Background: To evaluate cervical length changes as predictors of preterm delivery. Methods: Two hundred and fiftyseven pregnant women underwent transvaginal ultrasound examination at 16 and 24 weeks of gestation. Cervical length was measured and the difference between the 2 measurements was calculated. The sensitivity, specificity, positive predictive value and negative predictive value of cervical length and cervical length changes were calculated and these methods were compared by receiver-operating characteristic ( ROC) curve analysis. Results: Preterm delivery ( before 37 weeks of gestation) occurred in 19 patients ( 7.4%). The mean cervical length was shorter in the preterm group, the area under the ROC curve for prediction of preterm delivery was 0.914, ultrasound had a sensitivity of 84.2% to predict preterm delivery with a false-positive rate of 18.5%, and the relative risk was 4.56 at the 34.3-mm cutoff value at 24 weeks of gestation. In contrast, a cervical length change on transvaginal ultrasound had a sensitivity of 73.3% to predict preterm delivery with a false-positive rate of 18.1%, and the relative risk was 4.08 at the 6.6- mm cutoff value. Conclusion: A single cervical length measurement obtained at 24 weeks of gestation can predict preterm delivery as accurately as cervical length change. Copyright (c) 2007 S. Karger AG, Basel.