ACRIN: Virtual colonoscopy sensitive for flat polyps

Oct 12, 2009

Refuting the idea that virtual colonoscopy (also known as CT colonography or CTC) is as blind as love to flat polyps, a new study of data from the National CT Colonography Trial (ACRIN 6664) found that CTC prospectively detected more than two-thirds of flat adenomas, and nearly 90% were visible in retrospect.

The study of data from nearly 2,600 participants at 15 study centers across the U.S. also found a less-than-1% prevalence of flat adenomas -- lower than some previous reports, researchers said October 1 at the American College of Radiology Imaging Network fall meeting in Arlington, VA.

"There's been increased interest and awareness in nonpolypoid or flat adenomas of the colon -- but the significance of these lesions is controversial because of conflicting data regarding the prevalence and the significance of the lesions, the sensitivity of various techniques for their detection, and even what the criteria are to define these lesions," said Dr. Jeff Fidler, associate professor of radiology at the Mayo Clinic in Rochester, MN.

Analyzing ACRIN 6664

Fidler, along with Dr. C. Daniel Johnson and colleagues, analyzed the ACRIN 6664 trial data to determine CTC's sensitivity for flat lesions, defined as polyps having a height-to-width radio of ≤ 50% with elevation ≤ 3 mm. They examined the prevalence and incidence of flat histology in the lesion, answering a secondary question in the original trial, Fidler said.

Each screening subject in the original trial underwent both CTC and colonoscopy, with colonoscopy serving as the reference standard. Readers were required to have extensive experience, or undergo training that included the identification of flat adenomas, and pass a qualifying exam prior to participation. Cases were randomly assigned to be interpreted using primary 2D or primary 3D reading algorithms.

The analysis included location, lesion size, and morphology of each polyp, Fidler said. To compare reading approaches, each case was later assigned to a second experienced reader who reviewed the primary 3D analyses in 2D and vice-versa. All colonoscopy images were reviewed by a single gastroenterologist to determine their morphology. Polyps deemed flat by colonoscopy or CTC was reviewed retrospectively for lesion matching, Fidler said.

"We put together a truth panel to look at the polyp and assess the size, look at the conspicuity of the polyp on different window settings, at 2D versus 3D [primary] interpretation, and then determine if the lesions that were not detected prospectively could be seen in retrospect and, if so, what were some of the potential reasons for missing these prospectively," he explained in his presentation. "We performed a retrospective review of the endoscopic photos and a retrospective review of CTC and then classified them according to histology to make sure they really met the restricted [flat-lesion] criteria that we define in this study."

Screening revealed a total of 374 adenomas or adenocarcinomas ≥ 5 mm in size, of which 19 (0.75%; mean size, 9 mm) met the study criteria for flat lesions. Of these 19, eight (42.1%) were classified as advanced adenomas: six due to polyp size, two due to the presence of high-grade dysplasia, and three because of the presence of a villous component, Fidler said.

Sensitivity for flat-lesion detection for combined 2D and 3D interpretation was 68% (13/19; confidence interval [CI]: 0.43-0.87). Sensitivities for each individual reading technique were as follows: 47% (9/19) for 2D and 32% (6/19) for 3D. These numbers did not differ significantly (p = 0.37), but "when you combine 2D and 3D, it improved sensitivity compared to 2D and 3D alone," Fidler said.

Sensitivity for adenomas larger than 10 mm (67%) did not differ significantly from the overall detection sensitivity (68%), he said. No significant differences in polyp conspicuity rank were found between lung and soft tissue or colon window settings. Most lesions were seen either at 2D or 3D but not both techniques.

"The majority of lesions were either equally conspicuous between 2D and 3D lesions, or more conspicuous on the 2D algorithm," Fidler said. "We could see four of six polyps in retrospect that were not seen prospectively, so technically 89% (17/19) of these polyps could be seen by CTC."

Study limitations

Study limitations included the presence of 96 false-positive lesions found at CTC that were coded as morphologically flat. And because these findings were not re-examined retrospectively, their exact lesion height is unknown, he said.

"None of these [CTC false-positive] patients had follow-up colonoscopy, so we're really not sure if these represent false-positive exams by CTC, or false negatives by colonoscopy," he said. "It is likely that some of these cases were missed by colonoscopy."

As for other limitations, some publications have recommended even more restrictive criteria for flat lesions, such as polyp size ≥ 3 mm rather than ≥ 5 mm, or a height-width ratio of less than 50%, he said. "If we had used these criteria, we would have had even a lower prevalence than we had," Fidler said.

Finally, he said, some publications have suggested that the generally lower prevalence of flat lesions seen in Western countries stems from the typical colonoscopy technique that does not use dye spray to increase lesion conspicuity, in contrast to common practice in Asia. Because no dye spray was used for colonoscopy in this cohort, the study may have underestimated the prevalence of flat lesions, Fidler said.

In this screening cohort, "nonpolypoid flat adenomas had a very low prevalence, and 42% did have advanced adenomatous features," Fidler said. "These polyps are more difficult to detect [at CTC] than routine polypoid lesions, but the majority are visible," he said. "And an interpretation technique incorporating both 2D and 3D algorithm adds to sensitivity."

"I think this is a very important result," commented ACRIN chair Dr. Mitchell Schnall, Ph.D., professor of radiology at the University of Pennsylvania in Philadelphia. "I think the flat adenoma is not a reason -- there may be others people can argue -- but the flat adenoma is not a reason for CTC not to be a screening test for colorectal cancer."

As for why some visible flat lesions were missed, "you could see retrospectively that it wasn't prep or distention, so I think a lot of it is training. And I think a lot of it may be just individual idiosyncracies," Fidler told AuntMinnie.com. "Some people may have an eye for picking them up. But I think it's promising that they can see them."

By Eric Barnes
AuntMinnie.com staff writer
October 13, 2009

Virtual colonoscopy CAD improves flat-lesion detection, June 5, 2009

ACRIN trial shows VC accuracy comparable to optical colonoscopy, September 18, 2008

Minimal-prep VC may miss more flat lesions, April 17, 2008

Flat colorectal tumors relatively common in U.S. population, linked to cancer, March 5, 2007