The excellent paper by Arbab-Zadeh et al. (1) highlights the issues surrounding imaging coronary artery disease in the presence of a high calcium score; however, we are concerned that the outcome, at least in part, may result from statistical artifact.

First, the negative predictive value of cardiac computed tomography in the setting of a calcium score >600 was based on a subgroup of only 8 patients, with error margins far too large to draw robust conclusions.

Second, although it is implicitly assumed that measurement by invasive coronary angiography is both perfectly accurate and immutable, neither assumption is correct. The reproducibility of angiographic coronary stenosis measurement has a standard deviation of approximately 13% (2), whereas comparison with intravascular ultrasound results in a standard error of up to 26% (3). Failure to account for imperfections in the reference standard may result in significant error in the assessed accuracy of a comparator test (4). Even minor errors in the reference standard are greatly magnified when a high disease prevalence (5) or multiple lesions near the positive threshold value are observed. Consistent with this effect, computed tomography coronary angiography in patients with high calcium and patients with known coronary artery disease (prevalence 88% and 84%, respectively) had poor negative predictive values. If intravascular ultrasound were similarly compared with an imperfect quantitative coronary angiography reference in a population with an average of 2 coronary segments within the 40% to 50% stenosis range, the negative predictive value would be falsely estimated to be less than 50%. It would be incorrect to claim on this basis that intravascular ultrasound is inaccurate in populations with a high disease prevalence. Equally, estimates of the accuracy of computed tomography coronary angiography in comparison with unadjusted quantitative angiography should be treated with great caution.