ACCEL | Multivariable Risk Models: Framingham, Lifetime Risk, and Beyond
It’s probably no surprise that patients substantially over- and underestimate risk. In a study that looked at how accurate patients with hypertension or diabetes were in assessing their risk of stroke or MI, investigators found that patients were more apt to overestimate their 10-year risk.1 On one hand, overestimating risk may make a patient more amenable to risk reduction efforts; on the other hand, it could produce a sense of fait accompli. The study authors concluded that patients showed inadequate perceptions of their absolute risk of cardiovascular events and physicians should thus provide greater information about absolute risk when offering preventive therapy.
So, patients are lousy at accurately assessing their own cardiovascular risk, but let’s face it: that’s not their job! It turns out, physicians over- and underestimate risk, too—with similar, if not greater, swings in off-the-cuff estimates than among patients themselves. In a study of 79 physicians employed at all levels of three university hospitals, participants were asked to do cardiovascular risk assessment in 12 primary prevention scenarios.2 Only 24% of physician risk estimates were accurate, and physicians overestimated absolute risk 32–90% of the time.
Since global risk assessment is a binary proposition—either patients qualify for drug therapy or they do not—is there need for ongoing risk assessment once a patient qualifies for pharmacotherapy? According to Donald M. Lloyd-Jones, MD, chair of preventive medicine at Northwestern University, global coronary heart disease (CHD) risk assessment information taken over time increases the accuracy of perceived risk, and probably increases the intent to start therapy. Studies with repeated risk information and repeated doses of counseling show small but significant reductions in predicted CHD risk over 10 years. Conversely, providing global risk information at only one point in time is ineffective.
In brief, when doctors are presented with risk, there is a greater tendency to prescribe lipid and blood pressure medications more often and more appropriately. Routine use of global risk scores leads to greater use of guideline-based therapy and modest improvements in intermediate outcomes with no harm identified.3
But Which One?
There are a more than a couple of risk scores to help assess patients. Some risk assessment tools rely on CHD while others try to assess cardiovascular disease (CVD) risk (TABLE). Should we care about CHD versus CVD as our endpoint? Dr. Lloyd-Jones argues that CVD captures more events of interest, and notes that women are at risk of stroke and heart failure (HF) before CHD. Also, CVD captures more underlying risk factors and more young people at risk.
For example, the Framingham Risk Score classifies younger men and a large proportion of women as low risk, despite substantial risk factor burden. Additionally, the restriction of predicted endpoints to hard CHD only (to the exclusion of stroke and HF) also may lead to risk estimates that are unjustifiably reassuring to patients with significant risk factor burden.
There are some challenges, too, when choosing a risk score. Some newer scores use revascularization rates as a risk variable, yet such rates vary widely across the United States, so the value of revascularization as an endpoint is weakened by geographic variations in use. Some new risk scores target the endpoint of cardiovascular death, which is very restrictive. Also, it is difficult to consistently define or measure HF across different populations. Dr. Lloyd-Jones recommends focusing on assessing risk for atherosclerotic CVD and assuming all older adults with hypertension are potentially at risk for HF.
Similarly, considering cerebrovascular events (prior stroke or transient ischemic attack) may be a widely applicable addition to risk assessment, particularly for women who have a higher lifetime risk of stroke than men.
Authors have argued for predicting longer-term risk of CVD,4 and Dr. Lloyd-Jones has reviewed the current status and future directions of CV risk assessment.5 He points out that the concept of identifying younger individuals at low short-term but high lifetime risk was validated by comparison with subclinical atherosclerosis imaging data. In the Coronary Artery Risk Development in Young Adults Study, 90% of participants aged 32–47 years had a 10-year predicted risk <10%. However, among this group, more than one-third had high predicted lifetime risk (39% or greater). Even at these younger ages, participants with low 10-year but high lifetime risk estimates had significantly greater burden of carotid artery intima-media thickness and coronary artery calcification and greater progression of subclinical atherosclerosis compared to participants with low 10-year and low lifetime predicted risk.
1. Frijling BD, et al. Patient Educ Couns. 2004;52:47-53.
2. Pignone M, et al. BMC Health Serv Res. 2003;3:13.
3. Sheridan SL, Crespo E. BMC Health Serv Res. 2008;8:60.
4. Pencina MJ, et al. Circulation. 2009;119;3078-84.
5. Lloyd-Jones DM. Circulation. 2010;121;1768-77.
To listen to an interview with Donald M. Lloyd-Jones, MD, about multivariable risk models, visit youtube.cswnews.org. The interview was conducted by Robert A. Vogel, MD.
< Back to Listings