Fig. 2

Predictors of 2nd bDMARD drug discontinuation. Estimation was done using multivariable Cox regression analysis based on a backward inclusion methodology (all variables with p < 0.100 were included). The model was controlled for age, gender and previous prescribed medications (differentiated by 12 different ATC main groups). Excluded variables were: concomitant DMARD therapy, concomitant painkiller medication, concomitant corticosteroid medication, adverse events before index date as defined in Additional file 2. Results are shown as 95%-CIs