Yapay Zeka Hızlı Bakış & Çıkarımlar
AI InsightThese findings highlight the clinical value of integrating biologically meaningful fetal and maternal parameters to refine weight estimation and support more informed obstetric decision-making, while underscoring that future multicenter external validation studies are strictly required before broad clinical implementation can be recommended.
📌 3 Maddede Bu Makale
Abstract
Introduction: To compare the accuracy of the five Hadlock formulas (Hadlock I–V) for estimating fetal birth weight at term. In addition, to evaluate multiple fetal and maternal parameters and identify whether those with significant associations — including fetal anterior abdominal subcutaneous tissue thickness (FASTT), maternal gestational weight gain (GWG), and maternal body mass index (BMI) — improve predictive performance when used to calibrate the Hadlock IV model. Methods: This prospective cohort included 1,074 singleton pregnancies between 38–42 gestational weeks. Standard biometry — biparietal diameter (BPD), head circumference (HC), abdominal circumference (AC), and femur length (FL) — was obtained on the day of admission for delivery. Estimated fetal weight (EFW) was calculated using the five Hadlock formulas. Birth weight was measured within 1 hour postpartum. Calibration of Hadlock IV was performed using multivariable linear regression based on FASTT, GWG, and BMI. Model performance was assessed using mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), the proportion within ±10% of actual birth weight, and systematic error (bias). To safeguard against overfitting, a 100-fold random 70/30 split cross-validation was employed. Results: Among the five formulas, Hadlock IV demonstrated the highest baseline accuracy, with an MAE of 442.3 g, RMSE of 549.9 g, MAPE of 13.4%, and 44.3% of estimates within ±10% of actual birth weight; baseline bias was −47.3 g, indicating systematic underestimation, while Hadlock V demonstrated the lowest baseline performance. FASTT, GWG, and BMI were significantly associated with Hadlock IV residuals and were retained in the multivariable calibration model. Averaged out-of-sample internal validation metrics across the 30% test cohorts demonstrated that calibration markedly improved predictive performance, reducing MAE to 192.4 g and RMSE to 242.1 g, lowering MAPE to 5.9%, increasing the ±10% accuracy to 83.8%, and substantially minimizing baseline bias to a mean cross-validated validation bias of 0.1 g. Additional parameters — including prenasal thickness (PNT), placental thickness (PT), amniotic fluid index (AFI), umbilical vein diameter (UVD), and fetal presentation — did not provide significant independent contributions. Conclusion: Hadlock IV remains the most accurate baseline formula for term fetal weight estimation; however, its performance is modest and exhibits population-specific negative bias. Incorporating FASTT, GWG, and BMI into a calibrated model more than doubled accuracy, substantially minimized systematic error to a near-zero level, and substantially improved precision across the birth-weight spectrum for immediate pre-delivery assessment at term. These findings highlight the clinical value of integrating biologically meaningful fetal and maternal parameters to refine weight estimation and support more informed obstetric decision-making, while underscoring that future multicenter external validation studies are strictly required before broad clinical implementation can be recommended.