Keywords CT coronary angiography, Coronary calcium score, NICE guidelines
Introduction Current UK NICE guidelines recommend an initial CACS scan as a first line investigation in the assessment of patients with stable chest pain with a low pre-test probability of underlying coronary artery disease (CAD), with subsequent evaluation of the coronary tree with CTCA if the CACS lies between 1 and 400. We hypothesised that the CACS can be accurately derived from CT Coronary Angiography (CTCA) allowing a potential change to UK NICE guidelines for the assessment of stable chest pain.
Methods 503 consecutive patients undergoing conventional CACS and CTCA were included. A 0.1cm2 region of interest (ROI) was used to determine the mean contrast density on CTCA images either in the left main stem (LM), or – where the LM was heavily calcified - the proximal RCA.Axial contiguous 3mm CTCA images were then scored for calcium using conventional CACS software with a modified threshold calculated as: mean LM contrast density (HU) + 2 standard deviations. The results were compared to the traditionally acquired CACS (130HU threshold) and modelled using linear regression to derive a conversion factor subsequently applied to all CTCA-derived CACS. Bootstrapping (1000 samples) was used to calculate a shrinkage factor to adjust for model over-optimism.Accuracy of this method was determined using weighted Kappa for NICE recommended CACS groupings (0, 1–400, >400) and Bland-Altman analysis for absolute score.
Results A final conversion factor of 1.94 (95% CI: 1.88–1.99) was used with excellent agreement between methods both for absolute score (mean difference between scores 7.19 [95% limits of agreement -214.51, 228.89] Figure 1) and when categorising patients into risk groups (k = 0.83). There was excellent discrimination between high (>400) and low risk (<400) scores with a sensitivity and specificity of 83 and 99%, and a PPV and NPV of 91 and 98%, respectively. Both inter and intra-observer agreement were excellent with no significant difference between observers in median CTCA-derived CACS and risk group placement (0 [0–25.9] vs. 0 [0–33.0]; p = 0.49, k = 0.90) and (0 [0–24] vs. 0 [0–27]; p = 0.68), k = 0.94), respectively. There was a significant reduction in radiation exposure with exclusive use of CTCA both overall (4.2 [3.1–6.2] vs. 3.2 [2.3–5.3] mSv; p < 0.0001), and with high pitch single heart-beat acquisition protocol (2.0 [1.7–3.2] vs. 1.2 [1.1–2.3] mSv; p < 0.0001). This represented a reduction of 20.6 ± 12.5% and 33.7 ± 8.2%, respectively.
Conclusion Our proposed method allows a comprehensive assessment of coronary artery pathology through the use of an individualised, semi-automated approach. If incorporated into stable chest pain guidelines this protocol could lead to significant reductions in radiation exposure and the need for further functional testing or invasive angiography could be determined from CTCA alone. The study would support a potential change to guidelines (Figure 2).