Osteoarthritis of the knee is a major cause of pain, disability and the use of healthcare resources among middle-aged and older people.1 Although osteoarthritis is multifactorial, it is known to have a significant genetic contribution and a number of studies have attempted to dissect such a contribution (see Valdes and Spector2 for review).

The GDF5 gene encodes the growth differentiation factor 5, a bone morphogenetic protein involved in joint formation, expressed in different joint structures, which has been shown to ameliorate tendon, ligament and bone healing after trauma in mice.3 4

A promoter polymorphism (rs143383) in GDF5 has been found to be strongly associated with both hip and knee osteoarthritis in Asian individuals,4 and is the most widely replicated genetic association with knee osteoarthritis, although much less so for hip and hand osteoarthritis.5 This variant is functional, with the lower gene expression variant having increased genetic risk.4

A large-scale meta-analysis reported the association of the major (T) allele with knee osteoarthritis achieved OR 1.15 p=9.7×10⁻⁷ and achieved p=9×10⁻⁵ (OR 1.13, 95% CI 1.06 to 1.20) when Asian subjects were excluded.5

The genome-wide statistical significance level of p<5×10⁻⁸ is increasingly seen as the threshold at which genetic associations are considered credible.6 The aim of our study was to prove that common genetic variation in the GDF5 gene is important in knee osteoarthritis beyond reasonable doubt.

We genotyped 3303 controls and 2235 knee osteoarthritis cases from the UK, Estonia and The Netherlands, added published data from the Chingford Study (259 cases and 509 controls),7 and combined with published effect size estimates from the recent large-scale meta-analysis5 using both fixed and random effects models as described in Evangelou et al.5 A full detailed description of each study cohort on recruitment, radiographic and clinical assessment is found in Hofman et al8 and Valdes et al.9 The studies were approved by the relevant ethics committees and informed consent was obtained from all study participants. Genotypes were subtracted from the genome-wide association dataset of the Rotterdam Study III with methods described previously.10 DNA from UK and Estonian study participants was genotyped by Kbioscience (Hertfordshire UK) using methods described elsewhere.7 The total, including previously reported data, is 10 103 controls and 6861 knee osteoarthritis cases of European descent and 1844 controls and 718 cases of Asian descent, which has 86% statistical power to detect the association as genome-wide significant with an OR of 1.15 or higher for the T allele. The descriptive characteristics of new samples studied are shown in table 1 along with the OR for the T allele and CI. The individual study effect sizes are shown in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Forest plot of study-specific estimates and fixed-effects summary OR estimates and 95% CI for the association between the rs143383 polymorphism of the GDF5 gene and knee osteoarthritis (OA) in current and previously published studies. The summary statistics and corresponding p values for all studies and for sensitivity analyses excluding studies out of Hardy–Weinberg equilibrium (HWE; indicated by *) and/or Asian studies are shown. The between-study heterogeneity I² and the respective 95% CI are also included.

The results of the combined meta-analysis show that the T allele of GDF5 rs143833 is associated with a 17% increased risk of knee osteoarthritis (OR 1.17, 95% CI 1.12 to 1.23). When all data were analysed the genetic association reached genome-wide significance with p=6.2×10⁻¹¹ in all samples and with p=8.3×10⁻⁹ in European descent samples alone (figure 1). There was no significant between-study heterogeneity. The p value was even smaller when one study violating Hardy–Weinberg equilibrium5 was excluded (OR 1.18, 95% CI 1.12 to 1.23, p=4.1×10⁻¹¹). Stratification according to gender did not reveal differences in the effect size (OR 1.14, 95% CI 1.05 to 1.23, p=1.6×10⁻³ in men vs OR 1.19, 95% CI 1.10 to 1.27, p=3.7×10⁻⁶ in women by random effects).

We have shown that the association between a functional promoter single nucleotide polymorphism in the GDF5 gene and knee osteoarthritis achieves genome-wide statistical significance. The association is consistently replicated and no significant heterogeneity is detected between studies, further strengthening the robustness of GDF5 as a risk factor for knee osteoarthritis.