In Ultrasound in obstetrics & gynecology : the official journal of the International Society of Ultrasound in Obstetrics and Gynecology
OBJECTIVE : To develop a model based on factors available at diagnosis of twin-twin transfusion syndrome (TTTS) that predicts chance of dual twin survival following fetoscopic laser photocoagulation (FLPC) using machine learning algorithm.
METHODS : A retrospective analysis of data collected from two university-affiliated tertiary fetal centers between 2012 and 2021. The cohort included monochorionic twin pregnancies complicated by TTTS who underwent FLPC. Data were stratified based on survival rates 30 days after delivery, dual survival cases were compared to the rest. Following random forest, as an ensemble machine-learning algorithm, relative importance value was calculated for each parameter that presented statistically significant difference. Holdout method was applied to check overfitting of the random forest algorithm. A prediction model for having dual twin survival 30 days after delivery was constructed to a graphic nomogram based on the testing set.
RESULTS : The study included 537 women, of them 346 (64.4%) had dual twin survival at 30 days and were compared to 191 (35.6%) that had one or no survivors. Univariate analysis demonstrated no differences in demographic parameters between the groups. At time of diagnosis, the dual survival groups presented lower rates of donor's estimated fetal weight below 10th centile for gestational age (56.4% vs. 69.4% p=0.004), intertwin growth discordance above 25% (40.8% vs. 56.5%, p=0.001), and anterior placenta (40.5% vs. 50%, p=0.034). Doppler differences between the groups demonstrated lower rates of elevated pulsatility index (PI) above 95th centile, measured in the donor's umbilical artery and ductus venosus, as well as lower rates of decreased PI below 5th centile, measured in the donor's middle cerebral artery. Importance value for each of these 6 parameters was calculated allowing the construction of a prediction model with area under ROC curve (AUC=0.916, 95% CI= 0.887-0.946).
CONCLUSIONS : Incorporating six variables: donor's estimated fetal weight below 10th centile, intertwin growth discordance above 25%, anterior placenta, pulsatility index in the umbilical artery, ductus venosus and middle cerebral artery, obtained at time of diagnosis of TTTS into a predictive model for dual twin survival following FLPC has been developed. This clinically applicable tool may allow improved treatment plans and patient counseling. This article is protected by copyright. All rights reserved.
Krispin E, Mustafa H J, Espinoza J, Nassr A A, Cortes M Sanz, Donepudi R, Harman C, Mostafaei S, Turan O, Belfort M A, Shamshirsaz A A
Dopplers, fetoscopic laser photocoagulation, machine learning, prediction model, survivors, twin pregnancy, twin-twin transfusion syndrome