SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

Objective

To determine the prognostic significance of data collected early after starting certolizumab pegol (CZP) to predict low disease activity (LDA) at week 52.

Methods

Data from 703 CZP-treated patients in the Rheumatoid Arthritis Prevention of Structural Damage 1 (RAPID 1) trial through week 12 were used as variables to predict LDA (Disease Activity Score in 28 joints–erythrocyte sedimentation rate ≤3.2) at week 52. We identified variables, developed prediction models using classification trees, and tested performance using training and testing data sets. Additional prediction models were constructed using the Clinical Disease Activity Index (CDAI) and an alternate outcome definition (composite of LDA or American College of Rheumatology criteria for 50% improvement [ACR50]).

Results

Using week 6 and 12 data and across several different prediction models, response (LDA) and nonresponse at 1 year were predicted with relatively high accuracy (70–90%) for most patients. The best performing model predicting nonresponse by 12 weeks was 90% accurate and applied to 46% of the population. Model accuracy for predicted responders (30% of the RAPID 1 population) was 74%. The area under the receiver operating curve was 0.76. Depending on the desired certainty of prediction at 12 weeks, ∼12–25% of patients required >12 weeks of treatment to be accurately classified. CDAI-based models and those evaluating the composite outcome (LDA or ACR50) achieved comparable accuracy.

Conclusion

We could accurately predict within 12 weeks of starting CZP whether most established rheumatoid arthritis (RA) patients with high baseline disease activity would likely achieve/not achieve LDA at 1 year. Decision trees may be useful to guide prospective management for RA patients treated with CZP and other biologics.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

Many patients with rheumatoid arthritis (RA) exhibit a rapid clinical response to tumor necrosis factor α (TNFα) inhibitor therapy, most within 3 months of initiating treatment (1–10). However, limited information is available to determine which factors, measured at baseline or early after starting a new RA medication, will predict a good long-term response, and results are inconsistent. In one study of patients with recent-onset RA treated with methotrexate, a model based on sex, rheumatoid factor, smoking status, Disease Activity Score (DAS), and several genetic polymorphisms was able to predict treatment efficacy (11). However, a later study including patients with early inflammatory polyarthritis found routine clinical and laboratory factors to be poor at predicting the outcome of treatment with methotrexate (12). Several pharmacogenetic studies have suggested that a single nucleotide polymorphism in the promoter region of TNFα is significantly associated with response to TNF inhibitor therapy (13–15). However, no single marker has yet emerged that accurately predicts a response in a large cohort of patients. For that reason, clinical and laboratory information is likely to remain paramount in assessing the expected longer-term response to RA medications, ideally in conjunction with yet to be determined genetic or biomarker-based tests.

Even if predictors of long-term response for individual patients can be identified, they may not be easily applicable in clinical practice. Previous reports have shown that US rheumatologists rarely measure the indices for disease evaluation commonly used in clinical trials, e.g., American College of Rheumatology (ACR) response or the DAS (e.g., DAS in 28 joints [DAS28]) (16, 17). Predictive tools therefore need to be user friendly and should use data that are easily and consistently measurable in routine clinical practice.

Among several different methods to classify and predict future response, classification and regression tree (CART) analyses have been successfully utilized in a number of therapeutic areas (18–22) to categorize disease and to determine the likelihood of future events. Unlike some statistical methods, CART imposes no prior assumptions on the structure of the data or relationships among variables. This approach has the potential to identify patients likely or unlikely to respond to treatment and to present the predicted likelihood of response in a manner useful to inform RA clinical practice.

The objective of this analysis was to determine the prognostic significance of clinical and laboratory data measured at baseline and during the first 12 weeks of therapy with the PEGylated anti-TNF certolizumab pegol (CZP) to predict low disease activity (LDA) at 1 year. The goal of the prediction model was to categorize patients 12 weeks after starting CZP to facilitate clinical management at that time point. At 12 weeks, patients were categorized into 3 groups: 1) those predicted to have a low likelihood of achieving LDA at week 52 with CZP therapy and who might consider alternate treatment options at week 12, 2) those predicted to have a high likelihood of achieving LDA at week 52 with CZP therapy and who are likely to be reassured at week 12 about their future response, and 3) those with an uncertain likelihood of achieving LDA at week 52 who need longer than 12 weeks of CZP therapy (e.g., up to 6 months) before being able to accurately predict a 1-year outcome.

Significance & Innovations

  • Classification and regression tree (CART) analyses have been successfully utilized in a number of therapeutic areas to categorize disease state and to determine the likelihood of future events.

  • The ability to quickly predict achievement of outcomes in patients with rheumatoid arthritis at 1 year would enable physicians to tailor treatment early during the course of therapy.

  • Using CART, it was possible to predict 1-year response within the first 12 weeks of starting therapy with certolizumab pegol.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

Patients.

Participants included in this analysis were patients with active RA who were randomized to treatment with CZP in the Rheumatoid Arthritis Prevention of Structural Damage 1 (RAPID 1) clinical trial (23). In brief, eligible patients were ages ≥18 years and had active disease at screening and baseline as defined by ≥9 tender joints and ≥9 swollen joints and a erythrocyte sedimentation rate (ESR; Westergren method) ≥30 mm/hour or a C-reactive protein level >15 mg/liter. Patients were required to have received methotrexate for ≥6 months at a stable dosage of ≥10 mg/week for ≥2 months prior to baseline.

Protocol.

The design and primary results of the RAPID 1 trial have been published previously (23). The intent-to-treat population in RAPID 1 comprised 982 patients randomized 2:2:1 to receive treatment with subcutaneous CZP (at an initial dose of 400 mg at weeks 0, 2, and 4 followed by 200 mg or 400 mg every 2 weeks) in combination with methotrexate, or placebo in combination with methotrexate, for 1 year (i.e., 52 weeks). Patients who failed to achieve an ACR criteria for 20% improvement (ACR20) response at weeks 12 and 14 were withdrawn from RAPID 1 at week 16 (per the study protocol) (24). Patients who withdrew at week 16 or who successfully completed RAPID 1 were offered enrollment in an open-label extension (OLE) study of CZP 400 mg every 2 weeks (ClinicalTrials.gov identifier: NCT00175877). Institutional review boards or ethics committees approved the protocol at each center. All patients gave written consent and the study was conducted in accordance with the Declaration of Helsinki.

Patients initially randomized to placebo were not included in this analysis. A total of 783 patients were randomized to treatment with CZP, and the eligible patient population for this analysis (n = 703) excluded those who withdrew during the first year of the trial for reasons other than efficacy. For patients who withdrew due to efficacy and reconsented into the OLE, the OLE data gathered at week 52 were used.

For patients who withdrew due to efficacy but did not reconsent into the OLE or for any patients using rescue medication, nonresponder imputation was used for ACR20/50/70, and last observation carried forward imputation was used for all continuous data (e.g., DAS28, further categorized as LDA or not) in order to capture CZP response in a conservative way.

Data analysis.

The outcome of the prediction model was LDA at year 1 (i.e., week 52) after starting CZP therapy. LDA was defined as a 4-variable DAS28-ESR ≤3.2. Although other clinical outcomes are desirable (e.g., remission or DAS28 <2.6) (22), LDA was chosen to be a proxy for doing “well enough,” such that a patient and physician would likely choose to continue therapy. Achieving LDA is also in line with the recent European League Against Rheumatism (EULAR) recommendations that state that LDA is an appropriate goal for patients with longstanding disease and high disease activity because they are unlikely to achieve remission (25).

A probability plot based on logistic regression was constructed showing the absolute DAS28 at week 12, and the change in DAS28 at week 12, as a predictor of LDA at 1 year. With the intent to improve upon the use of this single measure assessed at only 1 time point, a decision tree–based approach was used for the prediction model. The advantages of a tree-based approach are that it 1) can be applied to the treatment of individual patients because each individual is categorized uniquely within the tree, 2) is easy to understand, 3) can incorporate multiple predictors measured at different times, and 4) can be described in a visual fashion without the need for calculations that would require a computer software tool.

An a priori model was constructed based only on clinical input that used DAS change from baseline of ≥1.2 units at weeks 4 and 12 as predictors. Several different prediction models were subsequently derived and tested using CART software (Salford Systems). Since any data-based empirical prediction model is likely to experience overfitting (i.e., the model adequately fits to the data used to develop the model, but the model fails to provide accurate predictions for new subjects or data sets) (26), a split-sample approach (i.e., separate training/testing data sets) was used. The prediction model was built on a two-thirds random sample of the data (training data set) and tested on the remaining one-third of the data (testing data set). In this article, we show only the more conservative performance from the testing data sets, and not the typically better performance from the training data sets. It has previously been shown that CART modeling can perform better than conventional logistic regression analysis, especially when the data contain nonlinear features, colinearity, and interactions (26, 27).

Within each tree, patients were categorized into “terminal nodes,” which were the lowest branches. Nodes with a likelihood of achieving LDA much greater than 50% were considered “predicted responder” nodes, and those with a predicted likelihood of achieving LDA much less than 50% were considered “predicted nonresponder” nodes. The misclassification rate for each node was therefore the proportion of patients who achieved LDA in a node classified as a predicted nonresponder and the proportion of patients in a predicted responder node who did not achieve LDA. The accuracy of classification for patients in each node was determined as 100 minus the misclassification rate. For example, if the proportion of patients who achieve LDA in a predicted nonresponder node is 14%, then the accuracy of classification in that node is 86%.

In the first model, described as CART model 1, demographic and disease characteristics at baseline and drug level and clinical and related laboratory data from CZP-treated patients from baseline and weeks 4, 6, 8, and 12 were used as candidate variables. Each variable was represented as the absolute value, absolute change from baseline, and percent change from baseline. The ability of model 1 to discriminate between patients who did or did not achieve LDA at year 1 was assessed using the area under the receiver operating curve (AUC). With the intent to reduce the misclassification of model 1, especially for patients predicted to be nonresponders, a related model (model 2) was constructed that implemented a 3:1 misclassification penalty. The aim of this model was to reduce the misclassification for predicted nonresponders to ≤10%. This penalty, available as a standard option within the CART software, allowed for misclassifying a nonresponder at week 12 who was actually a responder at 1 year to be set to be 3 times worse (in terms of model building) than misclassifying in the reverse direction. The goal of applying this penalty was to increase the certainty of prediction at 12 weeks for the patients predicted to be nonresponders.

Sensitivity analyses.

With the recognition that many physicians do not calculate DAS28 measures in real time, a “simplified” model (model 3) was derived. This prediction model was derived de novo, and more complex variables such as the DAS were disallowed. Only variables that were more easily measured in clinical practice, such as the Clinical Disease Activity Index (CDAI), were included as potential predictors. The outcome of this model was the same, i.e., LDA by DAS28-ESR at 1 year.

As patients who start with high disease activity might be satisfied enough if they reach an ACR50 response, even if they did not achieve LDA, we derived a final model (model 4). The construction of this prediction model was similar to model 1, but used a composite outcome (LDA and/or ACR50) at 1 year. Finally, to test the hypothesis that tree-based prediction models derived from RA patients with similar key characteristics (e.g., established disease) might generalize across anti-TNF therapies, we used RAPID 1 data to assess the performance of a recently published CART-based model that was derived using data from patients with established RA initiating etanercept (28).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

Patient characteristics.

Of 783 patients randomized to treatment with CZP in RAPID 1 (intent-to-treat population), 703 were included in this analysis; 80 patients who withdrew for reasons other than efficacy were excluded. The population of patients represented in this analysis was similar to the overall RAPID 1 population (Table 1). Disease activity at the start of the trial was high; patients had a median DAS28 of 7.0 and the mean RA disease duration was 6.1 years. A total of 486 randomly selected patients were used in the training data set to build the models. The remaining 217 patients were included in the testing data set to evaluate model performance, and these individuals form the basis of all results reported below. At year 1, 35% of patients (n = 76) included in the testing data set achieved LDA.

Table 1. Baseline characteristics*
 RAPID 1 ITT (n = 982)Analysis population (n = 703)
  • *

    Values are the mean ± SD unless otherwise indicated. RAPID 1 = Rheumatoid Arthritis Prevention of Structural Damage 1; ITT = intent-to-treat; VAS = visual analog scale; HAQ = Health Assessment Questionnaire; DI = disability index; DAS28 = Disease Activity Score in 28 joints; CRP = C-reactive protein; CV = coefficient of variation.

  • Keystone et al, 2008 (23).

  • The population represented in this analysis comprises 393 patients treated with certolizumab pegol (CZP) 200 mg plus methotrexate plus 390 patients treated with CZP 400 mg plus methotrexate in the RAPID 1 study; 80 patients withdrew due to reasons other than lack of efficacy and were excluded. RAPID 1 patients initially randomized to placebo also were excluded.

Patient demographics and characteristics  
 Age, years52.0 ± 11.651.7 ± 11.5
 Sex, % female83.282.6
 Disease duration, years6.1 ± 4.36.1 ± 4.3
 Concomitant methotrexate dosage, mg/week13.6 ± 4.213.6 ± 4.1
 Rheumatoid factor, % positive (≥14 IU/ml)81.881.5
Disease activity status  
 Tender joint count (range 0–68)30.7 ± 12.931.0 ± 12.9
 Swollen joint count (range 0–66)21.5 ± 9.821.8 ± 10.0
 Patient's assessment of arthritis pain, 0–100 mm VAS63.1 ± 18.962.7 ± 18.7
 HAQ DI1.7 ± 0.611.7 ± 0.60
 DAS28, median7.07.0
 CRP level, geometric mean (CV) mg/liter14.7 (144.2)14.2 (141.4)

Model performance.

The probability of achieving LDA at year 1 based on the absolute DAS28 and on the change from baseline in DAS28 scores at week 12 is shown in Figures 1A and B, respectively. Only patients in the right- and left-most tails of both of these figures could be predicted with high probability to be in LDA at 1 year.

thumbnail image

Figure 1. Graph showing the probability of low disease activity (LDA) at 1 year A, versus absolute Disease Activity Score in 28 joints (DAS28), and/or B, versus DAS28 improvement from baseline at week 12.

Download figure to PowerPoint

As an enhancement to this approach, results from the a priori tree-based model are shown in Figure 2. At week 12, 77% of patients had at least some clinical response to CZP therapy (i.e., they had a DAS28 change from baseline of >1.2 units at week 12). The patients who did not achieve a DAS28 change from baseline of at least 1.2 units at week 12 had a very low (6%) likelihood of LDA at 1 year. However, only 23% of patients could be classified as being nonresponders using this criterion. Moreover, this approach could not identify responder patients who were likely to achieve LDA, demonstrated by the observation that there were no nodes where patients had a >50% predicted likelihood of response. Overall, the a priori model correctly classified 65% of patients and had an AUC of 0.68.

thumbnail image

Figure 2. A priori model: constructed based only on clinical input using the Disease Activity Score in 28 joints (DAS28) changes from baseline of >1.2 units at weeks 4 and 12 as predictors. LDA = low disease activity; * = achieved LDA at 1 year; AC = node accuracy; PNR = % of patients predicted to be a nonresponder (data from predicted nonresponder node); MT = % of patients needing more time on treatment for classification.

Download figure to PowerPoint

Using CART model 1 (Figure 3), which included data from baseline and weeks 4, 6, 8, and 12 as potential predictors, LDA at year 1 was predicted with 65–86% accuracy for 88% of the population (n = 192) (Figure 3). This 88% included 54% of the study population (n = 117) predicted to be nonresponders and 35% (n = 75) predicted to be responders. Of the 54% of patients who still had a DAS28 >4.5 at week 12, 86% did not achieve LDA at year 1, and the remaining 14% were misclassified. Only 12% of patients required further time on treatment to accurately predict the outcome. Overall, 76% of patients were correctly classified, and the model had good discrimination with an AUC of 0.73.

thumbnail image

Figure 3. Classification and regression tree (CART) model 1: all potential predictors (measured at baseline and at weeks 4, 6, 8, and 12) included to predict low disease activity (LDA; Disease Activity Score in 28 joints [DAS28] ≤3.2) at 1 year. Variables and cut points were derived empirically using CART. Results shown are from the testing data set only, since performance of the models using the training data set was generally superior. * = achieved LDA at 1 year; AC = node accuracy; PR = % of patients predicted to be a responder (data from predicted responder node); PNR = % of patients predicted to be a nonresponder (data from predicted nonresponder node); MT = % of patients needing more time on treatment for classification.

Download figure to PowerPoint

Results from CART model 2 (Figure 4), which was a variation of model 1 that applied a penalty to reduce the misclassification rate of the predicted nonresponders, showed that the initial DAS28 threshold selected by CART at 12 weeks was higher (4.9). Model 2 was able to classify 46% of patients as predicted nonresponders with only 10% misclassification (i.e., 90% accuracy); however, the proportion of patients who could not be classified accurately at 12 weeks doubled to 24%. Of the 30% of patients predicted to be responders, 74% achieved LDA at 1 year. Model 2 was shown to have an improved discrimination ability compared with model 1 (AUC 0.76) and was able to correctly classify more of the overall patient population within 12 weeks of starting treatment (79% versus 76%).

thumbnail image

Figure 4. Classification and regression tree (CART) model 2: variation on model 1 (Figure 3) with 3:1 misclassification cost applied to maximize accuracy for predicted nonresponders. Outcome is low disease activity (LDA; Disease Activity Score in 28 joints [DAS28] ≤3.2) at 1 year. Variables and cut points were derived empirically using CART. Results shown are from the testing data set only, since performance of the models using the training data set was generally superior. * = achieved LDA at 1 year; AC = node accuracy; PR = % of patients predicted to be a responder (data from predicted responder node); PNR = % of patients predicted to be a nonresponder (data from predicted nonresponder node); MT = % of patients needing more time on treatment for classification.

Download figure to PowerPoint

CART model 3 (Figure 5) shows the results from the sensitivity analysis that included simpler predictors (e.g., CDAI and not DAS28). A total of 55% of patients (n = 119) were predicted to be nonresponders and 20% (n = 43) were predicted to be responders. The accuracy of the nonresponder prediction was 80%. For 25% of patients, further time on treatment was necessary to accurately predict the outcome. Model 3 had a similar discrimination (AUC 0.63) and accuracy to model 1 and was able to correctly classify 71% of the overall patient population within 12 weeks of starting treatment.

thumbnail image

Figure 5. Classification and regression tree (CART) model 3: selected predictors easy to measure in daily practice. Variables and cut points were derived empirically using CART but included only those that were more easily measured in clinical practice (e.g., Clinical Disease Activity Index [CDAI] and swollen joint count [SJC]; not Disease Activity Score in 28 joints). Results shown are from the testing data set only, since performance of the models using the training data set was generally superior. LDA = low disease activity; * = achieved LDA at 1 year; AC = node accuracy; PR = % of patients predicted to be a responder (data from predicted responder node); PNR = % of patients predicted to be a nonresponder (data from predicted nonresponder node); MT = % of patients needing more time on treatment for classification.

Download figure to PowerPoint

In CART model 4 predicting the composite outcome measure of LDA and/or ACR50 at 1 year (see Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658), 45% of patients were classified as responders with 85% accuracy, 39% of patients were predicted to be nonresponders with 78% accuracy, and the remaining 16% of patients were not able to be well classified at 12 weeks. Model 4 achieved similar performance to the best model (model 2) and had an AUC of 0.77 and an ability to correctly classify 77% of the overall patient population within 12 weeks of starting therapy.

Finally, the replication of a previously published CART-derived decision tree is shown in Supplementary Appendix B (available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658). The prediction model represented by this tree had a similar performance when applied to the RAPID 1 data as its original performance using the Trial of Etanercept and Methotrexate with Radiographic Patient Outcomes (TEMPO) study data (28).

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

A major challenge in the management of RA is the prediction of long-term response to therapy. At present, only limited information is available to determine which factors, if any, will predict a good long-term response. The ability to predict achievement of LDA at 1 year would enable physicians to tailor treatment early during the course of therapy, thereby improving outcomes and potentially minimizing patient exposure to ineffective therapies. Furthermore, achieving LDA is a treatment goal supported by the recent EULAR guidelines (25). Using clinically applicable models, we show that, 12 weeks after the start of CZP therapy, we could accurately classify the majority (∼88%) of patients from RAPID 1, a study population with mainly high baseline disease activity, as likely to achieve or not achieve LDA at 1 year. Across several prediction models, we also found that patients with an early response to treatment at weeks 4, 6, and 8 had an even greater likelihood of achieving LDA at 1 year. Approximately 12–25% of patients could not be classified accurately at week 12 and needed longer treatment to determine with a high degree of certainty the likelihood of achieving LDA at week 52. At least for the types of patients enrolled in RAPID 1, our results identify which patients can likely be switched at 12 weeks if they are predicted to be nonresponders (i.e., patients having a very low predicted likelihood of achieving LDA) with a relatively high degree of accuracy. Indeed, these data highlight the possibility to use such models as a negative predictability tool by identifying those unlikely to achieve LDA, and this is perhaps the patient population a treating physician would most like to identify early so that treatment can be altered. In our best performing model, model 2, this prediction was made with 90% accuracy. Whether a 90% level of certainty, or similar amounts of certainty, is sufficient to give physicians enough confidence to make treatment changes at 12 weeks is a matter of individual judgment. Other factors, including patient and physician preferences and access to alternative therapies, are also likely to play important roles in the decision to switch RA treatments (29–32).

Results using an alternate model in which the DAS was replaced by the CDAI (CART model 3) had somewhat lower discrimination and accuracy than model 1. This lower performance was likely to be a consequence of only moderate correlation between the CDAI (used for the predictor variables) and the DAS28 (used for the outcome). Performance of this model would likely have been better if we had used the CDAI to define LDA, rather than the DAS28. Finally, CART model 4 (similar to model 1, but with a composite outcome [LDA and/or ACR50] at 1 year) had the numerically best discrimination of all prediction models.

When comparing results across the models, we see a clear compromise between the accuracy of prediction and the proportion of patients who could be classified at week 12. In the a priori model (Figure 2), for example, the misclassification rate for patients predicted to be nonresponders was very low, at 6%. However, only 23% of eligible RAPID 1 patients could be classified with that high level of accuracy. The a priori model also had suboptimal discrimination and calibration for the entire study population. Using the CART-based data-driven approach, the misclassification rate for patients predicted to be nonresponders shown in model 1 (Figure 3) was slightly higher at 14%, but twice as many patients (54%) could be classified. Results from model 2 (Figure 4) were even better and had a misclassification rate of only 10% for a group of patients who comprised 46% of the eligible RAPID 1 population. These data show the concession between the certainty of classification and the proportion of all patients able to be classified.

Prior RA data evaluating single predictors at a fixed time point indicate that the level of disease activity at baseline and after the first 3 months of treatment is significantly related to the level of disease activity at 1 year (33, 34) (Figure 1). While probability plots offer important insights, they have limitations, i.e., they provide probabilities for only 1 specific time point and 1 variable. To improve upon this, using data collected within the first 12 weeks of CZP therapy, we constructed several models using CART to predict which patients would achieve LDA (DAS28 ≤3.2) at 1 year. The benefit of this approach is that it allows for inclusion of multiple predictors and allows patients to be further classified based upon whether they had a very early response (4–6 weeks). Furthermore, these early time points reflect visit intervals at which a patient could be reasonably assessed in clinical practice; measuring predictor variables of response at shorter time points (e.g., 2 weeks after initiating anti-TNF therapy) may not be realistically feasible outside of controlled trials.

This study has several limitations. Despite the use of a random split-sample methodology possible with the large RAPID 1 data set, with separate testing and training data sets, the potential for overfitting (i.e., the model fails to provide accurate predictions when applied to new subjects/data sets) remains. Furthermore, model accuracy reported herein may be specific to the biologic or biologic class examined (CZP, or perhaps only anti-TNF therapies), or applicable only to the types of patients recruited to RAPID 1 (i.e., those with high disease activity and established RA). To address this generalizability issue, we replicated a recently published decision tree prediction model that used clinical assessments at week 12 and earlier to predict LDA at 1 year derived from RA patients treated with etanercept in the TEMPO trial (28). The prediction model built from the TEMPO data appeared to perform similarly well using RAPID 1 data. Based on this empirical example, we would suggest that the prediction models shown in our results (that had much more data than in the previous analysis [28]) will perform well for established RA patients with high disease activity treated with other anti-TNF agents. Additional replication studies in RA patient populations receiving differing drug treatments will be useful to confirm whether these models can be applied more broadly. However, we suspect that different prediction models will be required for patients with early RA and for those who start with lower disease activity than RAPID 1 patients.

Despite these caveats, our results suggest that it may be possible to develop predictive tools that are user friendly and could be easily and consistently applied in clinical practice. Using an analytic framework such as CART, biomarker and pharmacogenetic information could be added to prediction tools and likely would complement clinical assessments to prospectively guide the management of individual RA patients. Designing a trial around the concept of predicting response at 12 weeks or earlier and altering therapy for patients predicted to be nonresponders would be an optimal approach and should be tested. Prediction models similar to the type we have developed also have the potential to improve treatment to target approaches (24) by identifying groups of patients that should be switched to an alternate strategy more quickly, which ultimately could improve patient outcomes.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Curtis had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Curtis, Luijtens, Kavanaugh.

Acquisition of data. Luijtens.

Analysis and interpretation of data. Curtis, Luijtens, Kavanaugh.

ROLE OF THE STUDY SPONSOR

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

All authors contributed to the study design, data collection, and data analysis, including Dr. Luijtens of UCB Pharma. UCB Pharma funded the costs associated with developing the manuscript and approved the content of the submitted manuscript. Publication of this article was contingent upon the approval of all authors, including Dr. Luijtens of UCB Pharma.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information
  • 1
    Keystone EC, Genovese MC, Klareskog L, Hsia EC, Hall ST, Miranda PC, et al. Golimumab, a human antibody to tumour necrosis factor α given by monthly subcutaneous injections, in active rheumatoid arthritis despite methotrexate therapy: the GO-FORWARD study. Ann Rheum Dis 2009; 68: 78996.
  • 2
    Weinblatt ME, Kremer JM, Bankhurst AD, Bulpitt KJ, Fleischmann RM, Fox RI, et al. A trial of etanercept, a recombinant tumor necrosis factor receptor:Fc fusion protein, in patients with rheumatoid arthritis receiving methotrexate. N Engl J Med 1999; 340: 2539.
  • 3
    Bathon JM, Martin RW, Fleischmann RM, Tesser JR, Schiff MH, Keystone EC, et al. A comparison of etanercept and methotrexate in patients with early rheumatoid arthritis. N Engl J Med 2000; 343: 158693.
  • 4
    Klareskog L, van der Heijde DM, de Jager JP, Gough A, Kalden J, Malaise M, et al. Therapeutic effect of the combination of etanercept and methotrexate compared with each treatment alone in patients with rheumatoid arthritis: double-blind randomised controlled trial. Lancet 2004; 363: 67581.
  • 5
    Smolen JS, Kay J, Doyle MK, Landewe R, Matteson EL, Wollenhaupt J, et al. Golimumab in patients with active rheumatoid arthritis after treatment with tumour necrosis factor α inhibitors (GO-AFTER study): a multicentre, randomised, double-blind, placebo-controlled, phase III trial. Lancet 2009; 374: 21021.
  • 6
    Bombardieri S, Ruiz AA, Fardellone P, Geusens P, McKenna F, Unnebrink K, et al. Effectiveness of adalimumab for rheumatoid arthritis in patients with a history of TNF-antagonist therapy in clinical practice. Rheumatology (Oxford) 2007; 46: 11919.
  • 7
    Abe T, Takeuchi T, Miyasaka N, Hashimoto H, Kondo H, Ichikawa Y, et al. A multicenter, double-blind, randomized, placebo controlled trial of infliximab combined with low dose methotrexate in Japanese patients with rheumatoid arthritis. J Rheumatol 2006; 33: 3744.
  • 8
    Ramiro S, van Tubergen AM, Landewe RB. RAPID and FAST4WARD trials: certolizumab pegol for rheumatoid arthritis. Expert Rev Clin Immunol 2010; 6: 71320.
  • 9
    Fleischmann R, Vencovsky J, van Vollenhoven R, Borenstein D, Box J, Coteur G, et al. Efficacy and safety of certolizumab pegol monotherapy every 4 weeks in patients with rheumatoid arthritis failing previous disease-modifying antirheumatic therapy: the FAST4WARD study. Ann Rheum Dis 2009; 68: 80511.
  • 10
    Keystone E, Curtis JR, Fleischmann RM, Furst D, Khanna D, Smolen JS, et al. Rapid improvement in the signs and symptoms of rheumatoid arthritis following certolizumab pegol treatment predicts better long-term outcomes: post-hoc analysis of a randomized controlled trial. Arthritis Res Ther 2011; 38: 9906.
  • 11
    Wessels JA, van der Kooij SM, le Cessie S, Kievit W, Barerra P, Allaart CF, et al, for the Pharmacogenetics Collaborative Research Group. A clinical pharmacogenetic model to predict the efficacy of methotrexate monotherapy in recent-onset rheumatoid arthritis. Arthritis Rheum 2007; 56: 176575.
  • 12
    Hider SL, Silman AJ, Thomson W, Lunt M, Bunn D, Symmons DP. Can clinical factors at presentation be used to predict outcome of treatment with methotrexate in patients with early inflammatory polyarthritis? Ann Rheum Dis 2009; 68: 5762.
  • 13
    Mugnier B, Balandraud N, Darque A, Roudier C, Roudier J, Reviron D. Polymorphism at position −308 of the tumor necrosis factor α gene influences outcome of infliximab therapy in rheumatoid arthritis. Arthritis Rheum 2003; 48: 184952.
  • 14
    Padyukov L, Lampa J, Heimburger M, Ernestam S, Cederholm T, Lundkvist I, et al. Genetic markers for the efficacy of tumour necrosis factor blocking therapy in rheumatoid arthritis. Ann Rheum Dis 2003; 62: 5269.
  • 15
    Seitz M, Wirthmuller U, Moller B, Villiger PM. The -308 tumour necrosis factor-α gene polymorphism predicts therapeutic response to TNFα-blockers in rheumatoid arthritis and spondyloarthritis patients. Rheumatology (Oxford) 2007; 46: 936.
  • 16
    Flowers N, Wolfe F. What do rheumatologists do in their practices? [abstract]. Arthritis Rheum 1998; 41 Suppl: S337.
  • 17
    Schoels M, Aletaha D, Smolen JS, Bijlsma JW, Burmester G, Breedveld FC, et al. Follow-up standards and treatment targets in rheumatoid arthritis: results of a questionnaire at the EULAR 2008. Ann Rheum Dis 2010; 69: 5758.
  • 18
    Bukkapatnam S, Komanduri R, Yang H, Rao P, Lih WC, Malshe M, et al. Classification of atrial fibrillation episodes from sparse electrocardiogram data. J Electrocardiol 2008; 41: 2929.
  • 19
    Mockel M, Danne O, Muller R, Vollert JO, Muller C, Lueders C, et al. Development of an optimized multimarker strategy for early risk assessment of patients with acute coronary syndromes. Clin Chim Acta 2008; 393: 1039.
  • 20
    Kohrt HE, Olshen RA, Bermas HR, Goodson WH, Wood DJ, Henry S, et al. New models and online calculator for predicting non-sentinel lymph node status in sentinel lymph node positive breast cancer patients. BMC Cancer 2008; 8: 66.
  • 21
    Duvvuri SK, Finkelstein J. Computerized decision support for asthma management [abstract]. Value Health 2006; 9: A1012.
  • 22
    Felson DT, Smolen JS, Wells G, Zhang B, van Tuyl LH, Funovits J, et al. American College of Rheumatology/European League Against Rheumatism provisional definition of remission in rheumatoid arthritis for clinical trials. Arthritis Rheum 2011; 63: 57386.
  • 23
    Keystone E, van der Heijde D, Mason D Jr, Landewe R, van Vollenhoven R, Combe B, et al. Certolizumab pegol plus methotrexate is significantly more effective than placebo plus methotrexate in active rheumatoid arthritis: findings of a fifty-two–week, phase III, multicenter, randomized, double-blind, placebo-controlled, parallel-group study. Arthritis Rheum 2008; 58: 331929.
  • 24
    Felson DT, Anderson JJ, Boers M, Bombardier C, Furst D, Goldsmith C, et al. American College of Rheumatology preliminary definition of improvement in rheumatoid arthritis. Arthritis Rheum 1995; 38: 72735.
  • 25
    Smolen JS, Landewe R, Breedveld FC, Dougados M, Emery P, Gaujoux-Viala C, et al. EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs. Ann Rheum Dis 2010; 69: 96475.
  • 26
    Breiman L, Friedman J, Stone CJ, Olshen RA. Classification and regression trees. Boca Raton (FL): CRC Press; 1984.
  • 27
    Steinberg D, Colla P. CART: tree-structured non-parametric data analysis. San Diego: Salford Systems; 1995.
  • 28
    Curtis JR, Yang S, Chen L, Park GS, Bitman B, Wang B, et al. Predicting low disease activity and remission using early treatment response to antitumour necrosis factor therapy in patients with rheumatoid arthritis: exploratory analyses from the TEMPO trial. Ann Rheum Dis 2012; 71: 20612.
  • 29
    Wolfe F, Pincus T, O'Dell J. Evaluation and documentation of rheumatoid arthritis disease status in the clinic: which variables best predict change in therapy. J Rheumatol 2001; 28: 17127.
  • 30
    Zhang J, Shan Y, Reed G, Kremer J, Greenberg JD, Baumgartner S, et al. Thresholds in disease activity for switching biologics in rheumatoid arthritis patients: experience from a large US cohort. Arthritis Care Res (Hoboken) 2011; 63: 16729.
  • 31
    Ledingham J, Deighton C. Update on the British Society for Rheumatology guidelines for prescribing TNFα blockers in adults with rheumatoid arthritis (update of previous guidelines of April 2001). Rheumatology (Oxford) 2005; 44: 15763.
  • 32
    Furst DE, Keystone EC, Fleischmann R, Mease P, Breedveld FC, Smolen JS, et al. Updated consensus statement on biological agents for the treatment of rheumatic diseases, 2009. Ann Rheum Dis 2010; 69 Suppl: i229.
  • 33
    Aletaha D, Funovits J, Keystone EC, Smolen JS. Disease activity early in the course of treatment predicts response to therapy after one year in rheumatoid arthritis patients. Arthritis Rheum 2007; 56: 322635.
  • 34
    Dougados M, Schmidely N, Le Bars M, Lafosse C, Schiff M, Smolen JS, et al. Evaluation of different methods used to assess disease activity in rheumatoid arthritis: analyses of abatacept clinical trial data. Ann Rheum Dis 2009; 68: 4849.

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. Acknowledgements
  10. REFERENCES
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ACR_21600_sm_SuppAppa.tif216KSupplementary Appendix A. CART model 4: all potential predictors (measured at baseline and at weeks 4, 6, 8, and 12) included to predict low disease activity (DAS28 ≤ 3.2) and/or ACR50 at 1 year. Variables and cut points were derived empirically using CART. Results shown are from the testing data set only since performance of the models using the training data set was generally superior.
ACR_21600_sm_SuppAppb.tif178KSupplementary Appendix B. Replication of a prediction model derived from RA patients initiating etanercept (from the TEMPO trial) (28) to the RAPID 1 population. No new statistical modeling was conducted. The predictors and cut points previously published were used and model performance evaluated using RAPID 1 data.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.