摘要目的:探讨人血液代谢物介导的人循环炎症蛋白与压力性损伤（PI）的因果关系。方法:该研究为基于中介孟德尔随机化（MR）分析方法的研究。获取91种人循环炎症蛋白（14 824个样本）、1 400种人血液代谢物（8 299个样本）、PI（467 794个样本）的全基因组关联分析数据，设置显著阈值，以单核苷酸多态性（SNP）作为工具变量并排除弱工具变量的影响。采用正向双样本MR（TSMR）分析循环炎症蛋白与PI之间的因果关系，以逆方差加权（IVW）法作为主要方法，用加权中位数法、MR-Egger回归、加权模式法、简单模式法对IVW法的结果进行验证（具体分析方法后同）。针对筛选出的循环炎症蛋白的SNP进行敏感性分析，包括行Cochran Q检验评估异质性，行MR-Egger截距检验、MR-PRESSO离群值检验评估水平多效性，采用留一法分析评估可靠性。基于反向TSMR分析，采用IVW法、MR-Egger回归、加权中位数法、简单模式法、加权模式法评估PI与筛选出的循环炎症蛋白之间是否存在反向因果关系。采用正向TSMR分析筛选出的循环炎症蛋白与1 400种血液代谢物之间的因果关系并筛选血液代谢物，对筛选出的循环炎症蛋白的SNP同前行敏感性分析。采用正向TSMR分析筛选出的血液代谢物与PI之间的因果关系并对筛选出的血液代谢物的SNP行敏感性分析（留一法分析除外）。计算筛选出的血液代谢物在筛选出的循环炎症蛋白与PI之间的介导作用中的中介效应值和中介效应比。 结果:筛选出5种循环炎症蛋白及59种血液代谢物符合暴露因素条件，达到显著阈值的SNP数量从16个到1 484个不等，且均为强工具变量。IVW法显示，白细胞介素-33（IL-33）、CUB结构域蛋白1、IL-5、干细胞因子、肿瘤坏死因子与PI之间均存在显著因果关系（比值比分别为1.29、1.20、1.25、1.16、1.23，95%置信区间分别为1.07~1.55、1.05~1.36、1.04~1.51、1.00~1.34、1.03~1.47， P<0.05）；经加权中位数法验证，IL-33、IL-5与PI之间均存在显著因果关系（比值比分别为1.37、1.37，95%置信区间分别为1.05~1.79、1.04~1.80， P<0.05）。其中IL-33与PI之间存在最为显著的因果关系（ P<0.01）。Cochran Q检验评估显示，与PI之间存在显著因果关系的IL-33的SNP不存在显著异质性（ Q=18.78， P>0.05）；MR-Egger截距检验（截距的绝对值<0.001， P>0.05）和MR-PRESSO离群值检验（RSSobs值为20.37， P>0.05）评估均显示，与PI之间存在显著因果关系的IL-33的SNP不存在显著水平多效性。留一法分析表明，IL-33与PI之间的显著因果关系在逐个剔除SNP后结果可靠。IVW法、MR-Egger回归、加权中位数法、简单模式法、加权模式法均显示，PI与IL-33之间不存在显著反向因果关系（比值比分别为1.00、1.00、1.00、1.00、1.01，95%置信区间分别为0.98 ~1.02、0.96 ~1.03、0.97 ~1.03、0.93 ~1.08、0.94 ~1.09， P>0.05）。IVW法显示，IL-33与59种血液代谢物之间均存在显著因果关系（比值比为0.79~1.20，95%置信区间下限范围为0.70~1.07、上限范围为0.89~1.37， P<0.05）；经MR-Egger回归、加权中位数法验证，IL-33分别与8、10种血液代谢物之间存在显著因果关系（比值比分别为0.63~1.70、0.82~1.21，95%置信区间下限范围分别为0.43~1.29和0.70~1.14、上限范围分别为0.94~2.25和0.97~1.42， P值均<0.05）。其中血液代谢物X-12798与IL-33之间存在最为显著的因果关系（比值比为0.79，95%置信区间为0.70~0.89， P<0.05）。Cochran Q检验评估显示，与血液代谢物X-12798之间存在显著因果关系的IL-33的SNP不存在显著异质性（ Q=24.94， P>0.05）；MR-Egger截距检验（截距的绝对值为0.012， P>0.05）和MR-PRESSO离群值检验（RSSobs值为27.45， P>0.05）评估均显示，与血液代谢物X-12798之间存在显著因果关系的IL-33的SNP不存在显著水平多效性。留一法分析表明，IL-33与血液代谢物X-12798之间的显著因果关系在逐个剔除SNP后结果可靠。IVW法显示，血液代谢物X-12798与PI之间存在显著因果关系（比值比为0.92，95%置信区间为0.84~0.99， P<0.05）；经MR-Egger回归、加权中位数法均验证，血液代谢物X-12798与PI之间存在显著因果关系（比值比分别为0.87、0.89，95%置信区间分别为0.77 ~0.98、0.80 ~0.99， P<0.05）。Cochran Q检验评估显示，与PI之间存在显著因果关系的血液代谢物X-12798的SNP不存在显著异质性（ Q=23.45， P>0.05）；MR-Egger截距检验（截距的绝对值为0.015， P>0.05）和MR-PRESSO离群值检验（RSSobs值为26.01， P>0.05）评估均显示，与PI之间存在显著因果关系的血液代谢物X-12798的SNP不存在显著水平多效性。中介效应值为0.02，中介效应比为8.27%。 结论:人循环炎症蛋白、血液代谢物和PI之间存在显著因果关系，血液代谢物X-12798在循环炎症蛋白IL-33与PI之间发挥介导作用。

abstractsObjective:To explore the causal relationship between human circulating inflammatory proteins and pressure injury (PI) mediated by human blood metabolites.Methods:This study employed a method of analysis based on mediated Mendelian randomization (MR). Genome-wide association study data of 91 human circulating inflammatory proteins (14 824 samples), 1 400 human blood metabolites (8 299 samples), and PI (467 794 samples) were retrieved. A significance threshold was established and single nucleotide polymorphisms (SNPs) were used as instrumental variables with the influence of weak instrumental variables excluded. Forward two-sample MR (TSMR) was employed to analyze the causal relationship between circulating inflammatory proteins and PI. The inverse variance weighted (IVW) method served as the primary approach, and the results were validated using the weighted median method, MR-Egger regression, weighted mode method, and simple mode method (the specific analytical methods were the same below). For the SNPs of selected circulating inflammatory proteins, sensitivity analysis was employed, including heterogeneity which was evaluated by the Cochran Q test, horizontal pleiotropy which was evaluated by the MR-Egger intercept test and MR-PRESSO outlier test, and reliability which was evaluated via leave-one-out analysis. Based on reverse TSMR analysis, the IVW method, MR-Egger regression, weighted median method, simple mode method, and weighted mode method were employed to evaluate whether a reverse causal relationship exists between PI and the selected circulating inflammatory proteins. Forward TSMR was employed to analyze the causal relationship between selected circulating inflammatory proteins and 1 400 blood metabolites and to select the blood metabolites. For the SNPs of selected circulating inflammatory proteins, sensitivity analysis was employed as before. Forward TSMR was employed to analyze the causal relationship between selected blood metabolites and PI. For the SNPs of selected blood metabolites, sensitivity analysis was employed as before (except for the leave-one-out analysis). Finally, the mediation effect values and mediation effect ratios of selected blood metabolites in the mediation effect between selected circulating inflammatory proteins and PI were calculated. Results:Five circulating inflammatory proteins and 59 blood metabolites were identified as meeting the exposure factor criteria, with the number of SNPs reaching the significance threshold ranging from 16 to 1 484. All the SNPs were confirmed as strong instrumental variables. The IVW method revealed significant causal relationships between interleukin-33 (IL-33), CUB domain-containing protein 1, IL-5, stem cell factor, and tumor necrosis factor and PI (with odds ratios of 1.29, 1.20, 1.25, 1.16, and 1.23, respectively, 95% confidence intervals of 1.07-1.55, 1.05-1.36, 1.04-1.51, 1.00-1.34, and 1.03-1.47, respectively, P<0.05). The weighted median method confirmed significant causal relationships between IL-33 and IL-5 and PI (with odds ratios of 1.37 and 1.37, respectively, 95% confidence intervals of 1.05-1.79 and 1.04-1.80, respectively, P<0.05). Among these, the most significant causal relationship was observed between IL-33 and PI ( P<0.01). The Cochran Q test indicated no significant heterogeneity in the SNPs of IL-33 which had significant causal relationship with PI ( Q=18.78, P>0.05). The MR-Egger intercept test (with intercept absolute value <0.001, P>0.05) and MR-PRESSO outlier test (with RSSobs value of 20.37, P>0.05) both indicated no significant horizontal pleiotropy in the SNPs of IL-33 which had significant causal relationship with PI. The leave-one-out analysis showed that the significant causal relationship between IL-33 and PI was reliable after removing the SNPs one by one. No significant reverse causal relationships were observed between PI and IL-33 through the IVW method, MR-Egger regression, weighted median method, simple mode method, or weighted mode method (with odds ratios of 1.00, 1.00, 1.00, 1.00, and 1.01, respectively, 95% confidence intervals of 0.98-1.02, 0.96-1.03, 0.97-1.03, 0.93-1.08, and 0.94-1.09, respectively, P>0.05). The IVW method revealed significant causal relationships between IL-33 and 59 blood metabolites (with odds ratios of 0.79-1.20, 95% confidence intervals lower limit range of 0.70-1.07 and upper limit range of 0.89-1.37, P<0.05). The MR-Egger regression and weighted median method confirmed significant causal relationships between IL-33 and 8 and 10 blood metabolites, respectively (with odds ratios of 0.63-1.70 and 0.82-1.21, respectively, 95% confidence intervals lower limit ranges of 0.43-1.29 and 0.70-1.14, respectively, 95% confidence intervals upper limit ranges of 0.94-2.25 and 0.97-1.42, respectively, P values all <0.05). Among these, the most significant causal relationship was observed between blood metabolite X-12798 and IL-33 (with odds ratio of 0.79, 95% confidence interval of 0.70-0.89, P<0.05). The Cochran Q test indicated no significant heterogeneity in the SNPs of IL-33 which had significant causal relationship with blood metabolite X-12798 ( Q=24.94, P>0.05). The MR-Egger intercept test (with intercept absolute value of 0.012, P>0.05) and MR-PRESSO outlier test (with RSSobs value of 27.45, P>0.05) both indicated no significant horizontal pleiotropy in the SNPs of IL-33 which had significant causal relationship with blood metabolite X-12798. The leave-one-out analysis showed that the significant causal relationship between IL-33 and blood metabolite X-12798 was reliable after removing the SNPs one by one. The IVW method revealed significant causal relationship between blood metabolite X-12798 and PI (with odds ratio of 0.92, 95% confidence interval of 0.84-0.99, P<0.05). The MR-Egger regression and weighted median method both confirmed significant causal relationship between blood metabolite X-12798 and PI (with odds ratios of 0.87 and 0.89, respectively, 95% confidence intervals of 0.77-0.98 and 0.80-0.99, respectively, P<0.05). The Cochran Q test indicated no significant heterogeneity in the SNPs of blood metabolite X-12798 which had significant causal relationship with PI ( Q=23.45, P>0.05). The MR-Egger intercept test (with intercept absolute value of 0.015, P>0.05) and MR-PRESSO outlier test (with RSSobs value of 26.01, P>0.05) both indicated no significant horizontal pleiotropy in the SNPs of blood metabolite X-12798 which had significant causal relationship with PI. The mediation effect value was 0.02 and the mediation effect ratio was 8.27%. Conclusions:Significant causal relationships are observed among human circulating inflammatory proteins, blood metabolites, and PI, with the association between circulating inflammatory protein IL-33 and PI being mediated by blood metabolite X-12798.