"Have your best baby." That was the 2025 marketing slogan for Nucleus, a California startup offering polygenic screening of human embryos. For somewhere in the low tens of thousands of dollars on top of the cost of in vitro fertilization, Nucleus and a handful of competitors will analyze the DNA of each embryo a couple produces, calculate scores for traits the company has built models for — diabetes risk, heart disease risk, height, intelligence — and recommend which embryo to transfer.

The pitch is straightforward enough that it bypasses most prospective parents' usual skepticism. Genetic testing of IVF embryos has been routine for over twenty years. Polygenic risk scores are a real technology that real research labs publish on every week. Why wouldn't you, given the choice, pick the embryo with the lower diabetes risk and the higher predicted IQ?

The answer is more interesting than either the marketing or the easy critiques suggest. The science doesn't say the technology is useless. It says it does much less than the marketing claims — by amounts that matter — while carrying risks and trade-offs that the consent forms don't always make obvious. In December 2025, the American Society for Reproductive Medicine's Ethics and Practice committees, after a comprehensive review, issued an opinion concluding that polygenic embryo screening, also known as PGT-P, "should not be offered clinically." That puts the largest U.S. reproductive medicine society directly at odds with a fast-growing industry.

This article walks through what PGT-P actually is, what it can and cannot do, and where the science draws the lines the marketing tries to blur.

What PGT-P is, in plain language

Preimplantation genetic testing has been part of IVF for decades. The standard panels — labeled PGT-A, PGT-SR, and PGT-M — look for chromosomal abnormalities (aneuploidy, structural rearrangements) or single-gene Mendelian disorders like cystic fibrosis or Huntington's disease. In these tests, the relationship between DNA and disease is direct: if a child inherits two copies of a known disease-causing variant, the disease will manifest. The clinical case for testing is unambiguous.

Polygenic risk scores are different. They aggregate the small statistical effects of hundreds or thousands of common genetic variants — single-nucleotide polymorphisms, or SNPs — to produce a probabilistic prediction for traits that aren't determined by any single gene. Most adult-onset diseases (type 2 diabetes, coronary artery disease, schizophrenia, breast cancer beyond the high-penetrance BRCA mutations) and most physical and behavioral traits (height, body mass index, educational attainment) are polygenic. The science of polygenic prediction is genuinely powerful, and is increasingly being deployed for adult risk stratification.

PGT-P applies these scores to embryos. The clinic biopsies a few cells from each blastocyst, sequences the DNA, computes polygenic scores for whatever conditions the lab offers, and ranks the embryos. The couple — guided by a genetic counselor, ideally — then chooses which embryo to transfer.

The technology relies on three things working: enough usable embryos to choose between, polygenic scores that accurately predict the trait of interest, and a meaningful difference between the candidate embryos on those scores. The science says all three are wobblier than the marketing acknowledges.

How much benefit, really

The single most important paper in this debate, published in Cell in 2019 by Ehud Karavani and colleagues, modeled exactly this question. Using empirical data on the genetic variation between sibling embryos, they estimated how much you could shift a child's expected phenotype by selecting from among a typical IVF cohort.

For the trait their analysis is most quoted on — height — the average gain from selecting the top-ranked embryo out of five was roughly 2.5 centimeters. For predicted IQ, about 2.5 points. These are real, but they sit far below what the marketing implies.

The gain shrinks for two reasons. First, sibling embryos are genetically similar by definition — they share half their genome with one parent and half with the other. The variation among them is much smaller than the variation across unrelated people, which is where most polygenic scores are validated. Second, even within that small range, the prediction interval around any single embryo's score is huge. A 2021 commentary in the New England Journal of Medicine on polygenic embryo selection put it bluntly: "the majority of children who are top-scoring for height in large nuclear families are not the tallest."

A 2024 review in Human Reproduction Update compiled the available evidence on PGT-P across both disease and trait selection and reached similar conclusions: the absolute risk reductions for common diseases through embryo selection are small (typically a few percentage points of lifetime risk), the prediction intervals are wide, and the clinical evidence base is thin. The same review pointed out that gains for non-disease traits like cognitive ability or height are at the low end of what is even biologically achievable with current technology.

The ancestry problem

Polygenic risk scores have a well-documented generalizability problem: most of the studies that built them — the giant genome-wide association studies of disease — were performed on people of European ancestry. A polygenic score that explains 9% of variance in schizophrenia liability in European populations may explain only 1% of the variance in African-ancestry populations. A 2020 review in Genome Medicine on polygenic risk scores as clinical instruments documents the same pattern across cancer, cardiovascular disease, and psychiatric conditions: predictive accuracy drops substantially outside the population the score was developed in.

For a couple of European ancestry, this is an annoyance in the consent process. For a couple of mixed or non-European ancestry, it is the central problem. The score you are being asked to pay for is, in the most literal sense, less accurate for you. A 2023 critical appraisal of the clinical utility of polygenic risk scores concluded that until GWAS catalogues are dramatically expanded across global ancestries, PRS-based clinical decision-making will systematically advantage European-descended populations and risks deepening existing health disparities.

In the embryo-selection context, this means the magnitude of any expected benefit varies by ancestry — but the price does not.

What the trade-offs look like in practice

Every IVF cycle produces a finite number of usable embryos. The number that survive to the blastocyst stage suitable for biopsy is typically small — often only one or two for older patients or those with diminished ovarian reserve. PGT-P can only do anything useful if there are enough embryos to differentiate between.

This is the first practical limit the marketing tends to soften. If you have three usable embryos, the polygenic-score differences among them may be smaller than the score's error bars. The "choice" being offered is statistical noise. The ASRM ethics opinion specifically flags this scenario: clinicians and patients may end up making transfer decisions based on score differences that are not clinically meaningful, with no way for the patient to know that.

A second trade-off is the cost of discarding embryos. Polygenic ranking inherently produces "better" and "worse" embryos. In settings where viable embryos are scarce — older patients, low responders — transferring the "worse" embryo is sometimes the clinical right answer. Patients who have paid for PGT-P and watched their embryos ranked may feel pressure to undergo additional IVF cycles to find a "better" candidate, exposing them to medical risk and significant additional cost for marginal expected gain. The 2024 Karavani-and-colleagues line in Human Reproduction Update makes this concern explicit: PES (polygenic embryo screening) should be offered, if at all, only within a research context until the balance of benefits and harms is better characterized.

A third trade-off is pleiotropy: the fact that the same genetic variants influence multiple traits, often in opposite directions. Variants associated with reduced schizophrenia risk overlap with variants associated with lower educational attainment. Variants associated with reduced bipolar risk overlap with creativity. Selecting against one disease can quietly select against traits a parent might value, in ways the consent form may not detail. A 2021 NEJM commentary on PGT-P used the term "unintended consequences" deliberately: with polygenic scores, the trait you are selecting against rarely travels alone.

Ethics: the part that's harder to outsource

The technical limitations are real. The ethical questions are unavoidable. The ASRM committee opinion identifies several:

The eugenics shadow. Selecting embryos for non-disease traits — height, intelligence, athletic potential — sits squarely within the cultural memory of twentieth-century eugenics. The 2024 Human Reproduction Update review notes that public, media, IVF patient, and clinician surveys all return "eugenics" as a top spontaneous association with PGT-P. The disease-versus-trait distinction is fuzzy at best, and clinics that offer disease screening today often add trait modules later.

Disability rights. Choosing against an embryo because it scores higher for autism spectrum risk, ADHD risk, or developmental delay risk is a statement about which kinds of lives are worth bringing into existence. Disability scholars and advocates have pointed out that the same logic, applied earlier, would have selected against many people now leading full lives — and that the assumptions behind the ranking are themselves contested.

Access and equality. PGT-P costs run from a few thousand dollars to the high five figures, on top of IVF. The science of polygenic prediction is most accurate for people of European ancestry. The technology is most accessible to people who can afford private fertility care. Each of these gradients compounds existing inequalities. The ASRM ethics opinion names this directly as one of the reasons PGT-P is "not ready for clinical use" — not because the questions can't be answered, but because they haven't been.

Informed consent. The information asymmetry between a startup with a marketing budget and a couple in the middle of a difficult IVF cycle is substantial. Consent forms tend to disclose the existence of uncertainty. They are less likely to convey what 2.5 IQ points of expected gain — within a 30-point prediction interval — actually means.

The interventions that actually move the needle in preconception

It is worth noticing what gets ignored when conversations about preconception genetics dominate. The interventions with the largest, best-documented effects on child health do not involve selecting embryos. They involve the months leading up to conception and the nine months that follow.

Folic acid supplementation before conception reduces the risk of neural tube defects by approximately 70%, an effect size larger than any plausible polygenic gain.

Controlling maternal blood sugar — through weight management, exercise, and, when needed, treatment of pre-existing diabetes — substantially reduces the risk of congenital anomalies, macrosomia, and pregnancy complications.

Stopping smoking and alcohol before pregnancy reduces risks across nearly every fetal outcome measured: birth weight, prematurity, congenital defects, and long-term neurodevelopmental outcomes.

Treating periodontal disease in pregnancy lowers the risk of preterm birth.

Achieving a healthy pre-pregnancy weight improves fertility, lowers gestational diabetes and preeclampsia risk, and improves long-term outcomes for the child.

None of these is marketed as a Silicon Valley innovation. All of them have larger, better-replicated effect sizes than anything PGT-P is currently capable of delivering. They share a different selling proposition: they involve sustained behavior change, they require honest tracking of habits, and they reward boring consistency. For people thinking about pregnancy, the routine work of monitoring weight, blood pressure, alcohol use, sleep, and mood in the months before conception — the kind of structured tracking WatchMyHealth is designed for — produces the kind of evidence-based gains the science actually supports. Preconception care, in other words, is one of the highest-leverage things you can do. It just doesn't make headlines.

What's likely to happen next

The ASRM opinion is not a regulation. It does not bind clinics legally. Several U.S. states are debating regulations on polygenic embryo screening, and the European Society of Human Reproduction and Embryology has issued similar cautions. In Russia, where IVF is widely available but PGT-P is not part of mainstream practice, the technology is more talked-about than offered. Internationally, the picture is patchy.

What is more or less certain is that the technology will get better. Polygenic scores will be calibrated across more ancestries. Larger family-based GWAS will sharpen estimates of the truly within-family genetic effect, separating it from the cross-population effects that inflate the apparent power of current scores. New tools — for instance, the 2024 "Promises and pitfalls of PGT-P" review in F&S Reviews — will reset expectations as the science evolves.

What is also more or less certain is that the marketing will outpace the science. The clinics that today claim to predict your future child's intelligence with meaningful accuracy are not going to back down voluntarily. Patients considering PGT-P — or being offered it as part of a fertility package — are largely on their own when it comes to assessing whether the technology delivers what the brochure says.

A few practical principles can help. If a clinic offers PGT-P, ask for the expected absolute risk reduction, not the relative risk reduction, for whatever disease you care about. Ask for the score's predictive performance in your specific ancestry. Ask how many embryos you'd need to make the selection clinically meaningful. Ask whether the clinic has any data on live-birth outcomes — which, for non-disease traits, will almost certainly be no. Ask whether the offering has been evaluated by an independent body like ASRM, and if so, what they said.

If you walk away knowing only one thing about polygenic embryo screening today, it should probably be this: the science is interesting, the technology is real, and the gap between what it does and what it is sold as doing is wide enough to drive a truck through. The marketing is selling certainty. The science is selling odds, and the odds are smaller than the slogans imply.