The role of fetal biometry in accurate prediction of gestational age: Is combination of parameters better?

Abstract

IS COMBINATION OF PARAMETERS BETTER? MARTIN CHAVEZ, CANDE ANANTH, JOHN SMULIAN, LAMI YEO, ANTHONY VINTZILEOS, UMDNJ-Robert Wood Johnson Medical School/Robert Wood Johnson University Hospital, Obstetrics, Gynecology and Reproductive Sciences, New Brunswick, New Jersey OBJECTIVE: To construct robust statistical models based on commonly utilized fetal biometric parameters for accurate prediction of gestational age (GA). STUDY DESIGN: Nomograms of gestational age based on fetal biometric parameters were constructed utilizing cross-sectional data from our antenatal testing unit (August 1993 to June 2003). Nomograms for predicting GA were constructed based on transcerebellar diameter (TCD), head circumference, biparietal diameter, abdominal circumference, and femur length measurements. Depending on what parameter was measured, there were between 24,000 to 50,000 well-dated singleton fetuses. These nomograms were developed using regression techniques by modeling the mean and variance separately. The models comprised of single biometric parameters as well as combination of parameters. The adequateness of these models was assessed using change in the proportion of variance explained (R). The performance of each nomogram in predicting GA was assessed in a separate validation data set (August 2003 to July 2005) comprised of approximately 10,000 ultrasound scans of well-dated singleton fetuses. RESULTS: We constructed several regression models using robust statistical models for predicting GA. The proportion of variance (R) in GA based on models with single parameter ranged between 97.1% based on the head circumference to 98% based on TCD. When multiple parameters were included in the model, the gain in R was both non-significant (PO0.05), and did not improve prediction. Based on models with a single parameter, prediction of GA within G3 days was feasible in 90-94% of pregnancies. The model based on TCD predicted GA within G3 days in 97% of pregnancies. CONCLUSION: Our sophisticated, innovative, and robust statistical models based on a single fetal biometric parameter provides accurate prediction of gestational age. More complex models do not appear to yield improved prediction in gestational age, suggesting that simple, parsimonious models using TCD only will suffice in accurate pregnancy dating.