【作者】 周云平；

【导师】 贾崇奇；

【作者基本信息】 山东大学 ， 流行病与卫生统计学， 2014， 博士

【摘要】 研究背景冠状动脉粥样硬化性心脏病即冠心病(CAD)是目前我国及大多数国家的主要死因与疾病负担之一。随着我国社会经济的快速发展、人们生活方式的改变以及人口老龄化的加剧,CAD的发病率呈逐年增长的趋势。因此及早对CAD危险因素的识别,为该病的人群防制提供依据,是预防和控制该病发生发展的重要措施。尽管CAD许多危险因素如高血压、高脂血症、吸烟等目前已被确认,但资料显示,这些传统因素的病因作用仅约占50%,说明其它危险因素在CAD的发病中同样具有重要作用。铁是一种维持生命与健康的重要微量元素,在氧运输、细胞呼吸、能量代谢、基因表达、DNA复制和修复等细胞过程中均发挥重要作用。机体中的铁约66%以二价铁(Fe+2)形式存在于血红素中,另外27%以三价铁(Fe+3)形式存在于组织中的铁蛋白(ferritin)。当铁量超过细胞内的存储能力时,多余的铁与含铁血黄素(hemosiderin)结合及部分存储于网状内皮系统。机体中游离的Fe+2可通过Fenton及Haber-Weiss反应产生羟自由基(·OH-),使DNA损伤、脂质及蛋白氧化导致氧化应激。并且铁能损伤内皮细胞功能,降低一氧化氮(NO)生物合成与利用,因而当机体铁超负荷时对机体产生致病作用。众所周知,绝经期前女性CAD发病率显著低于同龄男性,而绝经期后女性CAD的发病率逐渐增加,并逐步变得与同龄男性相似。针对这一现象以及有关研究结果,学者Sullivan于1981年首次提出了“铁假说”,认为女性经期周期性的铁丢失,导致机体铁水平较低,可能有助于女性CAD的预防；而男性由于机体铁进行性积累导致CAD危险性的增加。自“铁假说”提出以来,学者们对机体铁超负荷与包括CAD在内的动脉粥样硬化性疾病的关系从流行病学研究到临床试验做了许多探讨,但研究结果不尽一致,包括两者存在显著正相关、无显著相关及显著负相关。尽管造成研究结果不一致的原因众多,但是以下几个方面则是主要的影响因素：①反映机体铁负荷的指标不一：目前所应用的反映机体铁负荷的指标包括血清铁蛋白、血清铁、总铁结合力、可溶性转铁蛋白受体等,然而上述指标在独立反映机体铁负荷状态时皆存在一定缺陷,无法准确评价机体铁负荷状态。如血清铁蛋白是一种急性期反应物,容易受炎症、感染等因素的影响,因此其很难准确反映机体铁负荷水平。然而到目前为止,尚无准确反映机体铁负荷的特异指标,基于上述原因,Ramakrishna指出,对上述铁负荷指标应该同时予以检测,以反映机体铁负荷状态。②研究的结局变量不一：目前关于铁负荷与心血管病关系的人群研究所应用的结局变量主要包括心肌梗塞发病与死亡、CAD发病与死亡、心血管病总死亡等,虽然上述疾病皆以粥样硬化为病理基础,但由于其均为多因子疾病,所以在其它病因学方面可能不尽相同。③潜在的混杂与偏倚控制不全：由于以动脉粥样硬化为病理基础的心血管系统疾病包括CAD是多因子疾病,吸烟、高血压、高脂血症等皆为重要的危险因素,因此如果不能有效地控制有关混杂,很难准确揭示出机体铁超负荷的致病作用。针对上述目前研究现状及存在的问题,本课题采用以医院为基础的病例对照研究,以冠状动脉造影新诊断的CAD患者为病例组,以正常健康者为对照组,通过同时测定血清铁蛋白、血清铁、总铁结合力、可溶性转铁蛋白受体四个目前最常用的反映机体铁负荷的指标,利用多因素分析模型控制有关潜在的混杂后,探讨机体铁超负荷与CAD的关系,并探讨与CAD相关的反映铁负荷的相对敏感指标或相对敏感指标的组合。另外,由于CAD是受多种因素共同作用的多因子疾病,本文也初步探讨铁负荷相对敏感指标与环境因素的交互作用在CAD致病中的效应。并利用meta分析,进一步探寻机体铁超负荷与CAD关系的证据。通过上述研究,阐明机体铁超负荷与CAD的关系,为该病的预防、诊断与治疗提供科学依据。本课题来源于国家自然基金资助项目(81072357)。研究目的1.明确机体铁超负荷与CAD的关系；2.探寻与CAD相关的反映机体铁负荷的相对敏感指标或相对敏感指标的组合；3.分析反映机体铁负荷的相对敏感指标与环境因素之间的交互作用对CAD的效应。研究方法1.采用以医院为基础的病例对照研究,以新诊断的CAD患者为病例组,选择在医院体检的健康人为对照组。用酶联免疫吸附法测定血清铁蛋白、血清铁、总铁结合力及可溶性转铁蛋白受体的浓度。2.通过贝叶斯模型平均法筛选对CAD发病有重要影响的协变量,然后在logistic回归模型框架下,运用限制性立方样条函数估计各个机体铁负荷指标与CAD发病风险的剂量反应关系。3.通过受试者工作特征曲线(Receiver Operating Characteristic curve, ROC)下面积(Area Under the Curve, AUC)的比较,探讨与CAD关系最为密切的机体铁负荷的相对敏感指标或相对敏感指标的组合。4.利用广义多因子降维法,探讨反映机体铁负荷的相对敏感指标与环境因素之间的高阶交互作用。5.采用meta分析的方法,系统的检索以往发表的研究机体铁负荷与CAD关系的相关文献,根据异质性检验的结果采用固定效应模型(I2<50%)或者随机效应模型(I2>50%)。 Meta回归和亚组分析用来探索异质性的来源,影响性分析用来衡量单篇文献对总的合并结果的影响,应用漏斗图法和Egger检验法检测可能的发表偏倚。研究结果1.一般情况：本研究共调查258例CAD病例和282例对照,年龄、高血压、体质指数、吸烟、人8异前列腺素F2a及糖尿病史CAD组显著高于对照组(P<0.05),而高密度脂蛋白CAD组显著低于对照组(P<0.05)。2.单因素分析：结果表明,血清铁水平CAD组显著高于对照组(P<0.01),总铁结合力和可溶性转铁蛋白受体水平CAD组显著低于对照组(P<0.01)。3.协变量的筛选：在Occam窗中共筛选了25个模型用来估计贝叶斯模型平均法的参数估计值,最佳模型的后验概率为0.15,最佳模型对应的BIC值为-2950.08。选入的变量有9个,包括年龄、糖尿病史、总胆固醇、低密度脂蛋白、高血压、饮酒、性别、人8异前列腺素F2a及高密度脂蛋白。4.多因素分析：(1)血清铁蛋白：调整了年龄、糖尿病史、总胆固醇、低密度脂蛋白、高血压、饮酒、性别、人8异前列腺素F2a、高密度脂蛋白后,多因素分析结果显示,血清铁蛋白与CAD的相关关系具有统计学意义(P<0.01),且存在显著地非线性剂量效应关系(P<0.01)。剂量效应关系表明,血清铁蛋白的参照值为200ug/L时,血清铁蛋白浓度为100ug/L、150ug/L、250ug/L、300ug/L及350ug/L时对应的OR(95%CI)分别为：1.69(1.18-2.41)、1.15(1.00-1.31)、1.07(0.93-1.21)、1.40(0.99-1.97)及1.83(1.09-3.07)。(2)血清铁：调整了年龄、糖尿病史、总胆固醇、低密度脂蛋白、高血压、饮酒、性别、人8异前列腺素F2a、高密度脂蛋白后,多因素分析结果表明,血清铁与CAD的相关关系具有统计学意义(P<0.01),且存在显著地非线性剂量效应关系(P<0.01)。剂量效应关系显示,血清铁的参考值为19μmol/L时,血清铁的浓度为16μmol/L、22pmol/L、26μmol/L、30μmol/L时对应的OR(95%CI)分别为：0.36(0.21-0.60)、0.66(0.54-0.81)、1.41(1.20-1.66)、2.00(1.45-2.75)及2.52(1.67-3.82)。(3)总铁结合力：调整了年龄、糖尿病史、总胆固醇、低密度脂蛋白、高血压、饮酒、性别、人8异前列腺素F2a、高密度脂蛋白后,多因素分析结果表明,总铁结合力与CAD的相关关系具有统计学意义(P<0.01),且存在显著地非线性剂量效应关系(P<0.01)。剂量效应关系显示,总铁结合力的参考值为90μmol/L时,总铁结合力的浓度为50μmol/L、70μmol/L、110pmol/L、130μmol/L及150μmnol/L时对应的OR(95%CI)分别为：4.58(2.61-8.03)、2.03(1.58-2.62)、0.59(0.49-0.68)、0.41(0.30-0.53)及0.33(0.22-0.49)。(4)可溶性转铁蛋白受体：调整了年龄、糖尿病史、总胆固醇、低密度脂蛋白、高血压、饮酒、性别、人8异前列腺素F2a、高密度脂蛋白后,多因素分析结果显示,可溶性转铁蛋白受体与CAD的相关关系具有统计学意义(P<0.01),且存在显著地线性剂量效应关系(P>0.05)。剂量效应关系表明,可溶性转铁蛋白受体的参考值为10mg/L时,可溶性转铁蛋白受体的浓度为6mg/L、8mg/L、12mg/L、16mg/L及20mg/L时对应的OR(95%CI)分别为：1.33(0.97-1.84)、1.16(1.00-1.35)、0.87(0.79-0.96)、0.67(0.53-0.86)及0.53(0.35-0.79)。5.ROC分析：血清铁蛋白、血清铁、总铁结合力与可溶性转铁蛋白受体的ROC曲线下面积AUC(95%CI)分别为：0.53(0.48-0.58)、0.73(0.69-0.77)、0.74(0.69-0.78)及0.61(0.56-0.66),而血清铁和总铁结合力两个铁负荷相对敏感指标的组合的AUC(95%CI)为0.86(0.83-0.90),显著优于单个铁负荷指标或其他铁负荷指标的组合(P<0.05)。6.高阶交互作用：广义多因子降维结果显示,交互作用最优模型为血清铁与总铁结合力铁负荷敏感指标的组合、体质指数和高密度脂蛋白的三因子交互作用模型,其检验样本准确度为0.85,置换检验P<0.01,交叉验证一致性为10/10。另外,与多因子降维法相比,调整年龄与性别两个协变量后三因子组合模型的预测准确度要明显高于不调整协变量的情况。7. Meta分析：根据筛选排除标准,最终共有26篇包含血清铁蛋白、血清铁、总铁结合力及可溶性转铁蛋白受体四个铁指标在内的关于机体铁负荷与CAD关系的相关研究纳入meta分析。(1)对于合并标准化均差(SMD)的meta分析,总共包含4,410例CAD病例和7,357例对照。其中血清铁蛋白、血清铁、总铁结合力与可溶性转铁蛋白受体的合并SMD(95%CI)结果分别为：0.74(0.46,1.01)、-0.06(-0.28,0.16)、-0.24(-0.59,0.10)及0.13(0.02,0.24)。除了可溶性转铁蛋白受体,其他三个铁指标的研究均检测到较大的异质性,排除了对异质性有较大影响的研究之后,血清铁蛋白、血清铁与总铁结合力合并SMD(95%CI)结果分别为：0.12(0.07,0.17)、-0.17(-0.31,-0.03)及-0.23(-0.39,-0.07)。上述所有研究的影响性分析皆没有发现对总的合并结果有明显影响的单篇文献,Egger检验均没有检测到明显的发表偏倚。(2)对于合并OR值的1meta分析,累计纳入病例5,738例,对照98,362例。血清铁蛋白、血清铁、总铁结合力与可溶性转铁蛋白受体合并的OR(95%CI)值的结果分别为1.13(0.98,1.30)、1.21(0.95,1.55)、0.97(0.94,0.99)及1.54(0.85,2.77)。仅有血清铁的研究检测到较大的异质性,排除了对异质性有较大影响的研究之后,血清铁的合并的OR(95%CI)值的结果为1.19(1.08,1.45)。上述所有研究的影响性分析皆没有发现对总的合并结果有明显影响的单篇文献,Egger检验均没有检测到明显的发表偏倚。主要结论1.本研究人群中,血清铁蛋白浓度高于200ug/L是CAD的危险因素,且与CAD之间存在显著的非线性剂量效应关系；血清铁浓度高于19μmol/L时是CAD的危险因素,与CAD之间存在显著的非线性剂量效应关系；总铁结合力浓度低于90μmol/L时是CAD的危险因素,与CAD之间存在显著的非线性剂量效应关系；可溶性转铁蛋白受体浓度低于10mg/L时是CAD的危险因素,与CAD之间存在显著的线性剂量效应关系。因此,在调整了贝叶斯模型平均法筛选的重要混杂变量之后,多因素分析结果均提示,机体铁超负荷增加了CAD发病的危险性。2.血清铁和总铁结合力两个铁指标的组合与CAD关系最为密切,是反映CAD患者机体铁负荷的相对敏感指标。3.血清铁和总铁结合力两个反映机体铁负荷相对敏感指标的组合与体质指数、高密度脂蛋白在CAD发病中存在高阶交互作用。4.本项1meta分析结果显示,血清铁蛋白、血清铁及总铁结合力与CAD的相关性具有统计学意义,而可溶性转铁蛋白受体由于受到文献数量限制,尚无法得到明确结论。意义与创新1.本研究利用限制性立方样条函数拟合了各个机体铁负荷指标与CAD的剂量效应关系,特别是估计了血清铁蛋白、血清铁、总铁结合力与CAD的非线性剂量效应关系,提高了模型的预测能力,目前罕见报道。2.因目前国内外未见关于机体铁负荷敏感指标的筛选研究。本课题筛选出血清铁与总铁结合力两个指标的组合,是反映机体铁负荷对CAD发病影响的相对敏感指标,为今后有关机体铁负荷与疾病关系研究中铁负荷指标的选择提供了依据。3.运用广义多因子降维法,探讨了反映机体铁负荷相对敏感指标的组合与体质指数、高密度脂蛋白对CAD的高阶交互作用,该方面研究目前亦未见报道。更多还原

【Abstract】 BackgroundCoronary artery disease (CAD) is the leading cause of death and disease burden in China and other most countries. With the developing of economy, the change of life style and the population aging in our country, the morbidity and mortality of CAD has been increasing significantly. Therefore, identifying the risk factors and population at high risk early is very important for preventing and controlling CAD. Although many risk factors for CAD such as hypertension, hyperlipemia and smoking have been identified, however, data showed that these traditional risk factors only account for50%of the CAD etiology, which indicates that other factors also play an important role in the incidence of CAD.Iron is an essential microelement for life and health which plays an important role in the transportation of oxygen, respiration of cell, metabolism of energy, expression of gene, and repairation and replication of DNA in the cellular process. Approximately66%is found in hemoglobin in the ferrous (Fe2+) form. An additional27%of the body’s iron is incorporated as tissue ferritin inthe ferric (Fe3+) state. When the intracellular storage capacity is exceeded, additional intracellular iron is incorporated into hemosiderin and iron may besequestered in the reticuloendothelial system. Free Fe2+can produce hydroxyl radical (·OH-) through Fenton and Haber-Weiss reaction that promotes lipid peroxidation and causes severe damage to membranes, proteins, and DNA, and induces oxidative stress. Besides, iron may act as a nitric oxide (NO) scavenger and thus induce endothelial cell dysfunction.As we all know, the incidence of CAD in premenopausal women is significantly lower than men, and the CAD risk in postmenopausal women increased gradually to be similar with men. Based on the above-mentioned phenomenon and the relevant researches, Sullivan proposed the "iron hypothesis" in1981suggesting that body iron stores were positively related to CAD risk. The theory was that the low level of iron by the menstruation periodic iron loss could be helpfulfor the prevention of CAD in women, while progressive accumulation of the body’s iron in men lead to increased risk of CAD. Since the hypothesis was first proposed, extensive debates have been generated from epidemiological studies and clinical trials. The possible main reasons for the conflicting results are as follows:①Different biomarkers of body iron stores:So far, a good deal of serum iron indicators including serum ferritin (SF), serum iron (SI), total iron-binding capacity (TIBC), and serum transferrin receptor (sTfR) were used in the present publications to assess the relation of body iron stores. Among them, almost no one could independently and exactly reflect the body iron levelsfor various weakness. For example, SF, an acute-phase protein, was prone to be increased by infection and inflammation. Based on the above-mentioned reasons, Ramakrishna suggested that a well designed research with complete measurements of SF, SI, TIBC and sTfR should conducted to properly reflect the status of iron overload.②Inconsistent outcomes:Myocardial infarction, coronary artery disease and total death of cardiovascular diseases are all used in studies about association of body iron and cardiovascular disease as outcomes. Although atherosclerosis is the pathological mechanism of the above diseases, the etiology of them might be not the same as they are all multiple factor diseases.③Partial adjustment of covariates:As CAD is multiple factor and complex disease, smoking, hypertension and hyperlipemia are all important risk factors, thus, it is hard to indicate the effects of iron overload on CAD risk without complete adjustments.Above all, we conducted a hospital-based case-control study included newly diagnosed CAD patients by percutaneous coronary angiography and healthy controls and simultaneously measured SF, SI, TIBC and sTfR to evaluate the association of body iron and CAD and explore the ideal iron indicator or combination of iron indicators that has best effect on CAD risk. Besides, CAD is a chronic disease of multifactorial origin that develops from the interplay of lifestyle, this study explored the high-order interaction between iron overload indicators and enviroment factors on CAD risk. A meta-analysis was also performed to confirm the association of body iron overload and CAD risk. This study was sponsored by a grant from National Natural Science Foundation of China (81072357).Objectives1.To assess the association between body iron indicators and CAD risk;2. To explore the ideal iron indicator that has best effect on CAD risk;3.To study the interaction effects between iron indicators and enviroment factors on CAD risk.Materials and Methods1. A hospital-based case-control study was conducted with newly-diagnosed CAD patients. Healthy controls were recruited from the healthy persons who came to the Physical Examination Center for a medical checkup.The concentration of SF、SI、TIBC and sTfR were measured by enzyme-linked immunosorbent assay (ELISA).2. Bayesian model averaging (BMA) was applied to select the significant covariates that influence on CAD risk. Restricted cubic spline was performed to assess the concentration-risk association between each serum iron parameter and CAD risk.3. The areas under each Receiver Operating Characteristic (ROC) curve (AUC) were compared with each other toindicate the one showing strongest association with CAD risk.4. Generalized Multifactor Dimensionality Reduction (GMDR) was used to explore the high-order interaction between iron overload indicators and enviroment factors on CAD risk.5. A meta-analysis was performed with comprehensive search for relevant articles. Fixed or random effect pooled measure was selected on the basis of the results of homogeneity test, I2was used to evaluate the heterogeneity among studies. Meta regression and subgroup analysis were used to explore potential sources of between-study heterogeneity. An analysis of influence was carried out, which describes how robust the pooled estimator is to removal of individual studies. Publication bias was estimated using funnel plot and Egger’s test.Results1. Study characteristic:A hospital-based case-control study was conducted with258newly-diagnosed CAD patients and282healthy controls. The traditional CAD risk factors such as age, hypertension, BMI, prevalence of smoking, serum8-iso-prostaglandin F2a and diabetes in the CAD group are significantly higher than those in controls (P<0.05). Serum high density lipoprotein in the control group are significantly higher than that in CAD group (P<0.05).2. Univariate analysis:All iron parameters showed significant differences among the cases and controls except for SF(P=0.06). SI in the CAD group is significantly higher than that in controls (P<0.01), TIBC and sTfR in the CAD group are significantly lower than those in controls (P<0.01). 3. Covariates selection:There were25models in Occam’s window, and these were used to calculate the BMA estimates of the regression coefficients. The posterior probabilities and BIC value of the best model are0.15and-2950.08, respectively. Nine variables were selected by BMA on the selected models and they were age, history of diabetes, total cholesterol, low density lipoprotein, hypertension, alcohol, sex, serum8-iso-prostaglandin F2a and serum high density lipoprotein.4. Multivariate analysis:(1) SF:After adjusted for age, history of diabetes, total cholesterol, low density lipoprotein, hypertension, alcohol, sex, serum8-iso-prostaglandin F2a and serum high density lipoprotein, the overall (P<0.01) and non-linear (P<0.01) associations between SF and CAD were both signifcant in the multivariable analysis. The reference value for SF was200ug/L in the concentration-risk analysis, and the OR(95%CI) were1.69(1.18-2.41),1.15(1.00-1.31),1.07(0.93-1.21),1.40(0.99-1.97) and1.83(1.09-3.07) for100ug/L,150ug/L,250ug/L,300ug/L and350ug/L, respectively.(2) SI:After adjusted for age, history of diabetes, total cholesterol, low density lipoprotein, hypertension, alcohol, sex, serum8-iso-prostaglandin F2a and serum high density lipoprotein, the overall (P<0.01) and non-linear (P<0.01) associations between SI and CAD were both signifcant in the multivariable analysis. The reference value for SI was19dμmol/L in the concentration-risk analysis, and the OR(95%CI) were0.36(0.21-0.60),0.66(0.54-0.81),1.41(1.20-1.66),2.00(1.45-2.75) and2.52(1.67-3.82) for12μmol/L,16μmol/L,22μmol/L,26μmol/Land30μmol/L, respectively.(3) TIBC:After adjusted for age, history of diabetes, total cholesterol, low density lipoprotein, hypertension, alcohol, sex, serum8-iso-prostaglandin F2a and serum high density lipoprotein, the overall (P<0.01) and non-linear (P<0.01) associations between TIBC and CAD were both signifcant in the multivariable analysis. The reference value for TIBC was90μmol/L in the concentration-risk analysis, and the OR(95%CI) were4.58(2.61-8.03),2.03(1.58-2.62),0.59(0.49-0.68),0.41(0.30-0.53) and0.33(0.22-0.49) for50μmol/L,70μmol/L,110μmol/L,130μmol/L and150μmol/L, respectively.(4) sTfR:After adjusted for age, history of diabetes, total cholesterol, low density lipoprotein, hypertension, alcohol, sex, serum8-iso-prostaglandin F2a and serum high density lipoprotein, the overall (P<0.01) and linear (P>0.05) associations between sTfR and CAD were both signifcant in the multivariable analysis. The reference value for sTfR was10mg/L in the concentration-risk analysis, and the OR(95%CI) were:1.33(0.97-1.84),1.16(1.00-1.35),0.87(0.79-0.96),0.67(0.53-0.86) and0.53(0.35-0.79) for6mg/L,8mg/L,12mg/L,16mg/L and20mg/L, respectively.5. ROC analysis:The AUC(95%CI) were0.73(0.69-0.77),0.74(0.69-0.78),0.53(0.48-0.58), and0.61(0.56-0.66) for SI, TIBC, SF, and sTfR, respectively. After comparing the AUC with each other, the combination of SI and TIBC (AUC (95%CI):0.86(0.83-0.90)) was superior to other examined iron parameters or the combination of iron indicators (P<0.05).6. High-order interaction:The best prediction model identified in our analysis included SI-TIBC, BMI and HDL that scored10for cross-validation consistency,10for permutation test (P<0.01) and0.85for testing accuracy, suggesting that these three factors together significantly contributed to CAD risk. The accuracy of the best model with covariates (age and sex) adjustment was better than that without adjustment.7. Meta-analysis:A total of26studies on SF, TIBC, SI and sTfR were included in the meta-analysis.(1) For the pooled SMD analysis, a total of4,410cases and7,357controls were eligible for the meta-analysis performed on SMD. The SMD(95%CI) for SF, TIBC, SI and sTfR were0.74(0.46,1.01),-0.06(-0.28,0.16),-0.24(-0.59,0.10) and0.13(0.02,0.24), respectively. Substantial between-study heterogeneity were found in SF, TIBC and SI, and after excluding the studies exerting substantial impact on between-study heterogeneity, the SMD(95%CI) for SF, TIBC and SI were0.12(0.07,0.17),-0.22(-0.39,-0.07) and-0.17(-0.31,-0.04), respectively. No significant influence and publication bias were observed before and after sensitivity analysis.(2) For the pooled OR analysis, a total of5,738cases and98,362controls were included for the meta-analysis performed on OR. The OR(95%CI) for SF, TIBC, SI and sTfR were1.13(0.98,1.30),1.21(0.95,1.55),0.97(0.94,0.99) and1.54(0.85,2.77), respectively. Substantial between-study heterogeneity was found in SI, and after excluding the studies exerting substantial impact on between-study heterogeneity, the pooled OR(95%CI) was1.19(1.08,1.45) for SI. No significant influence and publication bias were observed before and after sensitivity analysis. Conclusions1. In this study, compared with SF concentrations less than200ug/L, SF was found significantly positively correlated with CAD risk in nonlinear in the concentration-risk analysis. Compared with SI concentrations less than19μmol/L, SI was found significantly positively correlated with CAD risk in nonlinear. Compared with TIBC concentrations more than90μmol/L, TIBC was found significantly negatively correlated with CAD risk in non-linear. Compared with sTfR concentrations more than10mg/L, sTfR was found significantly negatively correlated with CAD risk in linear. Thus, after adjusted for the important covariates screened by BMA, iron overload were positively correlated with CAD.2. The combination of SI and TIBCwere found to be the ideal iron indicator that have best effect on CAD risk.3. Our results showed that combination of SI-TIBC, BMI and HDL conferred a significant three factors interaction on CAD risk.4. In this meta-analysis, SF, SI and TIBC were found significantly associated with CAD risk. Meta analysis of sTfR can not be fully conducted because of insufficient studies.Innovations1. This study assessed the concentration-risk association between body iron parameters and CAD risk with restricted cubic spline, and indicated the non-linear association of SF, SI and TIBC and CAD risk which improved the predictive power of the model.2. Information is presently absent about better iron status biomarkers on CAD disease risk, this study explored the ideal iron indicator(SI-TIBC) that has best effect on CAD risk with ROC analysis and provided evidence for future studies related to iron overload metabolism research worldwide.3. This research explored the statistically significant high-order interaction among SI-TIBC, BMI and HDL associated with CAD risk, further studies with big sample are needed to confirm these novel results.更多还原