Pericoronary adipose tissue radiomics to improve risk stratification for patients with acute coronary syndrome: a multicenter retrospective cohort study.

Abstract

Background: Pericoronary adipose tissue (PCAT) radiomics derived from coronary computed tomography angiography (CCTA) for predicting major adverse cardiovascular events (MACE) in patients with acute coronary syndrome (ACS) remains unclear. This study aimed to assess whether PCAT radiomics could further provide complementary predictive value for the risk of MACE during long-term follow-up.

Methods: A multicenter retrospective study enrolled 777 subjects who underwent pre-intervention CCTA at 3 medical centers. Patients from one institution (n = 664) formed an internal cohort and were randomly split into training and internal test sets (7:3). Multivariable Cox regression models were developed using clinical scores, traditional CCTA, PCAT attenuation (PCATa) and PCAT radiomics, and were tested using the internal test set. Data from two additional institutions (n = 113) were reserved as an external test set to evaluate the applicability and generalizability of models.

Results: A total of 777 participants (61.0 ± 9.70 years; 506 males) were analyzed. During a median follow-up of 5.45 years (interquartile range: 4.03, 7.12 years), 177 (22.78%) cases experienced a MACE. Adding culprit PCATa or three vessels-based PCATa did not improve predictive ability for the model containing clinical scores and traditional CCTA, whereas further addition of PCAT_culprit Radscore (C-index: 0.721, 0.652, 0.645) and three vessels-based PCAT Radscore (C-index: 0.725, 0.660, 0.686) improved model predictive performance in the training, internal test and external test sets, without significant differences between datasets or models (all P > 0.05). Adding either the PCAT_culprit Radscore (training: IDI = 0.031, p < 0.001; NRI = 0.256, p < 0.001; external test: IDI = 0.094, p < 0.001; NRI = 0.339, p = 0.02) or the three vessels-based PCAT Radscore (training: IDI = 0.032, p < 0.001; NRI = 0.224, p = 0.02; external test: IDI = 0.126, p < 0.001; NRI = 0.480, p < 0.001) to a clinical model yielded a significant improvement in discrimination and reclassification ability in the training and external test sets, respectively.

Conclusions: PCAT radiomics can enhance long-term prediction of MACE in ACS patients beyond current clinical scores, traditional CCTA and PCATa. Addition of PCAT radiomics to a conventional risk assessment improves the identification of high-risk individuals with MACE.