Objective: To assess the intraobserver repeatability and interobserver reproducibility of transabdominal Doppler ultrasound measurements of ductus venosus blood flow in fetuses between 10 and 14 weeks of gestation.
Design: A prospective study with the following end-points: coefficient of variation, intraclass correlation coefficients within and between observers, repeatability coefficient and Cohen's kappa coefficient.
Subjects and methods: Doppler velocimetry indices (maximum peak systolic velocity during ventricular systole, end-diastolic velocity during atrial contraction, peak systolic velocity/end-diastolic velocity ratio and pulsatility index for veins) were successfully measured three times by the same trained observer in 67 fetuses. Of these, in 24 fetuses, the measurements were then repeated by a second trained observer. In addition, both observers classified qualitatively the blood flow as being normal or abnormal with regard to the presence or absence/reversal of flow during atrial contraction.
Results: The intraobserver repeatability of the pulsatility index for veins measurements was considered good with a coefficient of variation of 8.9% and an intraclass correlation coefficient value of 0.62. However, inconsistencies occurred in maximum peak systolic velocity, end-diastolic velocity and systolic/diastolic ratio measurements, which had coefficients of variation of 19%, 28.5%, and 25.4%, respectively. The corresponding intraclass correlation coefficient values were 0.70, 0.69 and 0.60. Interobserver reproducibility of these indices was unsatisfactory, the coefficient of variation varying between 11.5% and 47.2% and the intraclass correlation coefficient between 0.18 and 0.44. Agreement between the repeated observations with regard to interpretation of normal or abnormal blood flow varied from moderate (interobserver) to very good (intraobserver).
Conclusions: There was considerable variability in all Doppler indices except for the intraobserver repeatability of the pulsatility index for veins measurement. This makes the implementation of the semiquantitative Doppler indices in a screening setting difficult. On the contrary, qualitative classification of the flow velocity waveforms seemed to be a reproducible method.