Despite modern therapeutics and treatment strategies, a subset of rheumatoid arthritis (RA) patients remains insufficiently responsive to multiple therapies. Here, we identify predictors of such refractory RA (“reRA”).

Patients from a longitudinal academic clinical database with reRA (defined as failing to reach the treatment target of at least low disease activity with ≥3 DMARD courses, including ≥1 biological, over a total of ≥18 months) were compared to patients who did respond within the first two treatments (treatment amenable RA, “taRA”). We performed logistic regression analysis to identify risk factors for refractory disease, and several sensitivity analyses concerning different potential definitions for reRA to confirm the robustness of the results; key findings were also validated in an independent community cohort.

We enrolled 412 patients, of whom 70 were reRA and 102 taRA; 240 patients fulfilled neither definition. ReRA patients were more frequently female (92.9 vs. 70.6%, p < 0.001), younger (44.37 vs. 51.14 years, p = 0.002), and had higher CDAI levels at first presentation (26.06 vs. 15.39, p < 0.001). Treatment delay was significantly longer for reRA than for taRA (3.17 vs. 1.45 years, p = 0.001). In multivariable analyses, treatment delay, female gender and higher disease activity remained as independent predictors of refractory disease. Based on the identified predictors, we developed a matrix model for risk of future reRA.

Our data identified delay to initial treatment, female gender and higher disease activity as important predictors of a later refractory course of RA. Delay of treatment initiation is the single most important modifiable risk factor of refractory disease.