Early pregnancy biomarkers for gestational diabetes mellitus prediction: a systematic review and meta-analysis of routine laboratory, metabolic, and inflammatory markers

Objective To evaluate the diagnostic accuracy of early pregnancy biomarkers for predicting gestational diabetes mellitus (GDM). Methods We searched PubMed, Web of Science, Embase, and Cochrane Library from inception to December 2024 (specific dates: PubMed January 1980-December 31, 2024; Web of Science January 1980-December 31, 2024; Embase January 1980-December 31, 2024; Cochrane Library 1996-December 31, 2024) for studies evaluating first or early second trimester biomarkers for GDM prediction. ClinicalTrials.gov was also searched for ongoing studies. Eligible studies included prospective cohort, case-control, or cross-sectional designs reporting diagnostic accuracy metrics. Quality assessment was performed using QUADAS-2. Meta-analyses were conducted for biomarkers evaluated in ≥3 studies using random-effects models. Sample size adequacy was assessed based on the number of GDM cases, with a minimum of 100 events considered necessary for reliable pooled estimates. Diagnostic accuracy was assessed through pooled sensitivity, specificity, diagnostic odds ratio (DOR), and area under the receiver operating characteristic curve (AUROC). Results Twelve studies with sample sizes ranging from 75 to 41, 786 participants were included. Only one study comprehensively evaluated inflammatory markers, reporting a first-trimester model combining TNF-α with clinical parameters (maternal age, BMI, family history) and biochemical markers (fasting glucose, lipid profile). Pregnancy-associated plasma protein-A (PAPP-A) levels showed a trend toward lower values in women who developed GDM (SMD -0.558, 95% CI: -1.155 to 0.040). Metabolomic markers, particularly acylcarnitines C5 and C5:1, demonstrated excellent discrimination (AUROC 0.934, 95% CI: 0.873-0.995). The overall pooled sensitivity was 81.6% (95% CI: 68.7%-90.0%), specificity 93.4% (95% CI: 81.4%-97.9%), DOR 73.84 (95% CI: 5.65-964.94), and AUROC 0.881 (95% CI: 0.795-0.967). Substantial heterogeneity was observed across studies (I²=94-97%). Conclusions Early pregnancy biomarkers show moderate to high diagnostic accuracy for GDM prediction. However, high heterogeneity and limited data on integrated multi-marker panels constrain clinical application. Standardization of protocols and validation in diverse populations are needed.