PREDICT-1° CVD Risk Calculator, Based on Present-Day Patients, Beats ACC/AHA ‘Pooled Cohort’ Equations

New Zealand researchers say their CV risk tool more accurately reflects their socially and ethnically diverse population in primary care clinics today.

PREDICT-1° CVD Risk Calculator, Based on Present-Day Patients, Beats ACC/AHA ‘Pooled Cohort’ Equations

A new calculator for cardiovascular risk prediction developed from a large cohort of primary care patients in New Zealand suggests that risk equations based on earlier cohorts—now decades old—may substantially overestimate patient risk, a new study shows.  

Compared with the 2013 American College of Cardiology/American Heart Association (ACC/AHA) pooled cohort equation to calculate CVD risk, the New Zealand risk-prediction equation “performed better in predicting total cardiovascular disease events than the pooled cohort equations performed in predicting hard atherosclerotic cardiovascular disease events,” according to a study published online May 4, 2018, in the Lancet.

Overall, the ACC/AHA pooled cohort equation overestimated the incidence of atherosclerotic CVD events in New Zealand primary-prevention patients by approximately 40% in men and 60% of women.

“The main problem with most current equations is that they are based on old cohort studies that were conducted when CVD risk was much higher,” senior investigator Rod Jackson, PhD (University of Auckland, New Zealand), told TCTMD in an email. “Also, most previous studies do not include a sufficient range of ethnic groups and do not consider socioeconomic status.”

New Zealand was the first country in the world to develop national CVD risk factor management guidelines based on predicted rather than individual CVD risk factors, such as blood pressure or cholesterol levels, said Jackson. Historically, they have used the Framingham Heart Study prediction equations to make treatment decisions because there were no relevant New Zealand studies or risk-prediction models. Over time, however, “we were increasingly questioning whether the Framingham equations, based on a study of Americans who had their CVD risk factors assessed in the 1970s, were still relevant to the 21st century multiethnic and socioeconomically diverse New Zealand population,” said Jackson.  

Given those concerns, the researchers recruited patients in New Zealand from the PREDICT study to develop a new risk equation relevant to patients in primary care. Their study then compared the performance of the equation, dubbed PREDICT-1°, to the pooled cohort equations used by the ACC/AHA in the guidelines for the assessment of cardiovascular risk.

Measures of Socioeconomic Deprivation 

PREDICT is an ongoing, prospectively designed cohort study in New Zealand that includes participants automatically recruited from primary care practices when physicians complete standardized CVD risk assessments. Approximately 95% of New Zealanders are enrolled in primary health organizations as part of the country’s national healthcare service. To TCTMD, Jackson noted that PREDICT has near-complete follow-up of all hospitalizations and deaths because each person in New Zealand has a unique health identifier, which allowed the researchers to link data throughout the national public health system.

The variables in the new PREDICT-1° risk equation include those used for calculating CVD risk in the modified Framingham equations (sex, age, ethnicity, family history, smoking status, diabetes status, systolic blood pressure, and ratio of total cholesterol to HDL cholesterol) but add atrial fibrillation and medication use (antihypertensive, lipid-lowering, and antithrombotic drugs) in the 6 months before assessment. The PREDICT model also includes a modified index of social deprivation, which is a general marker of income, employment status, social support, and living space, among other variables.

The study population in the PREDICT-1° model included 401,752 primary care patients followed for a mean of 4.2 years. During this time, 4% of patients had a first major CVD event, with nonfatal MI being the most common outcome.

Using the PREDICT-1° equation, the mean estimated 5-year risk of total CVD was 3.2% in men and 4.6% in women. Overall, there was “excellent calibration” between the predicted versus observed 5-year risk of total CVD events. The risk of underestimation or overestimation did not exceed 0.5% in any of the predicted risk deciles. In contrast, the original ACC/AHA pooled cohort equation “significantly overpredicted the observed 5-year risk of hard atherosclerotic cardiovascular disease in the top seven deciles of predicted risk in both men and women,” report investigators.

The investigators conclude that measures reflective of health inequity, such as socioeconomic deprivation and self-identified ethnicity (if relevant) are important to add to the pooled-cohort equation for CVD risk prediction. Adding these variables can “help identify high-risk patient groups who might otherwise be undertreated,” they note.

Jackson said that the 2013 ACC/AHA guidelines for the assessment of cardiovascular risk—and their application in the ACC/AHA cholesterol and blood pressure guidelines—were the most important international developments in cardiovascular disease risk management of the last decade. However, when they are updated, our study makes clear that the pooled-cohort equations should be updated, too, he said. They need to be calibrated in the populations they are applied to, and they would benefit from the addition of new predictors.”

In an editorial accompanying the study, Johanna Damen, PhD, Lotty Hooft, PhD, and Karel Moons, PhD, (University Medical Center Utrecht, the Netherlands), applaud the use of a large contemporary data set to “update, validate, and report” a prediction model of CVD. The biggest problem with existing risk prediction models is that they typically overestimate risk, a finding that was confirmed in the present study.

“Whether their model leads to less overtreatment or undertreatment remains to be seen,” write the editorialists. “A more accurate prediction model is, unfortunately, no guarantee of improved patient outcomes.”

Sources
Disclosures
  • Jackson, Damen, Hooft, and Moons report no relevant conflicts of interest.

We Recommend

Comments