Intraobserver and interobserver reproducibility of three-dimensional gray-scale and power Doppler ultrasound examinations of the cervix in pregnant women

Objectives To determine intraobserver and interobserver reproducibility of three-dimensional (3D) gray-scale and power Doppler ultrasound examinations of the cervix in pregnant women. Methods Thirty-two pregnant women underwent transvaginal 3D gray-scale and power Doppler ultrasound examination of the cervix by two examiners. Each observer acquired two volumes, and they each analyzed their volumes twice using the commercially available software Virtual Organ Computer-aided AnaLysis (VOCAL (TM)). The variables analyzed were cervical volume (cm 3), vascularization index (VI), flow index (FI) and vascularization flow index (VFI). Intraobserver repeatability was expressed as the difference between two measurement results (mean difference +/- 2 SD, i.e. limits of agreement) and as intraclass correlation coefficient (intra-CC). Interobserver agreement was expressed as the difference between the results of the two observers (limits of agreement) and as interclass correlation coefficient (inter-CC). The contribution of various factors (examiner, acquisition, analysis of acquired volume) to intrasubject variance was estimated using different analysis of variance models. All statistical analyses were performed using log-transformed data. The results presented are those obtained after antilogarithmic transformation, i.e. the results are presented as ratios between two results of the same observer, or as ratios between the results of Observer 1 and Observer 2. Results All intraobserver and interobserver log-transformed differences were normally distributed. There was no systematic bias between the two observers. Both intra- and inter-CC values were high (0.93-0.98) for all variables except FI (0.63-0.88), despite the limits of agreement being wide, especially for VI (widest range 0.4-2.4) and VFI (widest range 0.3-2.6). Acquisition explained most of the intrasubject variance of the flow indices, the contribution of examiner and analysis being unimportant. Conclusions Given the wide range between the lower and upper limits of agreement, it would probably not be possible to detect anything but large differences or changes in cervical volume or cervical flow indices using current 3D ultrasound techniques. Because acquisition explained most of the intrasubject variance, the average of several repeated acquisitions should be used to enhance reproducibility. However, it is not worth doing more than one analysis of an acquired volume, because the effect of analysis on measurement results is small. Copyright (c) 2005 ISUOG. Published by John Wiley & Sons, Ltd.