Another Nail in the Dogma of Embryo Selection in IVF -beyond Basic Embryology
By Norbert Gleicher, MD, Medical Director and Chief Scientist, at The Center for Human Reproduction in New York City. He can be contacted though The Reproductive Times or directly at either ngleicher@thechr.com or ngleicher@rockefeller.edu.
Briefing: A recently published paper in Nature Medicine, accompanied by an equally good, signed Commentary put another nail into the sarcophagi of the dying dogma that embryo selection (ES) beyond embryo morphology will improve IVF cycle outcomes.
In comparing embryo selection (ES) in IVF (in vitro fertilization) between a deep learning artificial intelligence (A.I.) program and manual morphology, a recently published study in Nature Medicine found no significant difference (1). Because we have argued for years that ES beyond morphology is biologically illogical, this result was not a surprise to us. To many others in the field, however, it apparently was. If it hadn’t been, it likely wouldn’t have been accepted in such a prominent medical journal and given an accompanying commentary (2).
The concept of embryo selection is almost as old as in vitro fertilization (IVF) itself and must therefore be considered a founding dogma of IVF. It assumes that within a cohort of embryos produced in a single IVF cycle, there must be a “best”—or at least several “best”—embryos. While embryo cohorts do vary in quality (defined by pregnancy and live birth chances), no widely practiced ES methods in IVF have ever been shown to exceed the efficacy of manual morphology assessments.
A concept first explored when embryo morphology was recognized as a predictor of IVF cycle outcomes, embryo selection has since become the most widely investigated subject in IVF, with no other topic in fertility receiving more funding. Yet, hardly any other major investment has yielded such disappointing results. No wonder then that over more than 40 years of searching for better ES methods, the result was the introduction of unvalidated clinical IVF practices, such as routine embryo culture to blastocyst stage, time-lapse closed incubation systems, preimplantation genetic testing for aneuploidy (PGT-A), and others. All of these ES practices have been shown not to improve IVF outcomes and, in certain patient populations, to even worsen them (3,4).
With A.I. being ubiquitous (not just in medicine), it's no surprise that recent attempts at ES have involved deep learning systems. Start-ups from all around the world have reported claims of superiority over standard embryo morphology. Our review of these claims, however, left us skeptical, as none of them properly validated their statements. A recently published prospectively randomized multicenter study in Nature Medicine refuted this claim, demonstrating no significant outcome differences between A.I.-driven ES and standard manual embryology.
One can only hope that this recently published study (1) and its equally excellent accompanying commentary will finally convince the IVF field that the continuous pursuit of ES beyond basic embryology makes biologically little sense and has already wasted too much effort and investment. Resulting useless IVF practices have only served to increase the costs of an already expensive and, for many, unaffordable IVF process.
If a clinically relevant difference in embryo quality really did exist, good manual embryology would be highly efficient in selecting the “best” embryos. That was recently demonstrated by several papers, which reported no outcome differences between embryo transfers at cleavage and blastocyst stages in general populations, but suggested some outcome advantages for cleavage-stage transfers in poorer prognosis patients (mostly older women and those with fewer embryos).
To return to economics, one can only imagine how much cost could be saved for providers and patients in IVF if fertility clinics did not culture every embryo to blastocyst stage, but instead transferred at least a majority of embryos at cleavage stage on day-3 after fertilization, as used to be routine before one highly biased study alleged outcome benefits from blastocyst-stage transfers for all—a claim now clearly refuted.
The only outcome benefit that blastocyst culture may offer after embryo morphology has been established is a minor shortening of time to pregnancy if there is more than one obviously “best” embryo. But, so what if the first embryo doesn’t implant? The patient tries again the next month, and if that was indeed the better embryo, pregnancy will occur a cycle later. As the literature suggests, even this tiny benefit applies only to good-prognosis patients.
Is this worth the effort and additional costs of blastocyst-stage culture for everybody? We don’t think so!
And how about PGT-A for everybody, as many IVF clinics have made it almost a mandate? It is no longer possible to deny that PGT-A, in general populations, offers no established outcome benefit in IVF (though it may harm outcomes in some subpopulations). The most recent opinion paper from the Practice Committees of ASRM and SART has finally acknowledged publicly what everyone who can read the literature already figured out some time ago: such mandates or even recommended utilization of PGT-A no longer seem feasible (the recently filed class-action suits regarding this subject may also prove helpful).
Just imagine the cost savings for both clinics and patients if an IVF cycle did not require automatic blastocyst-stage culture or PGT-A. And then there’s so much more junk science being practiced in association with IVF—but that remains a topic for another time.
References
1. Illingworth et al., Nat Med 2024; 30:3114-3120
2. Kieslinger et al., Nat Med 2024;30:3059-3060
3. Gleicher et al., Nat Med 2022; 28::442-448
4. Kieslinger et al., Lancet 2023;401:1438-1446