The eligible studies from the time of database established to Oct. 2019 were collected from PubMed, Springer, Cochrane library, Web of Science, China National Knowledge Infrastructure (CNKI), Wan Fang. Two independent authors extracted the data and assessed case-control trial quality.
In the meta-analysis, we made use of the Newcastle-Ottawa Scale in Epidemiology (NOS) group [13]. The PubMed, Springer, Cochrane library, Web of Science, China National Knowledge Infrastructure (CNKI), Wan Fang library were searched (updated to October 20, 2019) with terms ‘differentiation factor 5’, ‘GDF5’, ‘rs143383 ’, ‘polymorphism’, ‘osteoarthritis’ and ‘OA’, as both medical subject heading (Me SH) terms and text words to find all papers that had studied the association of GDF 5 with OA. A manual search was applied to finding unknown references to additional studies. English and Chinese language restrictions were applied. Studies were selected if they satisfy the following criteria: (1) Case-control study; (2) Sufficient published data for calculating the odds ratio and 95% confidence interval; (3) The association of GDF 5 polymorphism with OA;(4) Matched Hardy–Weinberg equilibrium (HWE) in control cases; (5) Having five models’ data of allelic model; homozygote model heterozygote model; recessive model and dominant model.
Data Extraction and Assess of Quality
Two researchers (Lei Peng and Jiping Lu) conducted eligible studies based on the above inclusion criteria and collected information on each eligible study according to the inclusion criteria. The following items were extracted: first author, year of publication, country, population, genotype distribution, Hardy-Weinberg equilibrium (HWE), case, and control size. To avoid the wrong data, the researchers will examine the collected data and make a conclusion through discussion. The quality of studies was evaluated by two independent investigators (Peng and Lu) based on the Newcastle-Ottawa Scale (NOS) for case-control studies [14]. The study was considered high quality with the scores were ≥ 7. In the case of disputes, we settle disputes through discussion. A third investigator (Peng Wang) decided this on the basis of discussions.
Statistical methods
Pearson's X2 tested estimates deviation from HWE in the control group according to genotype distributions Crude OR with their 95% CI was estimated and used to assess the strength of association between GDF 5 rs143383 polymorphism and KOA. The pooled OR was calculated respectively for allelic effect of C versus T, homozygote comparison of CC versus TT, heterozygote comparison of CT versus CC+TT, recessive model (CC versus TT +CT) and dominant model (CC +CT versus TT). The significance of the pooled OR was determined by the Z-test (P ≤ 0.05). Q statistics (P <0.10 indicated the evidence of heterogeneity was used to assess heterogeneity between studies. When significant heterogeneity was achieved (P < 0.10), the effect size of the study was combined with the random effect model, otherwise the fixed effect model was used. Subgroup analysis was performed according to population, and sensitivity analysis was performed to determine the impact of individual studies on the aggregated results and to test the reliability of the results. The potential publishing bias was estimated by Begg's funnel plot and Egger regression test. All cases were analyzed by STATA 18.0 software (Stata Corporation, College Station, TX, USA). The P values were bilateral. This study followed the PRISMA standard.