Prediction of intrauterine fetal death in small-for-gestational-age fetuses: impact of including ultrasound biometry in customized models

Abstract

Objective: Customized growth charts derived from maternal demographic characteristics alone have been shown to improve the prediction of pregnancy complications compared to population growth curves. We sought to estimate the impact of adding ultrasound biometric parameters to the customized chart for the prediction of intrauterine fetal death (IUFD).

Methods: A retrospective cohort study was undertaken using an ultrasound database including singleton pregnancies followed between 16 and 20 weeks' gestation. After exclusion of preterm births, congenital anomalies, multifetal pregnancies and stillbirths (excluded only from derivation samples), we identified 59 016 births, divided into derivation (34 832) and validation (24 184) samples. Coefficients for significant physiological and pathological variables affecting fetal growth were derived using backward stepwise multiple regression (Cust-chart). The same process was repeated including second-trimester biometric parameters: biparietal diameter, head circumference, femur length and abdominal circumference in the regression models (Cust-plus-USS-chart). The association between small-for-gestational age < 10(th) centile (SGA) pregnancies, defined using the two customized charts or our population-based growth chart (Pop-chart) and IUFD, were compared. Statistical analyses including OR, 95% CI and screening accuracy using each chart were performed.

Results: The derived coefficients for fetal growth are comparable to those of previously published series. Of 24 184 pregnancies in the validation sample, IUFD was seen in 169 (0.7%). The pregnancies identified as SGA were: 2482 (10.26%), 2499 (10.33%) and 2634 (10.89%) using the Cust-chart, Cust-plus-USS-chart and Pop-chart, respectively. The OR (95% CI) for the association between SGA defined by the three charts and IUFD was: 7.0 (4.5-11), 6.5 (4.2-10.2) and 2.4 (1.6-3.6) according to the Cust-chart, Cust-plus-USS-chart and Pop-chart, respectively. Screening efficiency for IUFD using both customized charts was similar, with both demonstrating a higher sensitivity compared with the Pop-chart.

Conclusions: Customized charts are more efficient in identifying pregnancies at risk for IUFD compared with population-based charts. However, adding second-trimester ultrasound biometric parameters to the customized model does not improve the prediction of IUFD compared with using maternal characteristics only.