Ann Rheum Dis 2018; 77:650–57 DOI: 10.1136/annrheumdis-2017-212395Analysis of patient characteristics revealed that older and less healthy patients with RA were more likely to receive non-TNFi bDMARDs as a first bDMARD compared to other treatments.This study aimed to describe patient characteristics at initiation of bDMARD treatment at two-time points: first bDMARD initiation and switch to second bDMARD after TNFi treatment. The second aim of the study was to estimate the potential of treatment channelling to confound results in comparative treatment studies in RA. Data were collected from the Swedish Rheumatology Quality Register and the Nationwide Swedish Healthcare Registers. Baseline patient characteristics were used to assess factors considered to influence: therapy choice, safety or treatment outcome. Differences in patient characteristics were calculated and statistical analyses completed to compare treatments. Patients initiating non-TNFi therapy at first bDMARD initiation tended to be older and less well educated than TNFi initiators. Substantial differences in baseline medical history were noted at first bDMARD initiation – RTX and ABA initiators tended to have a previous history of assessed disease and had consumed more healthcare resources prior to treatment. Patient characteristics at switch showed that all non-TNFi initiators appeared to have a higher disease activity than TNFi initiators. Additionally, RTX and ABA initiators tended to be older with recent serious infections compared with TNFi initiators. Analysis of channelling revealed that age and sex were strong predictors of all outcomes except drug survival <1 year. Components of baseline activity and medical history were predictive of all outcomes, with concomitant therapy predictive of most outcomes.Authors concluded that patients who were older and less healthy tended to be given non-TNFi bDMARDs as a first bDMARD and at the switch from a first TNFi. The results also indicated that by accounting for differences age, medical history and disease activity, the likelihood of confounded results can be prevented for comparative bDMARD studies.