OBJECTIVE: Our purpose was to compare the accuracy of ultrasonographic and manual cervical examinations for the prediction of preterm delivery.

STUDY DESIGN: One hundred two singleton pregnancies at high risk for preterm delivery were followed up prospectively from 14 to 30 weeks with both serial cervical ultrasonography measurements and manual examinations of the length of the cervix. The primary outcome studied was preterm (<35 weeks) delivery.

RESULTS: Excluding six induced preterm deliveries, 96 pregnancies were analyzed. The mean cervical length measured by ultrasonography was 20.6 mm in pregnancies delivered preterm (n = 17) and 31.3 mm in pregnancies delivered at term (n = 79) (p = 0.003); the mean cervical lengths measured by manual examination were 16.1 mm and 18.6 mm in the same preterm and term pregnancies, respectively (not significant). The sixteenth- and twentieth-week ultrasonographic cervical lengths predicted preterm delivery most accurately (p < 0.0005). The 25th percentiles of ultrasonographic (25 mm) and manual (16 mm) cervical lengths showed relative risks for preterm delivery of 4.8 (95% confidence interval 2.1 to 11.1, p = 0.0004) and 2.0 (95% confidence interval 0.5 to 4.7, p = 0.1), respectively; sensitivity, specificity, and positive and negative predictive values were 59%, 85%, 45%, 91%, and 41%, 77%, 28%, and 86%, respectively.

CONCLUSION: Cervical length measured by ultrasonography is a better predictor of preterm delivery than is cervical length measured by manual examination. Cervical ultrasonography in patients at high risk for preterm birth seems to be most predictive of preterm delivery when it is performed between 14 and 22 weeks' gestation. (Am J Obstet Gynecol 1997;177:723-30.)