. They found a correlation of 0.46 between DAS28 and PG VAS (AIMS phrasing). From this lack of agreement they identified the importance to patients of not only having fewer tender and swollen joints, but also of feeling better in themselves. This highlights the fact that the patient opinion is vital for assessing the success of drug therapies as well as clinical signs. It is therefore paramount that the small contribution that patients make to the DAS is evaluated in an effective and standardized way. This should include addressing variation in phrasing, presentation and timeframe. Prior to standardization of this outcome measure, a consensus is required on what the PG-VAS should capture. Those questions relating to overall health or general well-being may be assessing the impact of disease, which may be a composite of disease activity, the patient's ability to self-manage and the importance of consequences of their condition (Sanderson et al., 2011).

Recent qualitative research, primarily to define well-being and flare, shows that patients do not consider the similarly phrased and commonly used PG-VAS version 5 (AIMS question) to capture disease activity, and would favour a more direct question if that is the intended target (Hewlett et al., 2012; Sanderson, 2009). In an observational study, it was found that pain was the main determinant (75.6%) for the PG-VAS when phrased ‘How do you estimate your disease activity today?’ (0 = no disease activity, 100 mm highly active disease) (Studenic et al., 2012). Therefore, further qualitative research with patients may be necessary to allow researchers to determine the remaining 25% of variability in the perception of disease activity that is not currently explained (van Tuyl and Boers, 2012) and how to construct the ideal phrasing to capture those determinants.

Strengths and limitations of the study

A limitation of the study may be the decision not to change the comparator ‘AIMS’ phrasing (PG-VAS version 5) to include a timeframe, which may have adversely influenced the study results. Also, the interpretation of the significance of the limits of agreement on the Bland–Altman plots is clinical rather than 2statistical, due to the distribution of the data collected. However, the range of DAS28 scores is comparable with those of Smith and colleagues (2007), who conducted a study with 679 patients examining the necessity of pre-assessment for anti-TNF therapy. This indicates that the results are generalizable to other patients attending biologic clinics. Recruitment of patients from a nurse-led biologic clinic is unlikely to have consequences for the representativity of the findings because care is shared with the rheumatologists. It was not possible to calculate test–retest reliability for the same measures in this study, and such calculations may indicate that a range of differences within a measure are the result of natural variation.

Conclusion

The present study has shown that different phrasing of the PG-VASs gives different DAS28 results; therefore, this outcome measure should be standardized. In the UK, the DAS28 is used to assess eligibility for and response to biologic therapies. National disparity in the implementation of the DAS28 could potentially affect patient access to these expensive and extremely effective treatments. Internationally, clinical trials use the EULAR response criteria (Smolen et al., 2010) and the ACR core set of outcome measures (Felson et al., 1995), which both include a PG-VAS; thus, the results of the study are applicable in all research settings in which these response criteria are used. First, greater clarity is needed about the underlying construct being assessed by patient global questions; then greater rigour should be used in the standardization and validation of patient global measurement instruments.

Acknowledgements

We thank the patient research partners who advised on the study: Bev Davies and Val Lloyd; Rosemary Greenwood for statistical advice; and Arthritis Research UK for an educational bursary (TF). The study was sponsored by Arthritis Research UK (educational bursary: grant no. 17974).

    The full text of this article hosted at iucr.org is unavailable due to technical difficulties.