The Multi-Ethnic Study of Atherosclerosis
CONTROL ID:
4144108
PRESENTATION TYPE:
Poster Only
CURRENT CATEGORY:
20.85 Biomarkers, Risk Assessment and Risk Prediction
AUTHORS (FIRST NAME, LAST NAME):
Morteza Naghavi, Anthony Reeves, Kyle Atlas, Chenyu Zhang, Thomas Atlas, Claudia Henschke, David Yankelevitz, Matthew J. Budoff, WENJUN FAN, Ruilin Yu, Andrea Branch, Ning Ma, Sion Roy, Khurram Nasir, Sabee Molloi, Zahi A. Fayad, Mike McConnell, Ioannis Kakadiaris, Javier Zulueta, David Maron, Prediman K. Shah, Kim A. Williams, Daniel Levy, and Nathan D. Wong.
INSTITUTIONS (ALL):
1. HeartLung.AI, Houston, TX, United States.
2. Cornell University, Ithaca, NY, United States.
3. Tustin Teleradiology, Tustin, CA, United States.
4. Mount Sinai Hospital, New York, NY, United States.
5. LUNDQUIST INSTITUTE, Torrance, CA, United States.
6. University of California Irvine, Irvine, CA, United States.
7. UCLA, Los Angeles, CA, United States.
8. UCLA Harbor, Santa Monica, CA, United States.
9. Houston Methodist, Houston, TX, United States.
10. University of California, Irvine, Irvine, CA, United States.
11. MOUNT SINAI MEDICAL CENTER, New York, NY, United States.
12. Stanford University School of Medicine, Stanford, CA, United States.
13. University of Houston, Houston, TX, United States.
14. Cedars-Sinai Medical Center, Los Angeles, CA, United States.
15. University of Louisville Medicine, Louisville, KY, United States.
16. National Heart, Lung, and Blood Institute, Bethesda, MD, United States.
Background:
The CHADs-VASc risk score is a clinical tool for stroke prediction. It is mainly used in patients with atrial fibrillation (AF) but is also applied to the non-AF population. We previously reported that artificial intelligence (AI)-enabled left atrial (LA) volumetry from coronary artery calcium (CAC) scans (AI-CAC) predicts AF as early as one year and outperformed CHARGE-AF and NT-proBNP. In this report, we compare AI-CACLA volumetry to the CHADS-VASc risk score and evaluate the incremental value of incorporating AI-CAC LA volume to CHADS-VASc for incident stroke prediction in the non-AF population.
Methods:
We applied AI-CAC LA volumetry to CAC scans of 5830 people without AF (52.2% women, age61.7±10.2 years) enrolled in the Multi-Ethnic Study of Atherosclerosis (MESA) baseline (2000-2002). We used the 15-year outcomes data for incident stroke (ischemic and hemorrhagic) and assessed discrimination using the time-dependent area under the curve (AUC) between AI-CAC LA volume vs. CHADS-VASc risk score. Notably, the CHADS-VASc score in this non-AF population ranges from 0 to 5, whereas in the AF population it typically ranges from 0 to 9 points.
Results:
252 cases of stroke accrued over 15 years. The median and mean ± SD of CHADS-VASc score at baseline were 1.0 and 1.58 ± 1.15, respectively. The cumulative incidence of stroke for the 95 percentile of AI-CAC LA volume (n=291) vs. CHADS-VASc 4 or 5 points (n=364) was 13.0% and 13.7%, respectively. AI-CAC LA volume significantly improved the AUC of CHADS-VASc for stroke prediction at 2-year follow-up(0.76 for CHADs-VASc vs. 0.81 for CHADS-VASc plus LA volume, p=0.03), 5-year follow-up (0.73 vs. 0.77,p=0.01), 10-year follow-up (0.70 vs. 0.75, p<0.0001) and 15- year follow-up (0.70 vs. 0.74, p<0.0001).
Conclusion:
In this multi-ethnic longitudinal cohort of people without AF, addition of LA volume significantly improved performance of the CHADS-VASc risk score for stroke prediction from 2 to 15 years follow-up. Further studies are needed to evaluate the clinical utility of adding LA volume to CHA2DS2-VASc risk score in people with and without AF.
Abstract Graphic/Image Description:
The cumulative incidence of stroke for the 95th percentile of AI-CACLA volume (n=291) vs. CHA2DS2-VASc 4 or 5 points (n=364) was 13.0% and 13.7%, respectively. AI-CAC LA volume significantly improved the AUC of CHA2DS2-VASc for stroke prediction at 2-year follow-up (0.76 forCHA2Ds2-VASc vs. 0.81 for CHA2DS2-VASc plus LA volume, p=0.03), 5-year follow-up (0.73 vs. 0.77,p=0.01), 10-year follow-up (0.70 vs. 0.75, p<0.0001) and 15- year follow-up (0.70 vs. 0.74, p<0.0001).
Keywords:
Artificial Intelligence, Stroke.
Comments