Ultrasonographic measurement of the dividing membrane in twin pregnancy during the second and third trimesters: a reproducibility study.

Abstract

Our purpose was to examine the reproducibility of intertwin membrane thickness measurements used to predict chorionicity in twin pregnancies. Twenty-seven twin pregnancies were scanned with high-resolution ultrasonography on 52 occasions during the second and third trimesters. Two observers, blind to other criteria of chorionicity, measured the dividing membrane twice in five different sites (total measurements 1040). The data were log-transformed and the coefficient of repeatability calculated as a measure of intraobserver variability. Interobserver variability was estimated by the Bland and Altman 95% limits of agreement. Random variation was assessed with the restricted maximum likelihood procedure in Genstat. The overall estimate of the coefficient of repeatability was 2.14, indicating that 95% of repeated measurements would be expected to be within 114% of each other. Measurements taken close to the placenta (up to 3 cm) were the most repeatable and displayed no bias when repeated. Coefficients of repeatability at this site ranged from 1.42 to 1.91, with no evidence of consistent differences between monochorionic and dichorionic twins. Gestational age was not significantly associated with membrane thickness for any of the models. The pregnancy type x subject x observer and pregnancy type x subject x site interactions were statistically significant (p < 0.001 and p < 0.005, respectively), implying that interobserver variability depends on the subject being measured, the site of sampling, and chorionicity. Ultrasonographic measurement of membrane thickness has high intraobserver and interobserver variability in the second and third trimesters. Our findings provide an explanation for the suboptimal accuracy reported with this method in determining chorionicity in the second and third trimesters.