Does slice thickness matter? Diagnostic accuracy of CT-derived fractional flow reserve (CT-FFR) compared with...

CT-derived fractional flow reserve (CT-FFR) has become a routine clinical tool with high accuracy and proven impact on outcome. However, the magnitude by which CT-specific reconstruction parameters influence the accuracy of CT-FFR in comparison to invasive FFR remains to be fully explored. We evaluated the influence of reconstructed slice thickness in coronary CT angiography (CCTA) on the accuracy of CT-FFR in comparison to invasively measured FFR in patients with suspected CAD.

A cohort of 50 consecutive patients (mean age 66 ± 9 years) in whom dual source CCTA had been performed due to suspected CAD and who were further referred for invasive coronary angiography with invasive FFR measurement within three months of the index CT examination was prospectively included in this analysis. Patients with left main coronary artery stenoses, bifurcation or ostial stenoses in CCTA were excluded. CCTA data sets were reconstructed with a slice thickness of 0.5 mm [increment 0.25 mm], 0.75 mm [0.4 mm] and 1.0 mm [0.5 mm] and all vessels were analyzed by an experienced observer blinded to the results of invasive FFR. CT-FFR was calculated using a locally hosted PC-based prototype (cFFR Version 3.2.0 and 3.5.0, Siemens Healthineers, Forchheim, Germany). Invasive FFR was measured using a pressure wire (CERTUS®, St. Jude Medical, Minnesota, USA or Verrata®, Volcano, San Diego, USA) with intracoronary adenosine injection. Stenoses with invasively measured FFR ≤ 0.80 were classified as hemodynamically significant.

A total of 81 vessels in 50 patients were included. Within these, 21 stenoses (25.9%) were hemodynamically significant according to invasive FFR. Median invasive FFR across all lesions was 0.89 with an IQR of 0.13. Median CT-FFR values across different slice thicknesses closely agreed with invasive FFR: 0.85 (IQR 0.13) for 0.5 mm, 0.87 (IQR 0.12) for 0.75 mm, and 0.88 (IQR 0.14) for 1.0 mm. CT-FFR reconstructed at 0.75 mm correlated most closely with invasive FFR (Spearman's ρ = 0.72, 95% CI 0.59-0.81, p ≤ 0.001). At a threshold of 0.80, the sensitivity of CT-FFR to correctly identify hemodynamically significant stenoses was 81% for 0.5 mm, 86% for 0.75 mm, and 76% for 1.0 mm slice thickness. Specificity was 85% for 0.5 mm and 88% for 0.75 mm as well as 1.0 mm reconstructed slice thickness. Positive predictive value (PPV) was 65% for 0.5 mm, 72% for 0.75 mm, and 70% for 1.0 mm; negative predictive value (NPV) was 93% (0.5 mm), 95% (0.75 mm), and 91% (1.0 mm). Overall diagnostic accuracy was 84% for 0.5 mm, 88% for 0.75 mm, and 85% for 1.0 mm slice thickness (p = 0.51 [0.5-0.75 mm], p = 0.75 [0.75-1.0 mm], p = 1.0 [1.0-0.5 mm]). As a function of cut-off threshold, the area under the curve (AUC) to predict hemodynamically significant stenoses was 0.88 ± 0.04 (95% CI: 0.80-0.95, p < 0.05) for 0.5 mm, 0.93 ± 0.03 (95% CI: 0.87-0.98, p < 0.05) for 0.75 mm, and 0.87 ± 0.04 (95% CI: 0.79-0.95, p < 0.05) for 1.0 mm slice thickness, with no significant difference between AUCs.

Across all reconstruction protocols, CT-FFR maintained high diagnostic accuracy, with no significant impact of slice thickness on the overall accuracy for the identification of hemodynamically significant coronary stenoses, but a tendency towards better performance for slice thicknesses below 1.0 mm.