NEW YORK — By applying machine learning to genetic data from different sources, an international research team has developed a genomic risk score (GRS) for ischemic stroke that performs as well or better than several established risk factors.
The findings, which appeared in Nature Communications last week, also suggest that current guidelines for stroke prevention may be insufficient for high-score individuals, and that patients could benefit from the inclusion of genetic information into risk management strategies.
GRSs have the potential to infer the risk of disease from birth, allowing for the initiation of preventative strategies before conventional risk factors appear. However, their predictive power for stroke has been limited due to a lack of genetic data available for the condition and the well-known heterogeneity of the stroke phenotype, the study's authors wrote.
To overcome this, investigators from the University of Cambridge, the Baker Heart and Diabetes Institute, and Ludwig Maximilian University used genome-wide association study summary statistics for 19 stroke and stroke-related traits to establish distinct GRSs for each trait. From these, they constructed a single metaGRS that was validated using data from the UK Biobank.
The metaGRS not only doubled the effect size of a previously published GRS, identifying a subset of individuals at monogenic levels of risk, but also showed similar or higher predictive power that established risk factors including family history, blood pressure, body mass index, and smoking.
Anticipating their metaGRS's potential use in early screening, the scientists estimated the reductions needed in modifiable risk factors for people at different levels of genomic risk. For those with the highest scores, they determined that achieving risk factor levels recommended by current guidelines may not be enough to mitigate risk.
"Previous research has demonstrated that intervening on modifiable risk factors can compensate for increased genetic risk of disease," they wrote. "However, these analyses relied on simply counting the number of elevated risk factors, which does not account for the differences in effect size between various risk factors. Our approach was flexible in that various combinations of risk factor reductions can lead to the same outcome in terms of risk."
The investigators proposed that their metaGRS could be used for early stroke prevention by identifying those at the highest genomic risk early in life for intensive lifestyle modifications. Later on when established risk factors are available, these can be combined with the metaGRS to give a more accurate view of a person's risk of incident stroke.
"It is time to contemplate whether future guidelines on primary and secondary stroke prevention should integrate genetic information when defining treatment goals for high-risk individuals," the team wrote. "Ultimately, the practical implications of these results to stroke risk screening in the general population will require public health modelling, taking into account what is considered 'high risk' of stroke in the context of each country and health system, and the efficacy of interventions or treatments that are available for risk reduction."