Effects of statin treatment on coronary plaque morphology: longitudinal radiomics-based precision phenotyping using coronary computed tomography angiography

Background Increasing evidence suggests that statins exert direct effects on coronary atherosclerotic plaques. Studies using coronary computed tomography angiography (CCTA) have shown statin therapy to associate with noncalcified plaque volume regression and calcified plaque volume progression. Radiomics enables deeper phenotyping of plaque morphology from CCTA and can be used to track temporal plaque changes. Purpose To examine the effects of statin treatment on coronary plaque morphology using CCTA-based radiomic phenotyping in patients with stable coronary artery disease. Methods Retrospective study of consecutive patients who underwent clinically indicated serial CCTA at an interscan interval of ≥2 years in a single tertiary centre between 2011-2023. Quantitative plaque analysis was performed using automated research software (Autoplaque v3.0). Plaque segmentations were exported to Radiomics Image Analysis software, where 1,103 radiomic features were calculated for each lesion. Linear mixed models were used to examine the association between statin treatment and lesion-level changes in radiomic features, with adjustment for cardiovascular risk factors, low-density lipoprotein cholesterol levels, and plaque burdens (plaque volume / vessel volume x 100%). Logistic regression was used to assess the incremental value of baseline radiomic features for predicting response to statin treatment (defined as annualised plaque burden progression below the median of the cohort). Results In total, 1,069 coronary lesions were analysed in 375 patients (mean age 58.9 ± 10.1 years; 52.3% male; 62.9% statin-treated). Over a mean interscan interval of 4.8 ± 2.2 years, lesions in statin-treated patients exhibited slower progression of noncalcified plaque burden (0.29 ± 0.92 vs 0.47 ± 1.06% per year; p=0.01) and more rapid progression of calcified plaque burden (0.38 ± 0.61 vs 0.28 ± 0.47% per year; p=0.005) compared with lesions in statin-naive patients. Among the 1,103 radiomic features, 68 (6.2%) were independently associated with statin treatment, including: 6 first-order features, 22 textural features, and 40 geometric features (Figure 1). Hierarchical clustering identified 4 distinct clusters among the 68 significant features. The top 4 radiomic features (1 feature with smallest p-value from each of the 4 clusters) provided incremental value for predicting statin response when added to a model containing visual assessment of plaque composition and quantitative plaque burdens (area under the receiver operating characteristic curve 0.71 [0.68-0.74] vs 0.63 [0.60-0.67], p=0.01; Figure 2). Conclusion Statin treatment is associated with temporal changes in radiomic features of coronary plaques on CCTA, independently of cardiovascular risk factors and plaque volumes. CCTA-based radiomic phenotyping may help identify novel imaging biomarkers of patient response to statin therapy.