Table 3.

Multivariable Cox proportional hazards model for analysis of retention of a first tumour necrosis factor inhibitor in r22axSpA and in patients with r3+axSpA versus nr-axSpA

Key variable	HR	95% CI	P value
r22axSpA versus nr-axSpA	0.82	0.57 to 1.17	0.27
r3+axSpA versus nr-axSpA	0.60	0.44 to 0.82	0.001
r3+axSpA versus r22axSpA	0.73	0.53 to 1.02	0.06

Analysis performed in 476 patients (282 events). Variables included in the model: age, diagnostic delay, sex, human leucocyte antigen-B27, presence of elevated C reactive protein level, presence of peripheral arthritis and of enthesitis, current smoking, Bath Ankylosing Spondylitis Functional Index and Bath Ankylosing Spondylitis Mobility Index, treatment with etanercept versus treatment with monoclonal antitumour necrosis factor inhibitor antibodies, body mass index (>30 vs <25 and 25–30 vs <25, respectively). The full model is presented in online supplemental table S2.

nr-axSpA, non-radiographic axial spondyloarthritis; r3+axSpA, unilateral/bilateral grade 3–4 radiographic sacroiliitis; r22axSpA, bilateral grade 2 sacroiliitis.