Application of machine learning to identify top determinants of fibrofatty plaque burden by CCTA in humans with psoriasis. (3rd October 2022)

• Record Type: Journal Article
• Authors: Hong, C; Li, H; Parel, P M; Berg, A R; Rodante, J; Keel, A; Teague, H L; Playford, M P; Chen, M Y; Zhou, W; Sorokin, A V; Bluemke, D A; Mehta, N N
• Is Part Of: European heart journal. Volume 43(2022)Supplement 2
• Publication Date: 2022-10-03
• Subjects: Cardiology -- Periodicals; Heart -- Diseases -- Periodicals; 616.12005
• DOI: 10.1093/eurheartj/ehac544.213
• Languages: English
• ISSNs: 0195-668X

Abstract

Abstract: Introduction: Fibrofatty plaque burden (FFB) is a high-risk, vulnerable plaque feature comprised of an atheromatous core and fibrous cap with increased risk of coronary artery disease (CAD) [1]. Psoriasis (PSO) is a chronic inflammatory disease linked with atherosclerotic risk and premature cardiovascular disease, driven in part by vulnerable plaque rupture [2, 3]. Machine learning (ML) previously showed the prognostic value of FFB in predicting 5-year risk of cardiac-related mortality in patients with CAD [4]. Whether ML can predict FFB in psoriasis is understudied. Purpose: To use ML to identify top determinants of FFB by CCTA in PSO. Methods: 320 consecutive participants with psoriasis were recruited as part of an ongoing cohort study, of whom 307 had FFB analyzed with coronary computed tomography angiography (CCTA) and quantified by QAngio CT (Medis, The Netherlands). 140 out of 182 potential determinants were subjected to ML algorithms analyzed by random forest and validated by 5-fold cross validation to select the top determinants based on R-square criteria. Lipid concentration and size were measured by nuclear magnetic resonance (NMR) and sdLDL-C was calculated by Sampson's formula. Results: The top 21 determinants of FFB at baseline were grouped into 3 categories: cardiometabolic risk factors (BMI, sex, DBP, mean arterial pressure, exercise, heart rate, glucose, anxiety, psoriasis disease duration), clinical measurements (basophils, platelets, hemoglobin, RBC, alkaline phosphatase, ALT, creatinine, neutrophil-to-lymphocyte ratio), and lipoproteins (LDL particle size, apolipoprotein A1, apolipoprotein B-to-A1 ratio, calculated sdLDL-C). Conclusion: ML confirmed that FFB strongly correlates with cardiometabolic risk factors, clinical measurements, and lipoproteins. Further investigations into these top determinants of FFB over time may provide insight into potential therapeutic interventions that decrease cardiovascular risk in patients with chronic inflammatory diseases and should be validated in larger studies. Funding Acknowledgement: Type of funding sources: Public Institution(s). Main funding source(s): This study was supported by the National Heart, Lung and Blood Institute (NHLBI) IntramuralResearch Program (ZIA-HL-06193). This research was made possible through the NIH MedicalResearch Scholars Program, a public-private partnership supported jointly by the NIH andcontributions to the Foundation for the NIH from the Doris Duke Charitable Foundation, Genentech, the American Association for Dental Research, the Colgate-Palmolive Company, andother private donors.