ABSTRACT
RESEARCH QUESTION: Does using an objective time-lapse imaging algorithm (TLIA) after IVF relate to conventional morphological assessment of human blastocysts as a prognosticator for live birth? DESIGN: Prospective use of a TLIA to select embryos in multicentre IVF clinics all using the same strictly controlled laboratory protocols. Each blastocyst was given a ranking from A to D, with the highest rank preferred for fresh transfer. This ranking was retrospectively compared with a given morphological score, which was blinded to the TLIA rank; all embryos were cultured under the same conditions. RESULTS: Using multiple variable logistic regression models, TLIA embryo rank enabled greater discrimination between cycles with and without live births than the conventional morphology grade, even when considered in isolation, and when adjusting for covariates related to treatment and patient criteria. Of the 1810 cycles of single blastocyst transfer, 894 (49.4%) resulted in a live birth. A Vuong non-nested test including covariates showed strong evidence of the superiority of the embryo rank model compared with the transfer grade model (P = 0.0008 [raw], P = 0.0003 [Akaike information criterion - corrected]). From the receiver operating characteristic (ROC) curves across all possible thresholds the TLIA rank showed better true positive and true negative rates and had a higher area under the curve [AUC] of 67.43% compared with 61.74% for the blastocyst morphology grade. The same analysis but excluding covariates demonstrated an AUC of 62.86% versus 54.02%, respectively. CONCLUSION: Objective TLIA is superior for selecting embryos for their propensity to generate a live birth over a conventional, subjective blastocyst morphology scoring system.
