SCORE2 algorithm improves 10-year CVD risk prediction in European populations
SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe
Introduction and methods
The ESC recommends use of risk prediction models, such as SCORE (Systemic Coronary Risk Evaluation), to identify people at high risk of CVD who would benefit most from preventive measures [1-3]. However, there are some limitations to the SCORE model. SCORE was developed from cohort data before 1986 and has not been updated with contemporary CVD rates. Furthermore, SCORE only estimates fatal CVD outcomes, thereby underestimating total CVD burden, which has shifted towards non-fatal outcomes, especially for younger individuals . Last, SCORE does not allow for considerable risk variations across countries from the same risk region.
Also, other risk prediction algorithms models recommended for other global regions cannot always be used in European populations, because these models usually include risk factors unavailable in routine European data sources [5-8].
A new risk prediction model (SCORE2) was developed, calibrated and validated to estimate 10-year fatal and non-fatal CVD risk in individuals (40-69 years) in Europe without a history of CVD and diabetes.
First, risk prediction model for fatal and non-fatal CVD outcomes was derived from individual-participant data from 45 prospective cohorts with 677,684 participants recruited between 1990 and 2009. Next, European countries were stratified to four different CVD risk regions using age- and sex-specific mean risk factor levels and CVD incidence rates. The model was subsequently recalibrated for these risk regions using CVD mortality rate and incidence data of 10.78 million individuals with 731,265 CVD events. The external validation of the model was performed using 25 prospective cohorts from 15 European countries (1,133,181 individuals). Finally, the recalibrated model was applied to contemporary populations to illustrate the variation of CVD risk across European regions. Primary outcome was CVD, defined as a composite of CV mortality, non-fatal MI and non-fatal stroke. Median follow-up in the model derivation cohort was 10.7 years.
- There were large regional differences in predicted CVD risk for men and women with a given age and combination of risk factors. For example, the estimated 10-year CVD risk for a 50-year old male smoker with a SBP of 140 mm Hg, 5.5 mmol/mL total cholesterol, and 1.3 mmol/mL HDL-c ranged from 5.9% in low-risk regions to 14.0% in very high-risk countries. A 50-Year old woman with the same risk factor profile had a 10-year estimated CVD risk of 4.2% in low-risk regions and 13.7% in very high-risk countries.
- External validation of the risk model demonstrated cohort-specific C-indices ranging from 0.67 (95% CI: 0.65-0.68) to 0.81 (95% CI: 0.76-0.86).
- The SCORE2 model, compared to the SCORE model, improved overall CVD risk prediction (C-index difference: 0.0100, 95% CI: 0.0085-0.0115, P<0.001), especially for individuals at younger ages (C-index difference for ages 40-50 years: 0.0213, 95% CI: 0.0162-0.0265, P<0.001) and for non-fatal CVD outcomes (C-index difference: 0.0113, 95% CI: 0.0097-0.0130, P<0.001).
- When the recalibrated SCORE2 model was applied to simulated data representing populations from each risk regions, the proportion of men aged 40-69 years with an estimated >10% CVD risk ranged from 3.4% in the low-risk region to 51% in the very high-risk region. In women, these proportions ranged from 0.1% to 32% respectively.
This study demonstrated that SCORE2, a new algorithm developed, calibrated and validated to predict 10-year risk of first-onset CVD, enhanced the identification of individuals with higher risk of fatal and non-fatal CVD events in contemporary European populations.