Ankylosing spondylitis in west Africans—evidence for a non-HLA-B*27 protective eVect

Dr Brown and his colleagues are to be congratulated for performing a logistically formidable, but necessary, epidemiological study testing the currently in vogue hypothesis that the B*2703 subtype of HLA-B27 is not related to ankylosing spondylitis (AS). They conclude that the B*2705 subtype, as well as B*2703, possesses a lower risk for developing AS in a group of B27 positive west Africans, the Fula from Gambia, when compared with B27 positive white subjects invoking the potential protective role of an environmental factor(s). This conclusion is based on an assumed risk of developing AS in B27 positive persons of 11.1% for men and 1.5% for women. No cases of AS were seen among 900 adult Fula men and 215 first degree relatives of 48 B27 positive Fula twin pairs. We would argue that the data warrant the more conservative conclusion implied in their discussion, namely, the risk for AS among B27 positive Fula subjects would need to be at least 2.7% in men and 1% in women to assign significance to the finding of no AS in this population. The risk of developing AS in HLA-B27 positive subjects clearly varies among diVerent ethnic groups, but it is now generally accepted that among white populations, the prevalence of AS is nearer 1–2% rather then 11.1%. 4 The Norwegian survey of 14 539 subjects quoted by the authors is in fact based on a highly selected sample of only 375 people responding positively to a questionnaire for low back pain or stiVness who actually turned up for examination and had x rays of sacroiliac joints. You arrive at an entirely diVerent conclusion if you apply the AS prevalence figures of 1.4% for B27 subjects from the Busselton population study or 1.3% of Dutch B27 positive subjects. The second study examined 2956 subjects older than 44 years who all had sacroiliac x rays and only three B27 positive subjects had AS according to the New York criteria leading to a prevalence of 0.1%. Recalculating the data of Brown et al according to these more generally accepted prevalence rates leads to the following conclusions. The probability of observing no cases of AS in 900 adult Fula men would be 46.9% (that is, p=0.47). The number of B*2703 persons expected to develop AS would be zero, as in fact observed in this population. Even assuming a risk of 2.7% for AS in B27 positive subjects, the likelihood that no cases of AS would be found in 900 adult Fula men is 23.2% (that is, p=0.232). Furthermore, we calculate that the prevalence of AS in the population of B27 positive adult Fula men would need to be at least 5.54% before the finding of zero observed cases of AS in 900 adult Fula men would be statistically significantly diVerent. We conclude that the issue of B*2703 and risk for AS remains an open question and in need of further more extensive population prevalence studies.