Cervical length in late second and third trimesters: a mixture model for predicting delivery

Abstract Objectives To examine the distribution of cervical length ( CL ) in the late second and third trimesters of pregnancy and construct survival models for spontaneous delivery. Methods This cross‐sectional study included 647 women with a singleton pregnancy who had routine CL measurements taken by transvaginal ultrasonography between 24 and 40 weeks' gestation. Only one measurement per patient was included in the analysis. Exploratory data analysis revealed that the distribution of CL measurements was a mixture of two Gaussian distributions, and subsequently a mixture model was applied to describe the distribution of CL . Changes in CL in relation to gestational age were examined by regression analysis and measurements were converted to Z ‐scores. Survival analysis was applied to the subgroups identified, to describe the probability of delivery throughout gestation. Results CL was best described by a mixture model of two subgroups with Gaussian distribution, one including women with a long cervix (73.85% of the study population, mean CL of 28.2 ± 4.45 mm) and the other including women with a short cervix (26.15% of the study population, mean CL of 12.3 ± 5.14 mm). CL was dependent on gestational age ( GA ), therefore conversion to Z ‐scores was employed in the analysis. Women with a short cervix had a higher probability for spontaneous delivery (hazard ratio ( HR ), 1.807; P < 0.001) than did women with a long cervix, after adjustment for GA at the time of measurement (GA HR , 1.115; P < 0.001). In both subgroups, the probability of spontaneous delivery was predicted by the GA and CL Z ‐scores. Survival models for the prediction of spontaneous delivery were constructed separately for the short‐cervix subgroup ( GA HR , 1.085; P < 0.001 and CL Z ‐score HR , 0.819; P = 0.003) and the long‐cervix subgroup ( GA HR , 1.130; P < 0.001 and CL Z‐score HR , 0.864; P = 0.005). Conclusions CL follows a mixture of two Gaussian distributions, one for a short and one for a long cervix. Late second‐trimester/third‐trimester CL measurement may be of value in estimating the probability for spontaneous delivery at any given gestational age. Copyright © 2014 ISUOG. Published by John Wiley & Sons Ltd.