350: Preterm birth prediction with transvaginal cervical length and uterocervical angle measurements

Abstract

To assess the performance of transvaginal cervical length (CL) and uterocervical angle (UCA) for predicting preterm birth (PTB) <37 weeks gestation. This is a retrospective cohort study of nulliparas with singleton gestations who underwent a routine transvaginal CL measurement between 18-23 6/7 weeks and delivered at a single tertiary center from July 2011-January 2014. For each woman with a short cervix (defined as a CL ≤25mm), 4 women without a short cervix were randomly selected from the remainder of the overall cohort, and together these women served as the study population. Sonographic images were reviewed and UCA was measured. Receiver operating characteristic (ROC) curves for the outcome of PTB <37 weeks were generated for CL and UCA as continuous measures, and the areas under the curves (AUCs) were compared. Based on cutoffs established in the existing literature, UCA was then dichotomized at 95 degrees. ROC curves for the outcome of PTB <37 weeks were generated for CL ≤25mm and UCA ≥95 degrees, and the AUCs were compared. Lastly, a ROC curve examining the prediction of PTB in women with either a CL ≤25mm or UCA ≥95 degrees was generated and the AUC was compared to each measure alone. A total of 405 women were included, of whom 81 (20%) had a short cervix, 204 (50.4%) had UCA ≥95 degrees, and 222 (54.8%) had either a short cervix or UCA ≥95 degrees. The AUCs for the ROC curves evaluating the predictive value of CL and UCA as continuous measures for PTB <37 weeks were not significantly different (0.72, 95% CI 0.63-0.80 vs. 0.66, 95% CI 0.57-0.75, p=0.27). Similarly, the AUCs for the ROC curves for CL ≤25mm and UCA ≥95 degrees were not significantly different (0.67, 95% CI 0.60-0.74 vs. 0.62, 95% CI 0.55-0.69, p=0.21). The AUC for women with either a CL ≤25mm or UCA ≥95 degrees was no better than that of a short cervix alone (AUC 0.64, 95% CI 0.58-0.70, p=0.32) (Figure). UCA has a similar predictive capability to CL for the prediction of PTB. Their combination does not significantly improve PTB prediction in nulliparas with singleton gestations.