Letters Section Editor: Robert M. Golub, MD, Senior Editor.
In Reply: Dr Dewey raises a concern about the effect of including patients with stents in our study. We addressed this by conducting analyses of the receiver operating characteristic curve separately for the whole group and for a subgroup that excluded the stent-bearing patients. Discriminative power for both groups was identical, demonstrating that there was no substantial influence of stent-bearing segments or patients on the study results.
There are a large number of studies assessing the variability and accuracy of quantitative coronary angiography. The article1 cited by Dewey assesses variability based on automatic edge detection with minimal user interaction. The article that we used as a reference for our study assesses the accuracy of automatic edge detection provided by a quantitative coronary angiography package and electronic calipers placed manually.2 The latter is more similar to our study design, which used manually placed electronic calipers for computed tomography (CT) evaluation and automatic edge detection for invasive angiograms and is therefore a more appropriate basis for comparison. It remains to be seen in studies using automatic edge detection for both CT and invasive angiograms whether stenosis grading of the intermodality comparison will match the interobserver variability of invasive angiograms. But conventional coronary angiography may not be the appropriate reference standard for this, as it has a propensity for false-negative results compared with true cross-sectional imaging.3
Hoffmann MHK. Coronary Angiography With Multislice Computed Tomography—Reply. JAMA. 2005;294(18):2298-2299. doi:10.1001/jama.294.18.2298-b