Article Text

Download PDFPDF

Virtual colonoscopy performed poorly in detecting colorectal neoplasia
  1. Robert H Fletcher, MD, MSc
  1. Harvard Medical School
 Boston, Massachusetts, USA

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

 Q In patients presenting for colonoscopy, what is the accuracy of computed tomographic (CT) colonoscopy (virtual colonoscopy [VC]) in detecting colororectal neoplasia?

    Clinical impact ratings GP/FP/Primary care ★★★★☆☆☆ IM/Ambulatory care ★★★★★★☆ Gastroenterology ★★★★★☆☆


    Embedded ImageDesign:

    blinded, non-inferiority comparison of VC with conventional colonoscopy.

    Embedded ImageSetting:

    8 clinical centres in the US and 1 centre in the UK.

    Embedded ImagePatients:

    615 patients (mean age 61 y, 55% women) presenting for colonoscopy because of overt and occult rectal bleeding, change in stool habit, abdominal pain, or surveillance after polypectomy. Patients who had had colonoscopy within the past 3 years were excluded.

    Embedded ImageDescription of tests:

    the colon was insufflated with room air or carbon dioxide. VC was done using 2 and 4 section CT scanners with nominal slice thicknesses of 2.5 or 5 mm and reconstruction increments of 1.5 or 1 mm, depending on equipment. Scans were read in 2 dimensional slices and 3 dimensional snapshot reconstructions when necessary. Radiologist interpretations were recorded in a sealed envelope for each colon segment. Conventional colonoscopy was done within 2 hours of VC. Endoscopists were blinded to VC results during insertion of the colonoscope. After each segment was examined and results recorded, the VC results for that segment were revealed, allowing the endoscopist to reexamine any discrepancy.

    Embedded ImageDiagnostic standard:

    initial VC results, additional findings on conventional colonoscopy after segmental unblinding to the VC results, and results of additional diagnostic tests done later when clinically indicated.

    Embedded ImageOutcomes:

    sensitivity and specificity of VC and conventional colonoscopy in detecting lesions ⩾6 mm.


    827 lesions were detected in 308 patients. The prevalence of lesions 1–5 mm, 6–9 mm, and ⩾10 mm was 79%, 14%, and 6.5%, respectively. The sensitivity of VC for detecting lesions of any size was much less than that of conventional colonoscopy (table).

    Test characteristics of virtual colonoscopy (VC) and conventional colonoscopy (CC) in detecting colorectal neoplasia*


    In patients presenting for colonoscopy, virtual colonoscopy was inferior to conventional colonoscopy in detecting colorectal neoplasia.


    Is it time for VC to be included among the screening options? If I had had only the Pickhardt study to guide me, I might have been tempted. But the Cotton study reminds us that the test is not yet ready for general use. Sensitivity and specificity in ordinary circumstances are not high enough. Also, cost and the consequences to patients with abnormal results have not yet been vigorously examined. Abnormal VC results must be followed up with another procedure (conventional colonoscopy), with its own demanding preparation and costs. As for the strength of the evidence of effectiveness, there are no studies of whether screening VC prevents colorectal cancer deaths. However, the medical community seems willing to accept that polyp detection by any means, followed by removal, leads to fewer cases colorectal cancer — by generalising from studies in which both polyp or cancer detection rates and colorectal cancer deaths have been reported.

    VC is already available in some centres and marketed to the general public. But it is not yet included in guidelines. As the technology continues to improve and if more studies of recent generation technology are as persuasive as the Pickhardt study, it may be just a matter of time before VC is added to the list of accepted screening options. The Pickhardt study suggests that the time might not be far away, and the Cotton study reminds us that the time has not yet arrived.

    All currently accepted tests for colorectal cancer screening—faecal occult blood tests, sigmoidoscopy, double contrast barium enema, and colonoscopy—are effective, but none is ideal. There is always room for another test with a different combination of such characteristics as accuracy, safety, convenience, comfort, cost, and availability. VC has been a promising option,1 but no rigorous evaluations of polyp detection have been done in people at average risk of colorectal neoplasia.

    Now there is good information on how well VC detects clinically important polyps in average risk people. 2 strong studies, published within 4 months of each other, come to different conclusions. The study by Pickhardt et al says “CT virtual colonoscopy ... is an accurate screening method ... and compares favorably with optical colonoscopy...” The study by Cotton et al says “computed tomographic colonoscopy ... is not ready for widespread clinical application.” I believe both are right. They ask different questions and get different answers.

    Pickhardt et al ask whether state of the art VC under ideal circumstances can detect polyps in average risk people as well as conventional colonoscopy, the current gold standard. The test they studied had technological features, such as “electronic cleansing” (computer based removal of residual fluid), that are not generally available. Interpretation relied primarily on a 3 dimensional, rather than a 2 dimensional, approach to the detection of polyps, which is not generally used. Also, the 6 radiologists were specially trained, having done ⩾25 (and in some cases >100) such studies. Cotton et al on the other hand, studied the performance of VC under more ordinary circumstances, the kinds of settings where most patients would have the procedure.


    View Abstract


    • For correspondence: Dr P B Cotton, Medical University of South Carolina, Charleston, SC, USA.

    • Source of funding: Office of Naval Research, US Department of Defense.

    • Abstract and commentary also appear in ACP Journal Club.