Dr. James suggests that the observed effect of birth order in AS might be due to the possibility that the younger siblings may not yet have developed AS. Despite the relatively high age of the patients included, as was described in our reply, this issue still might be considered to be confounding in the observation that the risk of AS in first-born siblings is increased compared with that in later-born siblings. Although, in theory, Dr. James is right, we consider the confounding effect negligible, because the probability that one of the younger siblings of our group did not yet pass the proper age for diagnosis is very small. Also, the later-born siblings in our study had a lower median age at diagnosis compared with the first-borns (see Table 1 in our study). Moreover, the fact that these siblings had an older brother or sister with AS increases the awareness for detection of the disease in these relatives, which makes detection in an earlier stage possible. When limiting the analysis to families with only 2 children, the number of first-born siblings with AS still outweighed the expected number of later-born siblings with AS, as was shown in Table 2 of our study. In families with 2 children, 26 first-born children with definite AS were found, whereas 20 were expected (χ2 = 3.6, P = 0.029, by 1-sided test).
Another concern of Dr. James was the fact that 1-tailed testing was based on false arguments. Our hypothesis was primarily based on an observation in an animal model that the risk of ankylosing enthesopathy was increased in earlier-born offspring in mice (Weinreich S, Hoebe B, Ivanyi P. Maternal age influences risk for HLA-B27 associated ankylosing enthesopathy in transgenic mice. Ann Rheum Dis 1995;54:754–6). This was the main argument for why we used 1-sided testing to determine whether first-born siblings had an increased risk of AS. Two-sided testing would have also resulted in a significant increase, for all sibship sizes together. We do agree, however, that the results of studies on the effect of birth order in related autoimmune diseases, such as rheumatoid arthritis and multiple sclerosis, are conflicting, as was mentioned in Dr. James' discussion of our article.
The doubts expressed by Dr. James that the statistical methods we used were not in accordance with the rules of the renowned statistician, Mr. Yule, have, in our opinion, small grounds. One of the possible biases concerns the change in population birth rates across time. However, we had the opportunity to include families with a large variation in numbers of children per family. The increase in AS in first-born siblings was observed in the small (2 or 3 children) as well as in the large (>3 children) families (see Table 4 in our study).
We agree that performing additional, prospective studies to test whether our hypothesis is correct would indeed be very interesting, but would take a very long time. However, to compare the maternal age at birth of a large group of AS patients with the maternal age at birth of age-matched controls would give us a clue as to whether birth order and/or maternal age influence the risk of AS.