Universal screening for preterm delivery by adding transvaginal ultrasound measurement of cervical length to routine second trimester ultrasound has been proposed. The aim is to estimate inter- and intra-observer agreement and reliability of second trimester transvaginal ultrasound measurements of cervical length performed by specially trained midwife sonographers.
This is a prospective reliability and agreement study performed in seven Swedish ultrasound centers. In total, 18 midwife sonographers specially trained to perform ultrasound measurements of cervical length and 286 women in the second trimester were included. In each center two midwife sonographers measured cervical length a few minutes apart in the same woman, the number of women examined per examiner pair varying between 24 and 30 (LIVE study). Sixteen midwife sonographers measured cervical length twice ≥2 months apart on 93 video-clips (CLIPS study). The main outcome measures were mean difference, limits of agreement, intra-class correlation coefficient, intra-individual standard deviation, repeatability, Cohen´s kappa, and Fleiss kappa.
The limits of agreement and intra-class correlation coefficient of the best examiner pair in the LIVE study were -4.06 to 4.72 mm and 0.91, those of the poorest were -11.11 to 11.39 mm and 0.31. In the CLIPS study, median (range) intra-individual standard deviation was 2.14 mm (1.40 to 3.46), repeatability 5.93 mm (3.88 to 9.58), intra-class correlation coefficient 0.84 (0.66 to 0.94). Median (range) inter-observer agreement for cervical length ≤25 mm in the CLIPS study was 94.6% (84.9% to 98.9%) and Cohen´s kappa 0.56 (0.12 to 0.92), median (range) intra-observer agreement was 95.2% (87.1% to 98.9%) and Cohen´s kappa 0.68 (0.27 to 0.93).
Agreement and reliability of cervical length measurements differed substantially between examiner pairs and examiners. If cervical length measurements are used to guide management there is potential for both over- and under-treatment. Uniform training and rigorous supervision and quality control are advised.
This article is protected by copyright. All rights reserved.