基于LASSO回归构建非妊娠期糖尿病孕妇分娩大于胎龄儿的列线图预测模型

Based on LASSO regression, a nomogram prediction model for large for gestational age infants delivered by pregnant women with non-gestational diabetes mellitus was constructed

摘要

目的 探讨非妊娠期糖尿病孕妇分娩大于胎龄儿(LGA)的危险因素,建立预测模型,为个体化孕期科学管理提供参考依据.方法 选择2021年1月至2023年12月于上海市第六人民医院临港院区在孕早期建档并正规产检的非妊娠期糖尿病孕妇1046例作为研究对象进行回顾性研究,收集孕妇的一般资料、孕期血糖血脂结果.将研究对象按7∶3比例随机分为训练集(n=732例)和内部验证集(n=314例).通过LASSO回归筛选非妊娠期糖尿病孕妇分娩LGA的危险因素,构建列线图预测模型.采用受试者工作特征(ROC)曲线、Hosmer-Lemeshow拟合优度检验以及Calibration校准曲线来评价列线图模型的预测效果,采用临床决策曲线预测模型的临床净获益.结果 共收集到非妊娠期糖尿病孕妇1046例,其中LGA 169例,非LGA 877例.LGA预测模型共纳入4个影响因素,分别为孕晚期体重(OR=1.063,95%CI:1.034~1.093,P<0.001)、孕早期甘油三酯(OR=3.360,95%CI:1.985~5.688,P<0.001)、孕晚期甘油三酯(OR=1.706,95%CI:1.285~2.263,P<0.001)和孕晚期高密度脂蛋白胆固醇(OR=0.227,95%CI:0.090~0.568,P=0.002).训练集和验证集的ROC曲线下面积分别为0.815(95%CI:0.766~0.864)和0.852(95%CI:0.779~0.925).Hosmer-Lemeshow检验结果显示模型的风险预测值与实际观测值之间差异无统计学意义(训练集P=0.423;验证集P=0.727),校准曲线和理想曲线几乎重合.临床决策曲线表明模型临床效用较高.结论 构建的非妊娠期糖尿病孕妇分娩LGA的预测模型具有良好的预测效果和临床应用价值.

Abstract

Objective To explore the risk factors of large for gestational age (LGA) babies in pregnant women with non-gestational diabetes mellitus, and to establish a prediction model to provide a reference for individualized scientific management of pregnancy. Methods A retrospective study was conducted on 1046 pregnant women with non-gestational diabetes who were registered in the first trimester of pregnancy and had regular prenatal examination at the Lingang Campus of Shanghai Sixth People's Hospital from January 2021 to December 2023, and the general data of pregnant women, blood glucose and lipid results during pregnancy were collected. The study subjects were randomly divided into the training set (n=732 cases) and the internal validation set (n=314 cases) in a 7∶3 ratio. LASSO regression was used to screen the risk factors of LGA in non-gestational diabetes mellitus pregnant women, and a nomogram prediction model was constructed. The receiver operating characteristic (ROC) curve, Hosmer-Lemeshow goodness-of-fit test and Calibration calibration curve were used to evaluate the prediction effect of the nomogram model, and the clinical decision curve was used to predict the clinical net benefit of the model. Results A total of 1046 pregnant women were collected, including 169 LGA infants and 877 non-LGA infants. The predictive model for LGA included four affecting factors: weight in the third trimester (OR=1.063, 95%CI: 1.034-1.093, P<0.001), triglycerides in the first trimester (OR=3.360, 95%CI: 1.985-5.688, P<0.001), triglycerides in the third trimester (OR=1.706, 95%CI: 1.285-2.263, P<0.001), and high density lipoprotein-cholesterol in the third trimester (OR=0.227, 95%CI: 0.090-0.568, P=0.002). The area under the receiver operating characteristic curve in the training set and internal validation set were 0.815 (95%CI: 0.766-0.864) and 0.852 (95%CI: 0.779-0.925), respectively. The Hosmer-Lemeshow test showed that there was no significant difference between the predicted risk value of the model and the actual observed values (P=0.423 in the training set; P=0.727 in the validation set). The calibration curve and the ideal curve almost coincided. The clinical decision curve of the nomogram model indicated a high clinical utility. Conclusion The proposed prediction model for LGA in non-gestational diabetes mellitus has good prediction effect and clinical application value.

关键词

非妊娠期糖尿病 / 孕妇 / 大于胎龄儿 / 危险因素 / 预测模型 / LASSO回归 / 列线图

Key words

non-gestational diabetes mellitus / pregnant woman / large for gestational age infants / risk factors / predictive models / LASSO regression / nomogram

参考文献

[1] 首都儿科研究所, 九市儿童体格发育调查协作组. 中国不同出生胎龄新生儿出生体重、身长和头围的生长参照标准及曲线[J]. 中华儿科杂志, 2020, 58(9): 738-746.

[2] Lei X, Zhao D, Huang L, et al. Childhood health outcomes in term, large-for-gestational-age babies with different postnatal growth patterns[J]. Am J Epidemiol. 2018, 187(3): 507-514.

[3] Suárez-Idueta L, Ohuma E O, Chang C J, et al. Neonatal mortality risk of large-for-gestational-age and macrosomic live births in 15 countries, including 115.6 million nationwide linked records, 2000-2020[J]. BJOG. 2023, 00(1): 1-12.

[4] Ewington L, Black N, Leeson C, et al. Multivariable prediction models for fetal macrosomia and large for gestational age: a systematic review[J]. BJOG. 2024, 131(12): 1591-1602.

[5] 郑薇, 张莉, 田志红, 等. 大于胎龄儿发生风险的人群归因危险度分析[J]. 中华妇产科杂志, 2019, 54(12): 833-839.

[6] Lin L, Lin J, Yan J. Association of maternal body mass index change with risk of large for gestational age among pregnant women with and without gestational diabetes mellitus: a retrospective cohort study[J]. J Matern Fetal Neonatal Med. 2024, 37(1): 2316732.

[7] Tinius R A, Blankenship M M, Furgal K E, et al. Metabolic flexibility is impaired in women who are pregnant and overweight/obese and related to insulin resistance and inflammation[J]. Metabolism. 2020, 104: 154142.

[8] Ikenoue S, Waffarn F, Sumiyoshi K, et al. Maternal insulin resistance in pregnancy is associated with fetal fat deposition: findings from a longitudinal study[J]. Am J Obstet Gynecol. 2023, 228(4): 455.e1-455.e8.

[9] Kelly A C, Powell T L, Jansson T. Placental function in maternal obesity[J]. Clin Sci (Lond). 2020, 134(8): 961-984.

[10] Ryckman K K, Spracklen C N, Smith C J, et al. Maternal lipid levels during pregnancy and gestational diabetes: a systematic review and meta-analysis[J]. BJOG. 2015, 122(5): 643-651.

[11] Pereira A P, Montero M F, Souza F D, et al. High-density lipoproteins-cholesterol (HDL-C) in women with gestational diabetes (GDM): a predictor for large gestational age (LGA) babies[J]. Cureus. 2024, 16(7): e65546.

[12] Huang D, Zhu H, Zhu Y, et al. Association of maternal HDL2-c concentration in the first trimester and the risk of large for gestational age birth[J]. Lipids Health Dis. 2022, 21(1): 71.

基金资助

上海市第六人民医院院级管理科研基金(lygl202212)

“市六G临港”紧密型健康联合体临床类科研项目(ynlglht202404)