To assess spinal inflammation by magnetic resonance imaging (MRI) before and after treatment with the tumor necrosis factor receptor fusion protein etanercept compared with placebo. As part of a recently published randomized, controlled trial, 40 patients with ankylosing spondylitis (AS) underwent MRI of the lower thoracic and lumbar spine at 4 different time points: baseline, 12 weeks, 24 weeks, and 48 weeks. Nineteen patients received subcutaneous etanercept twice weekly (25 mg twice daily) for 1 year, and 21 patients received placebo for 6 months before being switched to etanercept. The mean age of the patients was 39.7 years, 75% were male, 89% were HLA-B27 positive, and the mean disease duration was 13 years. MRI examinations included T1-weighted sequences before and after application of gadolinium-diethylenetriaminepentaacetic acid (Gd-DTPA) and T2-weighted fat-saturated sequences. MRI sequences were scored with an adjusted AS spinal MRI score, using predefined vertebral units (VUs) as a basis. After 12 weeks, spinal inflammation (as assessed by T2-weighted MRI with fat saturation) regressed by 54% in the etanercept group (mean score 1.33 per VU at baseline and 0.61 per VU at 12 weeks; P = 0.002) but worsened by 13% in the placebo group (0.94 at baseline and 1.06 at 12 weeks) (P < 0.001 between groups). After switching to etanercept, placebo patients improved similarly. T1-weighted Gd-DTPA MRI sequences performed equally well in detecting spinal changes. At baseline, >50% of all active lesions were detected in the thoracic spine. Deterioration of chronic changes was significant only in patients treated with placebo. Etanercept treatment of patients with active AS results in regression of spinal inflammation as assessed by spinal MRI. Inclusion of the thoracic spine in MRI examinations of patients with AS may be of particular importance.