Longitudinal change in cerebroplacental ratio in small-for-gestational-age fetuses and risk of stillbirth

Objective To investigate whether assessment of longitudinal change in Doppler variables in small-for-gestational-age (SGA) fetuses improves the prediction of those at risk of stillbirth. Methods This was a longitudinal study of two cohorts of singleton pregnancies, which included SGA and appropriate-for-gestational-age (AGA) fetuses, respectively. The inclusion criteria for the SGA cohort were singleton pregnancy at >= 20 weeks' gestation, classified as SGA (estimated fetal weight < 10th centile). The AGA cohort consisted of singleton pregnancies deemed at high risk of being SGA, which were followed up longitudinally but remained AGA. Fetal middle cerebral artery (MCA) pulsatility index (PI) and umbilical artery (UA)-PI were measured longitudinally and cerebroplacental ratio (CPR) was calculated, and values were converted to multiples of the median. The last two measurements prior to delivery were included in the analysis. Longitudinal models for Doppler variables were developed using linear-mixed models and their accuracy in the prediction of stillbirth was tested using generalized linear models. A Bayesian framework was employed to compare the accuracy of longitudinal and standard (last-scan measurement) models. Results In total, 1549 AGA and 941 SGA pregnancies were included in the analysis. There were 30 (3.2%) and no stillbirth cases in the SGA and AGA groups, respectively. Change in MCA-PI, UA-PI and CPR with advancing gestation was significantly different between liveborn AGA and SGA fetuses, with a less pronounced difference with advancing gestation. Using the last measurement, the best models for the prediction of stillbirth in SGA pregnancies were those based on CPR (accuracy, 75.0%; 95% CI, 72.6-77.2%) and UA-PI (accuracy, 71.0%; 95% CI, 68.6-73.4%). The posterior probability of the standard CPR model having a higher accuracy compared with the UA-PI model was 97.2% (magnitude of change (MC), 3.9%; 95% credible interval (CrI), 0.5-7.3%). The accuracies of the standard, compared with the longitudinal, models for UA-PI (71.0% vs 72.8%), MCA-PI (64.6% vs 63.8%) and CPR (75.0% vs 74.9%) in the prediction of stillbirth were not significantly different. The posterior probabilities for improvement in accuracy using longitudinal, compared with standard, assessment were 50.1% (MC, < 0.1%; 95% CrI,-3.3 to 3.3%), 35.2% (MC,-0.1%; 95% CrI, -4.5 to 2.8%) and 82.2% (MC, 1.9%; 95% CrI,-1.5 to 5.3%) for CPR, MCA-PI and UA-PI models, respectively. Therefore, change in Doppler parameters did not improve the accuracy of the prediction of stillbirth, compared with that of the last-scan measurement. Conclusion Longitudinal assessment of Doppler parameters was not useful in improving the detection of stillbirth in SGA pregnancies, as compared with a single-point assessment. Copyright (C) 2018 ISUOG. Published by John Wiley & Sons Ltd.