Article Text
Abstract
Background In previous studies it has been indicated that functional measures are less responsive in patients with established or late rheumatoid arthritis (RA) as compared with early RA, potentially because chronic irreversible functional damage is present. Therefore, they may not be useful as disease activity measures. We aimed to investigate whether this is specific to functional measures, or if it similarly also relates to other typical RA disease activity measures.
Methods We performed a pooled analysis of patient level clinical trial data of patients with RA. We investigated the effects of duration of RA on the responsiveness of all RA core set measures by using logistic regression analysis. We performed a number of sensitivity analyses to support our findings.
Results The probability of response in functional scores decreased from ∼60% in early disease to ∼30% in established/late disease (p=0.0023). No other core set variable or composite index behaved in this way. The effect of chronicity solely on functional responses was confirmed in all sensitivity analyses, particularly also when joint damage was used as a surrogate of chronicity, responsiveness decreased from >60% in patients with little structural damage to <20% in patients with severe joint damage (p<0.001).
Conclusions Physical function is among the most important outcomes of RA, but in contrast with other core set measures it is not a reliable measure for disease activity.
Statistics from Altmetric.com
Rheumatoid arthritis (RA) is the paradigmatical chronic inflammatory joint disease, with an estimated prevalence of approximately 1% of the population.1 Adequate care is pertinent for patients throughout all stages of their disease with the ultimate goal of maintaining physical function and quality of life.2 Clinical trials of interventions in RA are performed in early, established and late disease, and rely on the same outcome measures in all disease stages, which include a set of individual, so called, ‘core set’ measures, functional measures and structural measures.3 ,4
Although functional measures are tightly linked to quality of life in patients with RA, it has been suggested that they may not be useful in assessing response to treatment, particularly in non-early disease, because they reflect, to a varying degree, chronic damage.5 Thus, they are expected to be partly irreversible and this irreversible portion not to be amenable to treatment.5 ,6 Although the amount of irreversibility can be shown on the group level, it is rather challenging to predict it in the individual, although estimates can be made based on the degree of structural damage in scored radiographs.7 While the understanding and improvement of disability is of utmost importance, the mentioned difficulties have led to the claim that the measurement of physical function may not be an appropriate surrogate for disease activity, which is the target for disease modifying therapies, and the basis for potential changes in therapies.2 Most disease activity indices or response assessment tools in RA do not include functional measures,8–10 while some others do,11 and even some self-reported measures on function and pain were discussed as alternatives for other disease activity instruments.12
Despite the evidence for the effects of chronicity on physical function, it has not yet been proven that this ‘problem’ is inherent to functional measures, as it could theoretically also affect other disease activity measures. This is important to know, as these measures and their responses are commonly compared across patients and trials of patients of different stages of RA.
In the present study, we used a large dataset of patients with early, established and late RA, who had been enrolled in clinical trials in the recent past. The purpose of this study was to investigate whether core set measures other than functional measures are also affected by chronicity of RA. If this is the case, current considerations of disease activity responses need to be revised. If this is not the case, a clear point will have been made about the difference between the domain of function and of disease activity in RA.
Patients and methods
Data sources
We obtained and pooled data from pivotal clinical trials of RA of the past decade. For the purpose of this investigation, we pooled trials that had enrolled patients across a representative span of disease duration. We excluded all early RA trials, which only represent patients with less than 1–3 years of RA, because they would not be informative for a study modelling the effects of RA chronicity.
From the following trials we were kindly provided between 80% and 90% of the subject data as a random sample by the respective sponsors: the TEMPO trial of etanercept plus methotrexate (MTX) versus either alone in patients with established RA13 (the MTX and the combination arms were used for analysis); the European Leflunomide trial, which compared leflunomide (LEF) with MTX in patients with 10 years of RA or less14; the ATTRACT trial of infliximab+MTX versus placebo+MTX in MTX-insufficient responders (IR);15 and the DE019 trial of adalimumab+MTX versus placebo+MTX in MTX-IR.16 For the latter two, only the active arms were used, as the comparator arm was true placebo therapy in the beginning (placebo added to insufficient MTX). The year of publication from these studies ranged from 1999 to 2004. All patients had active RA at entry into these studies. We pooled the data of all (active) MTX monotherapy arms (including LEF) and all combination arms (tumour necrosis factor (TNF)-inhibitor+MTX). We separated these two groups only in one sensitivity analysis (see below) to assess whether the results differ between synthetic and biological disease-modifying antirheumatic drugs (DMARDs). The ATTRACT and the DE019 trials did not have erythrocyte sedimentation rate (ESR) available, which led to lower numbers for analysis of effects on ESR and Disease Activity Score 28 (DAS28 ESR).
Assessments
We obtained data on baseline demographics and measures of baseline RA disease activity. Disease activity variables were also obtained in three monthly intervals throughout Year 1 of all studies. They included the swollen joint count (SJC) or tender joint count TJC) (on a 28-joint scale), the patient global assessment (PGA) and evaluator global assessment (EGA) of disease activity (based on a visual analogue scale, from 0 cm to 10 cm), the C-reactive protein (CRP) and sedimentation rate (CRP in mg/dL, ESR in mm/h), pain scores (10 cm visual analogue scale), and functional scores (Health Assessment Questionnaire (HAQ)). From these variables, the Simplified Disease Activity Index (SDAI) and Clinical Disease Activity Index (CDAI) and the DAS28 were calculated as follows: SDAI=SJC+TJC+PGA+EGA+CRP; CDAI=SJC+TJC+PGA+EGA9 ,10; DAS28=0.56x√TJC28+0.28x√SJC28+0.70xlognatESR+0.014xGlobal Health.17 PGA and EGA in the SDAI and CDAI were used in cm (in DAS28 PGA was used as global health measure in mm); CRP in SDAI was used in mg/dL. Radiographic data at baseline using Sharp scores or van der Heijde modified Sharp scores (total sharp score; TSS) were available in all trials.
Main analyses
In the main analysis we aimed at comparing the responsiveness of RA disease activity core set measures and composite indices in patients with different levels of RA chronicity. In our main logistic regression analysis, we therefore investigated the effects of duration of RA on achievement of a 25% improvement over the first 6 months (dependent variable). The choice of the 25% cut-off and the 6 months’ time point for response, were arbitrary, and we will test the effect of these choices in respective sensitivity analyses (see below). The following 11 dependent variables were tested: SJC, TJC, CRP, ESR, PGA, EGA, Pain, HAQ, SDAI, CDAI, DAS28. In each model we adjusted for age (years), sex (male/female), originating trial (TEMPO, LEF, ATTRACT and DE019) and treatment regimen (MTX/LEF and MTX+TNF-inhibitor). We present the adjusted effects of RA duration on the dependent variable (along with 95% CIs across their scales); p values of <0.05 were considered statistically significant.
Due to a lack of ESR values in ATTRACT and DE019 trials, the analysis of ESR and DAS28(ESR) have much less power, particularly in the patients with late/very damaged RA, leading to larger CIs.
In one modification of the main model, we replaced ‘duration of RA’ as marker of chronicity by the extent of existing structural damage (‘structural’ chronicity).5 We then repeated all the models of the main analysis.
Additional sensitivity analyses
Since the selection of the dependent variable (ie, achievement of a 25% improvement over 6 months) in the main analysis was arbitrary, we also tested another time point of response (3 months instead of 6 months), and a more stringent response level (50% response at 6 months, instead of 25%). We also reanalysed the models and stratified by regimen (ie, MTX/LEF vs combination of MTX+TNF-inhibitor). We also estimated the probability of reaching a certain state (as opposed to achieving a response). For that purpose we—separately for each core set measure—used the median absolute level reached by all patients at 6 months as the dependent state variable. As all these sensitivity analyses were done to test the consistency of our findings across different assumptions, and not to test different hypotheses, we did not adjust their results for multiple testing.
Changes in individual disease activity and functional variables in patients with similar overall response
If chronicity has no impact on responsiveness, then it would be expected that in patients with similar overall responses, the response of individual variables would be comparable in early, established and late disease, irrespective of the trial. The most common way to determine patients’ ‘overall’ response in clinical trials of RA is by using the response definition of the American College of Rheumatology (ACR),11 by which patients need to improve by 20% in SJC and TJC, and in three of the remaining five (pain, PGA, EGA, ESR or CRP, HAQ). We first investigated the median changes in all core set variables in ACR20 responders, and compared these changes across patient groups with early (≤1 year duration), established (1–10 years) and late (>10 years) RA. We repeated these analyses in ACR50 and ACR70 responders.
We repeated all analyses of this part using structure as a measure of chronicity (little damage: total Sharp/van der Heijde-Sharp score, TSS, ≤3; moderate damage: TSS 3–30; high damage: TSS >30).
Contribution of HAQ scores to fulfilment of ACR responses in patient groups with different levels of RA chronicity
HAQ scores (or functional measures in general) are one item of the response definition of the ACR (see above).11 If HAQ scores become less responsive over time due to an irreversible component, then it would be expected that ACR responders with late RA fulfil the functional criterion less frequently than patients with early RA. We again used the definitions for temporal and structural chronicity as above to assess the effects of chronicity on functional contributions to the ACR responses.
Results
Patients
We included 1657 patients (74% female; 74% rheumatoid factor positive; mean age±SD: 55.4±11.6 years) from the four trials and five arms (table 1). Duration of RA was on average 6.84(±7) years with a range from 0.1 years to 52.1 years, and radiographic damage on the TSS was 42.7 (±50.7) on average with a range from 0 to 346. The average disease activity at baseline was high for all arms, but differed across the individual trials, with average SDAI levels ranging from 34.1 to 50.7; HAQ scores were relatively high at the outset, with their means ranging from 1.37 to 1.65.
Chronicity of RA consistently affects physical function, but none of the other disease activity measures or composite indices
In the main analysis testing for a 25% response in variables at 6 months after trial initiation, only HAQ scores showed a significant reduction in responsiveness to treatment with increasing duration of RA (figure 1; p=0.0023): the probability of response decreased from approximately 60% in early disease to approximately 30% in late disease (figure 1). No significant association with duration was seen for any of the other 10 measures tested. In addition, figure 2 depicts the effects of structural chronicity (using radiographic scores instead of RA duration) on all measures. The association of decreasing HAQ responses with chronicity was again highly significant (p<0.001). Here, also the CRP responses showed a significant association with chronicity (higher responses in patients with more structural damage). However, this was in contrast to the findings of the main analysis (temporal chronicity) and to the findings for ESR in the same analysis (as another acute phase marker). Since there were fewer patients with very late RA/severe and extensive radiographic damage, the respective CIs around the estimates are larger.
The significant effect of chronicity on the response of HAQ scores, and only on these, was robust against different definitions and assumptions of the statistical model; only at occasion did one or another measure show significance, but never for more than one of the many sensitivity analyses and thus not consistently. All of these sensitivity analyses are summarised in table 2.
Relative changes in functional scores are consistently reduced in chronic RA
We found clearly higher changes in functional scores among ACR20 responders who had a duration of RA of ≤1 year compared with 1–10 years or >10 years (62.8% vs 50.0% vs 43.0%; p<0.00001; table 3). The same was true for relative HAQ changes in patients with little radiographic damage (TSS of ≤3; 64.7%), moderate (TSS 3–30; 55.3%) and high radiographic damage (TSS >30; 41.2%) (p<0.0001). Acute phase reactants showed a significantly decreased association with temporal chronicity, pain scores with structural chronicity, but no measures except HAQ scores showed a consistent signal for temporal and structural chronicity (table 3).
Investigating the contribution of HAQ responses to achievement of ACR responses
Among ACR20 responders, we found that the frequency of fulfilling the HAQ criterion (HAQ change ≥20%) decreased across three groups by temporal chronicity (duration ≤1 year: 87.7%; 1–10 years: 81.1%; >10 years: 74.5%; p=0.019, χ2); similar results were seen when structural chronicity was used: TSS <3: 85.3%; TSS 3–30: 85.0%; TSS >30: 74.7%; p=0.012). For ACR50 responders the HAQ criterion (change ≥50%) was fulfilled in 71.8%, 58.3% and 63.8%, respectively, for groups by temporal chronicity (p=0.26), and in 76.7%, 70.4 and 58.7%, respectively, for groups by structural chronicity (p=0.014). Finally, for ACR70 responders, these numbers were 71.0%, 65.2% and 68.9%, respectively, for groups by duration (p=0.90) and 75.7%, 64.3% and 67.7%, for groups by structural damage (p=0.57).
Discussion
In our study we were able to show that improvement in physical function is blunted in RA when disease duration is long or significant joint damage has accrued. Importantly, this effect of chronicity was not seen for measures of disease activity, be it individual core set components or composite disease activity indices. This observation lends support to the clear distinction between measures of disease process (disease activity), measures of disease outcome (eg, radiographic damage) and measures that reflect both (eg, functional measures). Measuring function, despite it being a critical domain for patients with RA, cannot be regarded a reliable substitute for the measurement of disease activity in all patients, as the amount of disease activity that is represented in the level of disability varies considerably across patients.
The finding is notable, as many instruments that are heavily weighted on functional components, have repeatedly been proposed as a replacement of disease activity measures, particularly of the joint counts or laboratory parameters.12 Partly, this is based on the argument that patient self-report instruments are well correlated and consistent with composite disease activity indices, such as the DAS28, SDAI or the CDAI.18 However, the consistency is compromised if patients need to be treated to a specific target, as for function different patients will have different levels of reversibility, that is, different floors of achievable function.5 This clearly conflicts with the concept of disease activity, in which as per definition an active process needs to be—at least in principle or in theory—reversible and amenable to treatment. All current guidelines therefore recommend achievement of remission of disease activity.2 ,19 ,20
Our data support the comparability of disease activity data across trials of early, established and late RA, while at the same time, the ACR response attainment may be—even if only to a small extent—affected by chronicity. In fact, our data indicate that the contribution of HAQ responses decreases with temporal and structural chronicity for all levels of ACR response levels; for many patients with late RA, this means that the response options are reduced to a fulfilment of three of four, rather than three of five variables. Thus, assessment of ACR response in late RA may represent slightly different domains than it does in early RA. This may particularly make comparisons across study populations with different degree of chronicity different.
One limitation of our analyses is that they are solely based on clinical trial data, while patients and responses in real life may be different than in the selected and protected trial environment. Nevertheless, the trial setting allows testing of the principal hypothesis that there is an effect of chronicity on functional responses, because patients were enrolled by similar disease activity inclusion criteria, but at different stages of their disease. Indeed, baseline disease activity measures were quite similar regardless of trial or regimen. HAQ scores at baseline, however, ranged from 1.37 to 1.65 reflecting the impact of irreversible damage in more established disease. Moreover, the observations were made across almost all sensitivity analyses, implying that they ought to be valid also in clinical practice. In particular, since our main analysis used pooled data across different treatment regimens, the sensitivity analysis stratifying synthetic DMARDs (MTX and leflunomide) from biological+MTX was very important. In both strata the results from the main analysis were replicated, which indicates that the type of treatment is not an important determinant in this association.
It needs to be noted that the term ‘irreversible’, as used in this study, does not imply that disability in RA may not still be reversible upon physiotherapy, occupational therapy, orthopaedic surgery or other function-directed interventions. The point made in the present study is that disease activity directed interventions (as in most major trials of RA) will not affect the irreversible part of disability, and thus render functional measures unreliable for assessment of disease activity.
In summary, our data call for keeping the established conceptual distinction of RA measures into those capturing disease activity, those capturing functional disability and those capturing structural damage. While physical function remains to be a central outcome measure for patients with RA, and maintaining functional integrity is of utmost importance for patients and society, disease activity is the modifiable target for following patients with RA, and effective therapy aims to achieve its complete reduction (remission). Functional remission may not be an achievable target for many patients, as only parts of their disability are reversible. Finally, in an ideal scenario of early initiated and effective medication, structural progression may be completely prevented and functional disability may not acquire these irreversible components that lead to many of the issues addressed in the present publication.
Acknowledgments
We thank Abbott, Centocor (Janssen), Sanofi-Aventis and Wyeth (Pfizer), for providing the random sample of the original data from their trials.
References
Supplementary materials
Lay summary
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
-
Handling editor Tore K Kvien
-
Contributors DA: study design, data analysis, drafting of manuscript. FA: data analysis. JSS: study design, data analysis, drafting of manuscript.
-
Funding This study was supported through Coordination Theme 1 (Health) of the European Community's FP7; Grant Agreement number HEALTH-F2-2008-223404 (Masterswitch). This is a publication of the Joint and Bone Center for Diagnosis, Research and Therapy of Musculoskeletal Disorders of the Medical University of Vienna.
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.