The clinical applicability of polygenic scores in PGT

Polygenic scores can be of substantial interest in research, and at the population level. However, they have limitations when applied to 'predict' the health status of individuals (see BioNews 1302).

Heritability tells us how much of the variation we see in a trait, within a group of people, is explained by the genes of those people rather than by their environments. Polygenic scores often account for a modest fraction of the heritability of a trait (such as susceptibility to a complex disease), and this heritability will itself usually contribute up to about half of the various causally relevant 'influences'. We therefore cannot expect genetic testing of polygenic factors to have the same predictive power as genetic testing for a condition that is 'caused' by changes in a single gene of major effect.

We have to acknowledge that different polygenic assessments for the same condition will often give substantially different results, because such assessments cannot fully capture everything that's going on. Genes influence traits, but how they do so depends on how they interact with each other, the environment and what happens to an individual throughout their life. These further complexities cannot be straightforwardly captured and built into polygenic calculations.

Furthermore, differences between the circumstances of the original population(s) whose genetic data was used to develop the polygenic score, and the circumstances of the individual being tested – including secular cohort effects (for example, being born in 2026 rather than 1966) – will weaken any association between the effects of the 'polygenes'.

Heritability is often spoken of as if it were a fixed biological constant, but it also varies, depending on circumstances and history. Furthermore, if the individual being assessed is not drawn from the original population group (most polygenic scores are developed primarly from the genetic data of people with European ancestry), then the relevant associations will be weaker still.

Consequently, if there is any health benefit from polygenic score testing, this benefit will be diminished for people outside the original population group. Preexisting inequities may therefore be exacerbated (see BioNews 1231, 1232, 1293, 1297, 1300 and 1302).

One of the claimed benefits of polygenic scores is that they can provide useful information about lifestyle. However, such claims are grossly exaggerated – lifestyle advice is usually the same for everyone, and does not require any form of genetic assessment. There is no evidence that polygenic scores lead to sustained compliance with such advice, and there is great scope for unhelpful (indeed paradoxical) behavioural effects from polygenic score testing, through a sense of either invulnerability or fatalistic inevitability.

Given that polygenic scores have so many limitations in their clinical application, why would a professional seek to apply them to individuals? There are two plausible answers.

• Enthusiasm combined with a lack of understanding, trusting the marketing 'hype' about the benefits of the test.
• A cynical commitment to taking commercial advantage of people's curiosity... or their credulity.

Note that any polygenic scores reported will be probabilistic, so in practice they could never be definitively validated or falsified by events. Nor would there be much chance of the system learning from experience, given the likely long periods of time between a test and any outcomes. Even if an individual were to develop a particular condition, after being given a low score for their risk of developing that condition, this would not render the score 'incorrect'.

Now, turning to people going through fertility treatment – what about a health practitioner offering or suggesting preimplantation genetic testing for complex traits, known as PGT-P (and also sometimes referred to as 'polygenic embryo screening')?

PGT-P offers prospective parents an insight into their embryos' likelihood of developing complex traits or conditions, so that this information can be used to prioritise which embryo(s) to transfer. The range of possible scores for the offspring of a couple will be more constrained than the spread of possibilities found within a population, as the parents will only be heterozygous at a modest subset of the potentially relevant SNP loci found in the population (see BioNews 1130 and 1137).

Why would health professionals choose to offer PGT-P in this setting? There are – broadly – two possibilities, neither of them very flattering.

• Being genuinely enthusiastic, but without insight or understanding.
• Wishing to cash in on any parental concerns that can be stimulated by talk about 'doing the best for your baby', or 'choosing the best baby' (that is, embryo) 'from the batch' (or from the freezer).

This is why – when I was asked recently by the Guardian newspaper to comment on the news that some UK patients are now seeking PGT-P – I said that those who are offering PGT-P 'are dealing in murky science in an emotionally fraught context'.

I have no experience of offering PGT-P, so I have to resort to imagination – albeit informed by being a genetics practitioner and an observer and researcher of genetic counselling and fetal medicine consultations – to imagine what might be going on, if and when PGT-P is offered. I can imagine a practitioner with limited understanding of polygenic inheritance perhaps being caught up, enthusiastically, in offering this extra service to patients.

Patients will often already be paying substantial sums for IVF procedures. Why would they not want to run a few checks, to ensure that they take home 'the best baby possible'? Why allow chance to give them a second-rate, 'less-than-the-best' baby, when a few hundred more pounds or dollars could take care of it?

Imagine how prospective parents might feel, if the following 'offer' of just one more round of testing were made. 'Would you like to do all you can to ensure that your baby has (CHOOSE ONE, OR PERHAPS TWO, OF THE FOLLOWING OPTIONS) – the lowest risk of breast and bowel cancers, or the lowest risk of cardiovascular disease and Alzheimer's disease, or the highest IQ, or the greatest expected height, of the available embryos?'

At this stage, before a pregnancy has been established, the prospective parents' resistance to marketing ploys will be minimal. There is no baby there yet, to command the parents' loyalty. In an established pregnancy, patients who wish to decline the offer of more interventions and more testing will feel enabled to do so by an appeal to their love for their baby. But, at this point, there are only a few balls of cells in the petri dish or the freezer, and patients are being asked to choose one from among several.

If 'potential IQ' is what we hope to maximise through this process of selection, what else might we be selecting for at the same time, but unwittingly? We simply don't know what other personality traits or (in)capacities will also be selected. And in any case, how likely is it that the IQ of the resulting child will be greater than it would have been, if the embryo transferred had been selected on the simpler basis of the 'quality' of the embryos as they appear under the microscope? There is no way of ever knowing. If we 'optimise' the chances along one dimension of risk or quality, what tradeoff may be needed along some of the other dimensions?

We know so little about what we are doing (in terms of the limitations of selection in this way and the unwitting selection of other traits), and the potential for both emotional abuse and financial exploitation is so substantial, that I am unable to see how PGT-P can be offered – in good faith – by anyone who understands polygenic inheritance and what the testing process entails.

---

Professor Angus Clarke will be speaking at this week's PET Annual Conference – What Does Genomics Mean for Fertility Treatment? – at a session entitled 'Polygenic Risk, Polygenic Scores, Polygenic Indices: What Are They? What Should Be Done With Them?'.