Early detection of gestational diabetes mellitus (GDM) often relies on the oral glucose tolerance test (OGTT), which can be time-consuming and resource-intensive. A study published in Diabetes, Metabolic Syndrome and Obesity evaluated whether fasting serum metabolite profiling could support screening and risk assessment for GDM.

The study used a retrospective discovery cohort and a prospective validation design involving pregnant women who completed a standard 75 g OGTT. A total of 1,053 participants were initially enrolled. The discovery cohort included 435 participants enrolled between April and May 2021, while the validation cohort included 473 participants recruited between November 2018 and May 2021. Fasting serum samples underwent targeted metabolomic profiling.

Machine learning analysis using a random forest approach identified metabolites that differed between women with and without GDM. Eight metabolites showed significant differential expression between groups (false discovery rate <0.05). Based on feature importance rankings, a multivariable logistic regression model was developed incorporating seven metabolites: 2-hydroxybutyric acid, 1,5-anhydroglucitol, glycine, 3-methyl-2-oxobutyric acid, 3-methyl-2-oxovaleric acid, tyrosine, and oleic acid.

The composite model integrating fasting glucose, clinical risk factors, and the metabolite panel demonstrated discrimination with an area under the receiver operating characteristic curve of 0.78 in the discovery cohort compared with 0.62 for fasting glucose alone. External validation showed an AUC of 0.71. These findings describe a fasting metabolite-based model for GDM risk assessment using metabolomic profiling combined with clinical variables.