Letters When a test is too good

The accuracy of CT pulmonary angiography is not as high as purported

BMJ 2013; 347 doi: https://doi.org/10.1136/bmj.f5116 (Published 20 August 2013) Cite this as: BMJ 2013;347:f5116
  1. Harry Rogers, physician in independent practice1
  1. 1Melbourne, Australia
  1. DrHarryRogers{at}gmail.com

The principles presented by Wiener and colleagues are entirely correct, except for the assertion that computed tomography (CT) pulmonary angiography “is too good.”1 The authors contend that the test is exquisitely sensitive and detects small pulmonary emboli that may not be clinically relevant. Although this is true, a major confounder is that the test is not as accurate as it purports to be. This is evident from studies of inter-reporter variability showing a κ of only 0.47 for 3 mm CT pulmonary angiography.2 Such modest agreement is not compatible with an accuracy of over 90-95%—if the same study was shown to five radiologists, a variety of answers might be given. Many factors contribute, such as respiratory motion and contrast artefacts. In the real world, many radiologists fear missing a pulmonary embolism, which results in high sensitivity reporting. This is not “too accurate” but “over-calling.” A “small” pulmonary embolism was never there to start with. A key strength of functional imaging such as V/Q scintigraphy is its ability to visualise the physiological relevance of pulmonary embolism.


Cite this as: BMJ 2013;347:f5116


  • Competing interests: None declared.


View Abstract

Log in

Log in through your institution


* For online subscription