Why should two individuals with exactly the same risk factor profile for developing heart attacks or strokes be treated differently based purely on their race? This is one of the questions being asked in a new study by TU/e statistician Edwin van den Heuvel and Vasan Ramachandran, Professor of Medicine at Boston University School of Medicine. Together they investigated the potential impact of risk profiles on treatment decisions for cardiovascular diseases in the United States. They conclude that black people run the risk of being stereotyped as being at high risk purely based on the color of their skin. The results of the study have been published in Lancet Digital Health.
The American Heart Association/American College of Cardiology have formulated and endorsed so-called equations that can be used to predict the risk (probability or chance) that a person will develop heart attacks or stroke over the next 10 years. Physicians can plug in their patients' values for seven risk factors (age, sex, race, values of blood pressure, cholesterol (good and bad components), diabetes, and smoking status) to generate this 10-year probability of developing heart or brain attacks.
Race as proxy?
In the current form of the prediction equations, blacks and whites with exactly the same risk values on all other factors except race will have different probabilities of developing heart attacks and strokes. In these situations, doctors may treat their black and white patients differently, for instance by prescribing them statins, even when they have identical risk factors purely because of their race.
According to the authors, the use of race in risk profiles reinforces the concept of race - a social and political construct - as biologically meaningful, whereas it is more likely to be a proxy for actual biological factors, which may not be fully known.
"If other factors (instead of race) determine the risk differences, then the prediction equations should incorporate those factors that cause the differences in predicted risk between the races, rather than race itself. If we do not change our prediction strategy, there is a risk of labeling (stereotyping) black people as high risk purely based on the color of their skin," explains Vasan Ramachandran.
Calculating the risk factor
Van den Heuvel and Ramachandran examined 50,000 theoretically possible risk factor combinations using the risk factors noted above. They asked, if black and white patients had exactly the same (identical) risk factor combinations, by how much does the probability of heart and brain attacks predicted by the equations diverge so as to result in different treatment decisions across the two race groups. This analysis was done in men and women separately.
They observed that for 20 percent of the risk factor combinations in men and 22 percent of the risk factor combinations in women, black-white differences in risk predicted by these equations can result in different treatment decisions. For example, they found more often blacks would be prescribed a statin because they are deemed to be at higher risk. The difference in predicted risk (blacks vs. whites with identical risk factors) can be as large as 22.8 percentage points for men and 26.8 percentage points for women. According to the authors, these differences seem "biologically implausible" based on race alone.
Finding the actual causes
The researchers believe that by not treating the actual factors causing these race-related risk differences, physicians are at risk of medically treating the incorrect factor in the hope of lowering risk of heart attacks and strokes. "Since the equations are derived from historical cohort data, the black-white differences in predicted risk probabilities may reflect underlying race-related differences in health care access, structural racism or social determinants of health," says Ramachandran.
According to Van den Heuvel, the use of prediction equations to guide medical treatment should be based on causal factors only. "Furthermore, more research is needed to be able to determine if such causal prediction equations remain accurate after those at high risk are treated. In other words, we should investigate whether we can use the same prediction equations when risk factors are altered with interventions."
