基于孕中期代谢相关因素的巨大儿风险评估模型的建立

A prediction model for macrosomia based on metabolic markers in the second trimester of pregnancy

目的在不同孕前体重指数(BMI)分层的妊娠妇女中,通过孕中期的糖脂代谢指标建立巨大儿的风险评估模型,并探讨该模型的价值。方法为单中心、大样本的回顾性研究。收集2016年9月至2018年12月在北京协和医院产科门诊孕24~28周进行常规口服75 g葡萄糖耐量试验(OGTT)的孕妇共1781例,排除孕前糖尿病合并妊娠、高血压、甲状腺功能异常、肝肾功能疾病及多胎妊娠后,最终入组1114例,按照孕前BMI分为体重正常组(940例,BMI<24.0 kg/m^(2))和肥胖或超重组(174例,BMI≥24.0 kg/m^(2))。记录不同体重组患者的一般资料、孕期糖代谢状态资料,记录妊娠结局。采用单因素logistic回归分析寻找巨大儿发生的独立危险因素,采用二元logistic逐步回归建立巨大儿预测模型,采用受试者工作特性(ROC)曲线评估预测模型价值并寻找相应预测切点。用PROCESS 3.4进行中介分析。结果与孕前体重正常组相比,孕前超重或肥胖组发生巨大儿风险更高(OR值为2.72,95%CI 1.75~4.23)。调整混杂因素(包括妊娠年龄、糖尿病家族史、不良孕产史)后,孕中期的空腹血糖(FPG)、餐后1 h血糖(1hPBG)、妊娠期糖尿病(GDM)、甘油三酯(TG)、载脂蛋白B(ApoB)、孕重增加(GWG)均是巨大儿发生的危险因素,高密度脂蛋白胆固醇(HDL-C)是保护因素(均P<0.05)。二元logistic逐步回归法建立巨大儿预测模型,孕前体重正常组共940例,最终纳入预测指标为孕中期的GWG、ApoB及TG/HDL-C;孕前超重和肥胖组174例,最终纳入预测指标为ApoB和TG/HDL-C。孕前体重正常组预测巨大儿ROC曲线下面积为0.841,灵敏度为82.3%,特异度为75.2%;超重或肥胖组ROC曲线下面积0.727,灵敏度仅为50.0%,特异度为97.3%。在2组中,TG/HDL-C作为中介变量,在血糖和巨大儿的关系里中介效应显著。结论孕中期ApoB、TG/HDL-C联合GWG建立的巨大儿预测模型有较好的区分度和准确性,尤其在孕前体重正常的妊娠妇女中,提示血脂代谢紊乱对巨大儿的发生可能起着决定性作用。

Objective The purpose of this study is to establish a predictive model for macrosomia in different pre-pregnancy body mass index(BMI)groups.It also explores the values of this prediction model based on the metabolic markers in the second trimester of pregnancy.Methods The current study is a single-center,large-sample retrospective study.One thousand seven hundred and eighty-one pregnant women in Peking Union Medical College Hospital(PUMCH)during the 24th to 28th week of pregnancy were screened from September 2016 to December 2018.Subjects with pre-gestational diabetes,hypertension,thyroid dysfunction,renal or liver disease and twin pregnancies were excluded.Eventually 1114 subjects were divided into two groups according to pre-pregnancy BMI:normal weight group(pre-pregnancy BMI<24.0 kg/m^(2),n=940)and overweight/obese group(pre-pregnancy BMI≥24.0 kg/m^(2),n=174).Clinical data were collected,including general information,glucose-lipid metabolic parameters,and pregnancy outcomes.Univariate logistic regression analysis was conducted to reveal the potential prenatal risk factors for macrosomia.Stepwise logistic regression was conducted to establish prediction models of macrosomia.The receiver operating characteristic(ROC)curve was used to calculate cut-off points and assess the predictive ability of the models.Mediation analysis was conducted using PROCESS Macro version 3.4.Results Compared with the normal weight group,overweight/obese group has increased the incidence of macrosomia(OR:2.72,95%CI:1.75-4.23).Confounding variables include age,family history of diabetes and pregnancy loss history.After adjusting confounders,the fasting glucose(FPG),1 h post-prandial glucose levels(1hPBG),gestational diabetes mellitus(GDM),triglycerides(TG),apolipoprotein B(ApoB),and gestational weigh gain(GWG)were associated with increased risk for macrosomia.In contrast,high-density lipoprotein cholesterol(HDL-C)was associated with decreased risk for macrosomia(all P<0.05).Stepwise logistic regression was conducted to establish the prediction models of macrosomia.Serum ApoB level,TG/HDL-C ratio and GWG were independent predictors of macrosomia in the normal weight group(n=940).Only the combination of ApoB and TG/HDL-C has predictive value in the overweight/obese group.In the normal weight group,the area under the ROC curve(AUC)reached 0.841 with a sensitivity of 82.3%and specificity of 75.2%.In the overweight/obese group,the AUC reached 0.727 with a sensitivity of 50.0%and specificity of 97.3%.The TG/HDL-C ratio plays a mediating role in the relationship between glucose metabolism and the occurrence of macrosomia.Conclusions The prediction model combined with ApoB,TG/HDL-C ratio and GWG is a useful tool with good discrimination for macrosomia risk in the second trimester of pregnancy.Maternal lipid profile rather than glucose is an independent predictor of macrosomia,especially in pre-pregnancy normal weight women.