Examining the efficacy of six published time-lapse imaging embryo selection algorithms to predict implantation to demonstrate the need for the development of specific, in-house morphokinetic selection algorithms
A selection of currently available, published embryoselection algorithms cannot be applied externally, thus encouraging the development of in-house, specific embryo-selection algorithms.
Volume 107, Issue 3, Pages 613–621
Amy Barrie, M.Sc., Roy Homburg, M.B.B.S., Garry McDowell, Ph.D., Jeremy Brown, Ph.D., Charles Kingsland, M.D., Stephen Troup, Ph.D.
To study the efficacy of six embryo-selection algorithms (ESAs) when applied to a large, exclusive set of known implantation embryos.
Retrospective, observational analysis.
Fertility treatment center.
Women undergoing a total of 884 in vitro fertilization (IVF) or intracytoplasmic sperm injection (ICSI) treatment cycles (977 embryos) between September 2014 and September 2015 with embryos cultured using G-TL (Vitrolife) at 5% O2, 89% N2, 6% CO2, at 37°C in EmbryoScope instruments.
Main Outcome Measure(s)
Efficacy of each ESA to predict implantation defined using specificity, sensitivity, positive predictive value (PPV), negative predictive value (NPV), area under the receiver operating characteristic curve (AUC), and likelihood ratio (LR), with differences in implantation rates (IR) in the categories outlined by each ESA statistically analyzed (Fisher's exact and Kruskal-Wallis tests).
When applied to an exclusive cohort of known implantation embryos, the PPVs of each ESA were 42.57%, 41.52%, 44.28%, 38.91%, 38.29%, and 40.45%. The NPVs were 62.12%, 68.26%, 71.35%, 76.19%, 61.10%, and 64.14%. The sensitivity was 16.70%, 75.33%, 72.94%, 98.67%, 51.19%, and 62.33% and the specificity was 85.83%, 33.33%, 42.33%, 2.67%, 48.17%, and 42.33%, The AUC were 0.584, 0.558, 0.573, 0.612, 0.543, and 0.629. Two of the ESAs resulted in statistically significant differences in the embryo classifications in terms of IR.
These results highlight the need for the development of in-house ESAs that are specific to the patient, treatment, and environment. These data suggest that currently available ESAs may not be clinically applicable and lose their diagnostic value when externally applied.