What New York Magazine Gets Right About Embryo Screening, and What It Misses

Christopher Cox's New York Magazine piece raises the right anxiety about polygenic embryo screening: this category is easy to market badly. Probabilistic science can sound deterministic. Relative-risk numbers can make small absolute differences look dramatic. Many IVF cycles produce too few embryos for another layer of scoring to change the transfer decision. And ancestry calibration is still unfinished work across the field.

We should say that plainly. Herasight is a genetic testing company that provides rigorously validated polygenic embryo screening, also called preimplantation genetic testing for polygenic conditions (PGT-P). If a family has one or two embryos and no strong family-history concern, screening may not add much decision value. Herasight tells customers that. The responsible case for PGT-P is not "choose your best baby." It is narrower: when IVF already requires a choice among embryos, validated genetic risk information can sometimes help families make that choice with more context than morphology, PGT-A, or chance alone.

So the useful question after reading Cox is not whether every company in the category deserves defending. They do not. The question is what a responsible provider must prove before asking families to trust another number at one of the most emotionally loaded moments of their lives.

Where the article is right

Start with marketing. Subway ads that say HAVE YOUR BEST BABY should bother people. They bothered us too. That campaign was Nucleus Genomics' pickyourbaby.com campaign, not Herasight's, but the distinction only matters if Herasight behaves differently. PGT-P should never be marketed as a guarantee of a healthy child, a replacement for embryo viability assessment, or a way to escape uncertainty. It should be offered with published methods, visible limits, and careful counseling.

The embryo-count problem is one of Cox's strongest critiques. Lencz et al. 2021 modeled meaningful relative-risk reductions for schizophrenia and Crohn's disease, but at population baseline those can translate into absolute reductions under half a percentage point. That is exactly why absolute risk matters. Herasight reports estimated risk for each embryo with assumptions and uncertainty ranges attached, not only brochure-friendly relative-risk reductions.

The critique in Polyakov et al. 2022 is right: relative-risk language without absolute context can mislead. We report risk, not risk reduction. A 50% relative reduction can be a tiny absolute change when baseline risk is low. Families deserve the less dramatic number if it is the more honest one.

Ancestry is another real limitation. Most polygenic-score research has leaned heavily on European-ancestry datasets, and performance does not transfer evenly across populations. Herasight's Moore et al. 2025 reports ancestry-specific calibration and shows improvement for African and East Asian populations. That is not the same as saying the problem is solved. It means families should be shown the relevant performance numbers before deciding whether the information is worth adding to their IVF cycle.

Professional-society caution belongs here too. The ACMG points-to-consider statement raised serious questions about clinical utility, evidence, and counseling. Those questions should not be waved away. The answer is better validation, not louder marketing. Moore et al. 2025 tested seventeen disease scores on sibling pairs and found that sixteen held predictive performance within families. Widen et al. 2024 also argued that the original ACMG statement gave an incomplete picture of polygenic prediction. That does not end the debate. It shows the debate has moved from slogans to evidence.

It also helps to separate the genetic tests that often get blurred together. PGT-M (preimplantation genetic testing for monogenic disease) asks whether an embryo inherited a known family variant. PGT-A (preimplantation genetic testing for aneuploidy) asks whether the embryo has the expected chromosome count. PGT-P asks a probability question across many variants for conditions where there is no single yes-or-no gene. Different tests answer different questions, and the evidence standard has to match the decision the family is actually making.

What separates a responsible polygenic-screening provider

Once the marketing problem, the embryo-count problem, the ancestry limit, and the professional-society caution are on the table, the next question is what good work looks like when a family is choosing a provider. Most of it is methodology, not slogan.

Start with within-family validation. Embryos from the same IVF cycle are biologically siblings, so the relevant test is not only whether a score predicts risk among unrelated adults. It is whether the score can still distinguish risk among siblings who share the same parents. That is the methodological worry the ACMG statement names, and it is what Moore et al. 2025 set out to test. (More on within-family validation here.)

Outcome tracking is valuable, and any provider should want more of it. But it cannot be the only validation gate, because some outcomes take decades. The practical question today is narrower: do the variants and scores that predict disease in adults still rank siblings from the same family in the expected direction? That is the question within-family validation directly answers.

On ancestry, the standard is per-population reporting, not a single average number. A provider that hides ancestry performance behind an averaged statistic is asking families to trust a number they cannot inspect.

Absolute risk is the second rule. If a family's baseline risk is ordinary, the absolute differences between embryos will usually be small, and the report should say so. If the baseline is elevated by family history or ancestry, the differences can be larger, and the report should say that too. The conditioning has to travel with the number every time.

Interpretation is the third rule. A multi-page PDF is not counseling. A responsible provider should be able to explain what the absolute risk means for a specific family history, why a score is more or less informative for a specific ancestry background, and when the expected difference between embryos is too small to matter.

And then there is the willingness to say, in public, that PGT-P is unlikely to change the decision for many families. That is not a marketing risk to manage. It is part of the credential that gives a provider the right to do the work for families where the test can change a decision.

A reader who closes Cox's tab can take all of this with her into any provider conversation. Ask whether the provider has tested the scores on actual sibling pairs and published the result. Ask whether the report gives absolute risk with stated assumptions instead of only relative-risk language. Ask whether ancestry performance is reported instead of smoothed over. Ask whether the counselors on the call actually understand PGT-P. And ask whether the provider will tell you, plainly, when the test is unlikely to change your decision. The answers don't have to be perfect. They have to be specific.

On "designer babies" and the eugenics question

The phrase "designer babies" carries history that should not be brushed aside. The serious objection is not just the slogan. It is that even voluntary embryo selection still involves choosing among possible children, and pretending that this does not raise ethical questions would be dishonest.

Historical eugenics was coercive, government-sponsored, and targeted groups. Herasight offers voluntary, parent-directed disease-risk screening inside IVF, where families are already choosing among embryos using morphology and PGT-A. We believe parents should have access to honestly validated information. We also believe no one should be pressured to use it, and no one should be told the information can guarantee a child.

Cox closes on a customer who says he would use CRISPR editing if it were available. We cannot tell people what to want. We can be clear about what Herasight offers and what it does not. PGT-P at Herasight is preimplantation genetic testing for polygenic conditions, scoped to disease risk and reported in absolute terms with visible uncertainty. CRISPR-edited babies are a different technology and a different debate.

Some of the discomfort Cox surfaces should stay. It keeps providers from pretending this is only a product-design problem. The answer is not less scrutiny. It is better scrutiny.

If you want to see what a Herasight report actually looks like, request a sample. If you want to talk through what the current evidence means for your family, reach out to our counselors.