| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Eugene W. Caldwell Lecture |
1 Department of Radiology, Boston Medical Center, Boston University School of Medicine, 88 E. Newton St., Boston, MA 02118.
Received March 26, 2001;
accepted after revision April 24, 2001.
Presented at the annual meeting of the American Roentgen Ray Society,
Seattle, May 2001.
Abstract
 Top Abstract BackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
|
Current research is directed to improving the speed of image analysis, use of computer-aided methods for polyp detection, MR colonography, and oral contrast tagging of fecal contents to avoid the need for rigorous bowel preparation and thus improve patient compliance.
Radiologists introducing virtual colonoscopy for colon cancer screening should use several strategies: They should create evidence-based data by conducting well-conceived clinical trials, establish public accountability by developing performance standards and accreditation programs through the American College of Radiology, seek support from advocacy groups dedicated to prevention and cure of colorectal cancer, and foster new paradigms with gastroenterologists for improved colon cancer prevention strategies.
|
TopAbstractBackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
Nonetheless, radiologists have long been advocates of colorectal cancer screening using the double-contrast barium enema for detection of colorectal polyps. We have emphasized the accuracy, safety, availability, and low cost of the double-contrast barium enema, especially compared with the various endoscopic methods, particularly colonoscopy [4, 5]. There have been two annual orations at the Radiological Society of North America on colon cancer screening in the last decade [6, 7]. Barium enema technique and interpretation have been emphasized in innumerable refresher courses at national meetings and in review articles in our major journals. The American College of Radiology has a standing Committee on Colon Cancer that has sponsored two national workshops on colon cancer screening (1988, 1990), created a Practice Standard for performance of the barium enema, and produced an educational slide set [8]. Later this year, the American College of Radiology will pilot-launch a practice accreditation program in which the quality of double-contrast barium enema studies will be a critical element.
Yet, despite these earnest intentions, the message of radiologists and the use of barium enemas have both been largely ignored. My friends and colleagues in gastrointestinal radiology won't necessarily thank me for this, but it is increasingly clear that the use of barium enema as a primary screening test for colorectal neoplasms in healthy asymptomatic adults is nearly nonexistent. Similarly, lower gastrointestinal bleeding is almost always first investigated by colonoscopy. This is not to suggest that the barium enema is not used at all. Indeed, in patients in whom there is a question of large-bowel obstruction, or need for localization of colonic disease in a preoperative patient, or need to assess the status of a colon anastomosis, a contrast enema is still useful. In most practice environments, however, primary care physicians, their patients, and the general public have been persuaded to think first of sigmoidoscopy or, increasingly, of colonoscopy for colorectal cancer screening. Witness national television anchor Katie Couric's live on-air colonoscopy last year.
Why is this? There are several reasons. Our colleagues in gastroenterology have produced more and better scientific data on preventing colorectal cancer. They are more conversant with the natural history of colorectal polyps, the performance of various screening test strategies, and societal cost implications for reducing colon cancer mortality [9,10,11,12]. To take one recent example, in the highly regarded National Polyp Study, the double-contrast barium enema has been defined as an inferior examination for detecting colon polyps [13]. Despite some ex post facto criticisms of that study raised by some radiologists, we have produced no controverting contemporary evidence as to the efficacy of the double-contrast barium enema, nor is it likely we ever will.
The leadership of gastroenterology organizations works closely with national health policy groups, major research funding agencies, and colon cancer support groups. As a result, gastroenterologists have dominated most medical and public discourse on colon cancer prevention. In recent deliberations over national screening guidelines, they exerted considerable political influence. For example, in their lobbying efforts to add colon screening to Medicare benefits several years ago, gastroenterologists spent as much effort denouncing the barium enema as they did advocating colon screening. At one point, they even threatened to sue Medicare if payment for screening barium enema was approved—a stark display of turf rancor. Medicare ultimately did include the double-contrast barium enema as a reimbursable test, and the endoscopists did not sue. Nevertheless, the impression that the gastroenterologists dominated colon screening was unmistakable; and their defense of colonoscopy, an important source of income, was successful.
At the same time, the pace of revolutionary technologic developments in diagnostic imaging has been breathtaking, and the scientific energies of our young researchers have been understandably diverted from barium, with remarkable dividends for patient care. As a result—and not to put too heavy a point on it—radiologists are presently essentially out of the colon screening business.
In the last several years, however, a new imaging technique has evolved that could play an important role in the future of colorectal imaging—namely, virtual colonoscopy [14, 15]. Using conventional CT scanning and computer postprocessing of the data, a simple, safe, and truly novel method for examining the interior of the colon has emerged. Data from an increasing number of centers show results that are nearly equal to colonoscopy for detection of significant polyps and far better than the air-contrast barium enema [16,17,18,19,20,21]. We at Boston University, along with workers from many other institutions, have been actively investigating this technique for the last several years; and in this lecture, I would like to summarize its current status, future directions, and overall potential impact on the prospects for improved colorectal cancer prevention.
Technical and procedural details of virtual colonoscopy have been the subject of several useful reviews in the last 2 years [22,23,24,25]. However, there are ample reasons to suggest that a more general review of virtual colonoscopy is warranted at this time. The rapid dissemination of multidetector CT technology will make faster and more accurate colon imaging available at the local level worldwide. Greater computer processing power and cheaper data storage capacity will make image rendering faster, simpler, and more affordable. New as well as established commercial vendors are developing specialized software applications and novel bowel preparation products for virtual colon imaging, and the number of radiology publications on the subject has increased exponentially in recent years. Radiology researchers have obtained numerous National Institutes of Health and professional society research grants on virtual colonoscopy and nearly a dozen United States patents have been issued or are pending. Three large multi-center clinical trials of virtual colonoscopy are underway, one sponsored by the American College of Radiology. Two international symposia devoted exclusively to virtual colonoscopy have been held—both in Boston—in October 1998 and October 2000 [26, 27]. A third is scheduled for the spring of 2002, again in Boston. Although present results are still early and much of this activity and critical work remains to be done, the convergence of all these elements suggests that the development of virtual colonoscopy will follow a steep upward trajectory for the foreseeable future.
Virtual Colonoscopy: The Promise
|
TopAbstractBackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
Descriptive names for the procedure initially varied, including terms such as "CT pneumocolon," but "CT colonography" became the accepted scientific terminology [31]. However, "virtual colonoscopy" remains useful for marketing to patients, providers, and the public, just as the term "CAT scan" still conveys a familiar and easily understood meaning.
Technique
Broad consensus has evolved on the fundamentals of examination technique
[21,22,23,24,25].
Thorough colonic cleansing, essential for any direct examination of the
colonic mucosa, is readily achieved by a conventional 48-hr lowresidue diet
and an over-the-counter preparation of phospho-soda and bisacodyl (C. B.
Fleet, Lynchburg, VA). This is preferable to the usual conventional
colonoscopy preparation of polyethylene glycol electrolyte solution (Golytely;
Braintree Laboratories, Braintree, MA) because less residual fluid remains
[32]. The critical importance
of bowel preparation cannot be overemphasized. Although achieving it is the
cause of most patient reluctance, a clean colon is potentially more important
to ultimate success than the array of technical refinements in scanning and
data analysis that researchers are developing. After the insertion of a rectal
tube, the colon is gently insufflated with room air to the maximum level
tolerated by the patient (c. 50 bulb puffs) and colonic distention is
monitored by the CT scout radiograph. Although carbon dioxide can be
substituted to reduce crampy discomfort, this adds additional cost and
complexity. Similarly, IV antispasmodic agents (glucagon, Buscopan
[N-butylscopolammonium bromide; Boehringer, Ingelheim, Germany]) do
not routinely promote better distention of colonic segments and may cause
excessive reflux of gas into the small bowel, creating subsequent artifacts on
3D-rendered images. Thus, our method at Boston University stresses simplicity,
using room air only with careful attention to patient comfort.
Helical CT scanning is performed in a single breath-hold using 5-mm collimation and reconstruction intervals of 2-3 mm. Acquisition is repeated with the patient in the prone position to redistribute the gas into previously collapsed segments, which significantly increases the accuracy for polyp detection [19, 21, 33, 34]. CT tube currents can be operated at lower levels (100-140 mA) than with conventional abdominopelvic CT protocols (250 mA) because of the high inherent contrast between air and colon wall [35]. When multidetector CT scanners are available, data are acquired with tighter collimation of 2-3 mm and reconstruction intervals of 1-2 mm for higher spatial resolution, although cumulative radiation dose and data storage requirements may become limitations [36].
Image display parameters are also being rationalized, and the early attraction of simulated endoscopic endoluminal reviewing of rendered images has temporarily faded with the reality of long data processing times. Most groups rely initially on conventional 2D (or multiplanar reformatted) axial images, reserving 3D endoluminal views for problem solving [37,38,39]. The axial 2D images are displayed at lung windows to optimize air—soft-tissue interfaces. "Lumen tracking" is a term used to describe visually following the course of gas-distended colonic loops from slice to slice [17]. This analytic technique ensures confident inspection of the entire longitudinal course of the colon lumen. Trackball scrolling of axial slices improves the radiologist's efficiency, especially when assessing the larger numbers of slices produced by multidetector scanners. The 2D imaging features of polyps and the spectrum of colon pathology are common knowledge (Fig. 1).
|
For 3D rendering, the CT data are downloaded to a workstation equipped with specialized volume-rendering software. Data are reconstructed with sufficient overlap to optimize 3D image display (>50%). Volume rendering is generally preferred over surface rendering because of the inherent advantage of displaying the entire range of voxel attenuation values [19,20,21,22,23,24, 40]. Although the demands on computer processing power are greater than with surface rendering, the rapidly decreasing cost of computer power has largely removed this barrier. Volume rendering has the additional advantage of retaining data deep in relation to the colonic wall so that extraorgan abnormalities can be appreciated. The addition of perspective to the volume-rendering algorithm allows better differentiation of near-field from far-field objects essential for endoluminal navigation. The endoluminal fly-through can be performed at near-real-time speed, simulating the effect of conventional colonoscopy, and has the additional advantage of both antegrade and retrograde viewing to examine hidden surfaces of folds and flexures.
Interpretation
Image analysis uniquely blends information from several diverse sources,
including conventional double-contrast barium enema, abdominal CT scanning,
conventional colonoscopy, and computer graphic image processing. However, many
of the imaging features, especially at 3D endoluminal display, are entirely
novel and have been elaborated in recent reviews
[22,23,24,25].
On 3D endoluminal views, normal haustral folds appear as variably translucent
structures with well-defined opaque free edges. The degree of translucency of
normal folds can be altered by changing preselected opacity settings. As with
conventional colonoscopy, identification of position within the axial length
of the tubular colonic lumen depends on classic anatomic features. The
descending colon is recognized by its straight tubular course, and the
transverse colon, by its typical triangular fold configuration (Fig.
2A,2B).
The cecum is identified by recognition of the ileocecal valve as a fold or
ridge, and the appendiceal orifice is seen as a well-circumscribed ringlike
depression in the colonic mucosa. Accurate position can be determined at the
workstation by linked display of multiplanar reconstructed images to the
endoluminal fly-through views—a considerable advantage over conventional
colonoscopy. The illusion of depth perspective is achieved by the variably
darkened edges of mucosal surfaces as well as blurring of near-field objects.
Real-time movie loop display of the 3D data set allows for antegrade as well
as retrograde viewing, which is especially helpful in inspecting the hidden
surfaces of haustral folds for small polyps.
|
|
Polyps appear as well-defined rounded or oval intraluminal projections best seen in profile because of the prominent edge enhancement of their free margins (Fig. 3A,3B). Carcinomas appear as larger intraluminal masses with nodular surface configuration, but they often show loss of fine surface detail because of near-field blurring and the smoothing effects inherent in volume rendering (Fig. 4A,4B). Luminal narrowing is a common feature of constricting carcinoma. In such cases, a major advantage of virtual colonoscopy is the ability to examine the air-distended proximal colon on 2D and to navigate through the stenotic segment at 3D, especially when the lesion cannot be traversed by the endoscope [41].
|
|
|
|
Just as with conventional barium examination, sigmoid diverticular disease presents a special difficulty because of associated muscular hypertrophy, spasm, and poor distention [19, 22, 23]. Nondistended colon segments may simulate a constricting neoplasm (false-positive) or obscure a polyp (false-negative). Thus, the importance of complete air insufflation, as well as both supine and prone acquisitions to redistribute air, is now well accepted. Diverticular orifices appear sharply edged circumferentially, whereas polyps are sharply defined only along their free profile margin. Stool-filled diverticula may be confusing because of their presentation as localized polypoid lesions bulging from diverticular openings.
Combined 2D and 3D image analysis is particularly helpful in problem solving and lesion characterization [19, 22, 23, 25, 37,38,39]. Complex and bulbous haustral folds are particularly frequent at flexures and may simulate small polyps on 2D views, but are readily recognized on the endoluminal 3D view. Retained stool may simulate a polyp, especially on 3D, but on 2D axial images fecal material may show typical pockets of retained air and move to the dependent wall on prone views. Similarly, retained barium, lipomas, or ingested medications are readily recognized at standard 2D imaging by their characteristic CT attenuation features. It is important to reiterate that detection of a "polyp" on endoluminal 3D images does not allow differentiation from the several other causes of false-positive findings (stool, retained pill, lipoma, adjacent extracolonic mass, or bowel loop). Thus, referral to the corresponding 2D image for confirmation and characterization is essential [39]. Accordingly, most software rendering packages now offer some form of 2D-3D interactive display currently regarded as a requisite for efficient image analysis. One software vendor whose product emphasizes primary diagnosis by 3D endoluminal navigation is testing a method for tissue density analysis to directly characterize polypoid lesions without using traditional 2D image attenuation parameters.
Radiologists learning endoluminal interpretation soon recognize artifacts typical of 3D volume rendering [22, 23, 25]. These include retained colonic fluid appearing as "rivers" of illdefined opacities along the dependent colon wall, stairstep artifacts as a series of concentric rings where there is a rapid change in bowel contour, and shine-through artifactual "pseudolucencies" in the bowel wall caused by visualization of adjacent air-filled loops of small bowel. Endoluminal image display has a further complexity, that of optional opacity and color parameters. Presently, vendors of 3D rendering software show images with pink and orange tints, while many academic groups use the simpler two-tone black-on-white format. As yet, there is no consensus.
Results
Predictably, the performance of virtual colonoscopy depends primarily on
the size of the target lesion. In vitro, the threshold for reliable detection
of small lesions is approximately 5 mm
[42,
43]. In in vivo patient
studies with colonoscopic configuration, polyps greater than 1.0 cm are
detected with sensitivity and specificity approaching 90%, with sensitivity
falling to 50% for polyps at the 5-mm level. In the first blinded prospective
study, Hara et al. [16] from
the Mayo Clinic reported on 70 patients with 15 polyps greater than 1 cm with
sensitivity of 70%. In 1997, Royster et al.
[19] from Boston University
reported on 20 patients with 22 polyps greater than 1 cm, achieving
sensitivity for detection of 100%. Subsequently, authors from these same two
institutions reported much larger series with similar results. In a
prospective 1999 study with colonoscopic correlation of 100 patients at high
risk, Fenlon et al. [21] from
Boston University detected 20 (91%) of 22 polyps greater than 1 cm, 33 (82%)
of 40 that were 6-9 mm, and 29 (55%) of 53 that were 5 mm or smaller.
Importantly, these results were obtained with commercially available software
from a major CT vendor during 1997 and 1998, and, surprisingly, they remain
the best yet reported.
When analysis was extended from a perpolyp to a per-patient basis, sensitivity and specificity for lesions greater than 1 cm were 100%. The per-patient concept considers the clinical impact of detecting no, one, or more than one polyp in terms of management decisions for observation or intervention. In a still larger prospective study from the Mayo Clinic of 180 patients with 420 colonoscopically proven polyps, Fletcher et al. [34] reported sensitivity of 47% for identification of 142 polyps of 6-9 mm and 75% for 121 polyps greater than 10 mm. Per-patient sensitivities were 88% and 85%, respectively. In the largest and most recent series yet reported, Yee et al. [44], of the University of California at San Francisco, studied 300 patients and obtained a sensitivity of 90% in 82 polyps greater than 10 mm and 80% in 141 polyps 6-9 mm. The per-patient sensitivity for the 10-mm size was 100%. These important early studies have established a key benchmark for the performance of virtual colonoscopy and are summarized in Table 1. Several thousands of patients have undergone virtual colonoscopy at other academic medical centers worldwide, giving similar promising results, as recently reported at the Second International Symposium on Virtual Colonoscopy in Boston (October 2000) [27]. Publication of their results is awaited.
|
False-positive results (specificity percentage) have occasionally occurred in segments of the colon containing diverticular disease and poor distention, or where thickened and complex haustral folds were misinterpreted as polyps. Other causes of false-positive results may include undissolved medications, metal artifacts, and, of course, retained stool. In the Boston University study, there were only two false-positives (2%) for polyps over 10 mm and eight false-positives (7%) for polyps 6-9 mm. Because false-positives precipitate unnecessary follow-up tests and represent a potential limitation, the specificity of virtual colonoscopy will be carefully studied in the foreseeable future.
Hyperplastic polyps are far more prevalent than tubular adenomas when polyp size is less than 5 mm. Hyperplastic polyps are also less readily detected than adenomatous polyps of the same size, presumably because their soft consistency allows flattening against the surface of the air-distended colon [21]. Overlooking them, however, is of no clinical importance, because they are not precancerous, and removing them at colonoscopy, now a widespread practice, conveys no benefit. One area of concern, nevertheless, is the true "flat adenoma," which, although uncommon, may exhibit a more aggressive biologic behavior, evolving into invasive cancer more rapidly. Imaging flat adenomas could pose problems for CT colonography, but no published results are yet available.
Frank colon cancers can be detected with a sensitivity of 100% and no false-positives [17,18,19,20,21]. Moreover, in the specific instance of occlusive distal colon carcinoma, virtual colonoscopy provides the specific advantage of being able to identify synchronous neoplasms in the proximal colonic segments that could not be inspected by conventional endoscopy. In a study of 34 patients with occlusive lesions of the distal colon at endoscopy, Fenlon et al. [41] found 17 additional colorectal neoplasms in the proximal colon, including two adenocarcinomas and five polyps greater than 1 cm. In a similar subsequent study, Macari et al. [45] studied 20 patients referred because of incomplete colonoscopy and found virtual colonoscopy preferable to the standard technique of barium enema for completing the colon examination. The gas-distended colon in the colonoscopy patient is ideally suited for virtual imaging, whereas it often prevents adequate filling at barium enema. Thus, failed or incomplete colonoscopy has become the first established indication for CT colonography.
Evidence Basis
Medical decision making, especially involving new technologies, requires
substantiation by clinical evidence-based data
[46]. Hierarchal levels of
evidence have been defined, with the highest quality data obtained from
randomized clinical trials. Early workers typically produce observational
correlative studies, giving lower order evidence. Although the basic
principles are often established, design flaws and bias are common, and the
technology is usually immature.
Hypothesis-driven prospective randomized clinical trials give more robust data. At least three large trials of CT colonography are under-way in the United States, each with a different focus. In a study funded by the National Cancer Institute and sponsored by the American College of Radiology Imaging Network, interpreter performance is being assessed from a study case library of patients with polyps. This study includes three image rendering packages that are being studied separately, including one each from a major CT vendor, a software specialty company, and a university (a proprietary version). In the study, 18 radiologists, most with some prior experience in virtual colonoscopy interpretation, are reviewing images at a central reviewing station, with special emphasis on pretest training and posttest performance scores. Some results may be available by mid 2001. In more traditional randomized clinical trials sponsored by the Medical University of South Carolina and Wake Forest University, prospective studies of multislice CT performance are being carried out with polyps targeted to the 6-mm threshold level. In these trials, imaging studies will be closely correlated with conventional colonoscopy using videotape recording during endoscopy and then reexamination of suspect areas of the colon in real time when virtual colonoscopy shows suspicious findings. A third randomized clinical trial, also using multidetector CT technology, is being sponsored by Duke University.
Virtual Colonoscopy: Research Directions
|
TopAbstractBackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
20 min), patient
reluctance to undergo colon cleansing (poor compliance), and concerns over the
CT radiation dose when screening large populations. Researchers are
intensively seeking solutions to these limitations. The next section
summarizes these efforts.
Computer-Aided Diagnosis
Various computer processing techniques are being investigated to shorten
interpretation times and to increase diagnostic accuracy, both key
requirements for application in a population screening mode.
Surface mapping.—Computer methods for postprocessing volumetric data sets are being investigated to ensure that the entire colonic mucosal surface has been visualized [47,48,49,50,51]. Two-dimensional, zoomed axial multiplanar reformatted reconstructions of the air-distended colon have been mathematically unraveled, with the colon bisected open along its longitudinal axis. This permits more efficient inspection of the colon as a flat contour rather than a tubular structure to simulate a gross "virtual" pathology specimen [47]. However, this method lacks the attenuation information of standard 2D reconstructions, and supine and prone comparisons are difficult. In other studies, a panoramic endoscopic display of a colon phantom has been used in which an array of six image panels around the colonic circumference are laid side by side for a comprehensive view [50]. Yet another method for improved surface visualization has been drawn from cylindrical display Mercator techniques used in map making (cartography) [51].
Center-line navigation.—Because of the tortuosity of the colon, computer derivation of a colon central axis before analysis of endoluminal 3D data is being developed to decrease the time required for the interpreters' interactive fly-through review at the workstation (Fig. 5). A variety of advanced computer graphic methods are used, such as skeletonization techniques, steepest descent algorithms, and voxel growing methods [52,53,54,55]. These approaches have been adopted by several vendors of virtual colonoscopy software.
|
Automated polyp detection.—Mathematic computer algorithms measuring surface elevations or curvatures (shape-based) have been applied to the isotropic data sets obtained at virtual colonoscopy by computer scientists at several centers [56,57,58] (Fig. 6). Summers et al. [56] used computer-generated analysis to detect six of six polyps 10 mm or greater in a colon phantom. Paik et al. [58] applied similar edge detection algorithms to eight patients with 21 polyps at colonoscopy and found 76% sensitivity, with 7.7% false-positives per colon. Increased sensitivity was achieved as false-positive rates were allowed to increase. In a subsequent study of CT colonography data from 20 patients with 50 polyps, Summers et al. [57] reported 71% sensitivity for computer detection of polyps 10 mm or greater (10/14 polyps detected). Performance decreased for smaller polyps and with poor colon distention. Applications of neural network or artificial intelligence-aided detection have been clinically validated in other settings—for example, in studies of screening mammography [59, 60] and lung nodule detection [61]. In those instances, sensitivity for lesion detection was enhanced, although, again, a significant false-positive rate required direct radiologist assessment of positive studies. Similar effects are likely in colon polyp detection.
|
Stool Tagging
In an effort to reduce the unpleasantness of conventional colon cleansing
and offer a "prepless" study, several groups of investigators have
explored the potential for marking and identifying residual fecal material. In
early protocols, the patient ingests oral contrast agents over 2-3 days, and
the tagged fecal material is then "electronically cleansed" from
the soft tissue of the colon wall by computer algorithms
[34,
62,63,64,65,66,67].
Both barium sulfate and meglumine diatrizoate (Gastrografin; Bracco
Diagnostics, Princeton, NJ) have been used. On axial 2D images, the
contrast-impregnated stool or liquid content has distinctly higher CT
attenuation than polyps (Fig.
7A,7B).
Fletcher et al. [34] of the
Mayo Clinic administered iodinated contrast material orally the evening before
colonography to identify polyps submerged in high-attenuation colonic fluid.
In further studies at the Mayo Clinic, Callstrom et al.
[64,
65] used orally administered
barium sulfate to mark fecal material directly. A regimen of divided barium
doses over 48 hr without other dietary restriction labeled stool consistently,
and polyps were readily detected on standard 2D image analysis. Researchers at
the State University of New York at Stony Brook are incorporating oral barium
into a commercially prepared low-residue diet
[66]. Zalis and Hahn
[67] of the Massachusetts
General Hospital evaluated postprocessing algorithms that allowed subtracted
images to remain within the DICOM (Digital Imaging and COmmunications in
Medicine) standard and thus available on commercial display systems. The aim
of this work—to ultimately entirely eliminate the need for physical
colon cleansing—could be of momentous importance to the adoption of CT
colonography as an acceptable primary colon screening technique. However,
timetables for product development and testing are unknown.
|
|
Contrast Enhancement
Visible enhancement of colonic wall and polyps was observed by Morrin et
al. [68] after the IV
administration of iodinated contrast material. The effect was most helpful in
visualizing polyps submerged in retained colonic fluid; theoretically, the
benefit could obviate routinely obtaining prone acquisitions. However, the
risk and cost of IV contrast administration probably preclude its use as part
of a screening colonoscopy protocol for large populations.
MR Colonography
Driven by public concern about medical radiation exposure, European
investigators are developing robust, albeit complex, techniques for virtual MR
colonography
[69,70,71].
Colon distention and image contrast are obtained by on-table administration of
a 1.5- to 2.0-L saline-gadolinium rectal enema solution. Colon filling is
monitored by snapshot 2D gradient-recalled echo sequences. Because MR imaging
remains a motion-sensitive technique, bowel peristalsis is reduced by IV
administration of antispasmodic drugs. Fast breath-hold T1-weighted 3D
gradient-recalled echo sequences (similar to MR angiography) are used for data
acquisition and are analyzed by standard colonographic 2D, multiplanar
reconstruction, and 3D endoluminal fly-through methods. Results show accuracy
for polyps greater than 1.0 cm similar to CT colonography and conventional
colonoscopy—that is, approximately 90%
[70]. Fecal tagging techniques
are also being investigated, as with CT colonography
[71]. An oral gadolinium
contrast product is given orally 1 day before the study to be incorporated
into residual stool material. MR colonography remains a novel initiative, but
it has yet to compete with CT techniques in the United States. The quality of
3D-rendered MR data sets is inferior to CT for spatial evaluation, and, at
present, the method remains somewhat cumbersome (gadolinium enema, IV
medication), complex (MR imaging pulse sequences), and costly (MR vs CT). The
only group to study MR colonography in the United States to date has used room
air insufflation to simplify the technique. A pilot study of seven patients
examined by short-TE multislice half-Fourier acquisition single-shot turbo
spinecho (HASTE) breath-hold acquisitions gave good to excellent image
quality, detecting one 1.5 cm polyp, missing one polyp of 6 mm
[72].
The Significant Polyp: Defining the Target
|
TopAbstractBackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
Although it is widely accepted that most colorectal cancers arise from precursor tubular adenomas, colonic polyps are common after age 50 (50% of adults), increasing even more in prevalence with advancing age [73]. However, only about 3% of all adenomas ever become malignant [9, 10]. The likelihood of any given polyp being malignant (dysplasia, invasive carcinoma) is closely related to size, because only 1% of adenomas less than 1 cm will harbor invasive cancer [10]. This is a logistic dilemma for colon cancer prevention strategies, because more than 70% of all discovered polyps are less than 1 cm. Moreover, most of these small polyps are not even true neoplastic adenomas, because they show only hyperplasia histologically (50% of polyps <5 mm and 30% of polyps 5-9 mm) with no malignant potential [9, 10, 73]. Because the adenoma-carcinoma transformation is gradual, taking 10 years or more, and morphologic macrofeatures of smaller polyps are of little diagnostic value, the low likelihood that an individual subcentimeter polyp will evolve over time into a frank cancer becomes a limitation for achieving cost-efficient mortality reduction by most current screening strategies. Some clinical implications of these natural history features of colorectal polyps are summarized in Table 2.
|
As a result of these newer insights, the identification and removal of only the "significant polyp" in colorectal cancer prevention programs is an increasingly popular mantra. As Glick [4] has noted, the word "polyp" merely refers to a mucosal elevation or protuberance, not necessarily a precancerous neoplasm. Indeed, recent authors, including investigators in the National Polyp Study, now recognize the benefit of characterizing adenomas more explicitly as "advanced adenoma," defined as being polyps greater than 1 cm or containing villous or dysplastic elements at histology [12, 74]. The distinction between advanced adenoma and tiny innocent hyperplastic polyps is critical to the assessment of cost-benefit implications of various test strategies for finding and removing subcentimeter lesions. Thus, although it may be justifiable to remove all identified polyps once an invasive procedure such as colonoscopy is undertaken, the rationale for an aggressive strategy using endoscopy merely for detecting polyps is increasingly challenged. As Glick has stated, "reducing mortality from colorectal cancer should be accomplished by maximizing success at a minimum of expense and risk to the public" [4].
Screening by Imaging: New Paradigms
|
TopAbstractBackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
The first question is whether conventional endoscopy should even be performed for polyp removal. Endoscopic screening methods have focused on detecting small polyps and interrupting the polyp-carcinoma progression. But for patients having virtual colonoscopy, reassurance that there is no obvious occult cancer lurking in their bowel may be sufficient. Discovery of a small premalignant precursor is clearly important but may not require immediate action. When endoscopy is used as a primary test, its diagnostic and therapeutic functions are blended, often admittedly a benefit. Whatever polyps are found are removed. However, using virtual colonoscopy, patients and their physicians may now noninvasively obtain reliable data on the status of the colon mucosa before deciding whether to undergo formal colonoscopy. As a result, the diagnostic and therapeutic functions of colonoscopy can thus now be differentiated, and if patients can be apprised of the relative insignificance of small subcentimeter polyps detected on virtual colonoscopy, they may defer further interventions. In other diseases, patients are now accustomed to obtaining data and making choices. For example, men with newly diagnosed prostate cancer may research their prostate-specific antigen values, biopsy results, and imaging findings, and consult with several specialists before deciding among surgery, radiation, castration, or no treatment at all. The decision to do nothing about a newly discovered small colon polyp is already happening in our practice (Fig. 8). With objective documentation by imaging, the location and size of a visualized polyp can be recorded for comparison on a subsequent follow-up study, although appropriate intervals for repeated colon imaging have not been determined.
|
A further question is whether very small polyps (<5 mm) should even be reported at all. The higher resolution of multislice CT will likely show more of these tiny excrescences, far more than ever found with barium enema. Experienced radiologists are familiar with the daily dilemma of not formally recording small, clinically insignificant findings in official radiology reports, but few are likely to admit that on the record. Or, for that matter, should full reimbursement be provided when little tags of benign hyperplastic mucosa are excised? Here, obviously, the discussion might get more contentious.
But what if virtual colonoscopy shows a polyp that should be removed? Patients will not welcome the news that a second colon-preparation ordeal is in the offing. Data from a Veterans Administration colonoscopy screening study of 3500 asymptomatic adults showed a prevalence of advanced adenomas or cancers of 7.9% [74]. Adding in the rare potential for a false-positive interpretation of the image data set for a lesion at the 1.0-cm size range, the overall likelihood of a therapeutic colonoscopy being required is approximately 10% at the most.
Herein lies a possible blueprint for a new model for colorectal cancer screening programs of the future—a unique cross-referral collaboration between gastroenterologists and radiologists in real time. The occasional need for same-day CT colonography study on a patient with an incomplete colonoscopy and a well-prepared gas-distended colon requires immediate access to the CT scanner to avoid a second prep. Similarly, a positive virtual colonoscopy for a significant polyp might warrant an urgent conventional colonoscopy for the same reason. For obvious reasons, the potential 10% likelihood that a second colon cleansing will be required could be construed as a limitation of virtual colonoscopy. Provision for online or nearly immediate radiologic interpretation of virtual colonoscopy studies would be an inducement if not a clear necessity. Thus, design of clinical facilities, examination schedules, and staffing resources would be tightly integrated in a specialized multidisciplinary colon cancer screening program of the future.
An integrated multidisciplinary program will lack the usual access, speed, and convenience features of a conventional imaging-only screening examination. Moreover, there is the further analogy to screening mammograms, which have been widely accepted as a simpler and very different study than diagnostic mammograms. As with screening mammography, virtual colonoscopy could in theory also be done without a radiologist being present, in a high-volume technologist-only facility, with scan data batch-interpreted off-line and after the patient has been released. Because practical methods for population screening require simple, streamlined protocols to ensure patient acceptance and high throughput, the pure imaging model for screening will undoubtedly be appropriate in some situations. The two models are shown in Table 3. Predictions at this stage as to how clinical practices will evolve are entirely speculative.
|
Issues for Clinical Screening
There is consensus that CT colonography must give acceptable results in
populations considered at average risk (as opposed to high risk) for
developing colorectal cancer before its wide dissemination as a primary colon
screening strategy could occur. Assuming the high sensitivities and
specificities reported by early workers can be reproduced, other issues will
need refinement.
Incidental Extracolonic Findings
|
TopAbstractBackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
|
Cost—Benefit Issues
Outcomes of disease screening are measured with progressively strict
categories of reduction in cancer deaths, years of lives saved, cost per saved
life, and cost per quality of life years saved. Among the many factors
affecting cost—benefit analyses are procedure unit cost, false-positive
rates, test complications, patient compliance, and, of course, sensitivity for
the index or target lesion. Many different combinations of tests and test
strategies have been deemed acceptably cost-effective for colon cancer
screening [1,
76,77,78,79].
In classic work more than 10 years ago, Eddy
[1] assessed efficacy using
computer-based Markov modeling of a variety of tests and time-interval
combinations (e.g., fecal occult blood plus sigmoidoscopy every 5 years,
double-contrast barium enema every 10 years, and colonoscopy every 10 years).
Parameters for the cost-effective use of virtual colonoscopy as a primary
colorectal cancer screening method have been derived recently from similar
computer-based modeling by Sonnenberg et al.
[79]. Their computations
suggest that for virtual colonoscopy to be competitive with conventional
colonoscopy, it would have to be offered at approximately one half the cost
and secure a compliance rate among patients willing to undergo the test of
about 15-20% better than conventional colonoscopy. Not only does that
requirement appear eminently achievable, but such modeling techniques suggest
an appropriate reimbursement strategy for virtual colonoscopy of approximately
$500-750 combined technical and professional fees. Indeed, Saini et al.
[80] recently calculated the
actual cost of unenhanced abdominal CT at less than $200.
Radiation Dose
The prospect of population screening by CT requires careful consideration
of radiation dose, especially with the dissemination of multidetector CT and
its potential for "dose creep"
[81]. Because both the supine
and the prone requisitions are routinely used to redistribute colonic air and
fecal material for improved overall polyp detection accuracy, the use of
protocols with low tube current is receiving increased attention.
Investigators have examined low-dose protocols using tube currents of 50-70 mA
for both single-slice and multislice helical acquisitions. Using single-slice
acquisition, Hara et al. [35]
found no loss in sensitivity for detection of polyps at the 5-mm level. Those
authors calculated the radiation dose at 70 mA (187 mrad [1.87 mGy] for men,
285 mrad [2.85 mGy] for women) to be 75% lower than the dose for a standard
abdominopelvic CT examination. With multislice technology, the use of a cone
beam rather than fan beam geometry increases the width of the radiation beam
profile. However, the availability of thinner slices and higher
z-axis resolution appears to maintain satisfactory accuracies for
clinical polyp detection despite reducing tube currents to as low as 50 mA
[82]. Although image noise and
beam hardening are theoretic disadvantages of the lower dose, in practice they
do not appear to interfere with clinical interpretation. Multislice protocols
giving "same dose" have also been suggested, in which scan
parameters are adjusted to give the same dose, but no more, as with
single-slice acquisitions. Finally, as the relative clinical insignificance of
polyps smaller than the 0.5-cm threshold becomes understood, the need for
thinner slices could diminish, and there might be less reluctance to sacrifice
spatial resolution for lower dose.
|
TopAbstractBackgroundVirtual Colonoscopy: The PromiseVirtual Colonoscopy: Research...The Significant Polyp: Defining...Screening by Imaging: New...Incidental Extracolonic FindingsPoliticsConclusionReferences |
|---|
Indeed, practical real world political considerations will be essential for dissemination of virtual colonoscopy as a primary colon screening method. Radiologists have learned many of these political strategies from their long and successful efforts in reducing breast cancer mortality with screening mammography. As I have noted in a previous editorial, these lessons surely apply equally well to using virtual colonoscopy for colon cancer prevention [83].
Accountability.—The public and payors will demand guarantees of quality in performance, interpretation, safety, and patient communication, as has been evidenced by the Food and Drug Administration Mammography Quality Assurance program. The American College of Radiology, therefore, should once again move swiftly to develop formal parameters for the examination. Training is a less critical issue, because this will undoubtedly become available from a variety of sources, including our national scientific societies, university centers, and educational or scientific literature.
Allies.—Public and professional awareness of the significance of colorectal cancer has been increasing. New special interest groups have been created among patients, professionals, and the research community. As detailed by Levin [84], they are summarized here.
The National Colorectal Cancer Round Table was convened by the Centers for Disease Control and the American Cancer Society in 1997 to provide a national coalition of public, private, and voluntary organizations dedicated to reducing the incidence of and mortality from colorectal cancer in the United States. The ultimate goal of the round table is to increase colorectal cancer screening among the entire population by public education, provider education, and health policy.
The National Colorectal Cancer Research Alliance (www.nccra.org), founded in 1999, was established by celebrities such as NBC Today show coanchor, Katie Couric, and the Entertainment Industry Foundation, a California-based nonprofit organization. This group proposes to foster basic, clinical, and epidemiologic research conducted in leading centers.
The Colon Cancer Alliance (www.ccalliance.org) is an organization of colorectal cancer survivors, family members, others affected by the disease, and individuals with a hereditary predisposition to colorectal cancer.
The Colorectal Cancer Progress Review Group is an advisory panel convened by the National Cancer Institute to detail research priorities and recommendations for the next decade.
Through the efforts of these and other organizations, the month of March has been designated as National Colorectal Cancer Awareness Month. This will serve, on an annual basis, to highlight the importance of colorectal cancer screening, research, and the entire spectrum of preventive and therapeutic strategies. Obviously, the introduction of a technique such as virtual colonoscopy will be facilitated by the extent to which it is accepted and supported by these groups.
Advocacy.—Above all, a successful strategy for Radiology will be one that promotes the overall concept of colon screening to reduce death from colorectal cancer. Radiologists need to be advocates for the disease, not just for a test. Reducing deaths from breast cancer has clearly been a more compelling cause for the mammography community than merely promoting an imaging examination of the breast. Parochial salesmanship of virtual colonoscopy as a single test will ultimately be a short-sighted strategy unless the larger goal of preventing colorectal cancer mortality remains in clear focus for the radiology community.
Response of the Endoscopists
In contrast to the recent contentious exchanges between gastroenterologists
and radiologists over the relative roles of barium enema and endoscopy
(flexible sigmoidoscopy, colonoscopy) for colorectal cancer screening, the
dialogue developing in the context of colonography is more restrained,
expectant, and even mildly congenial. In public, endoscopists' editorial
endorsements are properly supportive but cautious
[85,86,87].
There are firm reminders that we need to show high accuracy in low-risk
screening populations and not in just the polyprich cohorts that have been
studied by early colonography researchers. They warn us that our
false-positive rates must be low to avoid unnecessary follow-up studies, and
that we must provide a fast and comfortable study with an affordable price
structure.
These endoscopic editorialists also highlight their own dual dilemmas of inadequate colonoscopy manpower to meet the potential demand and little or no reimbursement for purely screening diagnostic colonoscopy. They then acknowledge the potential benefit of a noninvasive method such as virtual colonoscopy to attract patients who might otherwise refuse altogether to be screened. Privately, colonoscopists also foresee a possible benefit in revenue productivity. If more of their colonoscopy procedures were performed on patients with known polyps, their rate of reimbursable therapeutic polypectomy cases would increase. In addition, the colonoscopy itself could be done much more expeditiously and accurately if the exact location and number of polyps could be clearly indicated beforehand. A diagrammatic annotated schematic of the virtual colonoscopy findings with text description could become a new standard of care for the radiologist's report (Fig. 10).
|
Opportunities, Challenges, and Caveats
Although historical advances in clinical practice tend to be evolutionary
rather than revolutionary, there is an increasing trend to fast-track
introduction of key treatments and technologies. Virtual colonoscopy has
arrived at a distinctly opportune moment in the colon cancer screening
initiative. Patient compliance rates remain low at appropriately 30%, multiple
test options are in use (fecal occult blood test, flexible sigmoidoscopy,
double-contrast barium enema, colonoscopy), and national policy organizations
are adopting a permissive and flexible approach, allowing patients and
physicians to choose among the options. Importantly, to date, colon cancer
screening by colonoscopy alone has not been allowed to dominate the debate or
to crowd out other approaches. Therefore, assuming the scientific results to
date are confirmed and extended, the challenges and caveats for clinical
dissemination of virtual colonoscopy would seem to include the following:
