Cervical assessment at 22 and 27 weeks for the prediction of spontaneous birth before 34 weeks in twin pregnancies: is transvaginal sonography more accurate than digital examination?

Abstract

This study compared the accuracy of ultrasound cervical assessment (cervical length and cervical index) and digital examination (Bishop score and cervical score) in the prediction of spontaneous birth before 34 weeks in twin pregnancies. In a prospective multicenter study, digital examination and transvaginal sonography were performed consecutively in twin pregnancies attending for routine sonography at either 22 weeks (175 women) or 27 weeks (153 women). The digital examination took place first, and the Bishop score and cervical score (cervical length minus cervical dilatation) were calculated. Ultrasound measurements were then made of cervical length and funnel length to yield the cervical index (1 + funnel length/cervical length). The association between each variable and delivery before 34 weeks was tested by the Mann-Whitney U-test. The receiver-operating characteristics (ROC) curves of the ultrasound and digital indicators were determined for both gestational age periods, and the areas under the ROC curves compared. The best cut-off values for each indicator were used to determine predictive values for delivery before 34 weeks. The median gestational age at delivery among the women included in the 22-week examination period was 36.0 (range, 21-40) weeks; 10.9% (19) gave birth spontaneously before 34 weeks. The median cervical length was 40 (range, 6-65) mm. All four parameters were predictors of delivery before 34 weeks. The areas under the ROC curves for cervical index, cervical length, Bishop score and cervical score did not differ significantly. The median gestational age at delivery among the women in the 27-week examination period was 36.0 (range, 27-40) weeks; 9.2% (14) gave birth spontaneously before 34 weeks. The median cervical length was 35 (range, 1-57) mm. All parameters except the Bishop score were predictors of delivery before 34 weeks. The likelihood ratio of the positive and negative tests for cervical length < or = 25 mm was 5.4 (range, 3.2-9.0) and 0.3 (range, 0.1-0.7), respectively, compared with 2.3 (range, 1.3-4.2) and 0.6 (range, 0.3-1.1), respectively, for cervical score < or = 1. The area under the curve for the cervical index was significantly larger than that for the Bishop score (P = 0.008) or cervical score (P = 0.02). Transvaginal sonography predicted spontaneous delivery before 34 weeks better than digital examination at the 27-week but not the 22-week examination.