In a recent study published in the Journal of American Heart Association, researchers assessed whether non-imaging data could be used to identify people with coronary atherosclerosis.
Study: Self‐Report Tool for Identification of Individuals With Coronary Atherosclerosis: The Swedish CardioPulmonary BioImage Study. Image Credit: Orawan Pattarawimonchai/Shutterstock.com
Introduction
Asymptomatic people with coronary atherosclerosis signs on imaging are deemed to be at risk for ischemic heart disease (IHD).
A computed tomography (CT)-derived coronary artery calcification score (CACS) ≥ 100 suggests the benefits of statin therapy in those with intermediate risk of IHD, irrespective of low-density lipoprotein levels.
Moreover, while coronary CT angiography (CCTA) shows even more promise, the limitations of these imaging modalities, including high costs, low availability, and risks due to contrast agents and radiation, are significant.
Therefore, developing tools using non-imaging data that identify people with elevated IHD risk could substantially reduce healthcare costs.
About the study
In the present study, researchers evaluated the utility of non-imaging data to identify people with moderate or severe coronary atherosclerosis. They used non-imaging data collected in the Swedish Cardiopulmonary Bioimage Study (SCAPIS), which mainly included 50–60-year-old people of European ancestry.
Data were collected from 2014 to 2018 at six university hospitals. Data from the SCAPIS pilot trial conducted at a single site in 2012 were used for external validation.
The analyses included individuals with high-quality CT or CCTA imaging without previous IHD. Questionnaires were administered for information on health, medication, family history, lifestyle, socioeconomic status, environmental or occupational exposure, and psychosocial well-being. Biochemical analyses were performed using blood samples.
Height, weight, physical activity, waist and hip circumference, blood pressure, and lung function were measured. Outcomes included CACS ≥ 100 and segment involvement score (SIS) ≥ 4.
A self-report tool was developed based on self-reported data, and a clinical tool was developed based on all SCAPIS data. They identified 105 and 127 potential predictors for inclusion in the self-report and clinical tools, respectively.
The performance of both tools was benchmarked against the pooled cohort equation (PCE) for a 10-year risk of atherosclerotic cardiovascular disease.
Furthermore, data reduction was performed using manual and data-driven techniques to include the most relevant factors. XGBoost was used to develop tools to identify CACS ≥ 100 and SIS ≥ 4. The area under the receiver operating characteristic curve was computed and validated.
Findings
Overall, the study included 25,182 individuals in the cohort assessing SIS; of these, approximately 12% had SIS ≥ 4. Besides, 28,701 subjects were included in the cohort examining CACS ≥ 100; of these, 12% had CACS ≥ 100. In both cohorts, there were fewer subjects with self-reported symptoms of angina.
The validation cohort comprised fewer individuals with university education and more people born outside Sweden compared to the SCAPIS dataset.
Fourteen factors were included in the self-report tool: age, sex, weight, weight at age 20, height, hip and waist circumference, smoking duration, cigarette pack-years, heredity of myocardial infarction, diabetes duration, hypertension, lipid-lowering medication, and anti-hypertensive medication.
By contrast, 23 factors were included in the clinical tool: heart rate, systolic and diastolic blood pressure, glycated haemoglobin, high-density lipoprotein cholesterol, plasma triglycerides, plasma glucose, creatinine, total cholesterol, and the 14 factors of the self-report tool.
The discriminatory capacity of the self-report tool for SIS ≥ 4 was high to excellent in the external validation cohort and was significantly better than PCE.
Age and sex were the most crucial variables in the self-report tool. The clinical tool performed slightly better than the self-report tool, with the most important predictors being systolic blood pressure, total cholesterol, and glycated hemoglobin.
Findings were largely similar for CACS ≥ 100. Both tools performed better in females and older individuals (age > 55).
Reassuringly, data reduction was successful as using all 127 factors yielded similar results as the clinical tool. Further, the team stratified the population into ten groups, ordered by the predicted risk.
Individuals with the top 30% of the mean absolute risk formed the high-risk group, while those with the bottom 30% represented the low-risk group. In the high-risk group, the self-report tool identified 64.6% of individuals with SIS ≥ 4 compared to 67.3% with the clinical tool.
Conclusions
The findings show that non-imaging data could be used to identify people more likely to have moderate/severe coronary atherosclerosis.
The self-report tool had a high to excellent discriminatory capacity in an external validation cohort and performed almost similarly to the clinical tool. Overall, the self-report tool could be the starting point to identify high-risk individuals needing imaging or further risk evaluation.
Cancer therapy may raise heart attack and stroke risks by disrupting immune regulation in arteries