External validation of a prediction model to select the best day-three embryo for transfer in in vitro fertilization or intracytoplasmatic sperm injection procedures

Objective To evaluate the multivariate embryo selection model by van Loendersloot et al. (2014) (VL) in a different geographical context. Design This is a retrospective external validation study of a 5-year cohort of women undergoing in vitro fertilization or intracytoplasmatic sperm injection. Setting Two outpatient fertility clinics. Patient(s) A total of 1,197 women who underwent 1,610 fresh in vitro fertilization or intracytoplasmatic sperm injection cycles with single embryo transfer were included. Intervention(s) None. Main Outcome Measure(s) The area under the receiver operating characteristics curve for diagnostic efficacy was used to assess the discriminative value of the model. Calibration for testing the validity of the VL model was performed using the Hosmer-Lemeshow goodness-of-fit test and a calibration plot. Result(s) Three hundred thirty-three patients (21%) achieved a viable pregnancy of at least 11 weeks. The area under the receiver operating characteristics curve using the VL model was 0.68. No significant difference between the predicted implantation rate and the observed implantation rates was showed using the Hosmer-Lemeshow (X 2 = 6.70). The calibration plot showed an intercept of the regression line of 0.34 and the estimated slope was 0.72. Conclusion The investigated VL model was able to distinguish between higher and lower implantation potential of embryos in our clinical setting.