Background CDAI (clinical disease activity index) and RAPID3 (routine assessment of patient index data 3) cut points defining responses that best match ACR20/50/70 response in rheumatoid arthritis (RA) patients (pts), are unknown.
Objectives To evaluate cut points in a study population treated with certolizumab pegol (CZP) plus methotrexate (MTX).1
Methods ACR responders through Week (Wk) 12 from 393 CZP treated pts (400 mg at Wks 0, 2 and 4 then 200 mg every 2 Wks) plus MTX in RAPID 1 (NCT00152386) were categorized by proposed response (R) cut points in CDAI (change from baseline [CFB] ≥6.7, ≥10.0, ≥13.9) and RAPID3 (CFB ≥1.8, ≥3.6). ACR20/50/70 responses were compared with proposed categorizations using cross-tabulations, AUC under the ROC curve (AUC-ROC), % correctly classified (%CC) pts and kappa statistics. CART2 (classification and regression trees) modeling identified CDAI and RAPID3 cut points defining responses most closely associated with ACR20/50/70 responses.
Results Sensitivities were very high; at Wk 12, 93–100% ACR20/50/70 responders achieved CFB in CDAI (≥6.7,≥10.0,≥13.9). Positive predictive value [PPV] was lower indicating CDAI-R is not as strongly associated with ACR-R as ACR-R is with CDAI-R; lower specificities indicate that ACR nonresponse (NR) does not correspond well to CDAI-NR (particularly for ACR50/70). However, CDAI-NR predicts well ACR-NR (very high negative predictive value [NPV]) (Table). CFB in CDAI≥13.9 was most closely associated with ACR20-R (κ=0.57). Association between proposed CDAI categorizations and ACR50/70 was weak (κ=0.05–0.29). A similar trend was observed for RAPID3. CART modeling identified CFB in CDAI ≥13.80/≥20.15/≥20.15 and CFB in RAPID3 ≥5.46/≥7.28/≥5.53 as categorizations most closely associated with ACR20/50/70 responses.
Conclusions CDAI and RAPID3 CFB thresholds were identified that defined the closest associations with ACR responses in pts with inadequate response to MTX and high disease activity at baseline (mean DAS28 6.9). Sensitivities/NPVs were very high overall. Association with ACR20 was relatively strong. Association with ACR50/70 was weaker. Further studies are needed to analyze different pt populations, including those with lower disease activity at baseline.
Arthritis Rheum 2008; 58(11):3319–29.
Classification And Regression Trees (CART) software, Salford Systems, CA, USA
Disclosure of Interest M. Schiff Grant/Research support from: UCB Pharma, Consultant for: UCB Pharma, K. Luijtens Employee of: UCB Pharma, O. Davies Employee of: UCB Pharma, Y. Yazici Consultant for: UCB Pharma
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.