Public Attitudes, Interests, and Concerns Regarding Polygenic Embryo Screening

Key Points Question Among US adults, what are the general attitudes, interests, and concerns regarding polygenic embryo screening (PES)? Findings In this survey study exploring public sentiment toward PES, there was high approval and interest among respondents despite high societal concerns. In a second sample, presenting concerns at survey onset (vs survey end) was associated with less public approval and more uncertainty but with only slightly higher disapproval. Meaning With the recent commercial availability of unregulated PES, the combination of high approval and high concerns observed among respondents in this study suggests the need for informed dialogue and guidance, particularly in addressing the divide between the public and health care professionals.


eMethods 1. Preregistrations
The survey preregistrations are uploaded at the following links: Sample1: https://aspredicted.org/JSJ_6BR Sample2: https://aspredicted.org/LM7_FZP eMethods 2. Survey Materials and Data Availability The survey materials are uploaded at the following links: Sample 1: https://researchbox.org/1646&PEER_REVIEW_passcode=XAFEJF Sample2: https://researchbox.org/1370&PEER_REVIEW_passcode=GLTTIB eMethods 3. Generation of Embryo Risk Scores for Introduction As part of the introduction, we informed participants about PES by presenting two embryos that varied in their genetic risk estimates across four conditions (see image below).Note that each risk/chance value was explained to participants in additional materials.
The difference between the standardized polygenic risk scores of the two embryos was computed based on their risk percentiles q1 and q2 using the following command in R: qnorm(q2/100) -qnorm(q1/100).For sibling embryos, the difference is expected to be distributed as a standard normal random variable.Thus, it would be rare for this difference to exceed 2 (or be less than -2).A difference within [-2,2] is feasible, but most of the time it would be smaller (i.e., [-1,1] or [-0.5,0.5]).We used the following inputs (K=; r=; q=) to compute each embryo's risk in R: q= percentile risk/chance is the proportion of variance in liability explained by the PRS.We used R 2 = 0.08 for all conditions as an upper bound on the current PRS effect size.This is based on values of 6% for Crohn's disease 5 , 7% for schizophrenia 6 and 9% for type 2 diabetes 7 Embryo Risk/Chance Output: We used the liability threshold model 5 .Given prevalence K, PRS percentile q, and proportion of variance explained r2, we used the following R code to compute the risk.
• zK=qnorm(1-K); s=qnorm(q/100,0,sqrt(r2)); risk=pnorm((zK-s)/sqrt(1-r2),lower.tail= F) Sample 1 (n=1427) was recruited by the sampling firm prolific and stratified to be nationally representative on the basis of gender, race/ethnicity and age.Sample 2a (n=97) had the concerns presented last (at the end of the survey) and was therefore the same survey as sample 1. Sample 2a was also recruited from prolific, but did not specify a nationally representative quota for gender, race/ethnicity and age.We compared mean approval, interest and concerns between sample 1 and sample 2a using a series of Welch's t-tests.
Results demonstrate that no significant differences in PES approval were observed between sample 1 (M=3.86,SD=1.02) and sample 2a (M=3.90,SD=0.941), t(111.90)=-.40,p=0.69; d=.04.Significant differences in PES interest were observed between sample 1 (M=3.13,SD=1.38) and sample 2a (M=3.53,SD=1.47), t(107.82)=-2.6,p=0.005; d=.29.No significant differences in PES concerns were observed between sample 1 (M=3.18,SD=0.95) and sample 2a (M=3.11,SD=0.77), t(116.58)=0.85,p=0.40; d=.08.We report these similarities and differences between the two samples for the sake of transparency, however, given the large differences in sample sizes and sampling methods, these results should not be overly interpreted as suggesting that there are either meaningful differences or a lack of differences between the two samples.Note: Group name "First" refers to participants from sample 2a randomly assigned to being presented with concerns first.Group name "Last" refers to participants from sample 2b randomly assigned to being presented with concerns first.(AVG)= Average of items.Note.As pre-registered, given that we did not find a significant difference in the average concerns between participants presented with concerns first (sample 2a) vs. last (sample 2b), we conducted exploratory Bayesian Interval-Null testing (equivalence testing) to determine how much evidence our data provide in favor of the null.Specifically, we tested whether the order of concerns (presented first vs.last) had no effect on the overall average of the concerns themselves.We set our equivalence region to -0.5 -0.5 and used the default Cauchy prior with a scale of 0.707.We tested the degree to which the data support the hypothesis that the parameter lies inside versus outside the equivalence region and found that the non-overlapping-hypothesis Bayes factor in favor of the interval-null was 171.Overall, these results demonstrate strong evidence for the null hypothesis when comparing participants in the concerns first condition with those in the concerns last condition.

eFigure 2 . 4 .
Sample 2: Concerns First vs Concerns Last-Equivalence Bayesian t Test on Concerns Plot eTable Sample 2: Concerns First vs Concerns Last-Descriptive Statistics

eTable 1 .
Sample 1: Political Ideology Weighted Sample on Polygenic Embryo Screening Approval Distribution We used Gallup's 8 data on political ideology from July 2023 (the month we ran the study) to calculate weights and re-analyzed approval with a weighted sample that matches the U.S. population distribution on political ideology.We measured political ideology on a scale from 1-7 and then aggregated values 1-3 to identify liberals, a value of 4 to identify moderates, and values 5-7 for conservatives.
Sample 2: Concerns First vs Concerns Last-Equivalence Bayesian t Test on Concerns