Contact: Ellen Bank
843.792.2626
April 13, 2004
A multicenter study led by the Medical University of South Carolina (MUSC) in
Charleston showed computed tomographic colonography, also known as virtual colonoscopy,
to be much less effective in detecting lesions in the colon than the standard
colonoscopy.
The results of the study will be published in the April 14 issue of the Journal
of American Medical Association (JAMA).
Six hundred subjects from nine medical centers across the United States and
England underwent both the virtual and the standard colonoscopy procedures.
The researchers found that the virtual colonoscopy detected only 39 percent
of subjects with at least one lesion greater than or equal to 6 mm and only
55 percent of subjects with at least one lesion greater than or equal to 10
mm. Conventional colonoscopy detected 99 percent and 100 percent, respectively,
for the two sizes.
The study was initiated and planned by MUSC’s Digestive Disease Center
and the Clinical Innovation Group who collected and housed the data from all
nine centers, and provided the statistical support to analyze the data. The
participating clinical centers were MUSC’s Digestive Disease Center, Wake
Forest University School of Medicine, Emory University Hospital, Indiana University
Hospital, the University of Texas Southwestern, the Medical College of Virginia,
St. Mary’s Hospital, London, MD Anderson Cancer Center and University
Hospital of Cleveland. The Office of Naval Research of the U.S. Department of
Defense supported the study.
Peter Cotton, M.D., director of MUSC’s Digestive Disease Center was principal
investigator of the multicenter study and Brenda Hoffman, M.D., was principal
investigator for the MUSC site. Valerie Durkalski, Ph.D., and Yuko Palesch,
Ph.D., of MUSC’s Clinical Innovation Group were study statisticians. Patrick
Mauldin, Ph.D., associate professor in MUSC's College of Pharmacy, served as
co-investigator with the Clinical Innovation Group.
Conventional colonoscopy was developed in the early 1970s and has been a routine
tool for examining the colon. It was initially used to diagnose individuals
who had symptoms such as bleeding. Screening for asymptomatic patients has become
popular in the last five to 10 years, as data has grown showing that detecting
and removing pre-cancerous lesions in the colon prevents them from growing into
cancer. Colonoscopy as a screening tool for individuals over the age of 50 has
been recognized by the major cancer organizations, and costs for the procedure
are reimbursed by insurance companies. Most lesions discovered during colonoscopy
can be removed during the same procedure. But because the test is invasive and
perceived as an unpleasant experience, many people who could potentially benefit
have resisted the screening.
Also, during the past 10 years researchers have begun investigating virtual
colonoscopy, a technique that uses a CT scanner and computer virtual reality
software, enabling a radiologist to “see” the inside of the colon
through the examination of computer-generated images of the colon constructed
from data obtained from an abdominal CT examination.
If a lesion was detected through virtual colonoscopy, the patient would have
a conventional colonoscopy performed to remove any lesions detected.
A number of recent clinical studies reported detection rates from virtual colonoscopy
to range from 84 to 94 percent. Most of these studies were initiated by committed
radiologists, many of them pioneers in the technique, and were restricted to
a single center. But to be valuable as a screening tool, virtual colonoscopy
must perform well in routine practice. The MUSC researchers’ objective
was to assess the accuracy of virtual colonoscopy in a large number of subjects
across multiple centers.
An interesting outcome of the study was that only one of the centers had a relatively
high detection rate with virtual colonoscopy, and this was the center that had
prior substantial involvement with the virtual techniques. That center had the
largest number of participants --184 -- and had an 83 percent detection rate.
The detection rate for all the other centers combined was 26 percent. The high-detection
rate center’s results were consistent with that center’s already
published results and those of several other single-center, small-scale studies.
The MUSC researchers noted the poor results from their study, with the exception
of the one highly experienced center, compared to the good results of earlier
published studies coming from centers with radiologists who were pioneers in
the technique. This led them to conclude that though previous studies have shown
virtual colonoscopy to be successful in the hands of experts, the technique
is not ready for routine use.
“I hope our study will stimulate more studies on improving both technique,
software, scanners and training, so that virtual colonoscopy will eventually
provide an accurate, non-invasive screening tool,” said Cotton. “If
virtual colonoscopies are improved so they are as effective as the conventional
colonosocopy, many more people would get screened, and more precancerous lesions
would be detected. These lesions would be removed by the conventional colonoscopy
and cancer would be prevented.” From the patient’s perspective,
the worst part of the procedure is the bowel prep. It is necessary to take medication
to clear out the bowels prior to either procedure. “If the computer techniques
can be improved to eliminate the bowel prep, this would revolutionize the whole
thing,” Cotton said. He said there are ongoing studies exploring the “virtual
bowel prep.” This entails giving patients something to drink that mixes
with residue in the colon until the material reaches a certain density. Then
a button can be pressed on the CT machine to ignore anything of that density.
Another future issue that would make a big difference in success of virtual colonoscopy is the development of software to recognize lesions. For each study, the radiologist must look at more than a thousand pictures, and in a screening situation you find something in only one in five patients. “You sit there all day, looking at thousands of images, most of which are normal, so it’s difficult to pick the ones that stand out,” Cotton said. “I’m not suggesting that computers would take over, but they can flag things and potentially make the process more effective.”
#####