Long-term impact of early treatment on radiographic progression in rheumatoid arthritis: A meta-analysis

INTRODUCTION

Rheumatoid arthritis (RA) is a chronic inflammatory disease that causes progressive joint destruction and disability and is currently without cure. It is the most common systemic rheumatic disease, affecting ∼1% of the population (1). Radiographic joint damage correlates strongly with long-term functional decline in patients with RA, and therefore preventing progressive joint damage has become a key treatment objective (2, 3). Until recently, the recommended therapeutic strategy was to start with nonsteroidal antiinflammatory drugs (NSAIDs) or low-dose glucocorticoids, and progressively introduce more potent antirheumatic therapies only if the NSAIDs and glucocorticoids were insufficient to control the disease (pyramid approach). In the last 15 years, treatment goals have evolved from a concept of symptom control to a concept of disease control (4). This has resulted in a more aggressive therapeutic approach with earlier introduction of disease-modifying antirheumatic drugs (DMARDs) (5), which have been proven to reduce structural joint damage.

The rationale for a prompt initiation of DMARDs in patients with RA is based on the idea that there is a critical period, a therapeutic window of opportunity, during early stages of the disease when treatment is more effective than later in the course of the disease. This concept covers a short-term effect with better disease activity responses and a long-term effect that would modify the disease permanently to a milder course. Randomized controlled trials (RCTs) have established the short-term efficacy of various DMARD combinations and anti–tumor necrosis factor (anti-TNF) therapies in RA (6-8), and earlier DMARD therapy is more effective than later therapy (9, 10). Although the short-term effects of early DMARD therapy on joint damage are established (11-15), the long-term benefits are controversial (16-18), with some trials demonstrating a persistent effect (19, 20) and others demonstrating no effect during a 5-year followup (21-23). To evaluate the long-term benefit of early DMARD therapy, we conducted a systematic review of the literature to test the hypothesis that early DMARD therapy reduces long-term radiographic progression in patients with RA.

MATERIALS AND METHODS

Conceptualization of selection criteria.

Rates of joint damage can be influenced by individual disease characteristics (i.e., genetic factors, inflammatory activity), efficacy of DMARDs (24), and the time between disease onset and the initiation of DMARDs. These factors must be taken into account when examining the effect of delayed DMARDs on progressive joint damage. Ideally, the most rigorous test to address this issue would be a double-blinded RCT comparing immediate versus delayed initiation of the same DMARD regimen with long-term followup (18). In our review, only 1 RCT with this design involving 23 subjects could be identified (25). Other RCTs addressed the issue but compared different DMARD combinations and sequences (11-13), making it difficult to clarify the effect of delayed treatment versus the effect of a given drug.

In the absence of definitive RCTs, we examined the effect of a delayed DMARD initiation on radiographic damage progression in cohort studies, the next best study design to address this issue (16). Two types of cohort studies have been used by researchers: open followup studies of randomized trials and longitudinal observational studies of patients with early RA. We assumed that long-term effects on radiographic disease progression of early versus delayed treatment could only be studied rigorously when both groups received similar antirheumatic therapy. Therefore data from followup studies of RCTs were suitable only if patients were free to receive any DMARD after trial termination, which effectively tends to balance DMARD regimen between the 2 arms.

The operational definitions of delayed DMARD initiation for the 2 types of cohort studies are contrasted and defined in Figure 1. Study inclusion criteria were as follows: subjects with RA according to ACR (formerly the American Rheumatism Association) criteria (26); disease duration <2 years at enrollment (early RA); cohort studies that had data on the time delay between disease onset and DMARD initiation, and followup studies of at least 1 year after termination of RCTs; duration of 3–24 months between early DMARD group and delayed DMARD group; comparable efficacy of DMARD regimen in treatment arms over followup period; and documentation of radiographic outcome. Exclusion criteria included duplicated data and DMARD regimen not comparable during followup.

RESULTS

Twelve studies, 6 followup studies and 6 cohort studies, with 1,133 patients were included (Table 1). The results of our literature search and the selection process are given in Figure 2. Of 362 candidate studies, 321 were excluded after reviewing the abstracts because they did not meet inclusion criteria (i.e., patients without early RA, no radiographic outcomes, incompatible study design). Twentynine additional studies were excluded after reviewing the articles, primarily because of noncomparable DMARD regimen during the followup period. Eleven investigators were contacted for additional information; of these, 10 responded and 8 provided additional data (see Acknowledgments). The kappa for inclusion and exclusion criteria fulfillment (κ = 0.92) and quality scores (κ = 0.81) showed good agreement.

The summary measure for the followup studies was −0.18 SMD (95% confidence interval [95% CI] −0.39, 0.02) and the summary measure for the cohort studies was −0.21 SMD (95% CI −0.48, 0.06). Because there was no significant difference between these study designs (P = 0.89), we combined them in the analysis. The pooled estimate of effects demonstrated a significant reduction of long-term radiographic progression (−0.19 SMD, 95% CI −0.34, −0.04) (Figure 3) in patients with early RA who initiated DMARD therapy early compared with patients who initiated DMARD therapy later. This finding corresponds to a −33% reduction (95% CI −50%, −16%) in long-term radiographic progression rates in patients receiving early compared with delayed DMARD therapy. Patients in the delayed DMARD group started effective therapy an average of 9 months later than patients in the early DMARD group. The median followup was 3 years.

No evidence for significant publication bias was apparent on a funnel plot or a specific test (Begg's test, P = 0.27). We also assessed the impact of our exclusion criteria by including all possible studies in the analysis and the summary measure was not significantly different (SMD −0.27; 95% CI −0.40, −0.14), suggesting that the results are not dependent on the studies selected.

Some heterogeneity was seen among the 12 studies (test for heterogeneity P = 0.16), but this was mainly due to 1 study (43). If this study was removed, heterogeneity would have disappeared (P = 0.75) and the summary measure would have been qualitatively unchanged (SMD −0.14; 95% CI −0.27, −0.01). We performed sensitivity analyses for possible sources of bias (Table 2) such as study characteristics, dissimilar risk factors of severe disease, or unequal exposure levels. Patients with more aggressive disease, as measured by higher estimated rates of radiographic progression at baseline, seemed to benefit most from early DMARD therapy (P = 0.04). Study characteristics and exposure levels did not significantly influence the rates of radiographic damage, but there was a tendency for studies with shorter disease duration at enrollment, with lower quality and using Larsen's scoring method, to report greater effect sizes. Other possible sources of heterogeneity such as efficacy of initial DMARD therapy (P = 0.82) or the proportion of female patients (P = 0.57) were not significant (data not shown).

DISCUSSION

The long-term rate of joint damage in early RA was reduced in patients who initiated DMARD therapy early compared with patients who initiated DMARD therapy later (−0.19 SMD; 95% CI −0.34, −0.04). An average delay of 9 months in starting DMARDs significantly increased radiographic progression subsequently. In patients with early RA, rates of radiographic progression average 4.3 Sharp score units/year (2). A reduction of 0.19 SMD represents a reduction of 1.4 Sharp units/year or a 33% decrease in radiographic progression rate and would prevent long-term functional disability (2, 44, 45), joint replacement surgery (29), and work disability (46). The effect size from early DMARD initiation was approximately half the effect size observed with methotrexate, which is considered standard therapy in RA (24). Of note, the effect of starting early DMARD therapy was observed several years after the intervention, regardless of subsequent treatment. Our conclusions are statistically robust, as demonstrated by multiple sensitivity analyses (Table 2). These data support the notion of a therapeutic window of opportunity early in the course of RA that results in a sustained benefit in radiographic progression.

To our knowledge, this is the first quantitative metaanalysis of the long-term effect of early DMARD therapy in patients with RA. Prompt DMARD initiation appears to be beneficial on other disease outcomes in patients with early RA. In cohort studies, indirect measures of disease, such as pain, functional disability, and work disability, are less severe in patients receiving early DMARD therapy (44, 47-50). Similarly, patients who receive early aggressive treatment are more likely to remain in remission (51) and have lower mortality rates (52) compared with patients who receive delayed DMARD therapy.

Why early introduction of DMARD therapy changes the long-term rate of structural damage is unknown. In an animal model of RA, anti-TNFα treatment demonstrated efficacy shortly after onset of the disease, but had little effect later in the course of the disease (53). It has been suggested that rheumatoid pannus is reversible in early disease but not later when it becomes self perpetuating (41). The results of the present meta-analysis support this hypothesis because the reduced progression was greater for shorter disease duration (SMD of −0.33 for disease duration <10 months versus SMD of −0.14 for longer disease duration).

There are potential limitations to this analysis related to the inherent quality of the data available. The period between disease onset and DMARD initiation was often ill defined and prone to error (54). However, similar definitions were used for all patients in a given study, so that the delay period was unlikely to be biased. Observational data are prone to confounding by indication, where patients who are more sick are likely to receive more aggressive therapy. However, this is unlikely to explain the observed effect, because patients treated earlier with DMARDs tend to have a more severe disease (47) and are more likely to have progressive structural damage than patients treated later. Therefore, confounding by indication would tend to underestimate the effect of early DMARD therapy. Another type of selection bias could result from completers-only analyses. Although attrition rates are always a concern, there is no evidence for systematic biases, which would imply differential dropout during followup. Confounding could also result from differences in DMARD regimen during followup (drug type, dose, duration). We have addressed this by excluding studies with noncomparable DMARD strategies during followup, but residual confounding is possible, in particular by glucocorticoid use. We defined comparable DMARD efficacy as being less than a 15% difference in the proportion of DMARDs in different efficacy levels. Another definition could have been used, but a sensitivity analysis showed similar results.

Patients with more aggressive disease, as measured by higher estimated rates of radiographic progression at baseline, seemed to benefit most from early DMARD initiation (P = 0.04). Neither different study designs (P = 0.89) nor disparate study quality (P = 0.36) contributed significantly to the study heterogeneity. One study had a significantly larger effect size then the overall summary measure (43). This smaller study was the only study to analyze DMARD initiation among patients with very early RA (≤3 months of disease duration), which could explain the larger effect size. To account for the heterogeneity among studies, we used a random effect model, which does not assume a uniform effect across all studies. Overall, these limitations underline the necessity of more research on this important clinical issue.

Radiographic end points in RA may not be normally distributed, and therefore the use of parametric statistics may not be appropriate. However, change scores of radiographic outcomes and yearly rates of radiographic progression are usually less skewed and have been used in meta-analysis (24, 55). Further imprecision could result from determining rates of radiographic progression from mean change scores instead of individual patient data. Two studies (22, 41) reported cross-sectional radiographic data only, but a sensitivity analysis of these studies compared with studies with longitudinal radiographic data demonstrated no significant difference (P = 0.84). Various scoring methods were used to measure structural joint damage. A sensitivity analysis of studies that used Larsen's scores versus Sharp's scores showed that both methods give qualitatively the same answer, even if the effect size in studies that used Larsen's scores tended to be nonsignificantly larger (P = 0.52).

In summary, a window of opportunity for initiating DMARD therapy appears to exist in early RA, which is associated with significantly improved long-term outcomes. Our results suggest that early DMARD therapy has a durable effect on the rate of radiographic progression of RA up to 5 years after DMARD initiation. To delay DMARD therapy for as little as 9 months after disease onset results in sustained radiographic deterioration, which underlines the importance of starting DMARD therapy as early as possible. Patients with more aggressive RA and early erosive disease seem to benefit most from prompt DMARD initiation. Delayed initiation of DMARDs may result from a variety of factors, from the organization and reimbursement of health services to patient and physician knowledge and preferences (56). Our study suggests that early referrals to a knowledgeable physician or a rheumatologist for definitive diagnosis, severity assessment, and early initiation of DMARD treatment could improve the long-term outcome of patients with RA. Physicians and patients should be aware of the importance of initiating DMARD therapy early in aggressive RA so the period between disease onset of RA and initiation of antirheumatic therapy is minimized.