For about $1000-less with a lung CT or coronary calcium scan-HealthScreen America offers consumers a day’s worth of laxatives and barium smoothies, followed by a CT scan of the colon in two 20-second breath-holds. On screen, a staff radiologist will scan through a three-dimensional reconstruction and two-dimensional views, looking for polyps, and search for aortic aneurysm, renal carcinoma, and kidney stones in the bargain.

Virtual colonoscopy is available in a handful of boutique screening centers around the country and in a number of hospitals where patients who don’t want or can’t tolerate conventional colonoscopy have the option of paying out of pocket for the noninvasive version. Most people receiving the screening, however, are high-risk patients in clinical trials designed to test whether CT colonography is ready for prime time.

At stake is a chance for radiology to reclaim turf in a field in which the double-contrast barium enema has been largely made obsolete as a result of gastroenterology literature touting colonoscopy. With gastroenterologists facing a labor shortage of their own, that specialty has become much more inclined toward the CT screen. Researchers in both fields, however, are waiting for definitive studies that show diagnostic equivalence between the scope exam and its noninvasive counterpart.

A standard for prepping and screening the colon with CT has not been established, but radiologists at Mayo Clinic, with more than 2500 of the scans under their belts, have a pretty good idea of what works. Prep is essentially the same as for a standard colonoscopy, with a clear diet and a gallon of polyethylene-glycol bisacodyl. Centers that use electronic stool subtraction require barium to tag liquid contents of the colon. Before the scan, the patient’s colon is inflated with room air or, more comfortably, CO2, which is absorbed into the colon faster. After a scout computed radiograph, the patient is scanned both prone and supine. A run-through takes about 40 sec with single-slice CT, 20 sec with a multidetector unit.

Radiologists disagree on optimal slice thickness. Mayo imagers often use 5-mm slices, said Dr. Robert McCarty, a professor of radiology at Mayo Medical School, but he and his colleagues like 2.5 mm contiguous slices as well. At the San Francisco Veterans Administration, CT and gastrointestinal radiology chief Dr. Judy Yee settled on 3-mm slices after initially trying out 5 mm. Best of all for reconstructing the colon accurately would be 1.5 mm, she said, but slices so thin can hamper an efficient read. At Mayo, scans for research purposes can produce up to 1400 slices.

“Taking (thinner slices) is likely better at getting smaller lesions,” Yee said. “But does that (appreciably) improve the outcome? If not, it just means more images to look at, and we need to be time-efficient. We need additional evaluation of optimal slice thickness.”

With a manageable set of slices, an experienced reader can complete a thorough read in 10 or 15 minutes, according to Yee. The key word is experience. Yee has read more than 600 scans and suggests that the learning curve requires at least 50 cases, just to know what looks normal.

At Mayo, readers start with the prone view on lung window settings, work quickly through the axial scans, then switch to soft-tissue views to determine whether abnormalities are folds, stool, or lipoma. The same process on supine follows, ending by running the two simultaneously to see whether the suspected stool shifts and whether all of the colon has been viewed.

Though they have 3-D views available, both Yee and McCarty prefer to read in 2-D, resorting to 3-D only for problem-solving. Although 3-D views mimic an endoscopic view, axial slices make it easy to see if a suspected polyp extends outside the colon, for example. Several commercial reconstruction software packages offer a 3-D fly-through, with axial view available as the reader travels up and down the colon. They can also measure a polyp’s distance from the rectum. Such innovations may make reading easier or faster, but virtual colonoscopy’s reputation will depend on whether it allows radiologists to find polyps as accurately as with scope-based colonoscopy.

Size Matters

Having set 1 cm as the threshold for clinical significance, radiologists who research CT colonography are starting to feel reasonably confident about its ability to pick up polyps of that size. With smaller polyps, however, accuracy decreases.

“At 5 mm or smaller, we don’t pretend to see them all,” McCarty said. “We have maybe 50% sensitivity at that size. Some are obvious, but some are just too small. It’s hard to put a firm number on it.”

How much that limitation matters has yet to be determined. Mayo researchers led by Dr. C. Daniel Johnson have completed an American College of Radiology Imaging Network (ACRIN)-sponsored trial, which directly compares virtual colonoscopy with the conventional version. Results have not been published. A 300-patient study led by Yee found that the CT technique matched colonoscopy for per-patient sensitivity, finding 90% of polyps 1 cm or larger and all carcinomas (eight out of eight) in a pool that included both symptomatic and high-risk patients. On a patient-by-patient comparison, overall sensitivity and specificity were 88% and 82%, respectively, for screening patients and 90.9% and 67% for symptomatic patients. The study was performed using single-slice CT scanners, but multislice units would likely improve performance, Yee said.

One hedge on performance standards will probably help virtual colonoscopy’s acceptance: There’s a growing consensus that it will need to be sensitive only on a per-patient basis rather than per polyp. One clinically significant polyp will be all that’s needed to send a patient to the gastroenterologist, who still must remove it and any others found along the way. Reaching that level of performance seems well within virtual colonoscopy’s grasp. Meeting standards set by nonradiologists is another matter.

“It’s debatable what a clinically significant polyp is,” said Dr. Christopher Gostout, a Mayo gastroenterologist and board member of the American Society for Gastrointestinal Endoscopy. “Setting 1 cm is a little too generous. Most gastroenterologists say 5 mm or larger is a cause for concern. Flat adenomas are becoming more common, and they can be several millimeters in size but have a high degree of dysplasia.”

Yet Gostout and many of his peers are cautiously optimistic about virtual colonoscopy. Aware that only 40% of eligible patients in the U.S. who actually have colon cancer get a colonoscopy, they see CT technology as a possible way to offer more palatable screening to the public. Faced with a personnel shortage, they may welcome a procedure that takes some screening off their hands, but they also could be overwhelmed if a flood of CT-detected polyp referrals came their way. That might not be bad news for them, however: Procedures to snip polyps pay better than screening.

Ideally, widespread use of virtual colonoscopy would create a collaborative process between the two specialties so that patients whose polyps are identified on CT can have them removed during the same appointment, avoiding another prep cycle. Gostout, McCarty, and colleagues already have such a system working at Mayo, as well as a reciprocal agreement that allows gastroenterology to send patients for CT if a colonoscopy fails.

“We’re already trying to anticipate patients who had a CTC and are prepped,” Gostout said. “We have several slots set aside daily for them, but if it takes off, we’ll face a serious issue of how to accommodate patients on the same day.”

Eliminating Elimination

Beyond accuracy, a number of factors could influence virtual colonoscopy’s future, among them incidental findings and ease of use. About one in three CT colonographies will show “incidentalomas” ranging from gallstones to aortic aneurysms. A Mayo study involving 264 patients found that 11% of patients had extracolonic findings that were clinically important, including large solid organ masses and pneumothorax. Such findings are a bonus that conventional colonoscopy can’t match and a prime reason that 2-D views will remain essential to a thorough reading.

“We saw a 1-cm pancreatic mass last week,” said Dr. Eduardo Balbona, chief medical officer at HealthScan. “Renal cell tumors are common, as are liver masses. There’s no solid benefit to extracolonic findings, but I suspect it’s not a bad idea.”

The problem, according to Dr. Seth Glick, is similar to that of polyp size: What determines a clinically significant finding? And what happens to patients who have borderline findings?

“People don’t understand that finding things isn’t necessarily good,” said Glick, a professor of radiology at the University of Pennsylvania. “If something shows up, the patients will get a serial study, but that won’t necessarily change their lives.”

If incidental findings include a lung nodule, for example, a patient may be sent for further scans to assess that mass. Even if it is benign, the finding may precipitate a series of follow-up CTs spread over years, increasing the patient’s lifetime exposure to radiation.

As with other scans, virtual colonoscopy’s incidental findings are duly noted and reported to the referring physician. Dr. Amy Hara, lead author of the extracolonic findings study, suggested that in practice the incidental findings don’t add up to an large burden in reading time or additional images.

“We report all extracolonic findings and send a note to the clinician if we find something we think is highly significant,” she said. “In our study, we did not find that an inordinate number of patients underwent additional unnecessary tests for insignificant disease. This result can vary widely from institution to institution, depending on the confidence of the reader and primary clinician.”

CAD

Within a few years, readers may get help from computer-aided detection algorithms designed to sort out folds and stool from actual polyps. As with human readers, CAD’s accuracy, as observed so far, decreases with smaller masses.

“For clinically relevant polyps, we’re getting 100% sensitivity, but our algorithm doesn’t detect smaller polyps as reliably,” said David Paik, a Stanford research fellow finishing his doctoral work on shape-detecting algorithm. “We know it can go at least as small as 5 mm. Last year, in a small trial, we got 100% sensitivity with 3.6 false positives per colon.”

So far, Stanford’s CAD algorithm is based only on surface shape: It detects spherical and hemispherical patterns in a 3-D reconstruction of the colon. Paik and colleagues are trying to reduce false negatives by comparing prone and supine images and by finding ways to minimize retained stool. Eventually, CAD could incorporate detection of other features, such as masses that extend outside the colon wall.

Dealing with the colon’s contents and finding ways to simplify or eliminate the prep process have become goals in themselves. The Viatronix system calls for a two-day cleansing diet prior to exam, along with barium to tag the remaining, liquified stool for “electronic cleansing.” At the 2000 RSNA meeting, researchers from Mayo presented preliminary data suggesting that a series of oral-contrast doses alone could tag stool adequately for digital subtraction in an otherwise unprepped colon. Such an advance, if workable, could generate interest from consumers unhappy not only with invasive exams but the broth-and-gelatin diet that precedes them.

With promising early results and some tentative support from outside radiology, virtual colonoscopy may face its greatest hurdle in the form of reimbursement. Many imaging centers that offer the service charge in the $800 to $1200 range for the procedure, but cost modeling has suggested that a combined professional and technical fee of $500 to $750 might be needed to make the scan competitive with its invasive counterpart.

The field still needs prospective studies in a general population rather than high-risk patients, along with solid evidence of high sensitivity for 1-cm polyps at the very least. It’s enough to keep researchers in the field busy for years, but private enterprise isn’t necessarily willing to wait that long.

“We’re working on studies that show equivalence to optical colonoscopy, and then we’ll push for reimbursement,” said Lt. Gen. Thomas Stafford, Viatronix’ chairman of the board. “I hope to see Medicare coverage within two years.”

Widespread popularity for the procedure could bring a set of problems radiology has seen before, according to Dr. Joseph Ferrucci, radiology chairman at Boston University.

“The prospect that virtual colonoscopy could become radiology’s next (step after) screening mammography is real,” he said during the Caldwell lecture at the American Roentgen Ray Society meeting in May.

It may not come to that. Gastroenterologists at Mayo last year published preliminary data on a stool test that may make cancer detection still cheaper and faster. Checking for 18 cancer gene markers in cells that slough off the digestive track, the test appeared to be as accurate as conventional colonoscopy. The population tested was small, however. A positive screen after such a test would still require imaging or other follow-up to find the source of the cells. Meanwhile patients could enjoy the same reassurance they now receive from colonoscopy, regular or virtual—without the prep.