Objective: Our purpose was to prospectively evaluate the interobserver reliability of digital and endovaginal ultrasonographic cervical length measurements.
Study design: Forty-three women were recruited from our antepartum clinic to participate in this study. Two independent and blinded digital cervical examinations were performed by the first author and a second examiner. Instructions were given to estimate the cervical length in millimeters. After micturition endovaginal ultrasonographic cervical length measurements were performed by two independent, blinded registered diagnostic medical sonographers. Cervical lengths were compared with the Student t test and Pearson's correlation coefficient. A kappa statistic was calculated for interobserver reliability at three levels of agreement +/- 1 mm, +/- 4 mm, and +/- 10 mm. Data are expressed as means +/- SD.
Results: Digital cervical lengths were not different between the two examiners (18.7 +/- 4.8 mm, 20.5 +/- 6.2 mm) nor between the two ultrasonographic measurements (38.6 +/- 6.1 mm, 39.2 +/- 5.4 mm). The digital cervical lengths agreed (+/- 1 mm) 35% of the time (R2 0.10, p = 0.02). The endovaginal ultrasonographic measurements agreed (+/- 1 mm) 74% of the time with a stronger correlation (R2 0.53, p = 0.0001). The kappa statistic for interobserver variability was marginal for both digital and endovaginal cervical length measurements when agreement was defined as +/- 1 mm. Endovaginal ultrasonography was significantly more reliable than digital examination when agreement between examiners was defined as either +/- 4 mm or +/- 10 mm.
Conclusion: Although both digital and endovaginal ultrasonographic cervical length measurements show correlation between examiners, endovaginal ultrasonography is significantly more reliable when agreement is defined as > or = +/- 4 mm. Serial cervical length measurements to predict preterm labor will be enhanced by the interobserver reliability of endovaginal ultrasonography.