Research on the Behavioral Impact of Polygenic Risk Scores: The Train Has Already Left the Station!

There has been a lot of discussion recently about the new generation of polygenic risk scores (information about a person’s disease risk based on many dozens, hundreds, thousands, or even millions of common DNA variants in their genome), and whether these new-and-improved genetic risk scores are going to turn out to be useful for disease prediction, prevention, and early detection.

In many ways, this is not a new discussion. For over two decades, questions have been asked about whether genetic information about a person’s disease risk might prove to be a way to empower people to take control of their health, and motivate them to engage in risk-reducing health behavior changes such as eating less, exercising more, or quitting smoking. There have also been concerns that genetic risk information might unnecessarily worry people, be misunderstood and cause confusion, or even that low risk results might make people falsely reassured and think it’s safe for them to eat and drink whatever and however much they want.

Several meta-analyses of the early studies on this topic have made clear that neither these hopes nor concerns have been borne out based on early evidence. Genetic tests for common DNA variants which only shift the risk needle by a very small amount alone have no ‘magical’ power to suddenly motivate people to engage in lifestyle changes. But is whether genetic risk information alone directly motivates people to change ‘lifestyle’ health behaviors (it does not) the only question to ask? Certainly, genome-wide association studies (GWAS) continue to be conducted with larger populations, and many groups are now working to develop polygenic risk scores and algorithms that use both genetic and non-genetic risk factors to improve identification of people at very high risk of disease. Also, ready or not, consumer genetic testing is a booming enterprise. Currently, millions of people have accessed direct to consumer genetic tests, including emerging genetic risk scores, both for health and genealogical purposes.

There are several sets of questions that need to be addressed of any genetic test, including genetic risk scores. We can no longer afford to answer these questions sequentially as the genomic train has already left the station!

(1) The first set of questions is about the analytic validity of the polygenic risk scores: analytic validity refers to whether the measured genetic information is accurate, and this is still work in progress. In spite of major progress in sequencing and genotyping platforms, there is a lot of concern about the validity of risk scores in non-white populations. As a result of underrepresentation of minority populations in genetic studies, several recent commentaries have raised alarm about the amount of errors introduced in inferring genetic information to subpopulations, thus impacting the analytic performance of such scores in different populations.

(2) The second set of questions is about the clinical validity of the risk information: does using polygenic risk scores alongside non-genetic (conventional) risk factors improve identification of people at very high risk of cancer, heart disease, or type 2 diabetes? There are many groups working hard to address this, such as the Cancer Research UK –funded CanRisk group in Cambridge UK. Given it is very clear that genetic factors do play a role in chronic disease etiology, then we might reasonably assume that the answer to this question is soon going to be ‘yes’, and there is some evidence that this is already the case, particularly in certain areas such as breast cancer.

(3) The third set of questions is relevant to clinical utility: does genetic risk information actually improve health outcomes? If so, do the benefits outweigh any harms? And for whom? Communication, behavior, and outcome research are all important here. We need to carefully assess whether risk stratified approaches to disease prevention and early detection using this genetic and non-genetic information to find high-risk people lead to changes in clinician and patient behaviors that ultimately have the potential to lead to improved health outcomes. Does risk stratification using genetic and non-genetic information change clinicians’ recommendations? How will all of this affect the patient receiving the information? Do they act positively on it in ways that do reduce their risk? What factors influence whether and which people have adaptive or maladaptive psychological and behavioral responses? If multi-disease risk assessments that include genetic and non-genetic info start to be offered, at what age would the offer be most effective? Birth? 18yrs? 30yrs? 50yrs? What are the relative benefits and harms of risk assessment at each of these life stages?

Embedded in the evaluation of clinical utility is clearly an important set of questions about how we communicate this complex genetic and non-genetic information to clinicians and patients so they will understand and facilitate shared and informed decision-making. This requires careful research and evidence in real-world clinical settings, and most likely the development of communication aids, electronic health record alerts, and decision support tools that are co-developed from the start with clinicians and patients. Crucially, how can we deliver the information in ways that maximally facilitate equitable access and benefit for all, not only those with the most resources?

The only way these questions will be answered is if we do the research on delivery, acceptability, communication, behavioral, psychological, and health impact alongside genetic discovery studies. We also need the development of population-specific risk algorithms and decision support tools for providers and the general public.

We cannot wait to have the “perfect” risk scores to do this research. Waiting means that the crucial implementation science and communication and behavioral and social research will lag behind discovery science by years if not decades. We should be doing the research simultaneously, and prioritizing both. And we should make sure we are asking the right questions when doing so. As the genomic train has already left the station, behavioral, communication, and social scientists have an important responsibility to make sure that it does not derail on its way to improved health benefits for all.