Problems With Using Polygenic Scores to Select Embryos

Embryo selection based on polygenic scores (ESPS) is the principle wherein one can select individualized embryos to transfer as part of in vitro fertilization based on predictions of health and other outcomes derived from genome-wide association studies (GWASs). The majority of human traits including physical characteristics and risk for disease are influenced by cumulative differences in genetic variants. A polygenic score intends to summarize the combined effects of these variations and imperfectly predict an individual's trait. This practice is marketed as being able to select for adverse traits as well as traits including cognitive ability, household income, and cognitive ability. The emergence of commercially available ESPS raises ethical as well as scientific questions regarding intended and unintended consequences of such practice, including alteration of population demographics, exacerbation of inequalities in society, and devaluation of certain traits. The extrapolation of data from GWASs to embryos is inappropriate. The authors of a recent article regarding polygenic scoring for education attainment found that the prevalence of college completion is 10% among those in the lowest quintile and 45% to 60% among those in the highest, concluding that the embryo with the favorable score should be selected. This is problematic given that the likelihood of exactly 2 viable embryos, one in the top quintile and one in the bottom, is less than 3%. In addition, the previously described article are based on persons from different families, all of whom have European ancestry, limiting the external validity. Indeed, sharing the same 2 biologic parents alone results in the expected gain from ESPS to be smaller. The "gene-environment correlation" suggests that those with a high polygenic score for education attainment were likely raised in family environments promoting education. The predictive power of a polygenic score is most accurate when the patient comes from the same environment as research participants who were involved in the GWAS study, and this applies for all studied traits. The differences in environmental context suggest that the expected gain from ESPS is smaller than one may infer from the study's conclusion. The use of polygenic scores for clinical decision making remains very limited, and more research is needed to derive clinical and personal utility. When a clinical outcome has a low baseline risk, small reductions in absolute risk can correspond to large reductions in relative risk and sound very promising. This is outlined in the study team's simulation, implying a relative reduction of 35% in type 1 diabetes risk with ESPS; however, given the average lifetime risk of 0.34% in the United States, the risk reduction amounts to only 0.12 percentage points. Outcomes for phenotypes defined using clinical cutoff points that represent a point on a continuum are even more limited by clear communication. Consider the simulation-derived ESPS reduction in risk of idiopathic short state of 1.8% compared with 10 embryos selected at random. The height of this child would be increased by approximately only 2.5 cm, a clinically irrelevant amount but one that would no longer earn the diagnosis of short stature. Given the widespread ramifications of such practice, the Federal Trade Commission should oversee information disclosure around ESPS, and there should be a greater society dialogue to decide whether regulatory frameworks should go beyond simply providing consumers with complete and accurate information.