Definition of improvement thresholds in juvenile idiopathic arthritis (JIA) using the JADAS

Background/Purpose: Evaluation of disease activity in JIA is fundamental in clinical assessment. The ACR paediatric response measure used in clinical trials are validated to analyse the response to a treatment in comparison to a baseline disease activity but does not judge about the absolute disease activity or the absolute improvement.

Methods: The JADAS has been calculated in patients of the BIKER registry newly starting treatment with etanercept or MTX. The JADAS10 was preferred because it values all 4 domains equally. Physicians + parents were requested to judge on treatment efficacy as very good, good, weak, none or worse. Improvement was assumed if judgement of both were very good or good. No improvement was assumed if at least one judgement was for none or poor. Inconclusive judgements or those with a difference >1 point were excluded from analysis.

Results: Initially, ANOVA of JIA categories showed no significant differences of mean DJADAS in all baseline classes and IQRs also showed good overall limits. So, all JIA categories were combined for a joint cutoff. Analysis was restricted to the 3 month evaluation because of a time dependence of the judgement of improvement in terms of the JADAS. Restriction to the 3 month results left 1340 patients. JADAS at baseline was finally put into 4 classes, class 0 for JADAS <5, “low” for 5≤JADAS<15, “moderate” for 15 ≤JADAS<25 and ”high” for 25≤JADAS≤40. An initial JADAS of <5 was assumed as only minor or no disease activity. An improvement cutoff was only defined for baseline classes “low”, “moderate” and “high”. Cutoffs for defining improvement were chosen by calculating interquartile ranges (IQR) of the judgement groups and considering accuracy as well as sensitivity/specificity of the resulting model. Analysis by baseline class revealed clear cutoff points. According to the baseline JADAS class the following minimum decreases of the JADAS (DJADAS) are proposed for definition of improvement: For baseline class ”low”: DJADAS of 4, for baseline class ”moderate”:  DJADAS of 10, for baseline class ”high”:  DJADAS of 15. Alternatively a relative decrease of the JADAS by 42% for JADAS class “low”, by 51% for JADAS Class ”moderate” and 56% for JADAS class ”high” were found to define improvement. (table1)

Conclusion: Disease improvement on therapy can efficiently be defined by the decrease of the JADAS depending on the initial JADAS score defining low, moderate or high disease activity. Our model demonstrates clear cut off values. After cross validation these cutoffs may be used in clinical trials and for decisions in clinical practice.

Table 1: Inter quartile ranges of variable DJADAS10 by improvement and baseline class, absolute and relative values. Chosen cutoff for improvement and goodness-of-fit parameters. Higher DJADAS10 indicate better treatment efficacy. Only integer cutoffs were considered.