Article Text
PDF
Abstract
Objective To study whether assessment of rheumatoid arthritis (RA) disease activity in individual patients using the disease activity score in 28 joints (DAS28) or other instruments excluding joints of feet may lead to misclassification of disease activity.
Methods A cohort of RA patients was classified into three ‘regional radiographic damage progression’ groups: predominantly progression in feet, similar progression in hands and feet and predominantly progression in hands; both in early (0–2 years) and later (2–5 years) disease. Baseline and mean DAS28, individual DAS28 variables and tender joint counts (TJC) and swollen joint counts (SJC) of the feet were compared between groups. The longitudinal relation of DAS28 with radiographic damage was investigated using a mixed model analysis with rheumatoid factor status, baseline joint damage and TJC and SJC of the feet as covariates.
Results Early (n=265) and later (n=200) in the disease course, by definition, the classification procedure resulted in 25% as predominantly foot, 25% as predominantly hand and 50% as similar progressors. In early RA predominantly foot progressors had higher TJC and SJC of the feet compared with predominantly hand progressors (p<0.001), but DAS28 was similar. This was not seen in later disease. The longitudinal relation between DAS28 and radiographic progression was influenced by the region of progression (predominantly foot progressors vs others, p<0.001), suggesting that DAS28 underestimates disease activity in predominantly foot progressors. In this group, joint counts for the feet were independently related to radiographic progression.
Conclusions DAS28 underestimates actual disease activity and expected joint damage of individual early RA patients predominantly with disease in the feet.
Statistics from Altmetric.com
The disease activity score based on 28 joints (DAS28) and other disease activity indices such as the clinical disease activity index and simplified disease activity index are often used in clinical practice to assess disease activity in individual patients with rheumatoid arthritis (RA),1 2 although they are primarily designed for and mainly validated at the group level in clinical trials. In tight control strategies, these instruments, especially DAS28, are used to define treatment goals, also assuming a relation of the score with long-term outcome in individual patients.3,–,5 However, these indices are based on 28 tender joint counts (TJC) and swollen joint counts (SJC), excluding joints of the ankles and feet. At a group level, these indices are highly correlated with more comprehensive joint counts and related to future joint damage,6 but at an individual patient level, this relation is less clear.7,–,9 The effectiveness of tight control strategies might be increased using a more personalised approach than those based on the assessment of 28 joints only.3 10
In early RA, radiographic joint damage starts earlier in the feet Fifth Metatarsophalangeal Joint (MTP5) than in the hands suggesting that joints of the feet are important to assess,11 especially within the ‘window of opportunity’. Moreover, if joints of the feet are involved in individual patients more than other joints, or less, this might lead to under or overestimation of disease activity when assessing 28 joints only, with negative consequences in tight control strategies.
We studied whether there are groups of patients, especially early in the disease, with predominant involvement of the feet and whether this impacts the estimation of ‘real’ disease activity and long-term radiographic outcome, when instruments excluding joints of the feet are used.
Patients and methods
Patients from the 2-year randomised, open-label prospective multicentre treatment strategy trial, the Computer Assisted Management in Early Rheumatoid Arthritis (CAMERA),12 were evaluated. Patients were included between 1999 and 2003. At study entry all patients fulfilled the 1987 revised American College of Rheumatology criteria for RA,13 had a disease duration of less than 1 year, and were disease-modifying antirheumatic drug and glucocorticoid naive. Patients were randomly assigned to either the tight control methotrexate-based treatment strategy or to the conventional methotrexate-based strategy; both aimed at remission. In the first 2 years, medication was restricted to the protocol and included oral methotrexate with possible next treatment strategy steps of subcutaneous methotrexate and/or ciclosporin,12 14 thereafter medication was free. During follow-up after the trial, approximately 11% and 13% of patients used glucocorticoids or biological agents, respectively.
Radiographic damage
Radiographs of the hands and feet were evaluated according to the Sharp van der Heijde score (SHS) 15 by two readers, blinded for treatment strategy and clinical information, but not for sequence. The final SHS was based on consensus between the two observers. Yearly radiographs were available and the mean yearly progression rate was calculated for the period of 0–5 years of disease and separately for early (from 0 to 2 years) and later (from 2 to 5 years) disease. Total SHS and separate scores for hands and feet were calculated.
Disease activity
To assess disease activity, TJC and SJC based on 28 and 38 joints, respectively, excluding and including feet, as well as TJC and SJC for the feet only (10 joints), erythrocyte sedimentation rate (ESR) and visual analogue scales (VAS) for general wellbeing and pain were used. As disease activity index, DAS28 (computed from TJC and SJC of 28 joints, ESR and VAS general wellbeing) was investigated,1 as DAS28 is generally used and European League Against Rheumatism (EULAR) response criteria are based on it.1 16
Classification according to regional radiographic joint progression
Patients were classified into three ‘regional damage progression’ groups: predominantly progression of feet (foot progressors); similar progression of hands and feet (similar progressors); and predominantly progression of hands (hand progressors); those with no progression were classed as non-progressors. These groups were determined based on radiographic progression rates (mean SHS units increase per year) in the hands and feet. For all patients with SHS greater than 0, the difference in radiographic progression of feet (SHSfeet) and hands (SHShands) was calculated: ‘difference score’ (SHShands minus SHSfeet). To do this, first hands and feet scores were normalised to a 0–100 scale, because in SHS more joints of the hands than of the feet are included. Based on the distribution of this ‘difference score’, patients in the first quartile (ie, smallest (negative) difference of SHShands minus SHSfeet) were classified as predominantly foot progressors, patients in the last quartile (ie, largest difference of SHShands minus SHSfeet) as predominantly hand progressors and all patients in between (25–75th quartile) as similar progressors. This classification was performed for 0–5 years and separately for radiographic progression within 0–2 and 2–5 years.
Time-averaged clinical variables
To summarise the clinical involvement of joints and disease activity in the different progression gruops, time-averaged values were calculated for DAS28, its individual components (TJC, SJC, ESR and VAS general wellbeing), and for TJC and SJC of feet. For the period of 0–2 years, 3-monthly measurements were used for all variables. An average value was calculated if seven or more out of 10 measurements were present, including the baseline value. For the period of 2–5 years only yearly measurements were available. An average value was calculated if three or more out of four measurements were available.
Statistical analyses
Baseline and time-averaged clinical variables (DAS28, its individual components, and TJC and SJC of feet) were compared between ‘regional damage progression’ and non-progression groups separately for early and later RA disease duration, with exploratory analyses using analysis of variance and post-hoc tests, not correcting for multiple testing.
To investigate whether the damage classification was stable over the two time periods (ie, whether this could be considered a patient characteristic or a time varying characteristic of RA) a four by four table was constructed showing the percentage of patients staying in the same category or changing from category.
The longitudinal relation of DAS28 with radiographic damage over 5 years was investigated within all patients using longitudinal regression analysis (mixed model analysis), using an unstructured correlation structure with a random intercept and a linear link function. For this analysis, repeated DAS28 values over time of all patients were used. The relation between DAS28 at time point ‘x’ was related to radiographic damage at time point ‘x’+1 year. This time interval was introduced because it takes time before damage resulting from (a period of) disease activity can be seen on radiographs.6 16 It is also in line with the rationale of using DAS28 in making treatment decisions in individual patients, directed at the prevention of future damage. In this model, rheumatoid factor status and baseline joint damage were used as covariates. The regression coefficient for the time varying independent variables in this model can be interpreted in terms of a cross-sectional relationship (for example, patients with a high DAS28 have, on average, a high radiographic score), as well as a longitudinal relationship (ie, an increase or decrease in DAS28 is related to an increase or decrease in the progression of joint damage). To explore the longitudinal relationship further, a first-order autoregressive analysis was performed, in which SHS as outcome variable was adjusted for SHS 1 year earlier. This analysis can be interpreted as modelling change (progression) scores calculated per time interval; its rationale is that a radiographic damage score at a point in time is determined both by radiographic damage present earlier and by changes in independent variables. To be in line with earlier publications regarding the prediction of progression of joint damage in relation to disease activity, all prognostic variables that have been described before (and were available in our data) were included in the analysis.6
We also investigated whether specific regional joint involvement (ie, ‘regional damage progression’ groups) had an influence on the progression of joint damage and whether this influenced the longitudinal relation between disease activity and joint damage, by including a variable ‘predominantly foot progressors versus others’ and the interaction term of this variable with DAS28 in the model.
In a separate analysis, SJC and TJC of feet were entered as covariates in the initial model to investigate whether this improved the prediction of radiographic damage in all patients and specifically within ‘regional damage progression’ groups.
Statistical software SPSS 15.0 and SAS 9.1 were used for analyses. p Values less than 0.05 were considered statistically significant.
Results
Clinical characteristics for patients in the different progression groups are given in table 1 for 0–2 years and for 2–5 years. By definition, the classification resulted in 25% of patients with any damage being classified as predominantly foot progressors, 25% as predominantly hand progressors and 50% as similar (hand and foot) progressors. At least one out of 10 patients had predominantly foot involvement in the first 5 years of the disease.
Clinical, demographic and time-averaged characteristics of the three ‘regional damage progression’ groups and non-progressors at 0-2 years and at 2–5 years
SHS at baseline was significantly lower in non-progressors compared with the three ‘regional damage progression’ groups at 0–2 years and 2–5 years (both p<0.02). Predominantly foot and hand progressors had higher radiographic damage (total SHS) compared with similar progressors, both at 0–2 years and 2–5 years (both p<0.02), except for predominantly hand progressors versus similar progressors for 2–5 years (p=0.21). Non-progressors and predominantly hand progressors were less frequently rheumatoid factor positive compared with predominantly foot and similar progressors. Also, in early RA TJC and SJC of feet were higher in predominantly foot progressors compared with predominantly hand progressors (p=0.048 and p=0.008, respectively; table 1). This difference was not seen in later disease (2–5 years).
Over the first 2 years, predominantly foot and hand progressors had a similar mean (SD) DAS28 of 4.0 (1.0) and 3.9 (0.9), respectively. Compared with predominantly hand progressors, predominantly foot progressors had higher time-averaged TJC and SJC of the feet: 3.4 (2.2) versus 1.5 (1.3), p<0.001 and 2.8 (1.9) versus 1.1 (1.2), p<0.001, respectively (table 1). This difference was not observed in the 2–5 year period (table 1). In both time periods, the similar progression group had lower or in between disease activity scores and involvement of feet compared with patients with predominantly foot or hand progression.
Classification of the ‘regional damage progression’ groups did not change between the two time periods for 47% of all patients; 48% of all patients with predominantly foot progression were in this category in both RA periods, versus 41% for similar progressors, 39% for predominantly hand progressors and 51% for non-progressors (table 2).
Stability over time of classification into ‘regional damage progression’ groups at 0–2 years versus 2–5 years of the disease course
Longitudinal relation between DAS28 and radiographic damage
This relation and the (possible) differences between the groups were explored by gradual model building. First, the effect of DAS28 on the (progression of) radiographic damage was explored in the total group. At a group level, there was a longitudinal relation between DAS28 and radiographic progression, in line with the literature (see supplementary file, model 1, available online only).6 Model 1 describes the longitudinal relation between disease activity and radiographic damage, corrected for time and predictive factors, excluding treatment strategy, because adding this variable to the model did not influence the relation between disease activity and radiographic damage. Regression coefficients can be interpreted as estimates of the association with the progression of radiographic damage for non-time varying independent variables and for time varying variables (ie, DAS28) as the association between a change in DAS28 and a change in the progression of radiographic damage.
In model 2, the additional influence of the groups (ie, feet vs other progressors) on radiographic scores and on the association of DAS28 with radiographic damage was investigated. Predominantly foot progression (versus other types of progression) was added as an independent variable together with an interaction term of this variable and DAS28. Predominantly foot progression was found to influence the relation between DAS28 and radiographic progression, ie, predominantly foot progressors showed more (change in) radiographic progression at the same (change in) DAS28 compared with the other patient groups (see supplementary file, model 2, available online only). In other words, disease activity when assessed by DAS28 seems to be underestimated in predominantly foot progressors, when compared with other ‘regional damage progression’ groups.
Model 3 explored whether the addition of TJC of the feet would diminish the underestimation of actual disease activity; in model 4, this was explored for SJC of the feet. At a group level, joint counts for feet were not independently related to the progression of radiographic damage, but in predominantly foot progressors TJC (p<0.0001) and SJC of the feet (p<0.001) were longitudinally related to radiographic progression (see supplementary file, interaction term TJCfeet*footprogressors or SJCfeet*footprogressors in models 3 and 4, available online only).
To understand better these modifying effects by the ‘regional damage progression’ groups, stratified analyses for the ‘regional damage progression’ groups of joint involvement were performed. These analyses showed that one unit increase in DAS28 is associated with 0.1–1.5 SHS units increase; this effect is largest in the predominantly foot progression group (p<0.001) and lowest in the hand progression group (table 3). In figure 1 a hypothetical patient, with a predefined DAS28 course and simulated radiographic progression is shown for all ‘regional damage progression’ groups.
Prediction of radiographic progression for ‘regional damage progression’ groups at the same values of DAS28 (ie, 5.5 at baseline, 2.6 at 1–3 years and 3.2 at 4 years), based on longitudinal (mixed model) regression, see supplementary file, model 2, available online only. ). The solid line represents predominantly foot progressors; the dotted line represents similar progressors; the dashed line represents predominantly hand progressors. DAS28, disease activity score based on 28 joints; SHS, Sharp van der Heijde score.
Stratified analysis for influence of ‘regional damage progression’ groups on longitudinal relation (see supplementary file, model 1, available online only) and stratified analysis for influence on TJC and SJC of the feet
TJC of the feet was associated with radiographic progression (p=0.059) in the predominantly foot progressors, but not in the other groups; a similar effect was found for SJC (p=0.16) (table 3). Analyses also showed that DAS28 underestimates the actual disease activity in predominantly foot progressors, ie, at the same DAS28 values, there is more progression of radiographic joint damage in predominantly foot progressors compared with similar and predominantly hand progressors, whereas DAS28 overestimates the actual disease activity in predominantly hand progressors (compare regression coefficient for DAS28 of 1.5 (95% CI 0.7 to 2.3) for foot progressors with 0.1 (−1.1 to 1.3) for hand progressors in table 3). Including TJC or SJC of the feet partly corrects this under and overestimation. Regression coefficients for DAS28 changed to 1.3 and 0.2 in predominantly foot and hand progressors, respectively (table 3).
Based on graphical inspection of predicted versus observed SHS scores, the model fit was found to be adequate with no systematic trends in residuals.
Discussion
DAS28 is a frequently used disease activity index, which has been validated at the group level as an outcome measurement in clinical trials. However, there are some drawbacks of DAS28 in general, also at a group level: a simple calculation is not possible and DAS28 gives no clinical insight into its individual components, ie, several combinations of individual components are possible at the same score.3 7 10 Moreover, at the individual patient level, an important issue is that joints of the ankles and feet are not included.3 7 18 This might lead to a lack of accuracy for remission (which is an important aim in tight control strategies).19 20 We show that in individual RA patients with mainly involvement of the feet, DAS28 underestimates actual disease activity and the expected joint damage, especially evident during the first 2 years of the disease. This could particularly be a problem in tight control strategies using DAS28 as a monitoring instrument, aimed at remission or low disease activity.3 5 Variance in other variables next to excluding feet could also lead to the misclassification of individual RA patients. Furthermore, for instance the presence of tender points disturbs the precise and valid classification of individual RA patients using DAS28.21 However, looking into all potentially disturbing factors of DAS28 was beyond the scope for this paper. Omission of the feet was evaluated as one of the most important issues when using disease activity indices in individual patient care in early RA patients.
We also show that in predominantly foot progressors, joint counts for the feet were independently related to radiographic progression, meaning that DAS28 does not capture the influence of foot involvement on radiographic progression in these patients. These problems also apply to other disease activity indices not including joints of the feet, such as the simplified disease activity index and the clinical disease activity index.2. Even small changes in joint damage early in the disease can have serious, clinically relevant consequences in the longer term, as radiographic damage is a surrogate variable predicting current and future clinical problems and with that also other important outcomes such as functional and/or work disability.
Of the patients defined as mainly having hand or foot involvement, ie, 25% of patients on the left or right side of the distribution of difference in damage progression between hands and feet, early in the disease almost half also had predominantly feet or hand involvement later on in their RA disease course. Thirty-nine to 51% of patients in the ‘regional damage progression’ groups did not change over time to another group. This suggests that the pattern of radiographic damage is relatively stable over time and seems (partly) a patient characteristic and less a time varying disease characteristic.
Looking in detail at the distribution of damage progression, it seems that patients with predominantly foot progression have more overall progression of damage compared with patients with predominantly hand progression. This might be related to some underestimation and thus undertreatment of disease activity in this group in clinical practice.
For a comparison of groups in randomised clinical studies, ‘regional damage progression’ is probably less important, because outcome parameters are evaluated at a group level, in contrast to tight control strategies in which treatment strategy is based on individual patient parameters. For follow-up of individual patients, assessment of all joints frequently involved in RA seems to be indicated. If feet are involved in the individual patient, one should be aware that DAS28 might underestimate actual disease activity. Including SJC or TJC of the feet in the joint assessment decreases this underestimation somewhat. When aiming for low disease activity or remission, our advice would be also to check regularly the joints of the feet clinically as well as radiographically (given the imperfect relation of disease activity and radiographic progression) particularly early in the disease, to detect predominantly foot progressors and to better interpret disease activity results as assessed with DAS28 or other indices excluding joints of the feet.
In conclusion, in RA patients with predominantly foot problems, DAS28 underestimates the actual disease activity (especially during the first 2 years of the disease) and the expected joint damage. This could particularly be a problem in tight control strategies, using DAS28 (or another disease activity index based on 28 joints) as monitoring parameter aiming at low disease activity or remission. For follow-up of individual patients, assessment of all joints frequently involved in RA seems to be indicated, especially early in the disease course.
Acknowledgments
The authors would like to thank all participating rheumatologists and research nurses of the Utrecht Rheumatoid Arthritis Cohort study group for data collection, AWJM Jacobs-van Bree for data entry, and AA van Everdingen, for scoring radiographs.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
-
Competing interests None.