To the Editor:

In 2004, Meulenbelt et al (1) reported an association of hip osteoarthritis (OA) with haplotypes composed of 3 markers from within the interleukin-1β (IL-1β) gene (IL1B) and the IL-1 receptor antagonist gene (IL1RN). Of the 8 haplotypes derived from their analysis, 2 occurred at significantly elevated frequencies (risk haplotypes), and the frequency of 1 was significantly reduced (protective haplotype) in patients with hip radiographic OA. Sixty-one patients and 653 controls were genotyped, and cases of hip OA were ascertained by radiography.

In 2005, Smith et al (2) reported that they had also detected an association of hip OA with 1 of the 2 risk haplotypes identified by Meulenbelt and colleagues. Furthermore, Smith et al were able to generate more significant data when they expanded their analysis to 8 markers. The study by Smith et al included 22 patients and 195 controls, and cases of severe hip OA were ascertained by the need for hip replacement surgery.

There are 2 ways of interpreting these reports. The first possibility is that the investigators have uncovered a 3-marker haplotype that has such a strong influence on the development of OA that it can be detected even with a small number of patients. The second possibility is that both studies constitute false-positive results arising from the small number of patients genotyped.

In 2002, we reported an association of 8 markers from within the IL1 gene cluster with knee OA (3). We detected only moderate evidence of an association, but we did not carry out the haplotype analysis performed by Meulenbelt et al. We have now revisited our data and carried out the 3-marker haplotype analysis. The markers used by Meulenbelt et al and by Smith et al were the IL1B single-nucleotide polymorphisms (SNPs) 3953 and −511, and the IL1RN variable-number tandem repeat (VNTR). We had also genotyped these markers in our previous study (in which SNP 3953 was given the number 3954). In their study of the IL1RN VNTR, Meulenbelt et al noted that this marker had 2 common alleles, with a combined frequency of 97%. They therefore treated the VNTR as a biallelic marker. In our 2002 report, we also made this observation. In addition, we noted that the VNTR was in complete linkage disequilibrium with the IL1RN SNP 9589. We did not, therefore, complete the genotyping of the technically more challenging VNTR, because SNP 9589 provided the same information. In the analysis reported here, our 3-marker haplotypes were composed of the following markers: polymorphisms 3953, −511, and 9589. These provide the same genetic information as that presented by Meulenbelt et al and by Smith et al when they used the 3-marker haplotype, polymorphisms 3953, −511, and the VNTR.

Table 1 lists the estimated frequencies of the 8 haplotypes for the 370 patients with hip OA and the 544 controls for whom complete data were available from our 2002 study of markers 3953, −511, and 9589. The haplotype frequencies were estimated using the EH-PLUS program ( The 2 risk haplotypes reported by Meulenbelt et al were 1-1-2 and 1-2-1. Smith et al replicated the association to the haplotype 1-1-2. In the study by Meulenbelt et al, the protective haplotype was 2-1-1. In our study, the 1-1-2 risk haplotype had a frequency of 0.10 in both patients and controls (P > 0.05). In the study by Meulenbelt et al, this haplotype had a frequency of 0.16 in patients versus 0.07 in controls. In the study by Smith et al, this haplotype had a frequency of 0.32 in patients versus 0.11 in controls. Therefore, the frequency of the 1-1-2 haplotype in controls in our study was comparable with the frequencies among controls reported by Meulenbelt et al and Smith et al (0.10 versus 0.07 and 0.11, respectively), whereas the frequencies among patients in these 3 studies were not comparable (0.10 versus 0.16 and 0.32, respectively). This pattern was also observed for the 1-2-1 risk haplotype, with frequencies in our patients and controls of 0.15 and 0.14, respectively (P > 0.05), compared with frequencies in the Meulenbelt study of 0.28 in patients and 0.15 in controls. For the 2-1-1 protective haplotype, we observed an identical frequency (0.19) in our patients and controls (P > 0.05). This compared with frequencies in the Meulenbelt study of 0.11 in patients and 0.21 in controls. Overall, we did not observe any significant differences (P > 0.05 for all comparisons) in frequency between our patients and controls for the 2 risk haplotypes or for the protective haplotype. The frequencies in our controls were comparable with the frequencies among controls reported by Meulenbelt et al and Smith et al, implying that any technical discrepancies or errors in marker genotyping or in the construction and estimation of haplotype frequencies between our groups are unlikely.

Table 1. Haplotype frequencies for the 3 markers of IL1*
  • *

    Values are the number (frequency).

  • Alleles are arranged in the following order: 3953, −511, 9589. Haplotype 1-1-1 is the C allele of 3953, the C allele of −511, and the A allele of 9589; haplotype 2-2-2 is the T allele of 3953, the T allele of −511, and the T allele of 9589.

All patients (n = 370)284 (0.38)72 (0.10)112 (0.15)110 (0.15)139 (0.19)8 (0.01)12 (0.02)3 (0.004)
All controls (n = 544)392 (0.36)111 (0.10)154 (0.14)168 (0.15)210 (0.19)28 (0.03)20 (0.02)5 (0.004)
Female patients (n = 228)172 (0.38)45 (0.10)75 (0.17)68 (0.15)79 (0.17)8 (0.02)7 (0.02)2 (0.005)
Female controls (n = 205)149 (0.36)46 (0.11)56 (0.14)62 (0.15)71 (0.17)16 (0.04)10 (0.02)0 (0)
Male patients (n = 142)112 (0.39)27 (0.10)37 (0.13)42 (0.15)60 (0.21)0 (0)5 (0.02)1 (0.004)
Male controls (n = 339)243 (0.36)65 (0.10)98 (0.14)106 (0.16)139 (0.21)12 (0.02)10 (0.01)5 (0.007)

Using the CLUMP program ( we compared the frequencies of all 8 haplotypes derived from markers 3953, −511, and 9589 between our patients and controls. All P values were greater than 0.05 (data not shown). We finally expanded our haplotype analysis to include all markers genotyped in our 2002 study. There were no significant differences between our patients and controls (data not shown).

In conclusion, we did not replicate the association with hip OA of the IL1 haplotypes that Meulenbelt et al had previously reported. Our patient sample size was 6 times greater than that used by Meulenbelt et al (61 versus 370) and 17 times greater than that used by Smith et al (22 versus 370); therefore, a lack of power cannot be responsible for our failure to replicate. We conclude that the haplotypes reported by Meulenbelt et al do not influence hip OA susceptibility to the degree suggested by Meulenbelt et al and Smith et al. As more independent studies are reported, an accurate assessment of the effect of the IL1 gene cluster on OA susceptibility will emerge.

Kay Chapman PhD*, John Loughlin PhD*, * University of Oxford, Oxford, UK.