Research indicates that patients with non-calcified coronary burden (NCB), particularly those with psoriasis, are at an increased risk of acute coronary syndrome, stroke, and cardiovascular mortality at a younger age. “NCB is composed of predominantly lipid-rich plaque—which patients with psoriasis have higher levels of when compared with those without the condition—that is less stable than other plaques and more prone to rupture,” explains Nehal N. Mehta, MD, MSCE. “Thus, these patients have an increased risk of myocardial infarction.” For a study published in the Journal of the American Academy of Dermatology, Dr. Mehta and colleagues sought to determine the main predictors of NCB to assist physicians in identifying it and providing interventions that could control or minimize risk factors.
The study team developed a machine learning algorithm using more than 250 patient records from 2013-2018. The algorithm had 92 phenotypic variables measured at baseline that were later condensed to 62 variables. To determine NCB measurements, all participants received coronary computed tomography angiography (CCTA). CCTA provided the team with information on patients’ lumen stenosis, arterial remodeling, and plaque subcomponents, including total, non-calcified, and calcified coronary artery burden. The researchers also conducted linear and logistic regression between NCB and the predictor variables, along with dichotomizing NCB by median NCB value for predictor variables with binary outputs.
At baseline the majority of participants were middle-aged men, with low cardiovascular risk by Framingham risk score, and with mild-to-moderate psoriasis. The top 20 predictors were identified using the random forest algorithm (Table). “Top predictors of NCB in psoriasis patients were markers related to obesity, dyslipidemia, and inflammation, demonstrating that these are potentially important comorbidities to treat in patients with psoriasis,” emphasizes Dr. Mehta. The algorithm indicates numerically the importance of each predictor in predicting NCB, with a highest possible score of 1.0. Predictors were confirmed by unadjusted linear regression models.
Dr. Mehta stresses the importance of not only treating the inflammatory skin condition but also treating comorbidities such as obesity, dyslipidemia, and inflammation. “There is increasing evidence that treating the underlying psoriasis may improve the patient’s cardiovascular disease profile in such patients, and the results of this study are in line with this hypothesis,” Dr. Mehta adds.
The study authors acknowledge the need for future studies to confirm their findings and reduce the risks of cardiovascular complications in patients with psoriasis. “Incorporating longitudinal data would provide additional insight into how changes in these top predictors modulate NCB,” explains Dr. Mehta. “Furthermore, incorporating the incidence of actual cardiovascular events will also help augment models for risk stratification. Research in these areas would help us better understand how modulation of NCB corresponds to clinical events in this population and would provide deeper insight into how clinicians can risk stratify patients with psoriasis in order to augment clinical care. With a better understanding of risk stratification in this patient population, we will be able to provide more individualized care and appropriate treatment strategies to reduce underlying cardiovascular disease risk.”
Application of Machine Learning to Determine Top Predictors of Non-calcified Coronary Burden in Psoriasis: an Observational Cohort Study