A British study has found that polygenic risk scores, which estimate a person’s disease risk based on thousands or millions of common genetic variants, perform poorly in screening and prediction of common diseases such as heart disease.
It has been claimed that polygenic risk scores will transform the prediction and prevention of common diseases, and already, companies that sell polygenic risk score testing services have been established.
Polygenic risk score testing is also one of the aims of Britain’s nationwide Our Future Health project.
In the latest study, published in BMJ Medicine, the researchers looked at 926 polygenic risk scores for 310 diseases, and found that, on average, only 11% of individuals who develop disease are identified, while at the same time 5% of people who do not develop the disease test positive.
Unaffected people usually outnumber those affected which results in far more false than true positive predictions.
Lead author Professor Aroon Hingorani (University College London Institute of Cardiovascular Science) said: “Strong claims have been made about the potential of polygenic risk scores in medicine, but our study shows this is not justified.
“We found that, when held to the same standards as employed for other tests in medicine, polygenic risk scores performed poorly for prediction and screening across a range of common diseases.”
For the study, the researchers looked at data available in an open-access database, the Polygenic Score Catalogue, to determine what the detection rate and false positive rate of the scores would be if used in screening.
For breast cancer and coronary artery disease, the risk scores identified only 10% and 12% of eventual cases respectively, using a cut-off that resulted in 5% of unaffected individuals testing positive.
The researchers also investigated how polygenic risk scores would perform if used alongside conventional screening methods.
They found that if used alongside conventional risk factors, several thousand people would need to have a polygenic risk score done to guide statin prescriptions to prevent one additional heart attack or stroke.
The researchers noted that using age alone as a guide to statin prescription would be simpler and more effective at preventing heart attacks and strokes, without the need for genetic testing.
They also found that adding polygenic risk scores as first stage screening to determine who should be prioritised for mammography would miss most women who later develop breast cancer and generate many false positives, adding to the burden on healthcare systems.
Co-author Professor Sir Nicholas Wald (UCL Institute of Health Informatics) said: "It has been suggested that polygenic risk scores could be introduced early on to help prevent breast cancer and heart disease but, in the examples we looked at, we found that the scores contributed little, if any, health benefit while adding cost and complexity."
In the paper, the researchers suggest regulation of commercial genetic tests based on polygenic risk scores to "protect the public from unrealistic expectations and already stretched public health systems from becoming overburdened by the management of false positive results."
The researchers said consumers of commercial polygenic risk score tests should be informed of the detection rate and false positive rate of the polygenic risk scores as well as the absolute risk with and without a polygenic score result so they can better judge whether the test is useful.
Co-author Dr Jasmine Gratton (UCL Institute of Cardiovascular Science) said: “Polygenic risk scores seem attractive because genotyping is now inexpensive, the same for all diseases, and is performed only once because a person's genotype does not change. However, these features are irrelevant if the test is not useful.”
Wald said: “Our results build on evidence indicating that polygenic risk scores do not have a role in public health screening programmes.”
The researchers said the performance of polygenic risk scores was unlikely to change much, as the variants with the strongest effect had already been identified.
Polygenic risk scores should not be confused with genetic testing for certain single gene mutations such as BRCA1 and BRCA2, which have an important role in screening for breast and ovarian cancer.
Discovering variants that are associated with a higher risk of disease is still crucial for drug development, the team emphasised, as the variants encode proteins that can be targeted with drugs that would be useful for everyone, regardless of their genetic makeup.
Study details
Performance of polygenic risk scores in screening, prediction, and risk stratification: secondary analysis of data in the Polygenic Score Catalog
Aroon Hingorani, Jasmine Gratton, Chris Finan, Floriaan Schmidt, Riyaz Patel, Reecha Sofat, Valerie Kuan, Claudia Langenberg, Harry Hemingway, Joan Morris, Nicholas Wald.
Published in The BMJ Medicine on 2023
Abstract
Objective
To clarify the performance of polygenic risk scores in population screening, individual risk prediction, and population risk stratification.
Design
Secondary analysis of data in the Polygenic Score Catalog.
Setting
Polygenic Score Catalog, April 2022. Secondary analysis of 3915 performance metric estimates for 926 polygenic risk scores for 310 diseases to generate estimates of performance in population screening, individual risk, and population risk stratification.
Participants
Individuals contributing to the published studies in the Polygenic Score Catalog.
Main outcome measures
Detection rate for a 5% false positive rate (DR5) and the population odds of becoming affected given a positive result; individual odds of becoming affected for a person with a particular polygenic score; and odds of becoming affected for groups of individuals in different portions of a polygenic risk score distribution. Coronary artery disease and breast cancer were used as illustrative examples.
Results
For performance in population screening, median DR5 for all polygenic risk scores and all diseases studied was 11% (interquartile range 8-18%). Median DR5 was 12% (9-19%) for polygenic risk scores for coronary artery disease and 10% (9-12%) for breast cancer. The population odds of becoming affected given a positive results were 1:8 for coronary artery disease and 1:21 for breast cancer, with background 10 year odds of 1:19 and 1:41, respectively, which are typical for these diseases at age 50. For individual risk prediction, the corresponding 10 year odds of becoming affected for individuals aged 50 with a polygenic risk score at the 2.5th, 25th, 75th, and 97.5th centiles were 1:54, 1:29, 1:15, and 1:8 for coronary artery disease and 1:91, 1:56, 1:34, and 1:21 for breast cancer. In terms of population risk stratification, at age 50, the risk of coronary artery disease was divided into five groups, with 10 year odds of 1:41 and 1:11 for the lowest and highest quintile groups, respectively. The 10 year odds was 1:7 for the upper 2.5% of the polygenic risk score distribution for coronary artery disease, a group that contributed 7% of cases. The corresponding estimates for breast cancer were 1:72 and 1:26 for the lowest and highest quintile groups, and 1:19 for the upper 2.5% of the distribution, which contributed 6% of cases.
Conclusion
Polygenic risk scores performed poorly in population screening, individual risk prediction, and population risk stratification. Strong claims about the effect of polygenic risk scores on healthcare seem to be disproportionate to their performance.
See more from MedicalBrief archives:
Prostate cancer risk higher for black men – US study
One-off DNA test could predict heart attack risk in childhood
Study identifies 579 genetic locations linked to anti-social behaviour, addiction