NEW YORK — Polygenic risk scores may provide people with a glimpse into their risk of developing a range of common diseases, but a handful of studies have suggested people may have difficulties understanding what their results mean.
PRS are largely research tools, though they have been making some headway into the clinic, particularly for predicting heart disease or breast cancer risk. But even as the clinical utility of PRSs continue to be debated, there are already a handful of services where someone could go to upload their data from 23andMe, Ancestry, or another direct-to-consumer genetic testing service to calculate their risk of developing a range of diseases or conditions.
One such service was Impute.me, which was acquired in July by the New York-based startup Nucleus Genomics. Nucleus plans to offer whole-genome sequencing and combined rare variant and PRS analysis to consumers.
In an intertwined pair of studies drawing on users of Impute.me, Jehannine Austin, a genetic counselor and professor of medical genetics at the University of British Columbia, found that not only did many users not fully understand the information their PRS provided — likely due in part to how the information was presented — but that those results could cause some individuals distress, occasionally to the degree of post-traumatic stress disorder.
"We found that 5 percent of people were experiencing reactions that [suggested] that maybe they were struggling quite hard with some of this stuff," Austin said.
Misinterpretation of results
For their studies, Austin and their colleagues worked with Impute.me, which at the time was a nonprofit tool. They first examined what motivated people to seek out third-party polygenic risk score analysis. Through a questionnaire, the vast majority of Impute.me respondents, 98.2 percent, said they were motivated by general curiosity, as the researchers reported in the European Journal of Human Genetics in 2021.
But that answer was not the whole story, according to the results of a follow-up study published this year, also in EJHG. in which they used a qualitative, interview-based approach. With these more open-ended questions, the third-party PRS service users provided more detail, with many saying their motivation stemmed from unsatisfactory interactions with medical professionals, such as having their concerns dismissed.
"This was a way for them to try to advocate for themselves in terms of their health," Austin said.
But it was also not clear whether the Impute.me users — who could broadly be described as early adopters — always understood the PRS results they were given. In the first study, Austin and their colleagues found that nearly a third of users misclassified one or more statements describing the results. For instance, a number of users agreed with the statement their results would definitively tell them whether or not they would develop a health condition. Further, more than 60 percent of users misinterpreted an example PRS result.
This misinterpretation of results, Austin noted, is likely in part due to how the information was presented. According to Austin, users of the site were presented with a bell curve graph that showed them the population distribution of the PRS and a line noting where along that graph their PRS fell.
"From the perspective of somebody who holds a Ph.D. in human genetics and somebody who's trained clinically as a genetic counselor, the first time I looked at one of these outputs, I sat there staring at it, going: 'So what's the X axis? What's the Y axis?'" Austin said. "It's just not entirely intuitive."
Misunderstanding of PRS results is not limited to this self-selected sliver of people who seek such scores from third-party sources. A recent study appearing in Genome Medicine examined this understanding of PRS reports in a different population: primary care providers (PCPs) and patients.
For this analysis, PCPs and patients interpreted mock PRS reports as part of the larger Electronic Medical Records and Genomics (eMERGE) Network study. These reports were based on ones previously in use — but, as first author Anna Lewis, a bioethicist at Harvard University, noted, unlike reports for monogenic diseases, what PRS reports include can vary widely — and were adapted for data to which the eMERGE trial would have access. The reports presented the results as either binary — "high risk" or "not identified as at high risk" — or continuous — with a percentile, odds ratio, and bell curve.
In this analysis, patients generally understood the key takeaway that the report indicated an increased risk of a disease and that steps could be taken to mitigate that risk, Lewis said. Patients in the study were provided information on a "disease X" rather than a specific one.
Patients also struggled to understand the meaning of the information provided. In particular, Lewis noted the classic misinterpretation of the percentile as the percent risk was common in her sample. This error was not limited to patients but was also made by some PCPs in the study.
"You say you're at the 98th percentile. You explain what that means, and people still say, 'Oh, this means a 98 percent chance of developing the disease,' which is very, very far from what the actual absolute risk would be," Lewis said. "It's very discouraging to see that kind of misinterpretation."
Response to results
For some of the Impute.me users, getting a PRS was empowering, but for others it could be distressing, Austin and their colleagues found. Some 60 percent of users in their 2021 study had a negative reaction to their PRS results.
Using two validated scales — the "Feelings About Genomic Testing Results" questionnaire and the Impact of Event Scale-Revised (IES-R) — the researchers assessed users' psychological reactions to their results. Of the participants who had a negative reaction, 14.4 percent said they were upset, anxious, and/or sad a good or great deal of the time. More than 5 percent had scores on the IES-R scale indicative of clinically significant distress and potentially post-traumatic stress disorder.
Likewise, in their qualitative follow-up 2022 EJHG paper, Austin and colleagues investigated how users reacted to PRSs they were not anticipating, to find that they responded with shock, confusion, and anxiety.
"People were experiencing distress when they got results that were unexpected in some way, that didn't match with their family history or their personal health," Austin said.
The two studies together further suggested that some individuals may be misinterpreting their results, which in turn leads them to a negative reaction. Indeed, in the earlier 2021 study, Austin and their colleagues traced factors influencing users' negative reactions, and they found a link between understanding of and reaction to PRS scores: Individuals who incorrectly answered some of the survey questions more frequently had negative reactions to their scores.
"The less people understand, the more anxiety is induced," study coauthor Lasse Folkersen, who was at Impute.me and now is CSO at Nucleus, said. "Does it affect me that, basically, we found that we induced PTSD in people? Yes, it does," he added.
How long these bad reactions last, though, is not yet clear. Harvard's Lewis noted that previous studies looking into the receipt of genetic testing results more broadly found that while people could be troubled when receiving a new result, that distress was typically transient.
"It's before you integrate it into your general view of yourself," Lewis said, adding that she suspected that if Austin and their colleagues redid their studies at a later time point, they would not find such a strong effect.
Indeed, Austin and their colleagues are now examining whether these negative reactions are transient or more long lasting.
Result presentation
Still, these studies suggest that the risk conveyed by PRS could be presented in a more understandable fashion. Risk, Austin noted, can be a difficult concept overall for people to grasp, and relative risk in particular. Instead, people have a better feel for the meaning of absolute risk. Patients in Lewis' study similarly expressed a preference for PRS reports to include information on absolute risk.
"Relative risk … doesn't actually tell you that much. Because at the end of the day, these [PRSs] are going to be useful insofar as they prompt action of some form or other," Lewis said. "And action is tied to absolute risk, for the most part."
Austin noted that PRS results could be better conveyed using improved visual aids. For instance, reports could include person-shaped icons representing 100 people and the portion of affected people could then be shaded in a different color. "This is absolute risk representation, and there's really good evidence that people understand this way of doing it best," they added. "And then if you also provide that in the context of what we know about risk for the condition in the general population, it allows you to make a much better appraisal."
Whether such changes in how data is presented translates to better understanding is something Austin and their team are working on. They worked with Impute.me prior to its purchase to change how the information is presented and then repeated the study to see whether better information leads to better understanding. The researchers have not yet finished the data analysis to know the answer to that question. "But the point of it is to see if we provide better quality information, do the results change?" they asked. "Do we have less bad outcomes, essentially?"
Nucleus Genomics, which acquired Impute.me, is also learning from these studies. According to CEO Kian Sadeghi, Nucleus is investing in communication, product design, and user interface and experience. The company, whose offering plans go beyond the PRSs of Impute.me to include wider whole-genome analyses, is also assembling a network of clinicians and genetic counselors who may provide pre- and post-test counseling to consumers.
"Let's actually improve [users' numeracy] so when they get back the results, they can understand as much as possible and then, importantly, act on them," Sadeghi said. "Because we don't want someone to be in a stress [state]."