This article has been updated to correct the current affiliation of Stephen Hsu, who no longer works at Michigan State University.
NEW YORK – As preimplantation genetic testing (PGT) rises in US fertility clinics, a small number of companies is betting on assays that incorporate polygenic risk scores (PRS) to gauge an embryo's lifetime risk of common and complex disorders such as schizophrenia, hypertension, and various cancers. But the use of these tests remains controversial, with many experts in the field saying they are unconvinced of their predictive value or the actionability of the results.
Within the US, Genomic Prediction currently offers a PGT for polygenic disorders (PGT-P) called LifeView, and Orchid Health recently began inviting couples to an early-access program for its test. Meanwhile, MyOme is preparing to launch its own test, pending further study results, technology development, and the finalization of a launch strategy.
Companies bringing PRS into the reproductive health field are also leveraging increasingly large genomic datasets from disease studies to support their claims, while initiating longitudinal trials to track health-related outcomes associated with PRS predictions in embryos.
The science of PRS
A PRS is a single value summarizing the estimated effect of multiple genetic variants — generally single nucleotide polymorphisms, or SNPs — on an individual's phenotype. They were first used in selective plant and animal breeding and were proposed as a way to identify disease risk in humans in 2007.
PRS is calculated by summing an individual's trait-associated alleles, where each allele is weighted by its effect size from a discovery GWAS and normalized to the total number of risk alleles and effect sizes evaluated within a relevant population.
Individuals with more alleles associated with a certain phenotype or disease — say breast cancer — would appear to have a higher genetic probability of acquiring that phenotype over time.
Researchers have identified a variety of PRS associated with disease and non-disease phenotypes, ranging from breast cancer and cardiovascular disease to height and IQ.
Polygenic embryo selection, or PES, represents perhaps their latest use of PRS, and while still a niche application, it has already raised headlines of "Gattaca babies," alluding to the 1997 movie about a dystopian future made possible through directed genetic selection.
Leaving aside questions of bioethics, many experts are unconvinced that the science of PRS supports the claims of PES.
"We don't have the long-term evidence to say that providing a PRS is better than what we already have," said Tatiane Yanes, a research fellow at the University of Queensland Diamantina Institute and an expert in PRS. Although she sees a lot of potential in PRS and suspects that evidence of their utility will come with time, "we're not there yet," she said.
But according to Stephen Hsu, cofounder of Genomic Prediction, a reproductive health company offering a PGT-P assay, large sibling datasets already provide sufficient information to identify polygenic signatures underlying various common and complex disorders.
"I could tell from the rate at which data was becoming available that the day would come sooner rather than later when we'd be able to make meaningful predictions [about individuals] using just DNA alone," said Hsu, a theoretical physicist by training.
Genomic Prediction says it works with approximately 200 IVF clinics on six continents, and Laurent Tellier, its CEO, estimates that about 172 couples have used the company's PGT-P, with about 40 live births so far.
PRS are already used for genetic risk prediction in adults, for example by Myriad Genetics, which offers RiskScore, a PRS-based breast cancer risk assessment tool offered alongside the company's MyRisk Hereditary Cancer test.
This isn't to say that PRS can predict with certainty who will and won't get a disease like breast or prostate cancer within their lifetime, Hsu explained, but that it can identify those who have much higher or lower genetic risk for certain conditions than the general population.
"There's really no serious question in this field that polygenic risk prediction does work at some level," Hsu said.
The level at which it does work to prevent future disease, however, appears open to debate. Jonathan Mosley, assistant professor of medicine and biomedical informatics at Vanderbilt University, questioned whether bigger genomic datasets necessarily translate to better risk predictions.
"[Greater] sample size allows you to measure smaller effects," he explained, but while expanded cohorts have been critical for discovering new genetic variants, the risks associated with the variants "are just getting increasingly small."
Essentially, he said, the argument that large datasets are needed to validate PRS boils down to saying that the magnitude of a PRS-based prediction is so small that we need very large numbers of people to even detect it, effectively enabling us to quantify smaller effects.
"A small study might say this risk accounts for four plus or minus 2 percent of the variability," he explained, whereas "a massive study will say, well, it actually counts for 3.5 plus or minus 0.2 percent of the variability. I think you're just giving a degree of precision to a smaller effect."
An example of this can be found in evaluating PRS for schizophrenia, one of the disorders included in Genomic Predictions' PGT-P assay.
Schizophrenia affects approximately 1.2 percent of Americans and has no known causative single-gene mutations. Research into PRS for schizophrenia has uncovered some genetic associations but these were generally slight and their validity as a diagnostic tool remains under investigation.
Alexander Charney, assistant professor of psychiatry as well as of genetics and genomic sciences at the Icahn School of Medicine at Mount Sinai, recently published one such study evaluating the prognostic value of PRS among 8,541 multi-ethnic adults with schizophrenia and related disorders. His team's analysis showed no predictive improvement with respect to PRS.
Charney's assessment in that study largely mirrors the research of other experts investigating the predictive value of PRS for a variety of disorders.
"The science behind PRS-based diagnostics remains highly investigational and it will take some time to prove any clinical worth," Charney said.
Although some studies have shown the risks for breast cancer due to pathogenic BRCA mutations and disease-associated PRS to be additive — that is, people who have a higher PRS value and a pathogenic variant appear to be at higher risk than those with the pathogenic variant and a lower PRS — the independent risk associated with the pathogenic variant typically outweighs the risk added by the PRS. In other words, the PRS may add little to the overall risk estimation.
But Myriad and others developing polygenic risk scores are betting that despite their relative weakness compared to strong pathogenic variants, tests like RiskScore will provide actionable risk information when large NGS panels screening for pathogenic variants in single genes turn up negative. They may also refine risk estimates when considered alongside mutations in moderate-penetrance genes.
"We see that for roughly one in five (17 percent) women, the RiskScore result may modify the plan for high-risk breast screening," Thomas Slavin, Myriad's chief medical officer, said via email.
The actionability of breast cancer PRS is further supported by some evidence linking the use of corticosteroids to heightened breast cancer risk for women already at increased genetic risk.
Although companies like Genomic Prediction are making similar bets, embryonic PRS for other disorders may be less actionable, and they come from less precisely defined populations, being available to any consumer who requests one, as well as coming from much earlier time points.
Myriad limits RiskScore's application, for instance, to women between the ages of 18 and 84 who have no personal history of breast cancer, hyperplasia, or atypical hyperplasia, and who test negative for known breast cancer gene mutations. In contrast, PGT-P aims to calculate a PRS for a greater number of disorders for embryos that have not even been implanted yet.
Also, some of the disorders tested for, such as hypertension, affect so many people — nearly 50 percent of US adults aged 50 and older — that all non-company experts interviewed for this article balked at the claim that a low PRS could meaningfully reduce one's risk of the disorder.
Research to support a hypertension-related PRS also leaves many questions unanswered. One recent study, for example, identified such a PRS, but the researchers were unable to link it with family history. Another identified a hypertension-related PRS among long-term survivors of childhood cancer but could only account for some 40 percent of the hypertension cases in the study.
Researchers skeptical of PGT-P also worry that the assay could inflate a couple's perceived risk of an embryo's lifetime health prospects.
Part of this has to do with estimations of relative versus absolute risk. Comparing an individual's PRS results to a population average provides a sense of relative risk. That individual's absolute risk — their personal probability of acquiring or avoiding some illness — could differ significantly.
In a special report published in the New England Journal of Medicine in July, Patrick Turley, a researcher at the Broad Institute, wrote that a couple's expected gain — the degree to which they can expect a PRS to make a difference in their embryos' health outcomes — may be much smaller than hoped for.
Via email, he explained that there is less variation in genetic risk between embryos born to the same parents than in the wider population. A couple who believes that the amount of genetic diversity in the population equals what they could expect for the genetic diversity of their own embryos, he explained, is like believing that each embryo's health risks represent a roll of an unweighted die.
Simplistically, if the chance of having a low-risk embryo was equivalent to rolling a one on a six-sided die, then a single roll gives that embryo a 16 percent chance of being low risk. Assuming that PRS-based embryo selection identifies such an embryo, then rolling 10 dice — selecting from 10 embryos — would seem to raise the odds to an 84 percent chance of rolling a one. Such a large relative improvement might make prospective parents excited about pursuing embryo selection by PRS (ESPS).
"But in reality, the couple is rolling weighted dice," Turley said, which could equate to only rolling threes and fours. "So the likelihood that they are going to roll a one is very small, whether or not they use ESPS. Understanding that may make ESPS seem less attractive."
In one example in Turley's NEJM article, for instance, he calculated a less than 3 percent chance of a couple having two viable embryos, one in the top PRS quintile and the other in the bottom. Although Turley favored using educational attainment as an outcome in that paper, the idea extends to health-related outcomes.
"On average," Turley and his colleagues wrote, "offspring will tend to have PRS scores near the mid-parental PRS value."
This implies that parents with a higher PRS — or risk for a disease phenotype — might have most to gain from reducing that risk in an embryo through PGT-P, while those with a lower risk might benefit very little.
The question of how much risk is acceptable or even meaningful is already coming up in fertility clinics that offer PGT-P.
Klaus Wiemer, lab director at Poma Fertility, a fertility clinic in Kirkland, Washington, that offers Genomic Prediction's LifeView assay, recounted a recent experience in which a woman opted for a second cycle of IVF in search of embryos with better risk scores.
"Even though the embryos are genetically normal," said Wiemer, who has a background in reproductive physiology, "she was just unhappy with the heritable scores that the embryos got for certain traits. So what happens to all those other embryos that are okay? That's the problem that we're starting to face."
Both Genomic Prediction and MyOme, a reproductive health company in the Bay Area preparing to enter the PGT-P field, said that genetic counseling is recommended to couples as a normal part of the IVF process. MyOme stated that it is evaluating consumer understanding of PRS in an ongoing trial. As Wiemer's anecdote suggests, understanding PRS — how it works, as well as its strengths and limitations — should be a key aspect of the technique's wider use.
Furthermore, it remains an open question whether prospective parents can avoid a negative health outcome for their child by selecting against a supposedly high-risk embryo.
While the health of a discarded embryo will never be known, Genomic Prediction and MyOme have each launched prospective, longitudinal studies to study the clinical outcomes of embryos selected based on PGT-P.
"We are continuing to enroll individuals who are already pursuing IVF and PGT in a free-to-participants prospective trial that we started in 2019, which provides additional analytical validation and also evaluates our approach to reporting and genetic counseling," Premal Shah, the company's CEO, explained via email, adding that study details are expected to be published early next year.
The company has so far conducted a retrospective study aimed at validating the analytical accuracy of its genetic predictions among people who underwent IVF, results of which are expected to publish soon.
Yet another US company, San Francisco-based Orchid Health, already offers PES services. Following a recent $5.4 million seed round, Orchid began inviting couples to join a waitlist for early access to the company's Couple Report, a " physician-approved" genetic test that estimates whether a couple can give their children higher genetic risks for 10 diseases: heart disease, stroke, atrial fibrillation, schizophrenia, Alzheimer's disease, breast cancer, prostate cancer, type 2 diabetes, type 1 diabetes, and inflammatory bowel disease. The company also advertises an Embryo Report, offering "advanced genetic screening" for embryos, through what appears to be a PRS-based analysis. Orchid did not respond to requests for further information.
Another issue with polygenic embryo selection is a lack of standards for how PRS are calculated and reported.
"It's not like another blood test that you just look at and can interpret," said Yanes. "There are different calculations and ways of doing it, so you may be getting different scores by different companies."
Although no company divulged the exact details of their polygenic risk assay, each maintained that theirs differs in some meaningful way from that of their competitors.
Two working groups, the Clinical Genome Resource Complex Disease Working Group and the Polygenic Score Catalog, recently proposed a set of standards for reporting PRS. Their proposal was published in March of this year, and it is not clear how widely adopted it has become since then.
All of this may be a moot point for consumers, however, who have described feeling empowered by the knowledge of their embryos' scores, according to Wiemer.
"We have several patients now that are pregnant," he said, referring to couples that had used LifeView. "They told us that it's very empowering to them to know that they really minimized or eliminated the risk that was in their family."
Whether or not, and to what degree, that risk has actually been minimized remains an open question. With products already on the market, more on the way, and a number of happy customers, PRS-based embryo selection appears likely to gain traction, at least in the near term, regardless of any questions of its underlying science, clinical utility, or ethical quandaries.