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

Abstract

Objective: To derive and validate decision trees to categorise rheumatoid arthritis (RA) patients 12 weeks after starting etanercept with or without methotrexate into three groups: patients predicted to achieve low disease activity (LDA) at 1 year; patients predicted not to achieve LDA at 1 year and patients who needed additional time on therapy to be categorised.

Methods: Data from RA patients enrolled in the TEMPO trial were analysed. Classification and regression trees were used to develop and validate decision tree models with week 12 and earlier assessments that predicted long-term LDA. LDA, defined as disease activity score in 28 joints (DAS28) ≤3.2 or clinical disease activity index ≤10.0, was measured at 52 or 48 weeks. Demographics, laboratory data and clinical data at baseline and to week 12 were analysed as predictors of response.

Results: 39% (67/172) of patients receiving etanercept and 60% (115/193) of patients receiving etanercept plus methotrexate achieved LDA at week 52. For patients receiving etanercept, 53% were predicted to have LDA, 39% were predicted not to have LDA and 8% could not be categorised using DAS28 criteria at week 12. For patients receiving etanercept plus methotrexate, 63% were predicted to have LDA, 25% were predicted not to have LDA and 12% could not be categorised.

Conclusion: Most (80-90%) patients in TEMPO initiating etanercept with or without methotrexate could be predicted within 12 weeks of starting therapy as likely to have LDA or not at week 52. However, approximately 10-20% of patients needed additional time on therapy to decide whether to continue treatment.

Figure 1

CART decision tree of DAS28 and CDAI response in patients receiving etanercept. CART decision trees with A) DAS28 ≤ 3.2 (responders) or > 3.2 (nonresponders) at 52 or 48 weeks as the outcome variable and with B) CDAI ≤ 10 (responders) or > 10 (nonresponders) at 52 or 48 weeks as the outcome variable are shown for patients receiving etanercept plus methotrexate combination therapy. Tree nodes of predicted responders, nonresponders, and patients with indeterminate outcomes are identified by borders of gray, black, and diagonal patterns, respectively. ETN, Etanercept; LDA, Low disease activity; DAS28, Disease activity score based on 28 joints; TJC, Tender joint count; CDAI, Clinical Disease Activity Index; HAQ-DI, Health Assessment Questionnaire Disability Index.

Figure 2

CART decision tree in patients receiving etanercept plus methotrexate with ACR/EULAR remission at 52 or 48 weeks as the outcome variable. Tree nodes of predicted responders, nonresponders, and patients with indeterminate outcomes are identified by borders of gray, black, and diagonal patterns, respectively. ETN, Etanercept; MTX, Methotrexate; SJC, Swollen joint count; TJC, Tender joint count; CRP, C-reactive protein (in mg/L).

Figure 3

CART decision tree of DAS28 and CDAI response in patients receiving etanercept plus methotrexate. CART decision trees with A) DAS28 ≤ 3.2 (responders) or > 3.2 (nonresponders) at 52 or 48 weeks as the outcome variable and with B) CDAI ≤ 10 (responders) or > 10 (nonresponders) at 52 or 48 weeks as the outcome variable are shown for patients receiving etanercept. Tree nodes of predicted responders, nonresponders, and patients with indeterminate outcomes are identified by borders of gray, black, and diagonal patterns, respectively. ETN, Etanercept; MTX, Methotrexate; LDA, Low disease activity; DAS28, Disease activity score based on 28 joints; CDAI, Clinical Disease Activity Index; SJC, Swollen joint count.

Figure 4

Application of prediction model shown in Figure 1A to 1000 bootstrap samples of TEMPO patients receiving etanercept plus methotrexate (n = 193 patients in each sample, sampled with replacement). The accuracy required of the prediction model was iteratively varied at set threshold levels between 50% and 100%. Each data point represents the proportion of each bootstrap sample that could be classified with that amount of accuracy, and a Loess curve line was fitted through the data points. This simulated model illustrates the tradeoff of misclassification rate versus the proportion of patients who can be classified with the prediction model. With the requirement for greater accuracy (left-hand side of figure), fewer patients could be classified at week 12.