A study on the first-order interaction of diagnostic performance of coronary CT angiography-derived fractional flow reserve

Abstract

Objective: To investigate the effect of image quality, degree of stenosis, calcification, and their first-order interactions on diagnostic performance of coronary computed tomography (CT) angiography-derived fractional flow reserve (CT-FFR). Methods: This is a reanalysis of data from a multi-center retrospective cross-sectional study of CT-FFR in China. A total of 522 patients with suspected or known coronary heart disease [mean age: 61.6 (34.0-83.0) years, 71.8% (354/493) were male] from 11 medical centers including the General Hospital of Eastern Theater Command from May 2015 to October 2019 were enrolled. All patients underwent coronary CT angiography (CCTA), CT-FFR, and invasive FFR examination. Subjective image quality scores of target vessels were recorded on CCTA images, and stenosis was visually assessed at the lesion level. Calcification arc and calcification remodeling index (CRI) were recorded for each lesion. Sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) were compared. Two-way analysis of variance was used to analyze the first-order interaction effects of image quality, degree of stenosis, and calcification. Results: A total of 493 patients with 629 lesions with invasive FFR as a reference were included in the study. The overall sensitivity, specificity, and accuracy of CT-FFR were 80.4%, 93.8%, and 88.6%, respectively. The specificity (95.0% vs. 87.3%, χ2=4.11, P=0.043); accuracy (90.1% vs. 81.9%, χ2=6.22, P=0.013); and NPV (89.7% vs. 80.9%, χ2=4.25, P=0.039) of the group with image quality ≥3 was higher than the group with image quality <3. The degree of stenosis affected the sensitivity, PPV, and NPV of CT-FFR and the calcification arc affected the specificity of CT-FFR (all P>0.05). The specificity (95.8% vs. 90.5%, χ2=4.23, P=0.040); accuracy (91.0% vs. 86.1%, χ2=4.01, P=0.045); and NPV (91.1% vs. 83.8%, χ2=5.10, P=0.024) of the group with CRI<1 were higher than that of the group with CRI≥1. In the subgroup of mild and severe stenosis, no calcification, and CRI<1, the accuracy of CT-FFR with image quality ≥3 points were higher than that with image quality <3 points. The accuracy of CT-FFR in the moderate stenosis group was mainly affected by CRI; the accuracy of CT-FFR in the group with CRI<1 was higher than that in the group with CRI≥1 (after Bonferroni correction, P values between groups were statistically significant). Conclusion: Subjective image quality, degree of stenosis, calcification of lesions, and their first-order interactions can all negatively affect the diagnostic performance of CT-FFR.