Patient-reported outcomes as early warning signs of flare following drug cessation in rheumatoid arthritis.
Gumber L., Rayner F., Bigirumurame T., Dyke B., Melville A., Kerrigan S., McGucken A., Naamane N., Prichard J., Buckley CD., Filer A., McInnes IB., Raza K., Siebert S., Wason JM., Ng W-F., Anderson AE., Isaacs JD., Baker KF., Pratt AG.
OBJECTIVES: Drug withdrawal in rheumatoid arthritis (RA) in remission can reduce toxicity, but with the risk of flare which requires close monitoring. We explored the potential of patient-reported outcomes (PROs) for flare detection among RA patients in sustained remission after conventional synthetic disease-modifying antirheumatic drug (csDMARD) cessation. METHODS: Four PROs (Factors that Limit sustAined Remission in rhEumatoid arthritis (FLARE-RA), EuroQol-5 Dimensions (EQ5D), Routine Assessment of Patient Index Data-3 (RAPID-3) and RA Flare Questionnaire (RA-FQ)) were captured at baseline and at sequential visits until time-of-flare or end of 6-month follow-up as part of the BIO-FLARE prospective cohort study. Flare was defined as any of (i) Disease Activity Score 28 (DAS28)-C reactive protein (CRP) ≥3.2 at any visit, (ii) DAS28-CRP≥2.4 on two visits within 2 weeks or (iii) resuming DMARD and/or steroid therapy despite DAS28-CRP<2.4. Cox regression models with time-varying covariates were fitted to evaluate associations between PRO changes and likelihood of flare. Receiver-operating characteristic (ROC) curves enabled discriminatory changes in each PRO to be compared as a means of identifying flare. RESULTS: 58/121 (47.9%) participants (70.1% females, mean age 64.8 years) experienced a flare. A 1-point change in each PRO score was strongly associated with flare development in the multivariate Cox regression model (p<0.001 in each case). ROC curve analysis confirmed that monitoring adverse changes in PROs from baseline offered robust discriminatory utility for identifying flare occurrence. This was most evident for RA-FQ and FLARE-RA (both areas under the curves 0.90, 95% CI 0.84 to 0.96; p=0.001); for example, an RA-FQ increment of ≥5.5 from baseline identified objective flare with positive and negative predictive values of 80% and 91%, respectively. CONCLUSIONS: Our data support the potential value of remote PRO monitoring of RA patients in drug-free remission to identify flare occurrence.