Novel Incidental Calcium Evaluation (NICE)

1 PROJECT SUMMARY 2 Coronary artery calcium (CAC), typically measured on gated computed tomography (CT) scans, is the 3 strongest predictor of atherosclerotic cardiovascular disease (ASCVD) events across all populations, including 4 groups underrepresented using traditional risk assessment methods. Despite the predictive ability of CAC 5 imaging, less than 1 million gated CT scans are performed annually in the US vs 19 million non-gated chest CTs 6 performed for reasons other than to measure CAC. Further, underrepresented individuals undergo fewer gated 7 CAC scans because these scans are not typically covered by insurance. When scored manually, non-gated CAC 8 scores predict ASCVD events as accurately as gated CAC scores, but grading CAC severity by visual estimation 9 is qualitative and inconsistent, and reporting varies substantially. Consequently, there is vital, lifesaving 10 information that has been collected but not used to guide preventive interventions for individuals unaware of their 11 increased ASCVD risk. 12 Stanford has developed a deep learning (DL) algorithm that quantifies CAC Agatston scores accurately on 13 routine non-gated chest CTs. We have shown that notifying patients and their clinicians about the presence of 14 incidental CAC dramatically increases statin prescriptions. The Novel Incidental Calcium Evaluation (NICE) 15 study will apply the now FDA-cleared algorithm on non-gated chest CTs performed across 3 geographically 16 diverse health systems (Stanford, MedStar Health, and Mayo Clinic). The study will include ~186,000 diverse 17 patients with non-gated chest CT scans without known ASCVD and follow-up within the 3 health systems. 18 Our team has expertise in preventive cardiology, radiology, epidemiology, health equity, DL, and qualitative 19 methods. First, the NICE study will evaluate the prevalence, epidemiology, and prognostic value of DL-CAC in 20 predicting ASCVD events across a real-world, diverse primary prevention cohort who underwent non-gated chest 21 CTs (Aim 1). Second, the algorithm will be extended to estimate CAC on lung cancer screening low radiation 22 dose CT scans (Aim 2). Automating CAC quantification would allow for the equitable and efficient implementation 23 of joint lung cancer and ASCVD screening programs for the 14.5 million eligible individuals in the US. Third, in 24 partnership with the National Minority Health Alliance, we will conduct focus groups with 100 diverse patient and 25 clinician stakeholders to identify facilitators and barriers to increasing preventive therapies following notification 26 of DL-CAC (Aim 3). 27 NICE will provide compelling evidence to support the immediate implementation of opportunistic screening for 28 and notification of CAC by leveraging routine, non-gated chest CTs already performed for other reasons. This 29 study fulfills the promise of data science approaches to equitably improve cardiovascular disease prevention.