Predictive Value of Second Trimester HbA1c for Adverse Perinatal Outcomes: A Comparative Study in Diabetic and...

Second-trimester glycated hemoglobin (HbA1c) has been proposed as a marker of adverse pregnancy outcomes, yet its predictive value across gestational diabetes mellitus (GDM), pregestational diabetes mellitus (PGDM), and non-diabetic pregnancies remains unclear. This study assessed whether mid-pregnancy HbA1c predicts adverse perinatal outcomes and identified a clinically meaningful threshold.

This retrospective cohort included 460 singleton pregnancies delivered between January 2021 and January 2025 at Bursa City Hospital. Participants were categorized as GDM (n = 203), PGDM (n = 58), or healthy controls (n = 199). HbA1c was measured at 24-28 gestational weeks. An ROC-derived cut-off of 5.3% for predicting macrosomia (>4000 g) was used for subgroup analyses. Perinatal outcomes included gestational age at delivery, delivery mode, birth weight, Apgar scores, and neonatal intensive care unit (NICU) admission. Correlation, ROC, and multivariate logistic regression analyses were performed.

Median HbA1c levels were highest in the PGDM group (6.50%), followed by GDM (5.36%) and controls (4.93%) (p = 0.038). Women with HbA1c >5.3% had higher rates of preterm birth (p = 0.005), cesarean delivery (p = 0.012), and macrosomia (p = 0.004). ROC analysis showed moderate predictive ability for macrosomia (AUC = 0.642), preterm birth (AUC = 0.602), and cesarean delivery (AUC = 0.562). In multivariate analysis, maternal BMI independently predicted macrosomia (OR = 1.07; 95% CI: 1.01-1.14; p = 0.02), whereas HbA1c showed a positive but nonsignificant association (OR = 1.35; p = 0.13).

Second-trimester HbA1c values above 5.3% were associated with increased risks of preterm birth, cesarean delivery, and macrosomia. Although HbA1c did not independently predict adverse outcomes after adjustment, it may serve as a useful adjunct biomarker for perinatal risk stratification and could contribute to multivariable prediction models.