Cardiovascular disease is the leading cause of death worldwide, according to a report from the World Health Organization, and moreover around 80 percent of premature heart attacks are considered preventable. The ability to accurately predict the risk of heart attack and cardiovascular death in an individual is therefore of significant value. Current clinical practice predicts the risk of heart attack through assessing a range of physiological parameters, such as blood pressure and cholesterol levels. Now improvements in genomics and transcriptomics have led to the development of genetic risk scores that when factored in to current risk estimation models may improve their accuracy. Predicting the likelihood of a heart attack however does not equate to predicting the risk of cardiovascular death. In a Genome Medicine study Greg Gibson from the Georgia Institute of Technology, USA, and colleagues deduce gene expression profiles associated with acute myocardial infarction and cardiovascular death. Here Gibson explains how their findings may improve current risk estimation models, and discusses the reality of their use in clinical practice.
Regarding current risk estimation models for coronary artery disease used in clinical practice, what factors are considered, and what are the limitations of these models?
The Framingham Ten Year Risk for Cardiovascular Disease considers age, gender, total and high density lipoprotein cholesterol, smoking status, and blood pressure (the NIH National Heart Lung and Blood Institute website hosts a convenient personal risk calculator). It is designed to predict the risk of having a heart attack, and has performed remarkably well for some time now, though most doctors recognise that there must be room for improvement. A big difference with our study is that we are predicting risk of a heart attack so adverse that it leads to death: the evidence suggests that these may be separable risks.
To what extent do you think the incorporation of genetic risk scores into these risk models will improve their accuracy?
I’d like to think they will, but those to date are targeted to cardiovascular disease in general, and in some cases risk of heart attack. They only explain a few percent of the variance in risk, but as we get samples of hundreds of thousands in the next few years, that will improve. My guess is that genetic risk scores, transcriptional risk scores, and biochemical ones will all add a few points to the Framingham score regarding risk of myocardial infarct. However I am not sure about the indicators of really adverse events.
Genetic risk scores have already been developed for coronary artery disease, however they are not predictive of adverse cardiovascular events. Why is this?
Current estimates are that as many as a half of all adults in America have sufficient heart disease that they are at elevated risk of heart attack relative to how they would be if they were in better shape. But heart attacks are by nature stochastic events, they occur when a plaque ruptures for example – when that happens, the consequences can be relatively benign, or they can lead to rapid complete occlusion of the vessel, which is more likely to be lethal, result in long-term damage, or otherwise cause hospitalisation that requires major surgery. I am not a cardiologist, but our study suggests that there is a high risk population who may be worth studying in greater depth to see what coronary features they have in common that predisposes them to these really adverse events.
What approach did you take in your study to discern whether genetic analysis could inform risk predictions of adverse cardiovascular events?
We actually set out to ask whether the differences that are observed between people having a heart attack or just experiencing incident coronary disease, are the same as those that distinguish people who have a heart attack at a young age, and those who experience disease relatively late in life. We did not see any differences in the blood that distinguish the early onset and later onset groups, but do have some evidence that the gene expression differences in the heart attack group may predict likelihood of a future myocardial infarct. Unfortunately, approaching ten percent of the sample died from an adverse event within three to four years after their visit to the Emory Cardiology clinic (directed by my colleague, Arshed Quyyumi who had the vision to archive blood samples for RNA analysis). At this point we were able to evaluate the additional risk of really adverse events.
What genetic differences did you identify that were suggestive of acute myocardial infarction? How did you determine whether these changes were causative or consequential?
The AMI (acute myocardial infarction) signal turns out to be largely related to neutrophilia, which was previously known, but also to downregulation of a particular arm of T-cell activity. A couple of years ago we showed that in healthy populations, there are nine or ten major axes of gene expression that involve hundreds to thousands of genes each, two of which are altered in the AMI individuals. It is difficult to tease apart the relative contributions of differences in cell abundance (say, neutrophil to T-cell ratio), and changes within these cell types, but we are fairly sure both contribute. One reason is that we performed a genetic analysis of the regulation of gene expression, so-called eQTL analysis. This showed that there is an excess of genes that are no longer under control of local regulatory polymorphisms during the myocardial infarct, whereas in healthy people the level of expression is a function of the genotype at that locus. This is one of the first demonstrations of a change in genetic regulation of gene expression in the disease state, a type of genotype-by-environment interaction.
Further analysis of gene expression profiles from acute myocardial infarcts revealed a subset of transcripts associated with significant risk of cardiovascular death. What further research is needed to determine whether these transcripts can be used to accurately predict risk of cardiovascular death?
We replicated the study in two phases conducted 18 months apart, but still have to recognise that the signature we describe is based on just over 30 people who have died. This clearly needs to be replicated in independent studies – there are a few prospective cardiology studies we know of, so hopefully the opportunity to confirm the findings will arise soon. Then after that, we need to think about mechanistic studies that establish what aspects of the profile are causal.
Do you foresee a future where routine clinical practice involves blood tests for biomarkers predictive of cardiovascular disease? What needs to happen for this to become a reality?
I actually think there is some weariness in the cardiology community toward biomarkers in general – they are not predictive in the sense that they are accurate enough to change people’s behaviour. So the main thing is finding out what motivates people, which in turn requires that there are concrete things that people can do to reduce their risk. If we can show that the profile can be reversed, presumably through exercise and diet in the main, then we would be a long way toward establishing the utility of predictive genetic tests such as this. If not, then at least we can work to ensure that the highest risk population have the support base and access to emergency care that they may need at any moment.