Applied Radiology

Dr. Schreiber received both his undergraduate and medical degrees from the University of Virginia, Charlottesville, VA. He is currently a third-year Radiology Resident at the Stanford University Medical Center, Stanford, CA. After completion of his residency, Dr. Schreiber will spend one year as a Fellow in Body Imaging at Johns Hopkins Medical Center, Baltimore, MD.

Computed tomography (CT) colonography has demonstrated efficacy for the detection of clinically significant polyps. Technologic advances, including multidetector scanners and computer-aided detection, will increase the sensitivity of detection and decrease the time for interpretation of this study. Fecal tagging may allow for the elimination of a bowel-cleansing regimen, which is currently a major factor in patient noncompliance. The combination of these expected advances promises to make CT colonography a viable option for population screening. The purpose of this article is to review the development of CT colonography, evaluate its efficacy, and to examine the future applications of this technique.

Colorectal cancer is the second leading cause of cancer-related death in the United States, and the third most common cancer. This neoplasm is diagnosed in 140,000 people annually in the United States, resulting in 60,000 deaths. ¹ It is widely believed that almost all cases of colorectal cancer begin with the development of benign colonic polyps, and early detection of these polyps has been documented to greatly improve patient survival. These facts have prompted the American Cancer Society to recommend yearly fecal occult blood testing and physical examination, as well as screening sigmoidoscopy every 5 years for all adults over the age of 50. Further recommendations include a double-contrast barium enema every 5 to 10 years or a colonoscopy every 10 years. The latter recommendations are based on the fact that nearly half of all colon cancers occur proximal to the sigmoid flexure.

Computed tomography (CT) colonography was first introduced in 1994 by Vining and Gelfand ² as they displayed three-dimensional (3D) endoluminal images of the colon. Since that time, there has been intense research and development in this field, greatly aided by such technological advances as the advent of multidetector CT and improved computer processing ability and software. Currently, CT colonography typically involves thin-section CT imaging of the cleansed colon with the data set reviewed both as two-dimensional (2D) and 3D endoluminal images of the colorectal mucosa.

The advantages of CT colonography over conventional colonoscopy include an improved safety profile and the ability to visualize the entire colon in both antegrade and retrograde directions with subsequent visualization of both sides of the haustral folds. Lesions can be accurately localized relative to extra-colonic landmarks. This technique has been proven to be effective in the evaluation of the proximal large bowel in patients with both incomplete conventional colonoscopy and obstructing distal lesions, as well as in frail, debilitated patients. ^3,4 Ultimately the goal is for this examination to be a viable option for population screening.

Technique

As of yet there has been no real consensus on precise techniques for CT colonography; techniques have been evolving over time to take advantage of improved CT technology. Current research is also focusing on fecal tagging to eliminate the arduous bowel preparation that is required.

At our institution, the following protocol is used. Bowel preparation consists of a phosphosoda in two doses, as commonly employed for fiberoptic colonoscopy. The total phosphosoda dose is 3 oz given in divided doses at approximately 2 pm and 7 pm the day before the exam. Other investigators have used polyethylene glycol and magnesium citrate. Thorough cleansing is critical to the success of the examination. Unfortunately, the preparation is also the least tolerated portion of the examination. Intense research is currently being done on what has been termed digital subtraction bowel cleansing that does not require bowel catharsis. ⁵ This technique will be discussed later.

The patient arrives in the CT suite the morning after the bowel preparation. A physician or nurse inserts a foley catheter into the rectum and insufflates air into the rectum while rolling the patient in multiple positions to facilitate colonic distension. Air is added to patient tolerance, approximately 40 puffs in our experience. A scout radiograph is then used to assess colonic distention. Additional air can be added at this time if necessary. Some investigators have also used spasmolytic agents such as glucagon to reduce bowel peristalsis and resulting motion artifact. The largest study to date addressing this subject showed no added benefit to the use of spasmolytics and their use remains controversial at this time. ⁶

The abdomen and pelvis are then scanned in both supine and prone position using a multidetector row scanner with the 4 * 2.5 mm detector configuration and images with reconstruction of 2.5-mm sections at 1.25-mm intervals. With this configuration, imaging at a pitch of 3 results in a 40-second acquisition time. Relatively low dose is obtained with the use of 120 kVp and 48 mAs (60 mA at a 0.8 second gantry rotation period). The development of 8-detector scanners with faster gantry rotation allows for image acquisition in approximately 20 seconds. Imaging at a pitch of 6 results in a 20-second acquisition time, suitable for older patients who may have difficulty holding their breath for 40 seconds. In such cases we use 120 kVp and 96 mAs (120 mA at a 0.8-second gantry rotation period). This protocol can be adapted to the ability of the CT scanner. While some investigators continue to perform CT colonography with 5-mm sections, there has been a trend toward thinner sections as scanners have advanced, with some now routinely using 1.0- to 1.25-mm sections.

Post-processing is performed in our 3D lab by RT and the study is reviewed using the axial 2D images, with the 3D images used for problem solving. The images are reviewed at lung window settings. Comparison of supine and prone images allows for discrimination between retained, mobile feces and stationary polyps. Repositioning also allows for visualization of the entire colonic surface as some degree of retained fluid is usually present. Research is currently being performed to evaluate the utility of computer-aided detection in the assessment of colonic polyps and will be discussed later.

Efficacy of CT colonography

Two factors must been considered when evaluating the efficacy of CT colonography. First, the sensitivity and specificity of this technique must be compared with conventional colonoscopy, the gold standard for detection of colon polyps. In addition, it is also critical to take into account the natural history of colon cancer to determine what represents a clinically significant lesion.

There has long been debate about what constitutes a clinically significant polyp. It is currently the practice of endoscopists to remove all polyps discovered at colonoscopy, regardless of size; but each removal has a small risk of perforation. The practical necessity of this removal approach has been called into question, however. In a presentation at the Second International Symposium on Virtual Colonoscopy, in October 2000, Glick ⁷ made several important points regarding this practice. Specifically, only about 1% of adenomas <1 cm in diameter harbor invasive cancer. Alternatively, polyps 1 to 2 cm are generally neoplastic with about 10% being malignant. ^8,9 Further data from a 1992 paper by Chantereau et al ¹⁰ further subdivides this category, demonstrating that most malignancies occur beyond the 15-mm threshold. Glick ⁷ therefore concludes that polyps >15 mm in diameter would be a worthwhile and productive target for detection and removal. As a more conservative goal, it seems critical for any detection method to detect polyps >1 cm with high sensitivity and specificity.

A review of the literature demonstrates that the sensitivity and specificity of CT colonography are directly dependent on polyp size, with the lowest sensitivity found for polyps <6 mm. Among investigators in the field, it is generally agreed that the lower size for lesion detection should be 1 cm and the available data will be described with regard to this threshold. In an early publication, Hara et al ¹¹ studied 70 consecutive patients who underwent both conventional and CT colonography. CT colonography had a sensitivity and specificity of 75% and 90%, respectively, for identifying patients with adenomas >1 cm. Conventional colonoscopy was used as the gold standard in this study. Apart from yielding promising early results, the authors point out that false negative results were caused by a combination of perceptive errors and suboptimal patient preparation. The authors ascribe many of the false positives in the CT colonography group to respiration artifact and perceptive errors, such as misidentification of stool. The authors conclude that with greater operator experience and improved scanning technology their results would be greatly improved.

The efficacy of CT colonography with regard to polyp detection has also been studied in other larger trials. ^12,13 One study, involving 300 patients, was designed to compare the results of CT colonography and traditional colon-oscopy with histologic findings. In this study, the researchers reported a sensitivity of 90% (74 of 82) for the detection of polyps >= 10 mm (figure 1). ¹² In a second study--this one involving 100 patients--Fenlon et al ¹³ reported a sensitivity of 91% for polyps >10 mm in size. They also reported a positive and negative predictive value of 96% each for polyps >9 mm in diameter.

There is growing evidence that CT colonography is also effective in evaluating colorectal masses in symptomatic patients. Lees and Gillams ¹⁴ submitted a cohort of 1250 symptomatic patients who underwent CT colonography, more than 1150 of whom had corroboration with colonoscopy, laparotomy, or a minimum of 6 months of clinical follow-up. This study included the use of intravenous contrast, which is not in general use at this time. The authors concluded that CT colonography is reliable for the detection of cancer in symptomatic patients with a sensitivity of 98.7% and a specificity of 98.2%. They detected 232 of 235 cancers in the cohort. One cancer was missed secondary to inadequate patient bowel preparation and was subsequently found on a follow-up CT. The other two were Duke's A "flat" cancers. These cancers, which arise in so-called flat adenomas, are important to consider. This pathologic sequence through a flat adenoma has an unclear contribution to the natural history of colon cancer but may be responsible for 1% to 20% of cases. ¹⁵

A second recent abstract presented at Radiological Society of North America Annual Meeting (RSNA 2001) in November 2001 by Laghi et al ¹⁶ studied 103 consecutive patients with risk factors for colorectal polyps who underwent both CT and conventional colonoscopy. Thirteen colorectal carcinomas and 42 polyps were found with conventional colonoscopy. CT colonography correctly identified all carcinomas and had a 92% sensitivity for polyps >10 mm. The per-patient specificity was 90%.

Lastly, long-awaited results were submitted at RSNA 2001 from the American College of Radiology Imaging Network (ACRIN) Study that conducted a multicenter performance evaluation of CT colonography. In a sample of 94 cases with 50% disease prevalence, the researchers reported that CT colonography demonstrated high accuracy for the detection of polyps >1 cm. Accuracy was measured by the average area under the receiver operator characteristic curve, which was 0.89. ¹⁷

While these recent results are very encouraging, they are difficult to extrapolate to a screening population given the high disease prevalence in the study groups. Further studies are still required to assess the performance of this study in large screening populations with low disease prevalence. In the meantime, CT colonography has proven efficacy in symptomatic patients, as well as patients with obstructing distal lesions as previously described. ^4,14 Belloni et al ¹⁸ demonstrated the potential efficacy of this procedure for following patients after polypectomy by conventional colon-oscopy, showing a sensitivity of 92.6% for polyps when comparing CT and conventional colonoscopy findings directly.

CT colonography as a screening tool

The case for colorectal cancer screening in the general population is well established and has been discussed earlier. Standard recommendations currently include annual testing for fecal occult blood and periodic sigmoidoscopy after the age of 50 for those at average risk for colorectal cancer. The federal government has endorsed these screening methods by establishing coverage for such procedures for all Medicare beneficiaries. A relatively recent editorial in The New England Journal of Medicine by Podolsky ¹⁹ eloquently makes the case that all persons age 50 or over should undergo comprehensive evaluation of the entire large bowel, which is currently best accomplished by conventional colonoscopy. His argument is strongly supported by large studies performed independently by Imperiale et al ²⁰ and Lieberman et al ²¹ who reported similar findings. These studies included large cohorts of asymptomatic persons who underwent colonoscopic screening. Greater than 50% of patients with pathology in both studies had advanced polyps or carcinomas only in the proximal colon, which would have been inaccessible to flexible sigmoidoscopy. The results of the National Polyp study were also striking, demonstrating convincingly that colonoscopic polypectomy resulted in a lower-than-expected incidence of colorectal cancer. ²² These findings support the well-established view that colorectal adenomas progress to adenocarcinomas, as well as the practice of finding and eliminating polyps to prevent colorectal cancer.

It is clear that comprehensive evaluation of the large bowel will be central to colorectal cancer screening in the near future. The question to be addressed is how CT colonography may be involved with this screening. We have seen that it is critical for a screening exam to detect polyps >1 cm. CT colonography has demonstrated efficacy in the detection of polyps >1 cm in patients with high disease prevalence but its utility can only be extrapolated in a screening population with low disease prevalence.

The current gold standard for complete evaluation of the large bowel is conventional colonoscopy, and this procedure is both diagnostic and therapeutic. Significant advantages of CT over traditional colonoscopy include complete evaluation of the colonic mucosa and patient safety. Lieberman et al ²¹ reported an overall complication rate of traditional colonoscopy of 0.3%. Significant problems related to gastrointestinal perforation and problems related to conscious sedation occur in approximately 0.2% of examinations. ²³ While these complications do have a low rate, the number of affected patients would rapidly rise if colonoscopy were to be used for population screening. Traditional colon-oscopy also fails to examine the proximal colon in 10% to 15% of cases. A study by Morrin et al ³ demonstrated the efficacy of CT colonography in evaluating these nonvisualized segments after incomplete colonoscopy. Additional advantages of CT colonography include decreased exam time, no sedation, and the ability to examine both sides of the bowel wall. Disadvantages include the need for very thorough bowel cleansing, problems of spasm and retained debris, resolution for flat neoplasms, and the fact that the exam is diagnostic only.

CT colonography is therefore established as being safer than traditional colonoscopy and will very likely prove to have similar diagnostic accuracy. The problem of flat neoplasms still exists and must be further evaluated. Widespread access to needed CT hardware and software is expected and must improve. A major barrier to the implementation of CT colonography for population screening is cost-effectiveness. A comprehensive study into this question by Sonnenberg et al ²⁴ concluded that for CT colonography to be cost-effective, it must be associated with an initial compliance rate 15% to 20% better or procedural costs 54% less than traditional colonoscopy.

Future developments

The major factor limiting patient compliance with traditional and virtual colonoscopy is the arduous bowel preparation required. False-positive and false-negative results in CT colonography have also, in some part, been attributed to inadequate bowel preparation or adherent stool. Investigators have been using contrast material to label stool that can then be subtracted from the images in the hopes of being able to eliminate this bowl preparation procedure. In this experimental technique, 24 to 48 hours prior to imaging, the patient ingests high-density contrast material that fully opacifies the colon. A single CT acquisition is then performed after colonic insufflation. Computer software is then used to subtract the ingested colonic contents and stool, leaving the colorectal mucosa unaffected. Promising early results were published by Zalis et al. ⁵ A larger study was published subsequently that suggested that sensitivity for polyp detection using this technique approaches that achieved in traditionally prepared colons. ²⁵ This technique also has the potential of a being performed in a single acquisition as opposed to the currently standard supine and prone scanning, reducing radiation by half.

Elimination of the bowel preparation is expected to greatly improve patient compliance. This increase in compliance would significantly contribute to making CT colonography cost-effective for population screening.

Interpretation time also has a significant impact on the cost-effectiveness of CT colonography. Postprocessing of supine and prone images is performed routinely to make virtual 3D movies that have the appearance of conventional endoscopy (figure 2). There is no consensus as to the best use of these images, but at this time they are generally used as a problem-solving tool in combination with the reference axial images. Given the huge volume of data generated, interpretation time is a significant concern. This has prompted the development of several techniques for computer-aided detection of polyps. In his abstract at the 2000 Virtual Colonoscopy symposium, Beaulieu ²⁶ described the Stanford approach to computer-aided detection and concluded that some form of computer-aided detection will eventually become the most effective means of making CT colonography efficient, accurate, and cost-effective (figure 3). This increase in cost effectiveness comes from significantly reducing the interpretation time for the examination. The most recent work on this subject, results from a series of 41 cases presented at RSNA 2001, concludes that computer-aided detection has the potential to detect polyps with a high sensitivity and acceptable false-positive rate. ²⁷ Novel 3D visualization techniques were also described by Vos et al ²⁸ and Wax et al ²⁹ who conclude that visualization with their techniques provides better visualization of the colonic mucosa then conventional colonoscopy.

The diagnostic ability of CT colonography has been shown to be similar to traditional colonoscopy, and future developments, such as computer-aided detection and fecal tagging, promise to significantly improve efficacy, patient compliance, and image interpretation time. These improvements, combined with the inherent safety of the examination, promise to make CT colonography a viable population screening method within the next decade. Radiologists, gastroenterologists, and internists must work together to improve screening compliance and patient education.

Acknowledgments

The author would like to thank Dr. Christopher Beaulieu and Dr. Robert Mindelzun for their guidance and mentoring during the completion of this project.