PubMed(http://www.ncbi.nlm.nih.gov/pubmed),EMBASE(http://www.embase.com),Weipu(http://www.cqvip.com/), CNKI(http://www.cnki.net/) and Wanfang(http://g.wanfangdata.com.cn/) databases were throughly searched by the authors (last search update, July 10, 2020）. The key words were ‘vitamin D receptor’ or ‘VDR’ and ‘osteoporosis’ or ‘fracture’ and ‘BsmI’ or ‘rs1544410’ in combination with ‘genetic’ or ‘polymorphism’ or ‘variant’.
Selection criteria of this meta-analysis were listed below: ①case-control or cohort studies. ②participants included postmenopausal women. ③assessment of the relationship of BsmI and osteoporosis or fracture. ④containing available genotype frequencies of BsmI. ⑤provided BMD values (mean and standard deviation) of lumbar spine and femoral neck, osteoporosis was defined as BMD ≤ -2.5 SDs (T-score).
Exclusion criteria of this meta-analysis were listed below: ①reviews, case reports, comments and letters. ②incomplete data. ③without full text. In addition, all relevant references were also reviewed. If there were duplicate data in papers published by the same author, only the most recent or complete study was included in this analysis.
Two independent investigators extracted data from eligible studies, the characteristics included: ①the 1st author. ②publication year. ③region. ④ethnicity. ⑤age range. ⑥sample size. ⑦allele frequency of cases and controls. and ⑧genotyping method. Any different evaluation results needs to be revisited until a consensus is reached.
The quality of eligible publications was assessed by the Newcastle-Ottawa quality assessment scales (NOS)14 . The scale contains three parts: the selection of groups (4 questions, 1 score each), the comparability of groups (1 question, 2 scores ), the ascertainment of exposure (3 questions, 1 score each). The scores ≥5 was regarded as a high quality study.
The observed genotype frequencies of the VDR BsmI polymorphism in control groups were assessed for Hardy-Weinberg equilibrium using the Χ2 test. The relationship between VDR BsmI gene and osteoporosis was accessed by calculating odds ratios (ORs) and 95% confidence intervals (CIs). The pooled ORs were performed for additive genetic model (b vs. B), dominant model (bb + Bb vs. BB), recessive model (bb vs. Bb + BB), and co-dominant model (Bb vs. BB, bb vs. BB) respectively. The subgroup analyses by ethnic groups also be performed. The statistically significant p value was set at 0.05. Heterogeneity assumption was evaluated by a chi-square based Q-test (p<0.05 indicated heterogeneity across studies ).The summary OR estimate of each study was calculated by the fixed-effects model if there was not significant heterogeneity. Otherwise, the random-effects model was used 15 16. The potential for publication bias was examined by a Begg's test (funnel plot method, p<0.05 considered representative of statistical significance) 17. All analysis were performed by Stata software (version 11.0).