The challenges facing widespread implementation of virtual colonoscopy for primary screening are significant but not insurmountable.
Virtual colonoscopy (VC), also referred to as CT colonography, is a rapidly evolving diagnostic tool for the detection of colorectal adenomas, which represent benign polyps that have the potential for malignant transformation. Colorectal cancer is a largely preventable disease through routine detection and removal of these adenomatous polyps. The need for additional effective screening options, however, is readily apparent as colorectal cancer remains the second leading cause of cancer-related deaths in the United States, accounting for almost 60,000 deaths each year. The prospect of widespread screening with VC is an exciting development that could lead to greater participation in screening, likely resulting in the prevention of many new cancers. This article will provide an overview of the results and clinical implications of the major clinical VC trials to date, with emphasis on the recent multicenter screening trial published in the New England Journal of Medicine. The various challenges facing widespread clinical implementation of VC screening will also be addressed, focusing primarily on the issues of cost effectiveness and reimbursement.
EARLY VC STUDIES
VC for colorectal polyp detection is an application of CT that is only 10 years old, but the field has grown considerably over that short interval of time. The earlier clinical VC trials primarily evaluated high-risk patient populations and essentially served as feasibility studies to show that the technique was indeed possible for polyp detection. It should be noted that for essentially all studies evaluating the diagnostic performance of VC, conventional or optical colonoscopy (OC), although not infallible itself, has necessarily served as the reference standard for comparison. Among the many earlier VC studies in polyp-rich cohorts, most of which consisted of fewer than 100 patients, two studies in particular have received the most attention. In 1999, Fenlon et al published on a series of 100 high-risk patients who underwent both VC and OC for the purposes of polyp detection. 1 The results were very encouraging, with a “by-patient” sensitivity and specificity for detecting polyps 10 mm and greater of 96% and 96%, respectively. This report was followed by a study from Yee et al in 2001 that consisted of 300 patients and also showed promising results. 2 Both studies employed a primary 2D reading approach, meaning that the 2D axial CT images were generally used for detection of polyps and the 3D endoluminal display was used for confirmation and problem solving. At this point, the future appeared bright for VC as a colorectal screening tool. All that remained was validation of the technique in a true screening population, where the prevalence of polyps is much lower.
The need to prove VC performance in a typical screening setting was critical at this juncture, since the results from polyp-rich cohorts, albeit encouraging, cannot simply be extrapolated to a low-prevalence population. The main reason for this is that achieving a high sensitivity for polyp detection while still maintaining an acceptable specificity becomes much more difficult in a low-prevalence screening setting, where significant polyps are somewhat akin to needles in a haystack. Unfortunately, two subsequent large VC trials in low-prevalence patients, both of which used a primary 2D approach similar to that used in the high-prevalence feasibility trials, yielded very disappointing results. It should be noted that neither of these studies was carried out in a true screening population, as symptomatic and high-risk patients were not excluded. One study, headed by a gastroenterologist, evaluated more than 600 patients and reported a by-patient sensitivity of only 55% for detecting 10-mm polyps. 3 The other study by Johnson et al, which consisted of more than 700 patients, reported a range of by-patient sensitivities of 35% to 72% for 10-mm polyps. 4 Such a low sensitivity for large polyps is clearly unacceptable for screening purposes, regardless of specificity, since the main charge of a screening test must be detection of the target disease. Consequently, the initial enthusiasm stemming from the early feasibility studies quickly evaporated and was replaced by a dark cloud of doubt over the viability of VC screening.
MULTI-CENTER VC SCREENING TRIAL
|Figure 1. Virtual colonoscopy for primary screening in an asymptomatic adult. A map of colon is automatically extracted from the CT data. The green line represents the automated centerline for virtual navigation through the colon.
On the heels of these disappointing developments, results from the first true VC screening trial were reported in the New England Journal of Medicine in December 2003. 5 This study was a prospective multi-center trial evaluating 1,233 asymptomatic adults in which strict exclusion criteria were applied to ensure a true screening population. The VC performance characteristics reported in this study represented a drastic turnaround from the VC studies in low-prevalence cohorts mentioned above. To wit, the by-patient sensitivity and specificity for adenomatous polyps 10 mm and greater were 94% and 96%, respectively. For adenomas 8 mm and greater, these values were 94% and 92%, respectively. In this trial, an enhanced reference standard was created through the use of a technique called “segmental unblinding,” which means that the colonoscopist was initially blinded to the VC results on prospective examination. After evaluation of a colonic segment was completed at endoscopy, the results from the preceding VC study were then revealed, allowing for immediate reexamination of discordant results. This practice turned many would-be VC false positives into true positives, and also allowed for assessment of false negatives at conventional colonoscopy, which is an accurate but not infallible “gold” standard. As such, the sensitivity of conventional colonoscopy for detecting colorectal neoplasms 8 mm and greater was actually slightly lower than VC sensitivity. At the very least, this trial established that VC is an accurate screening tool for detecting clinically relevant polyps, with a sensitivity comparable to conventional colonoscopy.
|Figure 2. A 3D image from an endoluminal perspective at virtual colonoscopy shows a sessile polyp in the colon.
So what allowed for this reversal in VC performance? The answer relates to a multifactorial blend of protocol departures from the prior VC trials. One pivotal deviation was the use of the primary 3D display for polyp detection instead of 2D images (Figures 1, 2). This likely had a significant impact on sensitivity since both polyp conspicuity and the opportunity for detection are greatly enhanced on the 3D endoluminal view. It is important to note that, at present, most available VC systems, including those used in previous trials, simply do not allow for time-efficient primary 3D reading. 6 Another crucial innovation was the use of oral contrast as part of the colonic preparation for the purposes of tagging residual stool and fluid. 7 The use of dilute barium significantly impacts specificity, since untagged stool represents the major cause of false positives on VC examination. Opacification of residual fluid with water-soluble iodinated contrast can improve sensitivity by uncovering submerged polyps that are not detectable in untagged fluid. Segmental unblinding increased both VC sensitivity and specificity by improving the reference standard as described above. The use of only 4-slice multi-detector CT scanners allowed for breath-hold acquisition and improved resolution through thinner collimation. The technique of patient-controlled air insufflation resulted in reliable colonic distention. Finally, a variety of novel 3D tools including translucency rendering improved interpretation and helped reduce overall reading times. 8
CHALLENGES TO IMPLEMENTATION
As we await further validation of VC for primary screening, actual clinical implementation should proceed without delay by those who are currently able to achieve an acceptable accuracy. The need for immediate action is self-evident, as far too many people are dying from a readily preventable disease. There are many important challenges that are likely to be encountered as VC screening makes the transition from the research realm to daily clinical practice. Examples include the development of a practical diagnostic algorithm for VC screening, establishing an effective relationship with gastroenterology, dedicated training of radiologists, and establishment of VC accreditation and practice guidelines. 9 Challenges that will be briefly considered further include demonstrating that VC screening is reasonably cost-effective and establishing reimbursement for VC screening by third-party payors.
|Figure 3. A digital photograph from conventional colonoscopy performed later that day shows the same colonic polyp.
For broad clinical implementation of primary VC screening to occur, it would be helpful to demonstrate that this approach is in fact reasonably cost-effective. Preliminary analysis using a Markov simulation model to examine the long-term costs and effectiveness of VC screening shows that, compared to no screening, VC is clearly both a clinically effective and cost-effective test (manuscript in preparation). In fact, VC screening every 5 years was actually the most clinically effective test of all the modeled strategies, and at an estimated cost per life-year gained that is comparable or better than those for other generally accepted medical interventions. Further analysis shows that VC every 5 years may actually be a dominant strategy over conventional colonoscopy with regard to cost-effectiveness, assuming that VC examination costs about one half of what the more invasive test costs. Of course, these complex cost-effectiveness analyses are heavily dependent on innumerable underlying assumptions and should not be viewed as absolute truth. The general conclusions, however, may help guide eventual reimbursement rates for VC screening.
REIMBURSEMENT FOR VC SCREENING?
At the time of this writing, no firmly established reimbursement structure exists for VC screening on the national scale. There are, however, some exciting developments that indicate that this situation may soon change. For one, a category III CPT code for VC screening (0066T) was recently created by the AMA that goes into effect July 2004. This coding will not lead to any direct change in the reimbursement status for VC screening, but it is an important first step and will provide a tracking system for how many studies are being performed. On a local level, the author has had early success in convincing at least one regional HMO to cover VC screening for average-risk asymptomatic adults under their care. This potentially landmark development is scheduled to officially begin later this month. If all goes according to plan, a report on the initial experience with VC screening in a reimbursed setting would be anticipated before the end of the year.
The prospects for primary colorectal screening with VC are now brighter than ever before. For successful performance of VC in a low-prevalence screening population, it may be necessary to closely follow the techniques employed in the recent trial published in the New England Journal of Medicine, at least until other methods are proven equally effective. In addition to VC study technique and interpretation methods, there are many other important issues that need to be addressed before VC screening can succeed. In the end, however, these challenges facing widespread implementation of VC for primary screening are not insurmountable.
Perry J. Pickhardt, MD, is a radiologist in the Department of Radiology, University of Wisconsin Medical School, Madison, and the Department of Radiology, Uniformed Services University of the Health Sciences, Bethesda, Md. The opinions and assertions contained herein are the private views of the author and are not to be construed as official or as reflecting the views of the Departments of the Navy, Army, or Defense.
- Fenlon HM, Nunes DP, Schroy PC III, Barish MA, Clarke PD, Ferrucci JT. A comparison of virtual and conventional colonoscopy for the detection of colorectal polyps. N Engl J Med.1999;341:1540-1542.
- Yee J, Akerkar GA, Hung RK, Steinauer-Gebauer AM, Wall SD, McQuaid KR. Colorectal neoplasia: performance characteristics of CT colonography for detection in 300 patients. Radiology. 2001;219:685-692.
- Cotton PB, Durkalski VL, Palesch YY, et al. Virtual colonoscopy: final results from a multi-center study. Gastrointest Endosc. 2003;57:AB174.
- Johnson CD, Harmsen WS, Wilson LA, et al. Prospective blinded evaluation of computed tomographic colonography for screen detection of colorectal polyps. Gastroenterology. 2003;125:311-319.
- Pickhardt PJ, Choi JR, Hwang I, et al. CT virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults. N Engl J Med. 2003;349:2189-2198.
- Pickhardt PJ. Three-dimensional endoluminal CT colonography (virtual colonoscopy): comparison of three commercially available systems. AJR Am J Roentgenol. 2003;181:1599-1606.
- Pickhardt PJ, Choi JR. Electronic cleansing and stool tagging in CT colonography: advantages and pitfalls encountered with primary three-dimensional evaluation. AJR Am J Roentgenol. 2003;181:799-805.
- Pickhardt PJ. Translucency rendering in 3D endoluminal CT colonography: a useful tool for increasing polyp specificity and decreasing interpretation time. AJR Am J Roentgenol. In press.
- Pickhardt PJ. CT colonography (virtual colonoscopy) for primary colorectal screening: challenges facing clinical implementation. Abdom Imaging. In press.