Disease activity early in the course of treatment predicts response to therapy after one year in rheumatoid arthritis patients

Authors

  • Daniel Aletaha,

    1. Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
    • Dr. Aletaha has received consulting or speaker's honoraria (less than $10,000 each) from Abbott, Sanofi-Aventis, Schering-Plough, and Wyeth.

  • Julia Funovits,

    1. Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
  • Edward C. Keystone,

    1. Rebecca MacDonald Centre for Arthritis and Autoimmunity, Mount Sinai Hospital, and University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
    • Dr. Keystone has received consulting fees and honoraria from Hoffmann-La Roche, speaker's fees from Abbott, and consulting fees from Amgen/Wyeth (more than $10,000 each).

  • Josef S. Smolen

    Corresponding author
    1. Medical University of Vienna, and Hietzing Hospital Vienna, Austria
    • Medical University of Vienna, Division of Rheumatology, Internal Medicine III, Waehringer Guertel 18-20, A-1090 Vienna, Austria
    Search for more papers by this author
    • Dr. Smolen has received consulting or speaker's honoraria (less than $10,000 each) from Abbott, Amgen, Centocor, Schering-Plough, Sanofi-Aventis, and Wyeth.


Abstract

Objective

To assess whether disease activity levels at treatment initiation or during the first 3 months of therapy predict disease activity at 1 year after treatment initiation.

Methods

Pooled patient data from early rheumatoid arthritis (RA) clinical trials (n = 1,342) of methotrexate (MTX), tumor necrosis factor (TNF) inhibitor monotherapy (adalimumab and etanercept), and the combination of the two (adalimumab or infliximab plus MTX) were used for the primary analyses. Pooled data from clinical trials of MTX and of TNF inhibitor plus MTX in late RA (n = 712) were used for validation of the results. Disease activity was primarily assessed using the Simplified Disease Activity Index (SDAI); in addition, we calculated the Disease Activity Score 28-joint assessment (DAS28) and the Clinical Disease Activity Index (CDAI). Associations of disease activity measures at baseline and at 1, 2, 3, and 6 months with disease activity values or disease activity states at 1 year were performed using Spearman's rank correlation, analysis of variance, and diagnostic testing procedures, including receiver operating characteristic (ROC) curve analyses, and probit analysis.

Results

Correlations with SDAI values at end point were significant (P < 0.0001) at baseline, and increased to r = ∼0.6 at 3 months. The area under the ROC curve indicated a high diagnostic test yield with respect to the 1-year outcome (area under the ROC curve ∼0.8). At all time points, including baseline, the group of patients who achieved remission at 1 year had lower average SDAI values than did those whose disease activity was high at 1 year. The groups achieving low or moderate disease activities at 1 year had SDAI values lying between. Baseline disease activity was less associated with disease activity at the end point for treatment with TNF inhibitor plus MTX, indicating its effectiveness over a broader range of baseline disease activity, but the association with end point disease activity was similar to that in the MTX treatment group at 1 month after treatment initiation. The data were similar when scores on the DAS28 and CDAI were used and were fully validated in the independent cohort of patients with late RA.

Conclusion

The level of disease activity at baseline and especially during the first 3 months of treatment is significantly related to the level of disease activity at 1 year. Patients who reach a moderate or low disease activity status after 3–6 months of therapy may require switching to alternative therapies. Our findings indicate that intensive and dynamic treatment strategies that include a closer look at disease activity at 3 months in patients with early and late RA is warranted.

The natural course of rheumatoid arthritis (RA) is highly variable (1–8). This natural course can be interfered with by disease-modifying antirheumatic drugs (DMARDs), the most effective of which today are high-dose methotrexate (MTX), biologic agents, such as tumor necrosis factor (TNF) inhibitors, and a combination of the two in particular (9–13). However, clinical trials and clinical practice reveal that despite all current treatment successes, many patients continue to have significant disease activity (14, 15).

Over the last decade, the principal assessment tool for treatment response has been the American College of Rheumatology (ACR) preliminary criteria for improvement in RA (16). These ACR response criteria evaluate the percentage of change from baseline irrespective of both the baseline disease activity and the disease activity status at end point. Likewise, the European League Against Rheumatism (EULAR) response criteria, which integrate the response (improvement) and the residual disease activity (state), have also been used (17). However, currently there are no established means to predict the clinical response to therapy (18).

In contrast to the inability to predict the clinical efficiency of therapeutic measures, it has been possible to predict joint destruction by assessing clinical data and levels of acute-phase reactants. Previous studies have shown that cumulative disease activity was associated with the degree of joint destruction (6, 19, 20). Moreover, in 2 clinical trials (11, 21), the C-reactive protein (CRP) level at the time of institution of DMARD therapy was predictive of radiographic outcome (22). Likewise, disease activity at ∼3 months from the start of therapy was found to be related to changes in radiographic scores at 1 year of treatment (22). These latter data indicate that disease activity assessed at a single time point (e.g., 3 months after treatment initiation or even before treatment initiation) is predictive of the progression of joint damage. Of significance, this relationship was found irrespective of the type of therapy used, although the magnitude of progression of joint destruction was significantly lower in those receiving TNF inhibitor plus MTX as compared with those receiving MTX alone (23).

However, reduction of functional disability is the most important treatment response to the patient and to society, and the reduction is greatest when overall disease activity is effectively reduced (3, 6, 24). In fact, in states of clinical remission, functional activity normalizes in patients who do not have joint damage. In patients with joint destruction, residual disability is primarily governed by the accumulated joint damage that currently cannot be reversed (24). Thus, improving clinical disease activity to as low a degree as possible is the pivotal aim of therapeutic intervention in RA because it will result in both the retardation or halting of joint damage and the maximal reduction of disability. While a variety of predictors of joint damage have been determined over the last decades, especially using biomarkers (25–28), it may be even more important for practicing clinicians to be able to predict the clinical response to therapy. Consistent with intensive and dynamic treatment strategies, the response to a therapeutic intervention should be predictable, ideally before the intervention, but at least after only little delay.

We hypothesized that levels of disease activity shortly after the start of DMARD therapy can be used to estimate the treatment response and the clinical disease activity state at end point. In this study, we evaluated this hypothesis using data from a large cohort of patients treated in double-blind randomized controlled clinical trials of TNF inhibitors.

PATIENTS AND METHODS

Data sources.

We obtained patient-level data from the sponsors of several RA clinical trials: Centocor (Malvern, PA) provided data from the Anti–TNF Trial in RA with Concomitant Therapy (ATTRACT) study of infliximab and MTX versus placebo and MTX in patients with an inadequate response to previous treatment with MTX (21) and from the Active-Controlled Study of Patients Receiving Infliximab for the Treatment of RA of Early Onset (ASPIRE) trial of initiating infliximab plus MTX versus MTX alone in MTX-naive patients with early RA of ≤3 years' duration (11); data from the DE019 trial in patients with active RA despite MTX therapy who were treated with MTX plus adalimumab versus placebo (29) and from the PREMIER trial of adalimumab versus MTX versus the combination of the two in MTX-naive patients with early RA (13) were made available by Abbott (Abbott Park, IL); Amgen (Thousand Oaks, CA) provided data from the Early RA (ERA) trial of etanercept versus MTX (30); and data from the MTX arm of the Trial of Etanercept and Methotrexate with Radiographic Patient Outcomes (TEMPO) study (12) were made available by Wyeth (Philadelphia, PA). We were provided data from the sponsors of those studies on a random sample of 80–90% of subjects from these trials.

All patients had active RA at enrollment in these studies, with requirements that they have more than 6–10 swollen joints and more than 6–12 tender joints (using a 68-joint count). With the exception of the ASPIRE trial, elevations in the levels of acute-phase reactants were also required (CRP ≥2.0 mg/dl and/or erythrocyte sedimentation rate ≥28 mm/hour). Patient demographics have been presented in the respective publications and are summarized in Table 1 for the cohort of patients in the present study.

Table 1. Baseline characteristics of the study patients*
 RA patient groupTreatment groupAll study patients (n = 2,054)
Early RA, analysis cohort (n = 1,342)Late RA, validation cohort (n = 712)MTX (n = 572)TNF inhibitor (n = 291)TNF inhibitor plus MTX (n = 1,191)
  • *

    The analysis cohort consisted of patients with early rheumatoid arthritis (RA) who were methotrexate (MTX) naive, and the validation cohort consisted of patients with longstanding (late) RA who had active disease despite MTX therapy (see Patients and Methods for details). Except where indicated otherwise, values are the mean ± SD. TNF = tumor necrosis factor.

  • Based on a 100-mm visual analog scale.

Female, %737576757274
Rheumatoid factor positive, %768177857878
Age, years50.4 ± 13.254.1 ± 12.050.4 ± 13.051.5 ± 13.052.3 ± 12.851.7 ± 12.9
Disease duration, years1.0 ± 1.010.0 ± 8.52.4 ± 3.71.1 ± 1.36.6 ± 8.35.0 ± 7.2
Tender joint count (68 joints assessed)15.1 ± 6.615.2 ± 6.915.1 ± 6.716.1 ± 6.714.9 ± 6.715.1 ± 6.7
Swollen joint count (66 joints assessed)12.0 ± 5.913.6 ± 5.912.6 ± 6.114.0 ± 5.912.2 ± 5.912.6 ± 6.0
Patient's global assessment of disease activity62.0 ± 23.556.5 ± 22.961.4 ± 23.063.9 ± 21.758.6 ± 24.060.1 ± 23.5
Evaluator's global assessment of disease activity64.5 ± 18.361.5 ± 16.962.6 ± 18.064.1 ± 18.363.7 ± 17.763.5 ± 17.9
Pain scores61.3 ± 22.858.9 ± 21.959.0 ± 23.661.4 ± 23.561.0 ± 21.960.5 ± 22.6
C-reactive protein (mg/dl)3.4 ± 3.82.4 ± 2.63.2 ± 3.73.5 ± 3.92.8 ± 3.33.0 ± 3.5
Radiographic score (Sharp/van der Heijde method)14.4 ± 18.060.2 ± 56.419.4 ± 30.414.7 ± 17.138.6 ± 48.329.8 ± 41.9
Disease Activity Score 28-joint assessment6.4 ± 1.06.3 ± 1.26.4 ± 1.16.6 ± 1.06.3 ± 1.06.4 ± 1.1
Simplified Disease Activity Index43.1 ± 14.242.9 ± 13.943.3 ± 14.446.4 ± 14.442.1 ± 13.743.1 ± 14.1
Clinical Disease Activity Index39.7 ± 12.640.6 ± 13.140.1 ± 12.942.9 ± 12.939.3 ± 12.640.0 ± 12.8
Health Assessment Questionnaire score1.5 ± 0.61.5 ± 0.61.5 ± 0.61.5 ± 0.61.5 ± 0.61.5 ± 0.6

Study variables.

For all of these trials, we identified patients who had complete sets of data at the baseline visit and the visits at 1, 2, 3, and 12 months (n = 2,054). Since visits were planned at slightly different intervals in the various trials, visits at 2 or 4 weeks were regarded as the 1-month visit, those at 6 or 8 weeks as the 2-month visit, those at 10 or 12 weeks as the 3-month visit, and those at 52 and 54 weeks as the 12-month visit.

Based on the individual instruments for the assessment of disease activity, we calculated values for the following composite indices: the Simplified Disease Activity Index (SDAI) (31), the Clinical Disease Activity Index (CDAI) (31, 32), and the Disease Activity Score with 28-joint assessment (DAS28) (33).

Primary analyses and validation.

To validate the findings of our study with respect to the different trial designs, we separated the early RA trials (analysis set) from the late RA trials (validation set). Thus, the analysis set consisted of patients from the ASPIRE, PREMIER, and ERA trials (n = 1,342), which were studies in early RA patients who were MTX naive, and the validation set consisted of patients from the ATTRACT, DE019, and TEMPO trials (n = 712), which were studies in patients with longstanding RA who had active disease despite MTX therapy. In addition, we performed subgroup analyses for all patients who received MTX monotherapy and those who received the combination of a TNF inhibitor plus MTX. The term “TNF inhibitor plus MTX” represents combinations of adalimumab or infliximab plus MTX, since data on etanercept plus MTX were not available for this analysis.

Statistical analysis.

For all statistical analyses, only patients with complete data at all visits (baseline and 1, 2, 3, and 12 months) were included. All followup data were used “as observed.” Wherever possible, we used parametric test statistics; nonparametric tests were performed when the respective data were not normally distributed. To simplify the presentation of our results, we focused on 1 of the above disease activity measures, the SDAI, as an easy way to calculate disease activity, and the remission criteria for the SDAI are stringent (34). All of the analyses were similarly performed using scores on the CDAI and the DAS28, and some of these results are shown for comparison.

We first correlated disease activity measures at baseline with disease activity at the various followup time points using Spearman's rank correlation. We next used a categorical approach based on the notion that achieving low disease activity and remission are ultimate goals of RA therapy (34–41). We therefore categorized the disease activity status of the patients at 1 year as follows: remission (SDAI ≤3.3), low disease activity (SDAI 3.3 to ≤11), moderate disease activity (SDAI >11 to ≤26), or high disease activity (SDAI >26). We then assessed the differences in average baseline SDAI scores across these 4 outcome groups using analysis of variance (ANOVA). Subsequently, we also used the 1-, 2-, and 3-month data in an analogous analysis.

To assess the value of disease activity scores at baseline, as well as at 1, 2, 3, and 6 months, as a predictor of low disease activity or remission at 1 year, we performed receiver operating characteristic (ROC) curve analyses and used probit plots to visualize the proportion of patients who will end up in remission/low disease activity or in moderate/high disease activity, depending on the SDAI value at 12 weeks. For all analyses, we used the SAS software package, version 9.1.3 (SAS Institute, Cary, NC).

RESULTS

Correlation of disease activity at baseline and at 1, 2, 3, and 6 months of therapy with disease activity at 12 months of therapy in MTX-naive patients with early RA.

All patients (n = 1,342) had active disease at baseline (Table 1). With therapy, the SDAI scores in early RA patients decreased significantly (P < 0.001), from a mean ± SD of 43.1 ± 14.2 at baseline to a mean of 13.9 ± 12.9 at 12 months. The corresponding baseline and 12-month values for the CDAI were 39.7 ± 12.6 and 12.8 ± 12.4, and for the DAS28, the values were 6.4 ± 1.0 and 3.6 ± 3.5, respectively. These mean values are consistent with a high level of disease activity at baseline and a moderate level of disease activity at 12 months for all 3 scoring systems (42).

The correlation between the SDAI scores at baseline and at 12 months was low but significant (r = 0.22); the correlation values increased considerably at subsequent visits, reaching an r value of 0.59 at week 12 after initiation of therapy. This indicates that while the response at 1 year is related to baseline levels of disease activity, the degree of association increases with the incorporation of information regarding early responsiveness to therapy.

Subgroup analyses for MTX and TNF inhibitor monotherapy (data not shown), as well as for TNF inhibitor plus MTX revealed similar results. At later time points, such as at 6 months, the correlations with disease activity at 1 year were further increased (Table 2). Overall, however, only a relatively small number of patients had newly met the ACR 20% improvement criteria (achieved an ACR20 response) between 3 months and 6 months of treatment (ACR20 response in early RA patients 55.1% and 61.8% and in late RA patients 39.9% and 45.5% at 3 and 6 months, respectively). In the TNF inhibitor plus MTX combination therapy group, correlations of baseline values with the values at 1 year were lower than those in the MTX monotherapy group (Table 2). Similar results were obtained for the CDAI (data not shown) and the DAS28 scores. Correlation coefficients for the DAS28 at 1 year were r = 0.28, r = 0.43, r = 0.52, and r = 0.59 for DAS28 scores at baseline, 1 month, 2 months, and 3 months of therapy, respectively.

Table 2. Correlations of disease activity at treatment initiation (baseline) and during early followup with disease activity at 1 year in patients with early and late RA*
AssessmentEarly RA (analysis cohort)Late RA (validation cohort)All study patients (n = 2,054)
All early RA patients (n = 1,342)TNF inhibitor plus MTX (n = 589)MTX only (n = 462)All late RA patients (n = 712)TNF inhibitor plus MTX (n = 602)
  • *

    The analysis cohort consisted of patients with early rheumatoid arthritis (RA) who were methotrexate (MTX) naive, and the validation cohort consisted of patients with longstanding (late) RA who had active disease despite MTX therapy (see Patients and Methods for details). Disease activity was determined according to the Simplified Disease Activity Index. Correlation coefficients were determined by Spearman's rank correlation. All correlations are significant at P < 0.0001.

  • Includes patients receiving tumor necrosis factor (TNF) inhibitor monotherapy.

Baseline0.2200.1410.2560.3400.3050.256
1 month0.4320.4180.4360.4460.4470.436
2 months0.5160.5130.5330.5350.5230.533
3 months0.5930.6000.5930.5530.5240.593
6 months0.7190.7090.6820.6060.5790.690

Similar levels of relative improvement in disease activity at 12 months regardless of disease activity levels at baseline.

The median percentage of change in the SDAI scores according to quartiles of SDAI scores at baseline (≤34.3, 34.3–43.3, 43.3–54.9, and ≥54.9) were 72%, 79%, 75%, and 79%, respectively, in all study patients. These similar relative changes suggested a trend toward a decreasing ability to achieve favorable disease activity status with increasing disease activity levels at the start of treatment. In fact, and consistent with the results on the correlation between baseline and 1-year SDAI values described above, disease activity at 1 year increased with increasing disease activity at baseline, with median SDAI values across the 4 quartiles of 7.6, 7.9, 11.8, and 13.5, respectively (P < 0.0001 by ANOVA). Moreover, the median 1-year SDAI levels for the first 2 quartiles were in the low disease activity category (SDAI ≤11), and those in the last 2 quartiles were in the moderate disease activity category (SDAI 11 to ≤26) (34).

Patients receiving a TNF inhibitor plus MTX had lower residual 12-month disease activity in relation to baseline quartiles than did patients receiving either MTX or TNF inhibitor monotherapy. The association of absolute and relative changes was similar to that in all patients and was not different when the CDAI or the DAS28 scores were used (data not shown).

Prediction of remission or low disease activity at 1 year based on disease activity during the first 3 months after treatment initiation.

The ROC curve analyses demonstrated an inverse relationship between disease activity and the possibility of remission. The results showed that baseline SDAI values were not useful as predictors of remission, although they still showed an area under the ROC curve (AUC) of 0.62 (95% confidence interval [95% CI] 0.54–0.69) for MTX monotherapy and 0.54 (95% CI 0.49–0.59) for TNF inhibitor plus MTX combination therapy, which is significantly better than a test without utility (Figures 1A and B). However, the ROC curves at any of the visits subsequent to treatment initiation demonstrated good utility with AUC values for 1, 2, 3, and 6 months of therapy, for example, as for the TNF inhibitor plus MTX group, of 0.71, 0.74, 0.77, and 0.85, respectively (Figure 1B). Consistent with these data, baseline SDAI values and subsequent values had a strong direct relationship to the residual high disease activity state at 1 year, with an AUC of 0.65, 0.74, 0.77, and 0.82, respectively.

Figure 1.

Receiver operating characteristic (ROC) curve analyses for achievement of remission at 12 months in patients with early rheumatoid arthritis (RA) (analysis cohort) and in patients with late RA (validation cohort). A, Patients with early RA receiving MTX (n = 462). B, Patients with early RA receiving TNF inhibitor plus MTX (n = 589). C, All patients with late RA regardless of treatment (n = 763). The increasing area under the ROC curve (AUC) corresponds to a higher diagnostic test yield; an AUC of 0.5 is a useless test, yielding 1 misclassified patient per 1 correctly classified patient. 95% CI = 95% confidence interval; SDAI = Simplified Disease Activity Index.

Figures 1A and B show the ROC analyses for the presence of remission at 1 year in patients with early RA treated with MTX and with TNF inhibitor plus MTX. It can be seen that for the TNF inhibitor plus MTX combination, the test performance of disease activity values during month 1 to month 3 was numerically higher than that for MTX monotherapies. In contrast, the test performance of baseline values was numerically higher for MTX than for TNF inhibitor plus MTX, which is consistent with the correlations presented in Table 2. This indicates that there was a greater range of responses in relation to baseline disease activity in the group receiving TNF inhibitor plus MTX, which therefore reduced the value of baseline disease activity as a test for remission at 1 year. However, the rapid onset of action of TNF inhibitor plus MTX therapy increased the value of the disease activity assessment at 1 month to 3 months after starting treatment.

These data were also visualized using a probit regression model that enables determination of sensitivities and specificities of various SDAI cut-point values at 3 months with regard to dichotomous outcomes (presence or absence of remission/low disease activity at 1 year). The results of this analysis (Figure 2) showed that the probability of achieving future remission or low disease activity decreased with higher levels of disease activity as determined by the SDAI at 3 months. In other words, the odds ratio for achieving a state of remission at 1 year was 4, comparing the first and the fourth SDAI quartiles.

Figure 2.

Accuracy of classification of disease activity status at 1 year based on the disease activity status at 3 months. Plots were derived from a probit regression model, using the Simplified Disease Activity Index (SDAI) values at 3 months to predict disease activity status at 1 year. At a given level of disease activity at 3 months, the curves indicate the probability of achieving remission (REM)/low disease activity (LDA) or moderate disease activity (MDA)/high disease activity (HDA) at 1 year in A, patients with early rheumatoid arthritis (RA) and B, patients with late RA. For example, if a patient has an SDAI of 40 at 3 months, the probability that this patient will have moderate or high disease activity at 1 year is >80% (red curve), and the probability that this patient will be in remission or have low disease activity at 1 year is <20% (green curve).

Relationship of lower levels of disease activity at baseline and during early stages of therapy to remission at 1 year.

We next evaluated the patients with early RA who achieved remission at 1 year. As can be seen in Figure 3A, among the patients whose RA was in remission at 1 year, 35% were in the lowest SDAI quartile at baseline, whereas only 13% were in the highest quartile at baseline (P < 0.0001 by ANOVA). In absolute values, patients achieving remission at 1 year had the lowest mean disease activity at baseline, but especially at 1, 2, and 3 months after initiation of therapy (Figure 3B). The average SDAI values over time were clearly different in patients who achieved remission or had low, moderate, or high levels of disease activity (Figure 3B). Moreover, as a group, patients who achieved low disease activity or remission at 1 year were already well into the moderate disease activity category (SDAI 11 to ≤26) at 3 months (Figure 3B).

Figure 3.

Disease activity at baseline and during the first year of study in patients with early rheumatoid arthritis (RA). A, Proportions of RA patients in remission (REM) or with low disease activity (LDA), moderate disease activity (MDA), or high disease activity (HDA) at 1 year, as classified by quartiles of disease activity at baseline (high in all patients). Disease activity status and quartiles were based on the Simplified Disease Activity Index (SDAI) values. For example, among patients in remission at 1 year, 35% were from the first quartile (lowest baseline disease activity), 30% were from the second quartile, 22% were from the third quartile, and 13% were from the fourth quartile (highest baseline disease activity). Results are shown for all patients with early RA (all treatment groups). B, Absolute SDAI values over time in patients who achieved remission or low, moderate, or high disease activity at 1 year. Throughout the course of study, patients achieving remission had the lowest average disease activity, patients remaining in a high disease activity state had the highest disease activity, and those reaching low or moderate disease activity had intermediate levels of disease activity.

Validation of results in patients with late RA as well as by comparison with additional disease activity scores.

To validate the data obtained in patients with early RA, we analyzed the data from patients with late RA from the ATTRACT, DE019, and TEMPO trials in the same manner as above. All of these analyses showed that the relationships between the SDAI values at baseline, 1 month, 2 months, 3 months, and 6 months and the SDAI values at 1 year in the patients with late RA were similar to those in patients with early RA. For reasons of simplicity, we focus here on selected analyses.

When ROC curve analyses for the presence of remission at 1 year were performed (Figure 1C), almost identical test performances were obtained in patients with established RA as were obtained in patients with early RA (Figures 1A and B). The AUC values for the patients with late RA were 0.62, 0.70, 0.78, and 0.83 for the SDAI values at baseline, 1 month, 2 months, and 3 months, respectively. Likewise, correlations between SDAI values at various time points, and especially baseline, 1 month, 2 months, 3 months, and 6 months with the SDAI value at 1 year were within similar ranges in patients with late RA as were found in patients with early RA (Table 2). These data not only validated the results obtained in the early RA population, but also revealed that disease duration does not bear a major influence on the associations obtained (i.e., these findings support the generalizability of our results).

Finally, we performed all analyses using the CDAI and the DAS28 scores. All associations obtained with the SDAI values were also seen with these 2 indices (data not shown). These findings further support the associations that were identified with the SDAI data.

DISCUSSION

One of the most tantalizing questions when treating patients with RA in clinical trials or in practice is, How can one predict a clinical response? While it has been shown that progression of joint damage is related to baseline and especially cumulative disease activity (3, 22, 43), reliable prediction of clinical responsiveness has not been accomplished hitherto. However, it is primarily the clinical response to therapy that is of major interest in the treatment of RA patients, since disease activity is directly and indirectly (via joint damage) associated with disability (3, 6, 24).

In the present study, we were able to evaluate a large set of patients from several clinical trials of TNF inhibitors. Using studies of patients with early RA from the ASPIRE, PREMIER, and ERA trials of infliximab, adalimumab, and etanercept, we first analyzed all study arms combined and separately for responses to MTX monotherapy and to the TNF inhibitor plus MTX. The results obtained revealed that the likelihood of achieving remission or low disease activity at 1 year after treatment initiation decreased with increasing disease activity at baseline. This association was observable despite the fact that the inclusion criteria required that all patients have high levels of disease activity at trial entry. More pertinently, the correlation with remission or low disease activity at 1 year increased considerably across subsequent visits after baseline, and clinical disease activity at 3 months from the start of therapy was highly correlated with disease activity at 1 year.

We used the SDAI as the main disease activity score. The SDAI and CDAI composite scores have been shown to be valid and reliable across cultures (34, 42, 44–46) and to relate best to the physician's decision regarding change of treatment (47, 48). Moreover, the SDAI currently constitutes the most stringent composite disease activity index for the definition of remission (34, 41, 49). However, all correlations were similar for the DAS28 and the CDAI scores as for the SDAI scores.

An important question in following treatment responses in RA is whether achieving a change in disease activity or a particular status (e.g., remission) should be targeted. Although this was not the focus of the present analyses, it was of interest that the proportional changes in disease activity at 1 year were quite similar (between 72% and 79%) when the 4 patient groups with increasing disease activity were compared. In other words, more patients in the lowest baseline quartile were likely to reach low disease activity or remission (the same relative change leads to lower end point values for the lowest quartile compared with the higher quartiles). There are 2 important implications of this finding. First, assessing patient response by determining proportional response rates, as is done with the ACR criteria (16), likely leads to comparable results regardless of the disease activity level at baseline. Second, lower levels of disease activity at baseline carry a higher probability of reaching remission at 1 year of treatment, and conversely, higher levels of disease activity at baseline bear a higher likelihood of having high levels of disease activity at 1 year of treatment, even if significant improvement was achieved.

One of the central findings of this study was the predictive capacity of disease activity achieved at 3 months. The ROC curve and correlation analyses revealed highly significant associations of SDAI values in patients who achieved remission at 1 year with their SDAI values at 3 months. While disease activity at later time points, such as 6 months, further increased the correlation with disease activity at 1 year, this is not surprising, given that the 6-month time point is frequently used as a primary end point for signs and symptoms in clinical trials. Therefore, the correlation of disease activity at 1 year with SDAI levels at 3 months is particularly remarkable. Moreover, 60% of patients who reached remission were already in the lowest SDAI quartile at 3 months (data not shown). This suggests that for the majority of patients, the response achieved within the first 3 months is highly predictive of the degree of clinical outcome at 1 year.

Thus, patients whose disease activity during therapy reaches at least a moderate level within 3 months are very likely to escape high levels of disease activity at 1 year. In contrast, patients who do not achieve a moderate or low disease activity level after 3 months of treatment, especially when taking traditional DMARDs such as MTX, are not likely to achieve a good outcome in terms of low disease activity or remission at 1 year. It is conceivable that such patients may benefit from rapid switching to alternative therapies. In fact, given that in patients with longstanding RA, the relationship between disease activity at baseline and especially at 3 months with disease activity at 1 year was similar to that observed in MTX-naive patients, any reduction of disease activity by a treatment regimen that may not be fully effective is likely to enable an improved outcome with a subsequent treatment regimen.

This is consistent with novel dynamic treatment strategies (50–52). Moreover, rheumatologists are always looking for ways to improve the benefit of therapies and reduce their risk. By predicting response at 3 months, we can improve therapeutic benefit by making early decisions and avoiding unnecessary costs and safety exposures. The data also indicate that a 3-month course of MTX may be appropriate in all patients regardless of baseline disease activity: those with good prospects for achieving low disease activity or remission will benefit from such therapy, while those with worse prospects may still experience a reduction of their disease activity from baseline, even if it is only a moderate reduction. This lower disease activity level resets the baseline for subsequent treatment with a biologic agent plus MTX and increases the likelihood of achieving a satisfying disease activity status.

Several limitations need to be borne in mind when interpreting the study data. First, these results were derived for groups of patients rather than for individual patients. Although the risk of not achieving remission is up to 4 times higher in patients with very high disease activity at baseline as compared with those with lesser degrees of high disease activity, the individual RA patient with very high activity may still respond well to treatment. Second, disease activity is a continuum, and dividing baseline disease activity into quartiles is an artificial approach (53). Third, these same limitations pertain to the categorization of disease activity states at end point (12 months). Fourth, data on etanercept plus MTX therapy were not available for this investigation and therefore the TNF inhibitor plus MTX analyses pertain to anti-TNF antibodies, although it is unlikely that examination of etanercept plus MTX therapy would have led to discrepant results, given the overall similarity of published trial data among these 3 TNF blockers. Fifth, with regard to traditional DMARDs, only MTX was analyzed here. Other traditional DMARDs, such as sulfasalazine or leflunomide, were not examined, and some of them may have different response rates than MTX and may take longer to achieve the maximum response. Nevertheless, in current clinical practice, MTX is the major anchor drug. Sixth, this is an analysis of responders at 1 year, rather than an analysis of all patients who entered the trial. However, imputation of 1-year data for the dropouts (e.g., by using the last observation carried forward method) would likely have overestimated the correlations of later data points with earlier ones. In addition, patients dropped out of these trials not only for lack of efficacy, but also for adverse events. Yet, even when we looked at the relationship between disease activity at 3 months and baseline activity in all patients, the results were very similar to those in the completers at 1 year (data not shown).

Importantly, the results initially derived for patients included in the trials of early RA were fully validated in patients who participated in the trials of established, late RA. Thus, the predictive capacity of disease activity in the early course of therapy constitutes a general phenomenon regardless of disease duration. Another essential observation of this study relates to the validity of the results to several types of therapy: MTX monotherapy, TNF inhibitor monotherapy, and combination therapy with a TNF inhibitor plus MTX. Interestingly, the association of baseline disease activity with end point disease activity was abrogated by treatment with the combination of a TNF inhibitor plus MTX. This indicates that treatment with a TNF inhibitor plus MTX reduces disease activity across a much broader range of baseline disease activity than does treatment with MTX alone. Moreover, the SDAI values at end point, as well as the frequencies of remissions, were higher among patients receiving a TNF inhibitor plus MTX than in those receiving MTX monotherapy, which is consistent with the published trial results11–13, 21, 29, 30.

In summary, we have shown that the disease activity status at 1 year of therapy is related to baseline disease activity and, in particular, to disease activity at 3 months of treatment. This is true for both early RA patients who were DMARD-naive and patients who had established disease and had had an inadequate response to MTX. Thus, baseline disease activity, and especially disease activity after 3 months of DMARD or TNF inhibitor therapy, determines the ultimate treatment response to an important extent. The findings from this study indicate that if a moderate or low disease activity state is not reached after 3 months of therapy, particularly if the patient is taking traditional DMARDs, switching to an alternative treatment regimen may provide a better outcome. This is especially true for MTX monotherapy, since radiographic damage will progress with active disease during MTX treatment, while therapy with a TNF inhibitor plus MTX still provides a significant benefit against joint damage, even if low levels of disease activity have not yet been achieved (22, 54). Using these simple clinical methods can be a valuable clinical guide in making treatment decisions, since they require neither specific equipment nor costly analyses. The data also imply that the aim to improve RA disease activity can potentially be achieved by using sequential drug therapy, taking advantage of the fact that disease activity may become progressively reduced with every new treatment course.

It is conceivable that there are quantitative and/or qualitative differences in pathogenetically relevant events between patients who start with highly active disease and continue to have highly active disease and those who achieve low disease activity or remission, especially when starting with lower levels of disease activity. Thus, segregating patients by the simple clinical methods outlined in this study may be of interest in the search for suitable predictive biomarkers.

AUTHOR CONTRIBUTIONS

Drs. Aletaha and Smolen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Aletaha, Smolen.

Acquisition of data. Aletaha, Smolen.

Analysis and interpretation of data. Aletaha, Smolen.

Manuscript preparation. Aletaha, Keystone, Smolen.

Statistical analysis. Aletaha, Funovits, Smolen.

Acknowledgements

We thank Abbott, Amgen, Centocor, and Wyeth for providing the data on their study patients, and we thank Drs. Oscar Segurado, Shao-Lee Lin, Dan Baker, Saeed Fatenejad, and David MacPeek for careful reading of the manuscript and valuable suggestions.

Ancillary