The Amgen-sponsored CV Academy 2021 invited Professor Rod Jackson, University of Auckland, New Zealand, to discuss his thoughts on CVD risk prediction equations in the primary care setting. Here we summarise his 4 key points.
In 2021, it’s unacceptable to make decisions about CVD preventive medications without first assessing CVD risk using a valid equation
Prof Jackson presented a graphical representation of absolute cardiovascular disease (CVD) risk over 5 years.1 He noted that impact of individual risk factors is largely determined by the presence or absence of other risk factors, and that “the benefits of CVD preventive medications are proportionate to pre-treatment CVD risk”. Prof Jackson explained, “Data from individual participant meta-analysis of 27 trials showed the predicted 5-year benefits of low-density-lipoprotein (LDL) cholesterol reductions with statin treatment at different levels of risk.2 The bottom line is that the benefits of blood pressure lowering, lipid lowering, and aspirin are all directly proportionate to pre-treatment cardiovascular risk.”
Currently, there are only 2 valid CVD risk prediction equations applicable to primary care patients available worldwide
When considering valid CVD risk prediction equations Prof Jackson stated, “A large cohort representative of the target population is required to be followed for 5–10 years to measure all risk predictors and major CVD outcomes.” Prof Jackson noted only 2 current studies are capable of generating accurate equations for primary care worldwide:
- QRISK: a retrospectively designed prospective cohort study in the UK with 7.89 million patients, aged 25–84 years, followed on average for 5–6 years with 363,565 CVD events identified.3 “They were able to do this by linking primary care risk factor data to national hospitalisations and deaths,” Prof Jackson explained.
- PREDICT: a prospectively designed prospective cohort study in New Zealand, with 401,752 patients, followed for about 5 years with 15,386 CVD events identified.4 “About a third of New Zealand GPs were given software for assessing cardiovascular risk using the Framingham equation and then using that data linked primary care data to national hospitalisations and deaths,” he said.
CVD risk prediction equations need to be regularly updated
“It’s important the equations are updated regularly due to two key reasons,” explained Prof Jackson. “Firstly, there are changes in event rates, like mortality trends for coronary heart disease, which peaked in the 1960s and then have been declining at a rate of 2–3% a year. The Framingham equation was developed in 1970s and 1980s and now coronary heart disease event rates and stroke event rates are way less than half of those rates,” he explained.
To further highlight how the Framingham equation is outdated, Prof Jackson showed a calibration graph showing that the Framingham equation overpredicts CVD risk by over 50% when applied to the PREDICT cohort. Prof Jackson also discussed changes in practice that can affect equations and showed the impact of universal diabetes screening in New Zealand as an example with the NZDC equation (2000-2006) vs PREDICT equation (2004-2016).6
Methods to add predictors to equations and metrics for assessing their impact are still poorly understood and a missed opportunity
“The ability to add new predictors and, indeed, assess just how good equations are is still in its infancy. We really have a lot of work to do in assessing how good these equations are,” said Prof Jackson. He discussed some variables from the PREDICT cohort, including ethnicity, atrial fibrillation, and blood pressure lowering medication and the impact on C statistic (the main performance measure used in cardiovascular risk protection equations).7 He noted that he doesn’t think coronary calcium is added to risk prediction scores because when looking at the C statistic it doesn’t do much to the performance.8
“We are beginning to look at new ways of predicting and patching in new risk factors, but this is a science in its infancy,”9 he said.
Prof Jackson concluded by emphasising how having a ‘license to link’ primary care records to hospitalisations and deaths would significantly improve CVD risk management in primary care and expand the science of precision medicine.
- Jackson et al. Lancet 2005;365:434-41.
- Lancet CTTC 2012; 380: 581–90.
- Hippisley-Cox J, et al. BMJ2017;357:j2099.
- Pylypchuk R, et al. Lancet. 2018; 391:1897-907.
- Pylypchup, et al. Lancet 2021; 397:2264-74
- Pylypchup, et al. Lancet 2018; 391:1897-907
- Detrano, et al. NEJM 2008;358:1336-45.
- Matsushita K, et al. Circ J 2019;83:1876-1882.