Postoperative pregnancy outcome prediction model based on 3-dimensional
ultrasound for intrauterine adhesion
Objective: The aims of this study were to establish a mathematical model
to predict intrauterine adhesion (IUA) prognosis based on 3-dimensional
(3D) ultrasound. Design：Derivate retrospectively and validate
prospectively. Setting: University affiliated hospitals Population: 668
IUA in the derivation cohort and 426 IUA in the validation cohort.
Methods: Logistic regression analysis was used to develop prediction
models to forecast the live birth rate following HA according to
postoperative 3D-TVUS characteristics and hysteroscopy characteristics.
The performance of each of the prediction models was compared by
calculating the area under the curve (AUC). A nomogram was then
constructed for the excellent model. Main Outcome Measures: General
clinical information, 3D-TVUS characteristics and hysteroscopy
characteristics. Results: Live birth was achieved by 242 IUA patients in
the derivation cohort (n = 668) and by 150 in the validation cohort
(n = 426). Logistic regression analysis revealed that the AUC for the
models based on 3D-TVUS characteristics in the derivation and validation
cohorts were 0.8506 and 0.8963, respectively, which was better than the
models based on hysteroscopy characteristics. The calibration curve of
the nomogram based on 3D-TVUS characteristics and a non-significant
Hosmer–Lemeshow test statistic showed good calibration (P = 0.3064).
Conclusion: Our findings indicate that postoperative 3D ultrasound
characteristics could be predictive factors for live births among IUA
patients and provide valuable reproductive guidance for women after HA.