The effect of day 5 blastocyst assessment timing on live birth prediction and development of a prediction algorithm
Reproductive BioMedicine Online
School of Nursing and Midwifery
Research question: Does variation in day 5 assessment timing confound live birth prediction using snapshot blastocyst morphology and is it possible to develop a numerical prediction algorithm? Design: Retrospective multicentre cohort study including 4851 autologous oocyte single day 5 fresh embryo transfers performed at 11 Monash IVF clinics between 2016 and 2020. Repeat cycles of the same patients were excluded to avoid clustering effects in regression analysis. Results: Hours post insemination (HPI) at day 5 assessment (115.9 ± 2.6 h) significantly correlated with blastocyst developmental stage (r = 0.118, P < 0.001). Independent association (expressed as adjusted odds ratio [aOR] and 95% confidence interval [CI]) was identified between live birth and HPI (aOR 0.950, 95% CI 0.925–0.976, P < 0.001) after accounting for blastocyst morphology and a range of patient/cycle characteristics. Algorithms were constructed using four significant live birth predictors: HPI at day 5 assessment, blastocyst developmental stage (aOR 1.347, 95% CI 1.217–1.491, P < 0.001), morphological grade (aOR 1.314, 95% 1.197–1.443, P < 0.001) and maternal age (aOR 0.922, 95% CI 0.907–0.936, P < 0.001). Receiver operating characteristic (ROC) analysis showed consistent predicting performance of algorithms via five-fold cross-validation, with similar area under the ROC curve (AUC 0.718, 0.715, 0.720, 0.712, 0.726, P < 0.001, respectively, in development subsets; and AUC 0.718, 0.731, 0.709, 0.741, 0.684, P < 0.001, respectively, in validation subsets). A score (ranging from 0.1 to 4.7) calculator based on the final algorithm was subsequently created. Conclusions: Day 5 assessment timing is a confounding factor for live birth prediction using snapshot blastocyst morphology. A numerical algorithm incorporating day 5 assessment HPI, blastocyst morphology and maternal age can be developed for live birth prediction.