SG Framingham Risk Score needs recalibration to improve prediction accuracy

Current guidelines that are being used in Singapore (SG) for predicting cardiovascular disease (CVD) must be updated to improve risk prediction accuracy. This can be done by recalibrating risk functions and employing wearable metrics that provide a huge amount of objective health data, suggests a study.

In addition, activity levels and prediabetic state must be considered in the stratification of coronary heart disease (CHD) risk, particularly in individuals at low risk.

“Updating [the] SG Framingham Risk Score (FRS) is necessary to account for the demographic chances and risk factors which contribute to higher CHD risk in the Singaporean population,” the researchers said.

In this study, improvements in SG FRS to predict myocardial infarction risk based on high or low classification of the Agatston score (surrogate outcome) were evaluated in healthy Singaporeans who were enrolled in the SingHEART study (absence of any clinical outcomes). The researchers performed logistic regression, receiver operating characteristic, and net reclassification index (NRI) analyses.

The area under curve (AUC) of SG FRS (AUC, 0.641) significantly improved following recalibration and incorporation of additional variables (ie, fasting blood glucose [FBG] and wearable-derived activity levels; AUC, 0.774; p<0.001). The recalibrated and improved SG FRS++ demonstrated significantly enhanced accuracy in predicting risk (NRI, 0.219; p=0.00254). [Singapore Med J 2024;65:74-83]

“By adopting the use of wearable health metric data in risk stratification, access can be obtained to a large pool of objective and continuously logged health data obtained from patients in the community setting,” the researchers said.

“To strike a balance between complexity and utility of risk prediction tools, we suggest that additional variables such as FBG be considered for CHD risk stratification in healthy cohorts,” they added.

Wearable device

Obtaining patient history on their physical activity can help improve risk stratification, since a sedentary lifestyle is known to increase the risk of cardiovascular and metabolic disease.

“Although self-reporting measures via questionnaires are useful, wearable technology measuring actual motion of the body rather than participant recollections and perceptions of activity, which are subject to bias, can provide more objective and accurate measures of activity levels,” the researchers said. [Prog Cardiovasc Dis 2016;58:613-619]

“Our findings suggest that the existing guidelines by [the Ministry of Health] Singapore, which utilize SG FRS for CHD risk prediction, need to be updated to improve our risk classification methods,” they added.

Several motives back this recommendation. One, patient demographics and underlying risk factors have changed since the last recalibration. Two, there are new types of data such as wearable-derived metrics that have been shown to improve existing risk prediction models. Finally, the prediabetic state, which has not been included before, is predictive of CHD and major adverse cardiovascular events risks.

“Further studies on the utility of proactive screening in healthy cohorts and follow-up studies using data from the SingHEART database to re-evaluate the improved Framingham risk model using actual outcome data are warranted,” the researchers said.