Objective To study the relationship of spinal inflammation and fatty degeneration (FD) as detected by MRI and new bone formation seen on conventional radiographs (CRs) in ankylosing spondylitis (AS).

Methods CRs at baseline, 2 years and 5 years and spinal MRIs at baseline and 2 years of 73 AS patients treated with infliximab in European AS Infliximab Cohort were available. Relative risks (RR) were calculated with a general linear model after adjustment for within-patient variation.

Results In a total of 1466 vertebral edges (VEs) without baseline syndesmophytes, 61 syndesmophytes developed at 5 years, the majority of which (57.4%) had no corresponding detectable MRI lesions at baseline. VEs with both inflammation and FD at baseline had the highest risk (RR 3.3, p=0.009) for syndesmophyte formation at 5 years, followed by VEs that developed new FD or did not resolve FD at 2 years (RR=2.3, p=0.034), while inflammation at baseline with no FD at 2 years had the lowest risk for syndesmophyte formation at 5 years (RR=0.8). Of the VEs with inflammation at baseline, >70% resolved completely, 28.8% turned into FD after 2 years, but only 1 syndesmophyte developed within 5 years.

Conclusions Parallel occurrence of inflammation and FD at baseline and development of FD without prior inflammation after 2 years were significantly associated with syndesmophyte formation after 5 years of anti-tumour necrosis factor (TNF) therapy. However, the sequence ‘inflammation–FD–new bone formation’ was rarely observed, an argument against the TNF-brake hypothesis. Whether an early suppression of inflammation leads to a decrease of the risk for new bone formation remains to be demonstrated.