OP05.02: Can transvaginal ultrasound at 7–9 weeks reliably diagnose chorionicity and amnionicity in twin pregnancies?

Abstract

Of 1000 women attending an early pregnancy unit over 3 months who had at least two ultrasound examinations confirming singleton viability, two classes of pregnancies are considered: ongoing pregnancies and early pregnancy losses. Our objective was to discriminate between these two classes to predict early pregnancy loss. We considered the rate of growth of the crown-rump length (CRL), the mean sac diameter (MSD) and the difference between MSD and CRL as a function of the gestational age (GA). Methods: Classical linear discriminant analysis (LDA) classifies data by maximizing the ratio of the between-class variation to the within-class variation. FLDA is an extension where the predictor variables are curves (e.g. a variable measured multiple times during pregnancy). Serial observations from each individual are modeled with a spline function (a curved line formed by two or more vertices), parameterized with a basis function multiplied by a 5-dimensional coefficient vector. A training set was used to estimate the mean coefficient vector for each class. New patients can then be classified by determining the class with the closest mean coefficient vector. Results: 270 patients had at least two measurements for CRL and GA. Nine of these subsequently resulted in early pregnancy loss. It was not possible to discriminate between classes with FLDA using CRL alone. Using MSD we could include 99 patients and mean growth curves of MSD as function of GA for ongoing pregnancies and pregnancy losses did not overlap (P < 0.0001). The MSD for early pregnancy loss was smaller at all gestational ages. Moreover, the difference between MSD and CRL could be used to discriminate between ongoing pregnancies and pregnancies destined to fail (P < 0.0001). Conclusions: Using FLDA it is possible to predict early pregnancy loss on the basis of MSD or difference between MSD and CRL for patients in whom MSD and CRL have been measured on at least two occasions.