Cross-validation and predictive value of near-infrared spectroscopy algorithms for day-5 blastocyst transfer

Abstract

Near-infrared (NIR) spectroscopic metabolomic profiling of spent embryo-culture media has been used to calculate a viability score for individual embryos. These scores have been found to correlate to the reproductive potential of cleavage-stage embryos. In this study, 137 spent blastocyst media samples were collected after single-embryo transfer and analysed by NIR spectroscopy to generate an algorithm and calculate viability scores. To blindly validate the algorithm development process, another algorithm was trained on 47 preselected samples from clinic 1 and then used to predict the outcome of 42 samples from clinic 2. The overall pregnancy rate from the two clinical sites was 50.4%. A positive correlation ( R 2 = 0.82, P = 0.03) was observed with the increasing viability score quintiles and their associated implantation rates. Cross-validation of an algorithm generated from NIR analysis of media samples at one clinical setting blindly was shown to predict implantation potential of blastocysts cultured at another clinic in a different culture media and culture volume. This study demonstrates that metabolomic profiling by NIR spectroscopic analysis of day-5 spent embryo-culture media can predict the implantation potential of blastocysts. Furthermore, this method may not be restricted to a specific set of culturing conditions. The successes of IVF treatment cycles are in part limited by the ability to select the best single embryo from a cohort of patient embryos for transfer back to the woman. Routine procedures of embryo selection are based on morphology, including cell number and size, and the timing of cell division. These methods are favoured because they are quick and easy to assess. Human embryos are grown in culture solutions, which are specific for their stage of development. Recent studies analysing the culture solution in which the embryo are grown, by near infrared (NIR) spectroscopic analysis, have been able to predict if an embryo will implant or not. As culture conditions often vary between IVF laboratories the questions remained if the NIR technique could be used to independently predict the implantation potential of an embryo cultured at one laboratory using an algorithm trained on embryos at a second clinic, a so-called cross-validation. The results of this study show that NIR spectroscopy can predict the ability of embryos to implant even when grown in different IVF laboratories and in two different culture solutions. This information supports the idea that NIR spectroscopy can be used globally not relying on specific culture conditions or media.