Genetic polymorphism (rs1800693) of TNFRSF1A (type 1 tumour necrosis factor receptor) encodes a potentially anti-inflammatory soluble truncated form of the p55 receptor, which is associated with predisposition to multiple sclerosis but protection against ankylosing spondylitis (AS). We analysed 2917 UK Caucasian cases by linear and logistic regression for associations of rs1800693 with disease severity assessed by the Bath Ankylosing Spondylitis measures of disease activity and function (BASDAI, BAS-G and BASFI) and/or responses to anti-TNF therapy. In contrast to predictions, rs1800693 GG homozygotes actually had significantly worse BASDAI (mean 4.2, 95% CI: 4–4.5) than AA homozygotes (mean 3.8, 95% CI: 3.7–4) in both the unadjusted (difference = 0.4, p = 0.006) and adjusted analyses (difference = 0.2–0.5, p = 0.002–0.04 depending on the adjustment model). We found no evidence that rs1900693 predicted functional status (BASFI) or global disease scores (BAS-G), and it exerted no influence on either the intention to treat with or efficacy of anti-TNF treatment.