Scholarlink

Objective <p>To develop an accurate prediction system for gestational diabetes mellitus (GDM) in women adhering to Institute of Medicine (IOM) weight gain criteria by analyzing early-pregnancy metabolic kinetics and identifying independent risk factors.</p> Methods <p>A prospective two-center cohort study enrolled 1,031 pregnant women meeting IOM guidelines. Clinical, anthropometric, and metabolic parameters (pre-pregnancy BMI, lipid profiles, inflammatory markers) were collected at 6–12 weeks of gestation. Machine learning models were trained on seven core variables identified via logistic regression, with performance evaluated by AUC, sensitivity, and specificity.</p> Results <p>The GDM group (<i>n</i> = 279) exhibited significantly higher pre-pregnancy weight, BMI, triglycerides (1.21 vs. 0.96 mmol/L, <i>p</i> &lt; 0.001), and inflammatory markers. Multivariate analysis identified parity ≥ 2 (OR = 4.37), family diabetes history (OR = 1.64), third-trimester weight (OR = 1.13), and triglycerides (OR = 1.49) as independent predictors. A neural network model achieved the highest AUC (0.732) with 65.1% sensitivity and 68.9% specificity, outperforming logistic regression and other algorithms. Dynamic weight trajectories and lipid dysregulation were critical risk drivers, even in IOM-compliant women.</p> Conclusion <p>In pregnant women adhering to IOM weight gain standards, ​early-pregnancy metabolic dysfunction​and specific weight trajectories ​persist as pivotal GDM predictors. Our neural network model integrating seven core variables—including triglyceride levels, dynamic weight patterns (6–10 weeks and 24–28 weeks), and clinical risk factors—achieved superior prediction accuracy compared to conventional weight-centric approaches. This system enables early risk stratification and personalized interventions during the first trimester, addressing a critical gap in prenatal care for guideline-compliant populations at residual GDM risk.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Establishment of an accurate prediction system for gestational diabetes mellitus based on the characteristics of metabolic kinetics in early pregnancy: a prospective two-center cohort study in population according to IOM criteria

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月6日

查看全文
  • Junxiang Gao,
  • Shuoning Song,
  • Yanbei Duo,
  • Xiaolin Qiao,
  • Yuemei Zhang,
  • Jiyu Xu,
  • Jing Zhang,
  • Xiaorui Nie,
  • Qiujin Sun,
  • Xianchun Yang,
  • Ailing Wang,
  • Wei Sun,
  • Yong Fu,
  • Mengmeng Zhang,
  • Yingyue Dong,
  • Zechun Lu,
  • Tao Yuan,
  • Weigang Zhao

摘要

Objective

To develop an accurate prediction system for gestational diabetes mellitus (GDM) in women adhering to Institute of Medicine (IOM) weight gain criteria by analyzing early-pregnancy metabolic kinetics and identifying independent risk factors.

Methods

A prospective two-center cohort study enrolled 1,031 pregnant women meeting IOM guidelines. Clinical, anthropometric, and metabolic parameters (pre-pregnancy BMI, lipid profiles, inflammatory markers) were collected at 6–12 weeks of gestation. Machine learning models were trained on seven core variables identified via logistic regression, with performance evaluated by AUC, sensitivity, and specificity.

Results

The GDM group (n = 279) exhibited significantly higher pre-pregnancy weight, BMI, triglycerides (1.21 vs. 0.96 mmol/L, p < 0.001), and inflammatory markers. Multivariate analysis identified parity ≥ 2 (OR = 4.37), family diabetes history (OR = 1.64), third-trimester weight (OR = 1.13), and triglycerides (OR = 1.49) as independent predictors. A neural network model achieved the highest AUC (0.732) with 65.1% sensitivity and 68.9% specificity, outperforming logistic regression and other algorithms. Dynamic weight trajectories and lipid dysregulation were critical risk drivers, even in IOM-compliant women.

Conclusion

In pregnant women adhering to IOM weight gain standards, ​early-pregnancy metabolic dysfunction​and specific weight trajectories ​persist as pivotal GDM predictors. Our neural network model integrating seven core variables—including triglyceride levels, dynamic weight patterns (6–10 weeks and 24–28 weeks), and clinical risk factors—achieved superior prediction accuracy compared to conventional weight-centric approaches. This system enables early risk stratification and personalized interventions during the first trimester, addressing a critical gap in prenatal care for guideline-compliant populations at residual GDM risk.