Computed tomographic colonography (CTC) was initially described by Vining [1] in 1994 and instantly created a huge enthusiasm amongst radiologists. This minimally invasive and patient-friendly technique introduced a totally unexpected and revolutionary method for imaging of the colon. CTC could be performed with little or no risk for the patient and most of all seemed easy to perform and interpret. Furthermore since then, the technique has improved dramatically with better preparation regimens, colonic distension, CT acquisition and spectacular developments in dedicated CTC software. With this capacity of reliable, nearly risk-free and patient-compliant detection of colorectal cancer (CRC) and the precancerous adenoma, it seemed likely that CTC would rapidly become the CRC screening test of choice. However, despite these improvements and innovations, CTC could not yet fulfil these high hopes and break through as a fully validated total colorectal screening examination. As with most innovations, CTC was subject to growing pains mostly due to lack of experience and underestimation of the complexity of both the examination technique and interpretation. The diagnostic test performance results were highly variable. Since then, ongoing research on individual and organizational levels has addressed many of the deficiencies in early CTC technique and interpretation to the point that many experts agree that CTC may be ready for widespread use. The moment appears to be appropriate to perform a critical appraisal of the technique with evaluation of its current benefits and limitations.

This article reviews the landmark studies performed to date, summarizes the accepted indications and considers critical areas of ongoing investigation that can facilitate the widespread implementation of CTC.

Results

The performance of CTC has been evaluated in several single center studies. Two of these studies merit particular interest. The first study to present results in a large patient cohort was published by Fenlon et al. in 1999 [2]. Using single slice CT technology in 100 patients, they obtained very good results of polyp detection with sensitivity on a per patient basis of 96% and 94% for lesions ≥1 cm, 6 mm–9 mm, respectively. They also made a distinction between the clinically important adenomatous polyp (which may be a precursor to cancer) and the hyperplastic polyp (which has no malignant potential). On a per polyp basis sensitivity was 71% and 90% for hyperplastic polyps and adenomas 6 mm–9 mm, respectively. All larger lesions were adenomas or cancers. In 2001 Yee et al. [3] made the same distinction on a per patient basis in a study of 300 patients with sensitivity for lesions ≥1 cm of 100% for both adenomas and polyps. For lesions 5 mm–9 mm sensitivity was 95% and 93% for adenomas and polyps, respectively. On a per lesion basis, sensitivity for lesions ≥1 cm was 94% and 90% for adenomas and all polyps, respectively. For lesions 5 mm–9 mm sensitivity was 82% and 80.1% for adenomas and polyps, respectively. Thus, the focus on the clinically relevant adenoma will usually demonstrate a higher sensitivity.

The single center studies represent relatively small patient cohorts. To evaluate the performance of CTC in larger populations, three meta-analyses of the published studies were performed. Besides combining the results of several studies, these meta-analyses also allowed to assess the methodology of the different studies. In a first meta-analysis, after surveying the entire literature, Sosna et al. [4] were able to include only 14 adequately performed and reported studies. In a total of 1,324 patients a pooled per-patient sensitivity of 88%, 84% and 65% was reported for lesions ≥1 cm, 6 mm–9 mm and ≤5 mm, respectively. Mulhall et al. [5] collected 33 studies representing a total of 6,393 patients. Pooled per-patient sensitivity was 85%, 70% and 48% for lesions ≥9 mm, 6 mm–9 mm and <6 mm, respectively. Overall sensitivity was 70%. A wide range of sensitivities of the individual studies was reported. This was related to different examination techniques with improved results in case thinner slice collimation and multi-slice scanners were used. In the third meta-analysis, 24 studies with a total of 4,181 were selected by Halligan et al. [6]. Pooled per-patient sensitivity was 93% and 86% for larger polyps and for larger and medium polyps together, respectively. Besides heterogeneities in the results, Halligan et al. detected shortcomings in methodology and quality of data reporting with unavailability of important data [6]. This finding underscored the need to define a minimum set of prerequisites to adequately report studies with the goal to obtain quality and consistency in research reporting. Dachman and Zalis confirmed the importance of reporting and comparing results of similar study cohorts (screening vs. non-screening) using comparable technical parameters (preparation, colonic distension, acquisition) [7]. Clear-cut standards for CTC reporting concerning lesion size measurements and morphology, colonic segment localization and careful matching with the optical colonoscopy findings together with per-patient analysis of the results were also considered mandatory in order to obtain homogeneity and generalizability of the results.

All three meta-analyses concluded that CTC is very specific for polyp detection with a good to very good sensitivity for the larger polyp and a medium sensitivity for the medium polyp.

However this was not enough to establish CTC as a reliable tool for primary widespread CRC screening. Two other multi-center trials reported poor sensitivity for CTC. Cotton et al. [8] reported sensitivity of 55% and 39% for lesions ≥1 cm and ≥6 mm, respectively, although when re-analyzing the results to include 2D and 3D techniques, by patient sensitivity for >1 cm lesions improved to 67% and for the ≥6 mm level to 56%. This trial was criticized for inadequate training of the reader who only had to have a ten case experience (with no minimum correct reading) in order to participate in the trial [9, 10]. Indeed, the one experienced group participating in that trial achieved an 83% sensitivity for at the ≥6 mm threshold, lending support to the need for adequate training. In the Rockey et al. [11] trial a by patient sensitivity of 59% and 51% was obtained for patients with lesions ≥1 cm and 6 mm–9 mm, respectively. Doshi et al. [12] recently re-analyzed some of the data from the Rockey et al. trial and found that when considering only adenomas, the by patient sensitivity at the 10 mm threshold was in fact 70%. When studying the sources of error they found that at the 10 mm threshold, 80% of the errors were potentially avoidable observer errors. Although the recent date of publication, both studies commenced in 2,000 and were hence not using state-of-the-art technique by current standards. Both trials could have benefited from preparation with fecal tagging, better software and possibly the use of automated colonic insufflation.

Some trials also evaluated the sensitivity of the double contrast barium enema (DCBE). Rockey et al. [11] reported a sensitivity for DCBE of 48% and 35% for lesions >1 cm and 6 mm–9 mm, respectively. In another study by Johnson et al. comparing both techniques, sensitivity of DCBE was 45% and 44%, while double-read CTC scored 81 and 72% for lesions >1 cm and 6 mm–9 mm, respectively [13].

In the third multi-centre trial, Pickhardt et al. [14] obtained excellent results in 1,233 asymptomatic patients with a sensitivity of 93.8%, 93.9% and 88.7% for CTC vs. 87.5%, 91.5% and 92.3% for optical colonoscopy for lesions ≥1 cm, ≥8 mm and ≥6 mm, respectively. Using a state-of-the-art approach with fecal tagging, electronic subtraction of tagged fluid, good colonic distension, 3D primary read and adequate training of the participating centres, CTC scored better than OC for lesions ≥6 and 8 mm. Since then two long-term single-center trials are ongoing at sites that participated in the Pickhardt et al. study. One of these, at the National Naval Medical Center (NNMC) has published interim results showing a high sensitivity similar to the original multicenter trial [15]. The performance of CTC in the different studies is listed in Table 1.

Table 1. Performance of CTC for large ( ≥1 cm) and intermediate (6 mm–9 mm) lesions, unless specified
Full size table

New large multi-centre trials are ongoing: the USA National Colonography Trial sponsored by the American College of Radiology Imaging Network (ACRIN) is a screening trial of asymptomatic average-risk patients which completed recruitment at the end of 2006 but has not yet reported results. In addition to overall accuracy, the trial has several sub-aims including comparing primary 2D vs. 3D reading methods with a primary aim of evaluating the per patient and per polyp analysis of lesions >1 cm and 5 mm–10 mm and identification of >5 mm and >1 cm polyps with villous features, high grade dysplasia, and carcinoma [16]; the SIGGAR 1 trial in the U.K. compares diagnostic efficacy of CTC to OC and DCBE in patients over 55 years with symptoms suggestive of CRC in two parallel randomised trials (CTC vs. DCBE and CTC vs. OC). Patient and physician preferences and the impact on daily practice of the radiologists are investigated with modelling of the health effects and costs of each examination. The aim is to examine 4,500 patients in ten participating centers; the IMPACT study in Italy assesses sensitivity and specificity of CTC on a per-patient basis to detect advanced adenomas for patients at increased risk for CRC. Additional aims are: assessment of sensitivity of CTC in a screening condition, frequency of lesions missed at OC, adverse events, acceptability and costs of OC and CTC and frequency of extra-colonic findings. Fourteen centers are participating with the aim of examining at least 900 patients with an expected closing date in the first half of 2007 [17]; the Munich Colorectal Cancer Prevention Trial in Germany compares adenoma detection in asymptomatic screening patients in 64-slice CTC to same-day OC and fecal occult blood test (FOBT) [18]; the French STIC multi-centre trial compares CTC performance with same-day OC in asymptomatic patients. Twenty-five centers, trained during a 2 days course, will examine a total of 1,500 patients.

Considering these results it can be concluded that: (1) CTC is second best after OC for detection of colorectal polyps and masses; (2) There is a problem of reproducibility or generalizability of published CTC results. In other terms, CTC obtains good results unless performed by an experienced team using meticulous technique and interpretation. However it seems difficult to realize this state-of-the-art performance on a large scale, in a large group of patients and by a large community of radiologists; (3) The learning curve and difficulty of performing and interpreting CTC should not be underestimated.

Indications

Accepted indications

Taking the results of the trials into account a European committee of CTC experts concluded in the ESGAR consensus statement on CTC that in the symptomatic patient CTC is the examination of choice if adequate local expertise and infra-structure is available [19].

Computed tomographic colonography is an accepted indication after incomplete OC because of an obstructing process (tumor, diverticulosis) or due to colonic tortuosity. CTC results in precise identification and localization of the obstructing lesion, reliably detects synchronous lesions and enables improved visualization of the proximal colon compared to DCBE.

Computed tomographic colonography is also indicated when OC is relatively contraindicated such as patients who must remain anticoagulated, patients with severe cardio-pulmonary disease as well as patients who are unwilling to undergo OC. Finally CTC is an indication for examining the frail and elderly patient for tumoral pathology and can be combined with a routine intravenous contrast enhanced CT for tumor staging and detection of metastatic disease.

Pending indication

Despite the availability of many tests, less than 50% of the eligible population have not undergone a screening test. This low compliance with accepted screening recommendations is multifactorial and can in-part be attributed to poor patient education, fear of the preparation or the exam itself (be it collecting stool sample or insertion of rectal catheters, tubes or endoscopes), inconvenience and even laxity on the part of the medical community to encourage screening [20, 21]. With the advent of CTC, the radiological community is again actively involved in the debate of what method should ideally be used. CTC is an appealing technique. It is a full structural colonic examination seemingly less invasive than the existing methods, OC and DCBE. Furthermore, it is based upon the latest technologies with multi-slice CT enabling covering the colon in only a few seconds and detailed two- and three-dimensional imaging. Interpretation can be performed using virtual reality making it attractive for the clinician and the patient. This creates the opportunity to develop a new CRC screening test with good performance and better patient attendance. Furthermore it can be expected that in the near future the cathartic part of the preparation will be eliminated, offering a laxative-free examination to the patient without interruption of the normal daily activity [22]. In reality the sky is not that bright yet. Before CTC can be considered a reliable tool for mass CRC screening, many issues have to be tackled.

The problems to solve

Generalizability

Computed tomographic colonography is faced with an important problem: the impressive results of the Pickhardt trial have not been confirmed by other trials. On the contrary, the steep ascent of CTC has been undermined somewhat by both the Cotton and Rockey trials. These two multi-centre trials caused a back-to-reality experience. It became clear that CTC was a difficult technique and that repeatability of the good results previously obtained was not evident at all. This problem is fed by many sources. First of all prevalence of disease is low in a screening population causing decreased performance when compared to a population at high risk for CRC. The reason for this is reader fatigue. Secondly, there is a need of defining a reliable and reproducible gold standard. OC considered the gold standard is indeed not infallible. In a systematic review of six studies by Van Rijn et al. [23] pooled miss rate of OC was 2.1% and 13% for adenomas ≥10 mm and 5 mm–10 mm, respectively. In a cohort of 183 patients examined by back-to-back colonoscopy, Rex et al. [24] had a miss rate of 6% and 13% for adenomas ≥1 cm and 6 mm–9 mm, respectively. Segmental unblinding as was introduced by Pickhardt et al. [14] is mandatory. However this does not guarantee the exact golden standard. There is a clear problem of matching the findings of CTC with those of OC, inducing the need of exact segmental localisation, exact description of morphology and exact measurement. This problem was revealed in a review of the Rockey trial by Doshi et al. [12]. Indeed, although detected on CTC, seven significant polyps also detected on OC were considered as false negatives because their measurements did not match.

Thirdly, to perform CTC in a reliable way, in a large population by a large community of “CTC-radiologists” a state-of-the-art approach is imperative. This has become clear after the three mentioned multi-centre trials. Before these trials, the importance of some technical aspects of CTC was underestimated: a preparation adapted to CTC and not to OC, optimal colonic distension, use of the right scanning parameters, meticulous interpretation. This state-of-the-art approach was rigorously applied in the Pickhardt trial with excellent results. On the other hand the Cotton and Rockey trials had several important limitations related to technique and interpretation which may in part explain their poor performance. Before these trials the importance of these technical aspects was not fully appreciated. As an example, in the first consensus statement on virtual colonoscopy, finalized in 2003 before the publication of the three trials, fecal tagging was not deemed necessary, while in the recent ESGAR consensus statement the use of fecal tagging in screening patients is recommended [19, 25]. In fact the positive aspect of these two trials was that it appeared that the use of meticulous technique and interpretation methods apparently improves CTC performance. This caused a shift in the general opinion with the definition of a state-of-the-art concept to perform CTC with a reasonable consensus between experts: bowel preparation with fecal tagging, colonic distension with a CO2-injector and if possible with hyoscine butylbromide, thin slice acquisition using multi-slice technique with low dose, detailed interpretation using dedicated and updated CTC software.

Computer-aided diagnosis (CAD) has the potential to improve CTC performance with better generalizability of results. Further research and development is however necessary to define its function in CRC screening [26].

Finally, and probably the key factor for realizing reproducibility and generalizability of CTC results is experience [27]. Indeed, it is now clear that CTC is not at all an easy technique requiring an experienced team applying the state-of-the-art approach. This brings us to the second important issue to solve before considering CTC as primary tool for CRC screening. How to overcome the steep learning curve and how to establish standards for training and widespread implementation?

Learning curve

Adequate training and experience in reading proven cases is necessary in order to gain competence in interpreting CTC and the extent of this experience is not trivial. The paradigm of interpretation, regardless of whether reading 2D, 3D or on novel views, requires specific training. Unfortunately, there are no evidence-based proven successful guidelines for training. Experts generally agree, that in addition to basic software knowledge and training in the basics of normal, abnormal and common pitfalls via lectures and sample training cases, a personal experience of reading 50–75 proven cases representing a spectrum of findings is necessary [28, 29]. Training opportunities exist in the form of courses held by several experts and by some radiologic societies. If reimbursement of CTC creates a broad interest, then these limited opportunities will need to be expanded in a systematic way to accommodate training a large number of individuals. Current research suggests that both training and attention to detail in a careful reading using both 2D and 3D is necessary [29–31]. Some investigators suggest that a primary 3D read of an electronically subtracted, fecal tagged CTC data set, is more intuitive and faster to read [14]. Others emphasize the value of novel views which although they may distort the colon, might speed-up interpretation [32, 33].

Given the long learning curve and sometimes tedious nature of the interpretation, investigators have embarked on studying the efficiency of training non-radiologists or technologists in reading CTC [34, 35]. Taylor et al. found the effect of training in a small study to be unpredictable and even caused deterioration in sensitivity for some observers [34]. In a multicenter trial of the ESGAR, nine experience radiologist, nine newly trained radiologists and ten trained CT technologists were evaluated [36]. Experienced observers interpreted CT colonographic images significantly better than did novices trained with 50 studies. On average, no difference between trained radiologists and trained technologists was found; however, individual performance was variable and some trainees outperformed some experienced observers. We have an ongoing trial for training novice readers and preliminary data showed that at least some non-radiologists can achieve high accuracy with a highly regimented training program (unpublished data).

The issue of training must be re-evaluated after the introduction of CAD and after the validation of novel views that make survey of the entire colon for polyps more intuitive. It is possible that novices can learn to read CTC with high accuracy with these software innovations.

Cut-off size of polyp detection

As CTC does not allow the possibility of removing lesions, a structured advice concerning an eventual follow-up or treatment of the detected lesion is necessary. This task is not evident. It is known that CTC does not score that well for the diminutive lesion (≤5 mm). Sending every patient with a possible diminutive lesion to OC would create too many false positives and unaffordable costs. Can therefore a cut-off size of polyp detection safely be established? In other terms, is there enough evidence to submit the small and even intermediate lesion to a regular screening or surveillance programme? What are the chances for a polyp to develop in an invasive cancer? As a consequence is it necessary reporting or communicating all findings to the clinician and ultimately to the patient? It is generally accepted that in CRC screening the target lesion is the advanced adenoma. This is a polypoid lesion ≥1 cm with or without dysplastic and/or villous components. It is known that for these lesions there is a probability of developing an invasive cancer in 10%–25% of cases [37]. So without any doubt these lesions need immediate diagnostic work up with OC followed by endoscopic or surgical removal. Unfortunately for the smaller lesions decision taking is not clear-cut. The natural history for lesions <1 cm is not entirely known [38].

What about the 6 mm–9 mm? About 30% of these lesions are hyperplastic. Only 3%–4% present with advanced dysplasia with 0.5%–1% of lesions developing invasive cancer. Distinction has to be made between patients presenting with 1–2 intermediate lesions and those with ≥3 lesions and between patients at average or increased risk for CRC [39]. It is accepted that patients with ≥3 lesions are at increased risk for developing invasive cancer.

What about the ≤5 mm lesion? These lesions are mostly hyperplastic with an increased risk for invasive cancer far below 1%. Distinction has again to be made between more or less than three lesions [39].

Decision making is a delicate task frequently needing a multidisciplinary approach. The opinion on how to take best care of the lesion is again not unambiguous and differs between radiologists and gastroenterologists. At first sight this difference might be dictated by a tendency of protectionism by the gastroenterologists. However it cannot be denied that for the lesion <1 cm the natural history is not entirely clear because until now most of these lesions were endoscopically removed. Even between gastroenterologists there is no consensus on how to best deal with the smaller lesion [40]. An overview of the different opinions is listed in Table 2. For lesions ≥1 cm there is no discussion: everybody agrees upon the necessity of removal of the lesion. The American College of Gastroenterology (ACG) [41] and Rex are advocates of removing the single ≥6 mm lesion and two lesions and three lesions of any size, respectively. According to the Working Group on Virtual Colonoscopy [42] and Ransohoff, >3 lesions 6 mm–9 mm are eligible for removal because of an increased risk for invasive cancer in these patients. Patients with 1–2 lesions 6 mm–9 mm can be submitted to a screening interval of 3 years (watchful waiting). Any lesion ≤5 mm is a candidate for a regular 5 year–10 year interval screening. According to the ACG only patients presenting with one or two 5 mm-lesions are eligible for routine screening. Rex is more stringent and considers this eligibility for routine screening if there is only a single 5 mm lesion [40].

Table 2. Recommendations for follow-up and removal of polyps detected at CTC
Full size table

The Working Group on Virtual Colonoscopy has elaborated a C-RADS system for reporting CTC with emphasis on the clinical importance of the detected lesion: C0 stands for an inadequate examination or examination awaiting comparison with prior examinations; C1 stands for a normal or benign lesion, eligible for routine screening; C2 stands for the 6 mm–9 mm lesion <3 in number, with surveillance or OC recommended; C3 stands for the ≥10 m lesion or ≥3 6 mm–9 mm lesions, with recommendation of immediate OC. Finally C4 refers to colonic mass, likely malignant deserving immediate treatment. It is clear that decisions will depend on the local expertise and guidelines and patient behaviour and age [42].

Risk profile

One of the prerequisites for a test to be suitable for screening is to have a low risk profile. First of all, there is the obvious intention to not harm the patient. Secondly, as a consequence the benefit-risk ratio will largely influence patient adherence. A low risk profile is required for CTC to become a reliable and cost-effective CRC screening tool. Major issues of adverse effects are inherent to the technique. First of all the use of ionizing radiation may compromise its use in an asymptomatic screening population. Secondly as optimal colonic distension is imperative for good performance, one should always be aware of a possible colonic perforation.

Radiation dose

As ionizing radiation submits the patients to an increased cancer risk, radiation dose should be minimal. What is the radiation dose of CTC with dual positioning? Is there an opportunity to reduce the radiation dose? Radiation dose is mostly influenced by mAs and pitch. Using the same acquisition parameters radiation dose may vary between scanners from different vendors with the same number of detectors. Using 4 × 1 mm collimation, a pitch of 6–7 and 50 mAS, 120 kV and dual positioning, Macari et al. [43] had a radiation dose of 5.0 and 7.8 mSv in male and female patients, respectively. Using the parameters described in a study by Johnson et al. [44] (8 × 1.25 mm, a pitch of 1.35, 65 mAs and 120 kV and dual positioning), Brenner et al. calculated that the radiation dose was 13 mSv [45]. In an inventory study among research institutions by Jensch et al. the effective dose of CTC in daily practice was estimated using the ImPACT patient dosimetry calculator. The median effective dose per acquisition was 5.1, 6.7 and 3.3 mSv for 4-, 8- and 16-slice scanners, respectively, using a median effective tube charge of 65.3, 83.6 and 55 mAs, respectively. This finding demonstrates that with the introduction of newer multi-slice scanners the radiation dose has not increased because of the tendency to decrease the tube current (mAs) [46].

Reducing radiation dose to a minimum increases the noise of the image. With CTC this is however not a real issue because of the high contrast difference between the colonic wall and the luminal air or CO2. As a consequence the idea was conceived to develop ultra-low dose CTC protocols. Using 4 × 2.5 mm, 10 mAs and 140 kV, Iannaccone et al. [47] could reduce the radiation dose to 1.8 and 2.4 mSv in male and female patients, respectively.

New developments as tube current modulation offer new opportunities to reduce the dose. Tube current modulation changes its dose according to the thickness of the part of the body to be scanned [48]. Using an updated version of this technique, Graser et al. could reduce radiation dose to 2.38 mSv using 16 × 0.75 mm, a pitch of 1, 120 kV and 40 mAs [49].

What is the radiation induced risk? Radiation may induce cancer. This risks decreases with the age of the patient. According to Brenner et al. [45], the estimated absolute life time risk for cancer induction from CTC at a normal dose (8 × 1.25 mm, a pitch of 1.35, 65 mAs and 120 kV) is 0.14% (1/700) in a 50-year-old patient. This risk is halved in a 70-year-old patient. With ultra-low dose protocols this dose can probably be reduced with a factor 5–10, making cancer risk small and smaller than the colon-cancer related risk of 6% and related mortality of 3%.

Perforation risk

Optimal colonic distension, with colonic inflation as key element, is a cornerstone of good CTC practice. Inflating the colon with room air or CO2 induces a risk of perforation. This probability is enhanced in presence of predisposing factors such as obstructing lesions.

In a review of 11,870 patients, Sosna et al. reported 7 perforations (0.059%) in 6 and 1 diagnostic and screening examinations, respectively [50]. Burling et al. [51] reported nine perforations (0.052%) in 17,067 patients, examined for diagnostic purposes only. Finally, for the Working Group on Virtual Colonoscopy, Pickhardt et al. reported two perforations (0.009%) in 21,923 patients with no perforations in 11,067 screening patients [52]. This resulted in an overall total of 18 perforations. Twelve patients were symptomatic with seven having an obstructive lesion and one having ulcerative colitis, indicating the high prevalence of pre-existing disease in these patients. Surgery was performed in six patients who all had an uneventful recovery. Manual inflation with room air was used in 16 cases of perforation. Overall symptomatic perforation rate was 0.005%–0.03%. This perforation rate is slightly lower than OC and similar to the DCBE with a high number of asymptomatic patients and low incidence of surgery in the symptomatic patient [53].

So far no prospective trial comparing safety between different examination techniques has been performed [53].

Further optimization of CTC technique and preparation is important to reduce the perforation rate [54].

Cardiovascular risk

Computed tomographic colonography has not been associated with any increased risk for cardiovascular complications. Furthermore CTC is frequently used in the frail and elderly patient [55].

Cost-effectiveness

To become a potential and efficient tool for CRC screening, CTC will need to be cost-effective. Large randomized controlled trials are necessary to assess the cost-effectiveness of a screening test. So far, only guaiac-based FOBT has been studied in large trials. In these trials unrehydrated guaiac-based FOBT, followed by OC for patients with a positive test, has proven to be cost-effective [56].

Based upon a model of the natural history of CRC and using the results of a meta-analysis, Vijan et al. could conclude that CTC is cost-effective compared to no screening. CTC is however not cost-effective and expensive compared to screening with OC. To become cost-effective CTC would need to be three to four times cheaper than OC and all adenomas detected on CTC should be removed by OC [57].

Using a similar model in Italy, Hassan et al. concluded that any screening is cost-saving compared to no screening. Colon cancer screening seems less expensive in Italy than in the US. This is due to the cost of the procedure [58].

Besides the relative cost of the examinations, other factors influencing cost-effectiveness of CTC are: test accuracy, impact of extra-colonic findings and patient adherence to the test.

Influence of accuracy

It is clear that to achieve cost-effectiveness CTC will need to obtain good results of polyp detection. Good results are possible if CTC is performed and interpreted by an experienced team using a state-of-the-art approach. To become a preferred option for screening over OC a sensitivity of 83% for 1 cm adenomas is needed [57]. However as demonstrated there is a problem of generalizability of results. So before widespread dissemination of CTC as a screening tool can be considered it will need to be performed reliably by a large group of experienced radiologists with a very probable role for CAD.

Influence of extracolonic findings

Extracolonic findings are an important potential advantage to CTC over OC and understanding the influence of these findings on the cost-effectiveness of CTC remains a continued area of investigation. Furthermore, the manner in which CTC is performed also impacts the ability to detect and characterize extracolonic findings; at the lowest doses used for CTC (e.g., 20 mAs) the noisy images may impair both detection of abnormalities and characterization of lesions as solid or cystic. One option to accommodate optimal detection of extracolonic findings is to perform one series at a slightly higher dose that would at least be sufficient to reliably characterize lesions as solid or cystic, yet the other view can be done at a very low dose.

Extracolonic findings have been stratified by their potential clinical significance into low, moderate and highly significant. While different criteria have been used to define these terms, the Boston Working Group [42] has suggested a categorization scheme as follows: E0, is defined as an exam in which evaluation of soft tissues is limited, e.g., by artefact; E1 normal or normal variant; E2, a clinically unimportant finding, e.g., hepatic or renal cysts or gallstones; E3, a likely unimportant or incompletely characterized finding whose work up is subject to personal preference, e.g., homogenously hyperattenuating renal cyst; and E4, a potentially important finding, e.g., a solid renal mass, aortic aneurysm or 1 cm lung nodule.

Prior studies reported an incidence of “significant” extracolonic findings of about 9%–12% [59–61]. This number may be higher, 10%–23% in symptomatic or high risk populations [14 ,59, 62, 63]. Flicker et al. [64] found a low rate of significance in a mixed cohort with only 7.7% as high risk extracolonic findings, less than half of the incidence reported by most others, with the exception of Pickhardt et al. who reported an incidence of 4.5% highly significant findings and 7.9% moderate significance findings [14].

When estimating the potential cost or benefit from reporting the extracolonic findings, it is important to keep in mind that all studies to date have been retrospective and follow up information has been limited. To know the true significance of the extracolonic findings, the investigator would have to be certain that information was culled from all patient care sources in order to prove that a finding was indeed not previously known and to determine if further tests were performed solely on the basis of the finding found at CTC. Gluecker et al. [59] estimated the total additional cost per exam for the work up of extracolonic findings in the range of $28–34 per exam. However they did not differentiate between “desirable” and “non-desirable” tests based on final diagnosis, nor did they consider follow up therapy.

In searching for extracolonic lesions, ideally a careful search should be made using soft tissue, liver, lung and bone window width settings. In particular, the narrow liver windows can be helpful for all the solid abdominal viscera. The CT data set can be reconstructed at thicker intervals such as 3 mm–5 mm to facilitate review of the axial images. Studies suggest that 2%–6% of extracolonic findings are missed on the initial interpretation [59, 61, 62, 64].

Influence of patient adherence

Besides the lack of awareness of possible disease and dedicated patient information concerning CRC, patient compliance influences participation to a CRC screening test. Patient compliance is related to: (1) pre-procedural inconveniences including the anxiety for outcome of the test, bowel preparation with interruption of normal daily activity; (2) Procedural inconveniences: discomfort, pain, embarrassment and interruption of normal daily activity the day of the procedure; (3) Compliance with immediate therapeutic intervention or routine follow-up examination [20, 21].

Comparing DCBE, CTC and OC has shown that DCBE causes most discomfort and pain. Studies comparing CTC and OC show discordant results with patients preferring CTC over OC and vice versa. This difference is probably caused by differences in patient questionnaires and different timing of these questionnaires. Furthermore OC is performed under conscious sedation in most patients resulting in an incorrect representation of the inconveniences caused by OC [65–68].

The inconveniences caused by the cathartic cleansing with side-effects and interruption of normal daily activity decrease patient’s adherence to a CRC screening programme significantly. Developing a robust and patient-friendly laxative-free CTC method will make CTC and most probably CRC screening more appealing [22, 69]. It is however clear that to achieve this goal, accuracy of CTC has to remain unchanged. Therefore reliable stool subtraction methods should be developed. Again the presence of an experienced CTC-team is imperative. Most gastroenterologists and radiologists accept that once this method is fine tuned the call for CTC will probably grow spectacularly and make the discussion of patient preference for any of the tests superfluous [70, 71].

This is important as patient compliance plays a preponderant role in achieving cost-effectiveness for a screening test. If a 15%–20% improved compliance could be achieved, CTC could become cost-effective compared to OC [72].

Other issues

The flat adenoma

Reported to represent 8.5%–42.7% of adenomas, the flat adenoma constitutes the ultimate diagnostic challenge for both the radiologist and gastroenterologist [73, 74]. A lot of controversy exists concerning their importance. Confusion sometimes occurs by using different morphologic characteristics to define the flat adenoma resulting in different prevalence and different sensitivities. The Working Group on Virtual Colonoscopy describes it as a lesion demonstrating plaque-like morphologic features, with less than 3 mm of vertical elevation above the colonic mucosa [42]. Others have defined it as a lesion with a height of less than one half of its width [75]. Most important are the flat adenomas >1 cm as they present with a higher frequency of dysplasia, with increased malignant potential of the depressed type tumor vs the protruded type [75].

In 1,233 patients Pickhardt et al. described 55 lesions ≥6 mm with 29 adenomas representing 13.8% of all adenomas. Sensitivity for adenomas was 82.8%. The use of primary 3D read was advocated for their detection [76]. In 547 patients, Fidler et al. detected 22 flat polyps with 8 adenomas using single slice CTC technique. With double reading they detected 68% of all flat lesions and 100% of flat adenomas. They used primary 2D read and advised to use both abdominal and intermediate window settings because of varying conspicuity of these lesions [77]. In a series of 18 flat lesions, Park et al. could detect 5 lesions using a 16 slice CT scanner. Poor technique hampered visualization of six lesions. Even in retrospect they were not able to detect seven lesions. All detected lesions had a height of ≥2 mm. They concluded that using 16 slice CT technology they were not able to detect lesions ≤1 mm in height [78].

Flat lesions are difficult to detect for both radiologists and gastroenterologists. A lot of training and experience is needed to adequately detect these lesions and to avoid too many false positives. Their importance remains controversial. Both primary 2D and 3D read can be used for their detection. Double reading seems to improve their detection. As suggested by Park et al. the use of immaculate CTC technique in experienced hands most logically will improve their detection.

Relationship with the gastroenterologists

From a practical point of view, good communication with the gastroenterologists is necessary. Whenever a lesion is detected in the colon, a report which includes exact measurement and detailed description of the lesion improves matching of lesions between CTC and OC. It is also helpful to alert the endoscopist to the presence of a long tortuous colon that might make their exam difficult or if a lesion is on the proximal side of a fold and potentially obscured to an end-viewing endoscope or is flat in morphology.

For patients with medium sized lesions a multidisciplinary approach is sometimes necessary in order to decide on immediate removal of the lesion or follow up with CTC or OC such as that described by Pickhardt et al. [79]. The same approach might be necessary in case of decision making with CTC performed after incomplete OC in a patient with a lesion which has to be removed (polyp ≥1 cm, lesion with suspected malignancy) and cannot be reached with the endoscope.

However, when it comes to the determination of the role of the radiologist and the gastroenterologist in a CRC screening program with CTC, the relationship becomes more delicate and complicated. In other terms, if ever accepted as screening test, what will be the impact of CTC on OC? The negative impact on OC suggested by Hur et al. [80] and based on a mathematical model is highly debateable. The combined factors of (1) generally busy endoscopy schedules (2) poor population compliance with CRC screening guidelines and (3) accurate screening CTC that can be repeated every 7 years–10 years, should form the basis of a cost-effective and mutually beneficial screening program as described by Pickhardt et al. [79].

The issue of who should perform and interpret CTC has come to the fore particularly with the recent position taken by the AGA Institute (American Gastroenterological Association) [81]. It may be possible for anyone who can be properly trained and certified to potentially read CTC. Nevertheless, as the discussion above regarding data to date on training, CTC reading is more complex than the casual observer realizes and is probably not time efficient for a non-radiologist to learn. High quality screening and diagnostic CTC requires quality assurance by the CT technologist while the patient is in the CT suite and sometimes additional scanning is different positions or with intravenous contrast is helpful. The reader of CTC must be able to respond to these technical questions. Somewhat analogous to certification consensus statement by the ACCF/AHA on cardiac imaging, the reader of CTC must be knowledgeable and certified in the use or radiation and use of intravenous contrast agents [82]. This is not intrinsic to the skills of a non-radiologist. If a non-radiologist cannot supervise a CT technologist doing a difficult CTC, then they best not get involved in interpreting the exam. Additionally, the clinically important extracolonic findings are more common in non-screening cohorts, but even in screening cohorts, represent a low-cost patient benefit. Even at low radiation doses, findings can be conspicuous and pose a medicolegal risk to any non-radiologist venturing to interpret CTC.

Conclusion

In summary, CTC is probably second only to optical colonoscopy as the best test to evaluate the colorectum for polyps and masses and data suggests that it can be integrated into an effective CRC screening program by working together with the endoscopists. In an institution with a radiologist who is adequately trained and experienced in performing and interpreting CTC, it is the best test for patients who can not undergo optical colonoscopy or who strongly prefer this less invasive exam. Reduced preparation or prepless CTC, if successful, will likely increase the publics demand for CTC screening and reimbursement.