Data_Sheet_2_Causal effect of polyunsaturated fatty acids on bone mineral density and fracture.docx

BackgroundPolyunsaturated fatty acids (PUFAs) are closely related to osteoporosis. To test their causal relationship, we conducted a Mendelian randomization (MR) analysis.
MethodsWe analyzed the causal relationship between four PUFAs measures, n-3 PUFAs (n-3), n-6 PUFAs (n-6), the ratio of n-3 PUFAs to total fatty acids (n-3 pct), and the ratio of n-6 PUFAs to n-3 PUFAs (n-6 to n-3), and five measures of osteoporosis, including estimated bone mineral density (eBMD), forearm (FA) BMD, femoral neck (FN) BMD, lumbar spine (LS) BMD, and fracture, using two-sample MR analysis. In order to verify the direct effect between PUFAs and BMD, we chose interleukin-6 (IL-6), tumor necrosis factor-β (TNF-β), and bone morphogenetic proteins 7 (BMP-7), three markers or cytokines strongly related to BMD, as possible confounding factors, and analyzed the possible causal relationships between them and PUFAs or BMD by MR. Inverse variance weighting (IVW), MR-Egger, weighted and weighted median were conducted. MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO) and MR-Egger regression methods were used to evaluate the potential pleiotropy of instrumental variables (IVs) and outliers were identified by MR-PRESSO. Cochran’s Q statistic was used to detect the heterogeneity among IVs. Leave-one-out sensitivity analysis was used to find SNPs that have a significant impact on the results. All results were corrected by the Bonferroni correction.
ResultsThe IVW results showed that n-3 PUFAs (OR = 1.030, 95% CI: 1.013, 1.047, P = 0.001) and n-6 PUFAs (OR = 1.053, 95% CI: 1.034, 1.072, P < 0.001) were positively correlated with eBMD, while n-6 to n-3 (OR = 0.947, 95% CI: 0.924, 0.970, P < 0.001) were negatively correlated with eBMD. These casual relationships still existed after Bonferroni correction. There were positive effects of n-3 PUFAs on FA BMD (OR = 1.090, 95% CI: 1.011, 1.176, P = 0.025) and LS BMD (OR = 1.056, 95% CI: 1.011, 1.104, P = 0.014), n-3 pct on eBMD (OR = 1.028, 95% CI: 1.002, 1.055, P = 0.035) and FA BMD (OR = 1.090, 95% CI: 1.011, 1.174, P = 0.025), n-6 to n-3 on LS BMD (OR = 1.071, 95% CI: 1.021, 1.124, P = 0.005); negative effects of n-3 pct on fracture (OR = 0.953, 95% CI: 0.918, 0.988, P = 0.009) and n-6 to n-3 on FA BMD (OR = 0.910, 95% CI: 0.837, 0.988, P = 0.025). However, these causal effects all disappeared after Bonferroni correction (all P > 0.0025). None of IL-6, TNF-β, and BMP-7 had a causal effect on PUFA and BMD simultaneously (all P > 0.05).
ConclusionEvidence from this MR study supports the genetically predicted causal effects of n-3, n-6, n-3 pct, and n-6 to n-3 on eBMD. In addition, n-3 not only associate with FA BMD and LS BMD through its own level and n-6 to n-3, but also link to fracture through n-3 pct.

研究背景：多不饱和脂肪酸（PUFAs）与骨质疏松症密切相关。为验证二者的因果关系，本研究开展了孟德尔随机化（Mendelian Randomization, MR）分析。
研究方法：本研究针对4种多不饱和脂肪酸指标与5项骨质疏松症指标间的因果关系展开双样本孟德尔随机化分析。4种PUFAs指标分别为n-3多不饱和脂肪酸（n-3 PUFAs）、n-6多不饱和脂肪酸（n-6 PUFAs）、n-3多不饱和脂肪酸占总脂肪酸的比例（n-3 pct）以及n-6与n-3多不饱和脂肪酸的比值（n-6 to n-3）；5项骨质疏松症指标包括估算骨密度（estimated bone mineral density, eBMD）、前臂骨密度（forearm BMD, FA BMD）、股骨颈骨密度（femoral neck BMD, FN BMD）、腰椎骨密度（lumbar spine BMD, LS BMD）以及骨折事件。
为验证多不饱和脂肪酸与骨密度间的直接作用机制，本研究选取与骨密度密切相关的3种标志物/细胞因子——白细胞介素-6（interleukin-6, IL-6）、肿瘤坏死因子-β（tumor necrosis factor-β, TNF-β）及骨形态发生蛋白7（bone morphogenetic proteins 7, BMP-7）作为潜在混杂因素，并通过孟德尔随机化分析三者与多不饱和脂肪酸或骨密度间可能存在的因果关系。
本研究采用逆方差加权法（inverse variance weighting, IVW）、MR-Egger法、加权法及加权中位数法进行统计分析。使用孟德尔随机化多效性残差和离群值（MR-Pleiotropy RESidual Sum and Outlier, MR-PRESSO）及MR-Egger回归评估工具变量（instrumental variables, IVs）的潜在多效性，并通过MR-PRESSO识别离群值。采用Cochran’s Q统计量检测工具变量间的异质性。通过留一法敏感性分析筛选对研究结果存在显著影响的单核苷酸多态性（SNPs）。所有统计结果均经Bonferroni校正。
研究结果：逆方差加权法分析结果显示，n-3 PUFAs（比值比OR=1.030，95%置信区间CI: 1.013~1.047，P=0.001）与n-6 PUFAs（OR=1.053，95%CI: 1.034~1.072，P<0.001）与估算骨密度呈正相关，而n-6与n-3的比值（OR=0.947，95%CI: 0.924~0.970，P<0.001）与估算骨密度呈负相关。上述因果关系在经Bonferroni校正后依然成立。
n-3 PUFAs对前臂骨密度（OR=1.090，95%CI: 1.011~1.176，P=0.025）及腰椎骨密度（OR=1.056，95%CI: 1.011~1.104，P=0.014）存在正向作用；n-3 pct对估算骨密度（OR=1.028，95%CI: 1.002~1.055，P=0.035）及前臂骨密度（OR=1.090，95%CI: 1.011~1.174，P=0.025）存在正向作用；n-6与n-3的比值对腰椎骨密度存在正向作用（OR=1.071，95%CI: 1.021~1.124，P=0.005）。此外，n-3 pct对骨折事件存在负向作用（OR=0.953，95%CI: 0.918~0.988，P=0.009），n-6与n-3的比值对前臂骨密度存在负向作用（OR=0.910，95%CI: 0.837~0.988，P=0.025）。但上述因果效应在经Bonferroni校正后均不再显著（所有P>0.0025）。白细胞介素-6、肿瘤坏死因子-β及骨形态发生蛋白7均未同时对多不饱和脂肪酸与骨密度产生因果影响（所有P>0.05）。
研究结论：本孟德尔随机化研究的证据支持遗传预测的n-3 PUFAs、n-6 PUFAs、n-3 pct及n-6与n-3的比值对估算骨密度存在因果效应。此外，n-3 PUFAs不仅可通过自身水平及n-6与n-3的比值影响前臂骨密度与腰椎骨密度，还可通过n-3 pct与骨折事件产生关联。