This study sought to determine whether 3-D transvaginal ultrasound (3D-TVS), compared with the 2-D transvaginal approach (2D-TVS), offers a better identification of some specific features of the uterine cervix that could be useful in the prediction of spontaneous preterm delivery (PTD). A total of 103 women with singleton pregnancies were prospectively evaluated with 2D-TVS and 3D-TVS in the second or third trimester of pregnancy. Statistical analysis was performed with Wilcoxon matched-pairs test, chi2 and Pearson test (p<0.05 was considered significant) and receiver operating characteristic (ROC) curve analysis. Significant differences between these approaches were found for cervical length (p<0.001). A significant correlation was identified between 3D-TVS cervical length and the interval between ultrasound examination and delivery as well as the gestational age at delivery (both p<0.001). ROC curve identified a threshold for 3D-TVS cervical length of 35 mm as an optimal predictor of PTD in the second trimester (sensitivity 100%, specificity 88%). 3D-TVS evaluation of the cervix in pregnancy seems to be an effective, noninvasive method for predicting PTD risk.