Cervical-length as a screening for spontaneous preterm birth in uncomplicated twins: one vs. serial measurements

While cervical length (CL) provides an effective screening tool for spontaneous preterm birth in singletons, the performance in twins is still controversial. Our aim was twofold. First, to compare the performance of a single CL measurement at mid-gestation (∼20 weeks) versus serial measurements as a predictor of spontaneous preterm birth < 34 weeks in uncomplicated twin pregnancies. Second, to describe the performance of a single CL at ∼24, ∼28 and ∼32 weeks.

cohort study of twins followed at Hospital Italiano de Buenos Aires from 2013 to 2017. Inclusion criteria were dichorionic or monochorionic diamniotic twins with CL measurement between 18 and 33 + 6 weeks with available data of the delivery. Exclusion criteria included any of the following complications: iatrogenic preterm delivery <34 weeks, cerclage, fetal growth restriction, fetal death, structural anomalies, polyhydramnios, twin-twin transfusion syndrome, selective fetal growth restriction, twin anemia-polycythemia sequence, and twin reversed arterial perfusion sequence. Spontaneous preterm birth was defined as spontaneous delivery < 34 weeks. Two different cutoffs were used to classify CL as short (positive screening) or normal (negative screening): (a) a fixed cutoff of 25 mm at any gestational age (GA). The screening was considered positive if any CL measurement was <25 mm; and (b) a GA adjusted cutoff to a 10% false positive rate (FPR). The 10% FPR for each GA was calculated and the screening was considered positive if any of the CL measurements were below this 10% FPR cutoff. We report sensitivity, specificity, positive and negative predictive value, positive and negative likelihood ratio and area under the ROC curve.

Among 777 twins followed in our Unit, 381 met exclusion criteria and 18 were excluded due to incomplete follow-up. We included 378 patients, 284 (75%) dichorionic and 94 (25%) monochorionic. The performance of one CL at 20 weeks showed a sensitivity ≤ 20% with an area under the ROC curve of 0.58 (95% CI, 0.45-0.70), while the performance of serial measurements showed a sensitivity of 58.8% (95% CI, 40.7-75.4) with an area under the ROC curve of 0.70 (95% CI, 0.61-0.79) (p < .001). The analyses of the performance of a single CL at ∼24, ∼28 and ∼32 weeks showed similar AUC than the serial measurements and, for a FPR = 10%, the performance of one measurement at 24 and 32 weeks showed a sensitivity of 30% (95% CI, 14.7-49.4) and 31.6% (95% CI, (12.6-56.6), while the measurement at 28 weeks showed a sensitivity of 48.3% (95% CI, (29.4-67.5).

Serial measurements showed a better performance than a single one in mid-gestation. Moreover, among single measurements the CL in mid-gestation showed the poorest performance, while the 28 weeks assessment detected half of the preterm deliveries. However, all the strategies showed modest performances.