To minimize phenotypic heterogeneity, we focused our initial genomewide association analysis on 306 subjects with RLS who also had periodic limb movements in sleep. Two markers, rs3923809 and rs6923737, in an intron of the BTB (POZ) domain–containing 9 (BTBD9) gene on chromosome 6p21.2 (Figure 1) showed genomewide associations that were significant (for rs3923809: odds ratio, 1.8; P=2×10–9; for rs6923737: odds ratio, 1.7; P=1×10–7) (Table 1, and Fig. 1 and Table 2 of the Supplementary Appendix). After adjustment for rs3923809, the association with rs6923737 was no longer significant (P=0.16), whereas the association with rs3923809 remained significant after adjustment for rs6923737 (P=0.001). None of the other 70 SNPs in a 600-kb region around rs3923809 remained significant after adjustments for rs3923809 and for multiple testing. The association with rs3923809 remained significant after adjustment for each of the SNPs individually (Table 2 of the Supplementary Appendix).
To validate these results, we analyzed a second Icelandic trial of 123 subjects with RLS and periodic limb movements in sleep and 1233 controls. The results with the second trial significantly replicated the original results for rs3923809 (odds ratio, 1.8; P=4×10–4) (Table 1, and Table 2 of the Supplementary Appendix). Extending the replication effort to a third trial of 188 subjects with RLS and periodic limb movements in sleep and 662 controls from the United States further confirmed the initial result for rs3923809 (odds ratio, 1.5; P=0.004) (Table 1). With all three samples combined, the association between the A allele of rs3923809 and RLS and periodic limb movements in sleep was highly significant (odds ratio, 1.7; P=3×10–14). There was no significant deviation from the multiplicative model, which assumed that the ratio of risk for homozygous carriers (AA) to heterozygous carriers (AG) was the same as the ratio of risk for heterozygous carriers to homozygous noncarriers (GG) in both the Icelandic subjects and the U.S. subjects (Table 3 of the Supplementary Appendix). The odds ratio for homozygous carriers was estimated at 3.2 for the Icelandic subjects and 2.3 for the U.S. subjects under the multiplicative model and at 4.3 for the Icelandic subjects and 2.0 for the U.S. subjects under the full model.
Among the 229 subjects who reported having RLS symptoms in the absence of periodic limb movements (35%), there was no association with the A allele of rs3923809 (odds ratio, 1.0; P=0.81) (Table 2). Conversely, among the 105 subjects who had periodic limb movements in sleep but who did not meet the RLS consensus criteria, there was an association with the A allele of marker rs3923809 (odds ratio, 2.3; P=2×10–6) (Table 2). The odds ratio for this group did not differ significantly from that for the group that had RLS plus periodic limb movements in sleep (P=0.19). With the combined data from all the Icelandic subjects who had periodic limb movements in sleep (i.e., those with and those without RLS), the strength of the association (odds ratio, 1.9; P=1×10–17) was greater than that for the group with RLS plus periodic limb movements in sleep alone.
We found that the frequency of periodic limb movements in sleep correlated with the presence of allele A of marker rs3923809 (Figure 2A) and that AA homozygotes had almost twice as many limb movements per hour of sleep as did noncarriers (P<0.001) (Figure 2B). The odds ratio for the group of subjects with the most severe symptoms (>20 movements per hour of sleep) was 2.0, whereas it was 1.0 for the group with the least severe symptoms (≤5 movements per hour of sleep) (Figure 2A). No significant correlation was observed between allele A and the severity of RLS symptoms, as assessed on the basis of the IRLSSG rating scale (P=0.35) or the self-reported age at the onset of RLS symptoms (P=0.73).
Female sex, advanced age, depletion of body iron stores, and western European ancestry were risk factors for RLS.38,39 To determine whether these factors interact with the at-risk variant, we analyzed them as covariates in conferring a risk of RLS. The risk of RLS and periodic limb movements in sleep conferred by allele A of rs3923809 in men was greater than that for women (odds ratio, 2.0 vs. 1.7), although the difference was not significant (P=0.28). A similarly insignificant trend was observed for the combined groups with periodic limb movements in sleep (odds ratio for men, 2.3; odds ratio for women, 1.7; P=0.09). The number of periodic limb movements in sleep was significantly higher after the age of 50 years than at a younger age (P<0.001 for both sexes), a finding that is consistent with a previous study.23 The difference was significantly less in women (P=0.04).
The principal clinical measures of iron availability are serum iron, transferrin iron-binding capacity, and ferritin. Serum soluble transferrin receptor, ferritin, total iron-binding capacity, and iron were assayed in 965 Icelandic subjects (subjects with RLS and their relatives). The ferritin index, a measure inversely related to body iron stores, was increased by 5.5% per A allele of marker rs3923809 (95% confidence interval [CI], 1 to 10; P=0.02). In line with this observation, serum ferritin levels were decreased by 13% per A allele (95% CI, 5 to 20; P=0.002) (Figure 3).