Contra the ACMG on Polygenic Embryo Screening
An opposing perspective to "Clinical utility of polygenic risk scores for embryo selection: A points to consider statement of the American College of Medical Genetics and Genomics (ACMG)"
There has been a growing interest in preimplantation genetic testing for polygenic disorders (PGT-P) following its recent commercial availability. Several scientists and professional organizations have published articles and statements critical of the practice. Turley et al. (2021) have called PGT-P an “unproven, societally risky service.” Forzano et al. (2022) have referred to it as “an unproven, unethical practice.” Polyakov et al. (2022) have said it is “not ready for prime time.” The most recent critique of this kind is a statement of the American College of Medical Genetics and Genomics (ACMG), in which Grebe et al. recommend that PGT-P “should not be offered as a clinical service,” citing several concerns such as lack of validation, questionable utility, and changing environmental influences.
While some of the authors’ concerns are legitimate, their recommendation against PGT-P is incorrect due in large part to their analysis not being framed in the context of the embryo choice problem—IVF couples who want a child must choose among the available embryos. Advising against PGT-P to non-carrier couples with multiple euploid embryos serves as an implicit recommendation for alternative methods, such as selection using embryo morphology, time-lapse imaging, or random selection. Listing reasons to question available evidence and utility is insufficient. The authors need to demonstrate that PGT-P is worse than all other acceptable alternative methods for selecting healthy embryos. They have neither demonstrated nor even claimed this.
After aneuploidy screening, the typical alternative to PGT-P is selection using embryo morphology, but this is not discussed in the ACMG statement. Although the intended purpose of selection using morphology is to improve the probability of successful pregnancy rather than health outcomes, this method is subject to some of the same criticisms leveled against PGT-P (e.g., measurement inaccuracy, proof of long-term safety). Moreover, some have proposed that selection via morphology could potentially perform worse than randomness in terms of health outcomes due to the “quiet embryo hypothesis.” The authors neither showed nor claimed that selection using morphology outperforms PGT-P in their areas of concern, so their asymmetric condemnation of PGT-P seems like an isolated demand for rigor.
Despite acknowledging multiple studies performing sibling validation with polygenic risk scores, the authors claim that the applicability of the evidence may be “limited if the embryos born today face a sufficiently different environmental context.” Unless they have a good reason to believe the improbable scenario that the environment of the future will change so drastically as to render PRSs worse than useless, then it would be reasonable to believe that PGT-P is still either equivalent to or better than randomness. However, their recommendation against PGT-P would seem to imply that PGT-P is a bad decision even when compared with mere random selection. Such a set of beliefs would seem to have the odd implication that decision-making could improve by intentionally disregarding relevant—although limited—information.
If polygenic risk scores are completely useless, they will produce outcomes equivalent to randomness. If PGT-P is actively harmful, selecting a top-scoring embryo on a health index will result in worse outcomes than random, and therefore, using PRS to avoid the top-scoring embryo would yield better outcomes. As far as I have seen, no PGT-P skeptic has advocated this strategy. Furthermore, there seems to be no concern that random selection and PGT-P would often result in the exact same embryo being transferred.
Grebe et al. claim that in many scenarios, “the risks [of PGT-P] outweigh the benefits, leading to concern for individual harm to either the prospective parent or the future child.” If transferring the same embryo is acceptable when unaware of genetic information but could be harmful to the future child when informed with PGT-P, then it would appear the supposed harms of PGT-P are strangely not related to actual outcomes but making informed choices. There is no “harmful” outcome of embryo choice that PGT-P could produce that random selection could not. If using PGT-P is dangerous, then so is random selection. In actuality, PGT-P is safer and more effective than alternatives.
Grebe et al. propose other potential harms, saying, “[e]ven in the best scenarios, there remains the risk of harm in the form of false assurance and monetary loss for unclear gain.” Neither of the leading providers of PGT-P —Genomic Prediction and Orchid Health—claim to wholly eliminate the chance of disease. Regardless, the possibility of some level of misunderstanding or miscommunication is not a good reason to oppose providing a service. We would not think that someone incorrectly believing they cannot contract a disease because they are vaccinated is a valid reason to oppose vaccination generally. Additionally, if we consider psychological costs, we cannot ignore the benefits that may come from having a healthier child.
The authors claim that in many scenarios, “the risks [of PGT-P] outweigh the benefits,” but they do so without providing a cost-benefit analysis that weighs numeric estimates against monetary harms and the probability of various outcomes. Rather than just claiming the possibility of “monetary loss for unclear gain,” it would be illustrative to see the authors’ estimated return from PGT-P on a broad health index and the value they believe parents should put on their child’s health in monetary terms.1 A universal recommendation against PGT-P as a clinical service seems unwarranted if not even a rough cost-benefit calculus is performed. This case against universal recommendations is especially true considering that the expected returns of PGT-P vary depending on couples’ family history, life circumstances, risk tolerances, and personal values.
Grebe et al. reaffirm a statement by the ACMG Board of Directors that “prenatal testing for disorders that exhibit multigenic or polygenic inheritance is not yet appropriate for clinical use and should not be offered as direct-to-consumer testing.” These are recommendations for restrictions on the ability of couples to learn information about their own embryos and constitute violations of reproductive liberty. Women should have the right to learn information about the embryos they are going to put into their bodies, and couples should have a right to know about the child they are eventually going to raise. Surely, some couples at an elevated risk of certain conditions would not even have a child without PGT-P. If a couple does not have a child or has a significantly less healthy child due to being denied this service, it could amount to the most impactful restriction on personal liberty they face in their entire lives.
In conclusion, the primary shortcoming of the article is that it does not consider the ethics of PGT-P through the lens of the embryo choice problem—an embryo must be chosen somehow.2 A list of “points to consider” is not adequate to reach the conclusion that PGT-P should not be provided; it needs to be shown that PGT-P is worse than alternatives. Such an analysis would warrant some use of cost-benefit calculus with numeric estimates of expected return as well as some effort to quantify the utility of the health benefits, but this is lacking. If PRS for embryos are completely worthless, then their utility would be equal to random selection, but no plausible case has been made that PGT-P are worse than randomness. Ultimately, this case will be impossible to prove because wilfully disgarding relevant information cannot improve decision-making. Eventually, the medical community will recognize that choice prevails over chance.
Such an analysis would demonstrate that there are strong reasons for believing PGT-P passes the bar of cost/risk-benefit analysis. Without considering outlier cases, expected average return for European couples is quite high. Using UK Biobank data, Widen et al. (2022) estimates an expected return from selection among 10 individuals at ~4 disability-adjusted life years (DALYs) and among 5 embryos at ~3 DALYs. Selection among genetic siblings—equivalent to embryos—would produce returns between approximately 3 to 4 DALYs for selection among 10 and approximately 2 to 3 for selection among 5. For couples already undergoing aneuploidy screening, Genomic Prediction charges $1000 plus $400 per embryo. It is extremely reasonable to claim a disability-adjusted life year of your child’s life is worth more than a few thousand dollars. Even if PGT-P ended up being 50% as effective for reasons of ancestry or environmental changes, the benefits remain clear. Skeptics’ claims that PGT-P is not ethical seem to not use cost-benefit analysis but instead largely revolve around semantic arguments with loaded language (e.g. “unclear gain,” “unproven,” “insufficient evidence”) and claims about potential low-probability risks. We can make analogous arguments to reveal the inadequacy of this method: “We have yet to prove the COVID-19 vaccine safe in the longterm. There is a risk that everyone suddenly develops adverse outcomes at 5 years.” Such an analysis would fail to mention that the probability of this is extremely low (<1%) and the alternative of everyone going unvaccinated also carries many potential harms. Such a one-sided analysis would not be an adequate way to evaluate the COVID-19 vaccine, and obviously so.
There are certain less common cases in which couples may undergo IVF solely for the purpose of using PGT-P. Although this complicates the ethical analysis, it should still be defended for reasons of reproductive liberty. Furthermore, once again, whether this is worthwhile depends on the unique life circumstances of the couple. Consider this hypothetical scenario: is a 50% relative risk reduction of schizoprenia, a 20% RRR in Alzheimers’, 25% RRR in breast cancer worth a 100% increase in the relative risk of preeclampsia? How much money is it reasonable for a parent to spend for a 20% absolute risk reduction in type 1 diabetes? If one is advocating for restricting reproductive liberty, they should be performing this sort of analysis.
Seems like an excellent opportunity for you to do a such cost-benefit analysis.
GM, I have a question: is someone studing if we are selecting for the WRONG poly-genes?