What Is Polygenic Embryo Screening?

Polygenic embryo screening, sometimes called PGT-P, can sound more certain than it is. In the counseling room, the first step is usually to slow the language down.

PGT-P does not diagnose an embryo. It does not tell you what a future child will or will not develop. It gives a risk estimate for selected complex conditions, based on many genetic markers and the way those markers have behaved in population studies.

This distinction is not just word choice. In its 2026 Ethics Committee opinion, the American Society for Reproductive Medicine describes PGT-P as a nascent, unproven technology and says it is not recommended for clinical use at this time.

What "polygenic" means

Some genetic conditions are mainly about one gene. In those situations, the question may be whether a specific variant was inherited. Polygenic risk is different. It is about many genetic markers, each usually contributing a small amount to the model.

A simple way to think about it is a dimmer panel, not a single light switch. No one marker turns a common condition on or off. The model adds many small signals together, then estimates whether the inherited pattern looks higher or lower risk compared with a reference group or sibling embryos.

Why a "high" score is not destiny

A polygenic score is the sum of thousands of tiny inherited effects. A classic way to picture how many small, random factors add up is a Galton board — a peg-studded box where balls bounce left or right at each peg and pile up into a bell curve at the bottom. Most balls land near the middle; only a few reach the edges.

Your genes place you somewhere on that curve. But where you land is a predisposition, not a prediction. Whether a condition actually appears also depends on environment, chance, and factors no score measures. Plenty of people toward the higher end never develop the condition, and plenty near the middle do. A polygenic result nudges the odds; it does not decide them.

What PGT-P is trying to estimate

Many common conditions are influenced by a long list of factors: inherited genetics, age, environment, access to care, lifestyle, exposures, and chance. Type 2 diabetes, coronary artery disease, some cancers, and some psychiatric conditions are examples.

A polygenic score looks at many inherited genetic markers at once. The result is then translated into a statistical risk estimate for a specific condition. The estimate may be shown as a percentile, a relative comparison, an absolute risk estimate, or some combination of these.

This is different from looking for one known family variant. For common conditions, the genetics usually acts less like a single on/off switch and more like many small nudges that can add up in different directions.

For embryos from the same IVF cycle, the most careful framing is often relative inherited risk. That means one embryo may have a lower or higher modeled inherited risk estimate than another sibling embryo for the condition being reported.

The heritability gap: why even good scores are incomplete

Here is a piece most reports leave out. Twin and family studies suggest many common conditions are strongly heritable — for a lot of traits, roughly half or more of the variation between people traces back to inherited genetics (Polderman et al., Nature Genetics, 2015). That makes it sound as if genetics should predict a great deal. But the polygenic scores available today capture only a fraction of that inherited signal.

Height is the easy case: it is roughly 80% heritable, and current scores capture something like half of that. For most diseases the gap is far wider — scores for conditions like type 2 diabetes or psychiatric conditions explain only a small slice of the inherited variation, even when those conditions are highly heritable. "Highly heritable" does not mean "accurately predictable today."

The reason is in how the scores are built. They add up many common variants, each with a small effect, and largely assume those effects act independently. They are not designed to capture rarer variants, interactions between genes, or how the genome is regulated. There is real signal in a polygenic score — but it sits well below the ceiling that heritability implies.

Why the number of embryos matters

PGT-P can only compare embryos that exist, are tested, and are still clinically eligible for transfer. If there is one embryo, there is no sibling embryo ranking within that cycle. If there are several, there may be more genetic differences to compare, but the differences may still be small or uncertain.

Modeling studies suggest that possible risk reduction depends strongly on embryo count, the condition, the score's accuracy, and how the information is used. The eLife modeling paper by Lencz and colleagues found that excluding only very high-scoring embryos usually gives limited benefit, while choosing the embryo with the lowest score can produce larger modeled relative reductions when there are enough viable embryos. A Human Reproduction Update review emphasizes that these estimates come from statistical models, simulations, and sibling analyses, each with important assumptions.

In plain language: more embryos may create more room for comparison. It does not make the result certain. The transfer conversation still has to include embryology, PGT-A or PGT-M results when relevant, reproductive history, medical judgment, and family values.

A cautious way to read a report

A careful report should avoid phrases that sound like certainty. A more appropriate sentence might be: "This embryo has a lower modeled inherited risk estimate for this condition than the other tested embryos in this group."

That phrasing is a little longer, but it protects the meaning. It keeps the focus on the condition being reported, the comparison group, and the uncertainty around the model.

It is also reasonable to ask for natural frequencies whenever possible. The CDC Clear Communication Index notes that people generally understand natural frequencies and absolute risk more easily than relative risk alone. A report that says "30% lower risk" should also help you understand what that might mean in absolute terms, such as an estimated change from 10 in 100 to 7 in 100, if the model can support that conversion.

A question worth asking: does the score pass the sibling test?

Embryos from the same two parents are siblings. So the question that matters for embryo selection is not whether a score can tell strangers apart across a whole population — it is whether it can tell apart siblings, who already share much of the same genetic background.

This is more than a technicality. A score can look impressive on population data for reasons that have little to do with the genetics being inherited — differences in ancestry, patterns in who tends to have children together, and the fact that parents' genes also shape the environment a child grows up in. A score can pass standard tests across a population and still fail to predict differences within a family. Because embryos are siblings, within-family validation is the harder, more honest standard, and it is fair to ask any program whether its scores have been tested that way.

Questions to ask before using PGT-P information

These are the kinds of questions that can make the clinic conversation more concrete:

1. Is this result being used for research, counseling context, or a clinical recommendation?
2. Was the score validated in people with ancestry similar to our family?
3. Does the report show both relative risk and absolute risk, or only a percentile?
4. How much uncertainty is shown around each estimate?
5. How will this information be weighed against embryo morphology, PGT-A, PGT-M, age, medical history, and our priorities?
6. Who will review the report with us, and what are they qualified to counsel on?

The bottom line

PGT-P is best understood as a nascent, unproven way to estimate inherited risk for selected complex conditions. If you are considering it, ask what the result can reasonably add to your specific IVF decision, and what it cannot answer.