retinolbinding protein were measured in these patients and in 46 healthy control individuals (mean age 53 years, 45% males). A larger population of patients (n = 123) and controls (n = 131) underwent genotyping for the presence of an exonic, single nucleotide polymorphism in the CYP26C1 allele, which has a role in retinol metabolism. The mean serum retinol level was signifi cantly lower in patients than controls (2.39 ± 0.88 μmol/l versus 3.34 ± 1.01 μmol/l, P <0.0001), as was the mean retinolbinding protein level (4.65 ± 2.10 μg/l versus 7.48 ± 4.87 μg/l, P <0.001). These levels did not significantly differ between patients with active and inactive disease, and retinol levels did not correlate with any clinical or serological indi ces of disease activity. There was no difference in the frequency of the minor CYP26C1 allele between patients and controls. The authors conclude that ankylosing spondylitis is associated with abnormally low serum levels of retinol. Further studies in large populations are warranted to elucidate the genetic determinants of retinol metabolism.