William Francis McCarthy PhD, Presenter: Nothing to Disclose

Douglas Thompson PhD, Abstract Co-Author: Nothing to Disclose

Bruce Barton PhD, Abstract Co-Author: Nothing to Disclose

Contrast-enhanced multi-detector row spiral computed tomography (MDCT) has been introduced as a method for non-invasive visualization of coronary artery stenosis. To determine the diagnostic accuracy of MDCT coronary angiography, as compared to the “gold standard” invasive coronary angiography, sensitivity and specificity are estimated (95% Confidence Intervals). Three separate levels of estimation are computed: at the patient level, at the coronary artery level, and at the coronary artery segment level.

We review the methodology for the estimation of sensitivity and specificity of non-clustered binary data (patient level analysis) and present a methodology for the estimation of sensitivity and specificity that considers the patient as a cluster and the coronary arteries (or coronary artery segments) as the diagnostic units of the study (DUOS) within each cluster. We also present how to estimate the weighted kappa for the comparison of ordinal measures of stenosis when non-clustered and clustered data are considered and the mean difference for the comparison of continuous measures of stenosis when non-clustered and clustered data are considered.

If the clustering effect is not taken into account in this example, the calculated variance (thus the calculated confidence limits) is inappropriately small because the calculation ignores the positive correlation among the coronary artery segments nested within the patient. Even if the amount of correlation among the coronary artery segments nested within the patient is small, the variance (confidence limits) will be biased if the clustering effect is not taken into account.

Analyses at the coronary artery and coronary artery segment levels must take clustering into account (coronary arteries are nested within patients, and coronary artery segments are nested within coronary arteries).

Analyses at the coronary artery and coronary artery segment levels must take clustering into account (coronary arteries are nested within patients, and coronary artery segments are nested within coronary arteries). Assessments will be biased if clustering is not accounted for.

McCarthy, W, Thompson, D, Barton, B, How to Analyze Data Generated by Contrast-enhanced Multidetector Row Spiral Computed Tomography (MDCT) When the Outcome of Interest Is the Noninvasive Visualization of Coronary Artery Stenosis.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5000491.html