Prediction of the Statin responders with radiomics

Background: Statins are a cornerstone medication for coronary atherosclerosis. This study assessed whether radiomic analysis of coronary computed tomography angiography (CCTA) could predict patient response to statin therapy. Methods: Patients from a multinational registry with serial CCTA (≥2-year intervals) on statin therapy were analyzed. Radiomic scores were calculated, categorizing patients as statin responders or non-responders (≥1.0% increase in percent atheroma volume (PAV) per year indicated non-response). Data were split into training (79%) and test (21%) sets based on sites. Four predictive models were developed: Model 1 used clinical risk factors (CRF), Model 2 included CRF, calcified and non-calcified PAV, and number of high-risk plaques. Model 3 used only the radiomic score, and Model 4 combined Models 2 and 3. Results: A total of 386 statin responders (mean age 61.2 ​± ​8.4 years, 60.1% male) and 177 statin non-responders (mean age 63.2 ​± ​9.0 years, 44.1% male) were analyzed. Model 3, based solely on the radiomic score, demonstrated superior predictive power compared to Model 1 (area under the receiver operating characteristic curve [AUC] [95% confidence interval (CI)]: 0.75 [0.67–0.83] vs 0.54 [0.45–0.63], p ​< ​0.05) and was comparable to Model 2 (AUC [95% CI]: 0.82 [0.74–0.88], p ​> ​0.05) in the test set. Model 4 exhibited the highest power (AUC [95% CI]: 0.84 [0.77–0.90], all p ​< ​0.05 compared to Model 2 and Model 3). Conclusion: CCTA radiomic features show proof-of-concept for predicting statin response, warranting further validation. Clinical trial registration: ClinicalTrials.gov NCT0280341.