Juvenile Idiopathic Arthritis (JIA) is an autoimmune condition characterised by persistent flares of joint inflammation. However, no reliable biomarker exists to predict the erratic disease course. Normally, regulatory T cells (Tregs) maintain immune tolerance, with altered Tregs associated with autoimmunity. Treg signatures have shown promise in monitoring other autoimmune conditions, therefore a Treg gene and/or protein signature could offer novel biomarker potential for predicting disease activity in JIA. Machine learning on our nanoString Treg gene signature on peripheral blood (PB) Tregs generated a model to distinguish active JIA (active joint count, AJC≥1) Tregs from healthy controls (HC, AUC=0.9875). Biomarker scores from this model successfully differentiated inactive (AJC=0) from active JIA PB Tregs. Moreover, scores correlated with clinical activity scores (cJADAS), and discriminated subclinical disease (AJC=0, cJADAS≥0.5) from remission (AUC=0.8980, Sens=0.8571, Spec= 0.8571). To investigate altered Treg fitness in JIA by protein expression, we utilised spectral flow cytometry and unbiased analysis. Three Treg clusters were increased in active JIA PB, including CD226highCD25low effector-like Tregs and CD39-TNFR2-Helioshigh, while a 4-1BBlowTIGITlowID2intermediate Treg cluster predominated in inactive JIA PB (AJC=0). The ratio of these Treg clusters correlated to cJADAS, and higher ratios could predict inactive individuals that flared by 6-month follow-up. Thus, we demonstrate altered Treg signatures and subsets as an important factor, and useful biomarker, for disease progression versus remission in JIA, revealing genes and proteins important in Treg fitness. Ultimately, PB Treg fitness measures could serve as routine biomarkers to guide disease and treatment management to sustain remission in JIA.