Intra- and interobserver reproducibility of ultrasound measurements of cervical length and width in the second and third trimesters of pregnancy.

Objectives: To determine intra- and interobserver reproducibility of ultrasound measurements of cervical length and width in the second and third trimesters of pregnancy.

Design: Twenty healthy women in the second or third trimester of pregnancy underwent transvaginal ultrasound examination of the cervix by two examiners. Three replicate meaurements of cervical length and width were taken by each observer for each woman. Intraobserver repeatability was expressed as the difference between the highest and lowest measurement value obtained by one observer, and as the repeatability coefficient and intraclass correlation coefficient. Interobserver reproducibility was expressed as the difference between the mean of the three measurements of each observer, limits of agreement, and interclass correlation coefficient. The repeatability coefficient/limits of agreement define the range within which 95% of the differences between two measurements by the same observer/two observers are likely to fall.

Results: Intra- and interobserver differences did not vary in any systematic way over the range of values measured. For cervical length measurements the repeatability coefficient was +/- 5.4 mm for Observer 1 and +/- 5.9 mm for Observer 2. The intraclass correlation coefficient for cervical length measurements was 0.93 for both observers. The corresponding coefficients for cervical width measurements were +/- 3.9 mm and +/- 7.3 mm, and 0.97 and 0.91. The interclass correlation coefficient for cervical length measurements was 0.76, the mean interobserver difference was 0.4 mm, and the limits of agreement were -10.0 mm to 10.8 mm. For cervical width measurements the interclass correlation coefficient was 0.74, the mean interobserver difference was -0.02 mm, and the limits of agreement were -12.4 mm to 12.4 mm. There was no systematic difference between the first, second and third cervical length measurements and no systematic bias between the two observers, and the results were not affected by which of the two examiners started the examination (three-way analysis of variance). The same was true of cervical width measurements.

Conclusions: Our results do not support the contention that the first measurement of cervical length is the longest. There is substantial intra- and interobserver variability in the results of measurements of cervical length and width, even when experienced observers perform the measurements under standardized conditions. It is important to consider the possibility of 'measurement error' when ultrasound measurements of cervical length and width are used in clinical practice to monitor women at high risk of delivering preterm or to screen for preterm birth.