667: Population vs. customized growth curves and composite neonatal morbidity: a secondary analysis of 10 MFMU trials

Abstract

Publications differ on the improved accuracy of identifying adverse neonatal outcomes based on population nomograms (birthweight [BW] and gestational age [GA]) vs. customized (for maternal height, weight, ethnicity, parity, GA and BW) growth curve. Our objective was to assess the diagnostic accuracy of population vs. customized growth curves (PopGC vs. CusGC) in identifying newborns with composite neonatal morbidity (CNM), including stillbirth and neonatal mortality. Non-anomalous singletons delivered at ≥ 24 wks. in 10 MFMU trials were analyzed. PopGC we based on Alexander’s nomogram while CusGC utilized software from http://www.gestation.net/index.htm, to categorize newborns as SGA (BW < 10% for GA), AGA (BW at 10-89%) or LGA (BW ≥ 90%). Random-effect logistic regression was used to estimate adjusted odds ratio (aOR) with 95% confidence intervals (CI). Receiver operator characteristic (ROC) curves (identifying CNM) with area under the curves (AUC) were constructed. Of 86,809 singleton pregnancies enrolled in 10 trials, 90% (N=77,987) met inclusion criteria. Using the PopGC, 12% were SGA and 10% LGA; with CusGC, rate of SGA and LGA was 15% and 16%, respectively. Compared to PopGC, aOR for SGA based on CusGC was higher for CNM, stillbirth and neonatal death; the same pattern was seen when evaluating aOR for CNM and stillbirth among LGA. Sensitivity for identifying stillbirth among SGA was higher with CusGC than PopGC (54%; 95% CI 47% - 60%) vs. 36%; 30% - 43%), as was for neonatal mortality (48%; 95% CI 44% - 53% vs. 27%; 95% CI 24% - 32%). Comparison AUC for SGA to identify CNM, stillbirth and neonatal mortality is significantly better with customized than population nomogram (P < 0.01; Table). For LGA, customized nomogram was better at identifying cases of stillbirth than population nomogram (P=0.03) but not for CNM (0.64; 95% CI, 0.63 - 0.64) vs. 0.64; 95% CI, 0.63 - 0.65) or neonatal death (p=0.06). Though CusGC was better than PopGC at identifying deliveries with CNM, stillbirth and neonatal mortality, it is not useful to obstetricians for BW is unknown during pregnancy. Thus, a randomized trial comparing customized fetal growth curve with population curve is warranted to determine which method mitigates morbidity and mortality with aberrant fetal growth.