Specificity and Sensitivity of Computed Tomographic (Virtual) Colonoscopy versus Conventional Colonoscopy for Colorectal Lesion Detection in a Middle Aged or Older Population

Specificity and Sensitivity of Computed Tomographic (Virtual) Colonoscopy versus Conventional Colonoscopy for Colorectal Lesion Detection in a Middle Aged or Older Population: A Systematic Review of the Literature

Todd Schettini, CMR

University of Medicine and Dentistry of New Jersey

School of Health Related Professions

Craig L. Scanlan, EdD, RRT, FAARC

IDST 6400 – Critical Literature Review and Scientific Writing

Spring 2005

Introduction

Computed tomographic or virtual colonoscopy (CTC) is a promising new non-invasive diagnostic tool for detecting colorectal lesions. Originally developed in 1994 by Vining et al, CTC has progressed over the past decade to challenge standard colonoscopy in specificity and sensitivity for colorectal lesion detection and as a primary screening tool (Pescatore, 2000). Although CT colonoscopy technology has improved the data from reported sensitivities has been widespread (Barish 2005). Currently CT colonoscopy is used to evaluate the colon after incomplete colonoscopy, an obstructing carcinoma, or poor candidates for conventional colonoscopy (Macari, 2004). Improvements in CTC colorectal lesion detection have been suggested when transitioning from single to multi detector computer tomography (MDCT), decreasing collimation size, increasing the number of channels for MDCT, and 3 dimensional (3D) versus 2 dimensional (2D) imaging has been used (Chung, 2004; Cohnen 2004; Fenlon 1999). Inconsistency and disagreement are still present in recent clinical literature regarding the specificity and sensitivity of CT colonoscopy versus conventional colonoscopy in detection of clinically appropriate colorectal lesions. Concern was expressed for the ability of CTC to detect various sizes and shapes of lesions and the eventual use of standard colonoscopy for lesion removal. Computed tomographic colonoscopy has a clinical limitation in that ionizing radiation is used. Techniques have been perfected to allow optimal scanning quality with extremely low doses radiation for scans (Cohnen et al, 2004; Vogt et al, 2004). Other potential limitations include CTC scans requiring significant experience and training to interpret and the procedure is not reimbursed by many managed care organizations. However, computed tomographic colonoscopy continually provides a benefit over the limitations of standard colonoscopy.

Conventional or standard Colonoscopy (CC) by definition has been a mainstay in healthcare. Colonoscopy is regarded as the current most effective primary screening method for lesion detection in the large intestine and rectum. CC has evolved into a “gold” standard for colorectal lesion detection, especially for colorectal cancer. Other indications for use of standard colonoscopy include: gastrointestinal hemorrhage, inflammatory bowel disease (Crohn’s disease & ulcerative colitis), high risk patients with hereditary factors for colorectal cancer and unexplained drops in hematocrit in older patients (Cappell, 2002).

Before the use of conventional colonoscopy or any derivative there of (proctosigmoidoscopy or sigmoidoscopy), the only way to examine the large bowel and excise any abnormal lesions was through extensive abdominal surgery (Brenna, 1990). The diagnostic prototype that lead to a widespread use of invasive standard colonoscopy was flexible sigmoidoscopy. Despite the convenience flexible sigmoidoscopy provides, colonoscopy enables a more complete picture of the colon with improved lesion detection attributed to examination of the other two colonic sections (Elwood, 1995). Flexible sigmoidoscopy still remains an alternative choice for colorectal lesions screening according to 2003 American Gastroenterological Association guidelines. These guidelines also hold that conventional colonoscopy is still the most valued and preferred test for colorectal diagnostic evaluation and screening (Winawer, 2003). However, standard colonoscopy has limitations compared to CTC. First CC is an uncomfortable, time consuming, invasive and regarded as a higher risk procedure that usually requires sedation (Chung, 2004). Second it fails to show the entire colon in 5-10% of cases (Macari, 2004; Vogt 2004; Yee, 2001). Third it does not show extracolonic lesions and has potential blind spots near colonic folds (Chung, 2004). Lastly, colonoscopy can be quite expensive compared to CTC in the US (Marcari, 2004; Vogt 2004).

Despite the many applications of conventional colonoscopy for assessing diseases and disorders, the main focus for utilization regresses back to identifying abnormal colorectal lesions. Many colorectal diseases and disorders eventually lead to a higher risk for developing colorectal cancer. Malignant cellular transformations to colorectal carcinomas occur mostly in the ascending colon with Crohn’s disease (Bernstein, 1996). Ulcerative colitis and mechanisms for non-hemorrhoidal gastrointestinal hemorrhage significantly increase the risk of colorectal cancer diagnosis in patients with a longer duration of disease (Itzkowitz, 2002). For younger populations with rare hereditary colorectal cancers (Familial Adenomatous Polyposis-FAP & Hereditary Nonpolyposis Colorectal Cancer-HNPCC), the primary concern is for early detection and surgical removal. These disorders are usually rare, anatomically complex and varied which would allow for a confounding bias in this review. Thus their formal analysis was not considered for the scope of this review.

The current lines of research show promise that CTC may eventually replace CC as an effective colorectal lesion screening measure for both average and high colorectal cancer risk patients. Screening these average and high risk patients for precursors of colorectal cancer accurately and cost effectively is a push of current guidelines (Winawer, 2003). Hence, it makes sense to focus on this data set in order to effectively organize comparisons of CTC versus CC for lesion detection.

This review paper will discuss how CTC lesion detection compares to CC for sensitivity and specificity along with any specified positive and negative predictive values for colorectal lesions. The targeted age for this comparison was 45 years old or greater.

Methods

Studies included in this literature review came from a computerized search of several databases. Review databases included Cochrane Database of Systematic Reviews, ACP Journal Club, Database of Abstracts of Reviews of Effects and Cochrane Controlled Trials Register and were searched via EBM Reviews in early March 2005. Original research articles were found through Ovid MEDLINE(R) <1996 to March Week 1 2005>. This initial search was not restricted on the basis of study language or any other secondary modifier. Medical subject headings (MeSH) were derived from the key terms of colonoscopy, sigmoidoscopy, tomography, sensitivity and specificity. Additional support studies and background information were captured from selected references in the original research articles for the pre-1996 era. These articles were found and accessed via PUBMED. In addition, clinical practice guidelines from relevant associations were considered during a search conducted via the National Guideline Clearinghouse (NCG) in February 2005. Appendix B at the end of this review has information about the specific search criteria used.

The first search of Ovid-MEDLINE for original research studies had yielded nineteen articles with the specified criteria. Fourteen of these studies appeared to be relevant to this review topic. A second search of Ovid-MEDLINE with slightly modified criteria compared to first search yielded about twenty three relevant studies. Most of these were duplicates from the first search, but an additional three articles with relevance were captured. A combined total of seventeen controlled trials were initially deemed fit for analysis based on face validity of titles and/or abstracts.

Inclusion and Exclusion Criteria

Inclusion criteria for the selected articles included a preferred design template for diagnostic studies. This template included cross sectional surveys or prospective, blind comparisons to a gold standard. All studies discussed in the Review of Literature section had a prospective design, standard colonoscopy as the comparative reference to computed tomographic colonoscopy, sensitivity and/or specificity data reported, populations with an age range over 45 years old and a sample size greater than 50 patients. Specific inclusion criteria also consisted of data for all shapes of lesions at least 10mm in size, virtual colonoscopy with 2D or 3D scans and any channel single or multidetector scanner, same day colonoscopy following CTC, any radiation dose was acceptable, any collimation size was allowed, any computer enhancement software was appropriate, rater-blind or “segmental unblinding” and proper bowel preparation for both types of colonoscopy. Exclusion criteria included any study non-translatable to English, specific studies of younger populations significantly below a mean of age 45 (FAP and HNPCC included in this exemption as screening starts in the teens and early 20s along with some IBD disorders) and indirect comparison of CTC to CC to anything other than specified colorectal lesions (ie. gastric hemorrhage). According to these criteria, four of the original seventeen articles were excluded from the literature review. All integrative articles or clinical guidelines are presented in Appendix B, along with specific review and consensus papers in References section.

Original research articles retrieved from both preliminary searches	37
Articles retrieved after duplicates excluded and face validity assessed	17
Articles left after application of Review’s Inclusion / Exclusion Criteria	13

Current gastroenterology and endoscopy guidelines delineate the age of 50 for initial colonoscopy and colorectal lesion screening. As of late March 2005, guidelines by the American College of Gastroenterology lowered the African American screening age from 50 down to 45 due to a general increase in risk for this population. This awaited change is due to data on African Americans having the highest incidence of colorectal cancer of any racial/ethnic group. Thus the decision to look at patients that were 45 years old or greater was two fold. First, conveniently MEDLINE has search subheadings that include age 45 years (Middle Aged) or older (Aged) for studies. Second, this age cut off would capture the most representative patients of all appropriate risk categories when screening for colorectal lesions. These criteria would also help to reduce any selection bias that may skew results more in favor of one modality over another.

Background

During the mid 1990s, an increasing number of initial studies were being published comparing virtual colonoscopy and conventional colonoscopy. Many of the studies were feasibility studies for early techniques of CTC. During the first five years of this century, CTC has made tremendous strides for improving the accuracy of colorectal lesion detection. A consistent basis for comparison of different studies in computed tomographic colonoscopy versus conventional colonoscopy is sensitivity and specificity data. Virtually all comparative studies of CTC to CC have presented an overall study sensitivity and/or specificity. Most comparative studies have even broken down lesion categories into type and size. The category of “type” can be stratified further into cancerous and non-cancerous lesions. Classification of non-cancerous lesions excluded diverticula and fistulas. The lesion category of “size” was separated into three groups. Some studies explicitly expressed positive and negative predictive values for the overall and specific lesion size categories.

Many trials in the 1990’s started publishing data on three size intervals for lesion detection. Despite some variation of a millimeter in one direction or another, the categories distinguished were large (10mm or greater), medium (6-9mm), and small (less than 6mm). A consensus paper on virtual colonoscopy in March 2005 stated that a lesion size of 5mm or below is a clinically unimportant find and need not be reported (Barish, 2005). Thus detection of a colorectal lesion 10mm or greater is clinically significant from a health risk and excision standpoint. Pressures for accurate detection of colorectal lesions sized 6mm to 9mm are growing as evidence for clinical utility and risk continues to build. There is still debate in the literature on the shape of a colorectal lesion and the clinical significance (Pickhardt, 2004).

Review of the Literature with Discussion

Computed tomographic colonoscopy now has a significant wealth of prospective cohort study literature behind it. Unfortunately there is not much structure or consistency to this large volume of information. The data supporting CTC is consistently specific when compared to CC but the sensitivities reported are inconsistent. Many of the sensitivities and specificities reported were for the overall studies which incorporated clinically significant and insignificant lesion detection data. Extracting clinically relevant data of lesion detection for 10mm or greater (large) and 6-9mm (medium) groups should improve overall sensitivities for CTC as this is more practical and valid. Assessment of common pitfalls in determining sensitivity should be characterized as well. Major mechanical factors thought to influence results of a CTC scan results are the type of scanner used, collimation, and the dimensional mode of imaging (Cohnen, 2004; Gluecker, 2002).

Sensitivity and Specificity Data for Lesion Sizes of ≥10mm (large)

A lesion size of 10mm or greater has clinical significance for the long term health risks of a patient. Almost all comparisons of CT colonoscopy to conventional colonoscopy use this size group as a benchmark. Cotton et al (2004) cites a sensitivity of 90% for lesion sizes of 10mm or more with single center studies, while other somewhat recent studies reported sensitivities from 61-78% for this large lesion group. Data extracted from all studies in this review reported an average sensitivity of 86% and a specificity of 85% for the 10mm or greater lesions. An individual numerical summary of each study for a lesion size of 10mm or greater can be found in the Appendix A under TABLE 1.

Each study included in this review had expectations of CT colonoscopy performing at least equivalent to conventional colonoscopy for colorectal lesion detection. When looking at the 10mm or greater lesion group, five studies had sensitivity results significantly below expectations along with sensitivities less than 85%. Many if not all of these studies according to improvements in mechanical technology are be expected to have sensitivities in the 90% range. Interestingly enough the timeframe of all five studies spans from the year 2000 through present day 2005.

Three out of the five studies had sensitivities in the 70-80% range, not too far from an acceptable and expected 90-100% sensitivity range. The earliest study of Mendelson et al (2000) demonstrated a sensitivity of 73% but had incorporated mostly supine only scans. Patients started to receive what is held to be the standard in CT colonoscopy of both supine and prone scans later into the study. This was also one of the few non-US based early CTC studies. Scanning with only one position can decrease the sensitivity of the scan by increase blind spots on the CT scan (Gluecker et al, 2001; Chung et al, 2005). Both Cohnen et al (2004) and Gluecker et al (2002) had solid methodologies, improved technologies and valid study designs, but a pattern was seen with the missed lesions. A flat shape or a resemblance to more of a thickening on the colon wall than a pronounced colorectal lesion may be to blame. All but one lesion was found on retrospective analysis of the scans.

The last two studies had demonstrably lower sensitivities in the mid-50 percent range. Aside from Pickhardt et al (2003), Cotton et al (2004) had one of the more rigorous and explicitly designed studies for comparison and repeatability when examining CTC to CC colorectal lesion detection. Cotton et al (2004) and Rockey et al (2005) both used large patient sizes, MDCT scanners, appropriate image quality and collimation, high risk populations and multicenter study design. This made both studies appear to have high internal validity. One factor that was ascribed to in both trials as a potential confounder was the interpretation of the radiological scans. Cotton et al (2004) accounted for image quality but not diagnostic accuracy as study radiologists only had to perform 10 CTC procedures prior. This was the lowest investigator acceptance criteria for radiologists with in the past five years. In addition the only site that had significant prior experience with the CTC technique had a sensitivity of 82%. Ironically in the Rockey et al (2005), the mention a cut point of 50 CTC procedures for experienced radiologist interpreters. Anyone else below a lifetime threshold of 50 CTC scans underwent non-specific scan interpretation training. Half of the readers were below the 50 CTC threshold. Both of these studies did not provide feedback to the radiologists for the accuracy of scan interpretation.

Sensitivity and Specificity Data for Lesion Sizes of 6-9mm (medium)

Clinical significance for lesions greater than 10mm in size has been established by the medical community. Many researchers and clinicians have conflicting views on how to interpret the clinical utility of lesion sizes within the 6mm to 9mm range. Some papers are suggesting a lesion size threshold of 8mm for clinical utility and 5-9mm as a size to be followed with surveillance (Barish, 2005; Pickhardt et al, 2003).

When assessing the 6-9mm size lesion group, a small drop in sensitivity was expected. Overall sensitivity dropped significantly to 68% but specificity remained fairly high at 87%. A numerical summary of results for the 6-9mm lesion size group is located in Appendix A under TABLE 2. A pattern was noted from the larger lesion size group to this medium size group. The same five studies that had significantly lower sensitivities in the 10mm or greater lesion group also had extremely lower sensitivities in the 6-9mm lesion group. One study by Cohnen et al (2004) depicted an almost paradoxical improvement in sensitivity to 86% for the 6-9mm lesion size group. Reasons for a lower sensitivity from the 10mm or greater lesion group, resulted from flat lesions or lesions that resembled the colonic wall initially. Two of the three lesions were found on retrospective analysis, which would have brought the sensitivity for the large lesion group for that study to 93%.

Extracting out the five studies that had poor sensitivities in the larger lesion size groups, the overall sensitivity rises to 83% for the medium size lesion group. There was an expectation that detection differences from the larger size lesion group to carry over to the medium sized group if there was consistent systematic error. The 6-9mm lesion group did have proportionally smaller sensitivities in four of the five trials, Cohnen et al (2004) excluded. Another dissonant observation was that the sensitivity for Marcari et al (2004) had fallen by almost half to 53%. Some considerations included a sample size of only 68 patients, all males and the study population represented only average risk patients. The impressive sensitivity of 100% found in the 10mm or greater lesion size group may have been due to only three large lesions to detect. There were seventeen lesions that were present in the 6-9mm category.

Sensitivity and Specificity Data for Type of CTC Scanner

Initial chronological studies used a single detector CT (SDCT) scanner for colorectal lesion assessment. Three studies looked at single row detection for lesions sizes of 10mm or greater. Fenlon et al (1999) and Yee et al (2001) show that SDCT can produce comparable sensitivities ranging from 90-96% and specificities from 74.5-96%. The 6-9mm lesion group also enjoyed relatively high specificity and sensitivity. One paper with SDCT by Mendelson et al (2000) provided conflicting evidence that CT colonoscopy was not a useful diagnostic tool with a modest sensitivity of 73%. This sensitivity fell significantly to 19% for lesion sizes below 10mm. A use of mostly supine scans in the paper by Medelson and colleagues, unlike both supine and prone scans in the prior two studies may have effected sensitivity. Pooled positive and negative predictive values for the SDCT studies were high. The positive & negative predictive values averaged 89% and 96% respectively for the 10mm or greater lesion group along with 81% and 92% respectively for the 6-9mm lesion group.

The rest of the literature had used some variation of multidetector CT scanner (MDCT). Most of the MDCT scanners use in this study had 4 rows or channels. Current technology allows number of row choices from 2, 4, 8 and 16. Switching from a single to multidetector CT scanner is supposed to increase image resolution, with more rows of MDCT adding to the resolution quality (Cohnen, 2004). The majority of articles had used 4 row MDCT (4-MDCT). Five other studies had used various other row MDCT scanners from 2-16. The sensitivity and specificity for the l0mm or greater lesion group with the 4-MDCT was extremely high at 94% and 95% respectively. Data on positive and negative predictive values in this group was limited but in the low 90% range for both values. When assessing the 6-9mm range for this group the results were less than satisfactory with positive and negative predictive values barely in the 77-80% range for each.

Looking at the rest of the MDCT sensitivity and specificity ranges, any increase in MDCT rows correlated with an increase in sensitivity from 55% to 100%. Specificity was high around 90-97% for all ranges of MDCT and both lesion size groups as well. Conflicting results from Rockey et al (2005) showed a poor sensitivity at 59% for a 4/8 MDCT scanner, while Pickhardt et al (2003) demonstrated at 94% sensitivity with a 4/8 MDCT. The magnitude of difference was maintained in the 6-9mm group as well. Rockey and colleagues surmise that their primary use of 2D analysis versus the 3D analysis in Pickhardt et al (2003) was mainly responsible for the difference. The only study that looked at 16MDCT had a sensitivity of 100% in the large lesions and 94% in medium lesions. These results proved superior to colonoscopy in the 10mm or greater lesion size group.

Sensitivity and Specificity Data for Collimation Size

Collimation refers to focusing a radiation beam to a particular area in a specific width or band. The lower the collimation the higher the amount of radiation per area but there would be less noise distortion and improved image quality. There were approximately three different groups of collimation sizes of 1mm, 2-3mm and 5mm. There was one study that had collimation at 0.75mm. Collimation size decreased with chronological order of the studies, with the lowest collimation of 0.75mm by Chung et al (2005) that was included in the 1mm collimation size category.

Most of the three studies with 5mm collimation were earlier in the timeline and had an overall sensitivity of 84% and a specificity of 94% for the large lesion group. However, sensitivity dropped to 50% with the medium lesion group. The next lowest collimation of 2-3mm group has an overall sensitivity of 82% and specificity was 93% for the 10mm or greater lesions. Again a sizable drop to 69% sensitivity was seen for the 6-9mm lesion group. The smallest collimation size of 1mm demonstrated the highest sensitivity at 95% and a specificity of 94% for the largest lesion size. The medium lesion size group also had a fairly respectable sensitivity at 81% while specificity was strong at 89%.

Positive and negative predictive values were mostly reported for the 2-3mm collimation size studies. Positive predictive values were heterogeneous averaging 74% and 58% for large and medium lesion size groups respectively. Negative predictive values for both lesion size groups were 91% to 98%.

A collimation size of 1mm or smaller appeared to correlate with a high sensitivity and specificity for colorectal lesion detection.

Sensitivity and Specificity Data for 2 Dimensional (2D) and 3 Dimensional (3D) Imaging

Only three studies had used primarily used 3 dimensional imaging with or without fly through. The rest of the studies employed a 2D imaging mode with 3D used to confirm a lesion or resolve and 2D image discrepancies. The imaging mode available depended on the computer software and processing capabilities.

The three studies that utilized primarily 3D or fly through (simultaneous combination of 3D-2D) imaging demonstrated an extremely high sensitivity at 98% and a specificity of 90% for the 10mm or greater lesion group. Impressive results were also seen in the 6-9mm lesion size group with a sensitivity of 91%. Pickhardt et al (2003) was one of the first to employ this imaging technique and reported a sensitivity of 94% for the 10mm or greater lesion group. Both Vogt et al (2004) and Chung et al (2005) in future studies had reported 100% sensitivity with this type of imaging for the same large lesion group. The remaining articles had a varied range for their sensitivity averaging 83% for the large lesion group and 61% for the medium lesions. Positive and negative predictive values were not reported for many of these analyses.

Although only a small amount of studies had primary 3D imaging software available, this appears to have an advantage of improving sensitivity over the conventional 2D imaging, even with 3D confirmation.

Sensitivities for Colorectal Cancer Lesions of Any Size

Colorectal cancer (CRC) has moved up from the third leading cause of cancer-related death to ranking second. 60,000 CRC deaths still occur in the US annually (Hoppe, 2003). Conventional colonoscopy has demonstrated a reduction the risk of colorectal cancer death via early detection and polyp removal. Colonoscopy is currently considered the reference standard for detection colorectal cancer in symptomatic patients (Vogt 2004). This invasive procedure is not without risk and limitation. Comparable alternative primary screening methods are needed to overcome the limitations of colonoscopy. Computed tomographic colonoscopy has been one of those alternative methods presented. Studies that would have examined CTC versus CC in an exclusive colorectal cancer population have been left out of this paper. This was done to limit any potential bias for the utility of CTC lesion screening compared to CC. Studies included in this review had mixed populations of average risk (screening) patients as well as high risk patients.

CT colonoscopy has progressed in many different directions for colorectal lesion detection in the past decade. Separate data on sensitivities for specific detection of colorectal cancer carcinomas was provided by seven studies in this analysis. Fenlon et al (1999) was one of the first larger scale studies with over 100 high risk patients to comment on the detection of three carcinomas picked up by CT colonoscopy. This provided the CTC with a sensitivity of 100%. However, the number of colorectal carcinomas for detection was small at only three. In 2001, Yee and colleagues assessed an even larger sample size of 300 patients with a mixed risk profile. This study was the first reported use of CT Colonoscopy for detection of colorectal neoplasm in average risk patients. A sensitivity of 100% was reported for all eight colorectal carcinomas. Both of these initial studies had used a single row detection CT colonoscopy. This implies that the lesions were most likely larger than 10mm in size if sensitivity was this high with single row CTC.

Single row detection CT colonoscopy was being phased out by the new multidetector technology in 2002 that provided better image quality in comparison. When considering low dose multi-detector row CT (MDCT) colonoscopy, data from Iannaccone et al (2003) verified high sensitivities for colorectal cancer detection with CTC. A predominantly high risk symptomatic sample of 115 patients yielded 22 colorectal carcinomas. Initial multidetector CTC had found all 22 colorectal carcinomas with a sensitivity of 100%. However, Hoppe et al (2004) had reported that only seven of the eight colorectal carcinomas were detected with comparable MDCT colonoscopy. This single institutional study generated a sensitivity of 88% from a symptomatic sample population. Hoppe and colleagues were the first to indicate specific carcinoma size for cancer detection with CTC in this review. All seven of the eight carcinomas detected were over 10mm in size. The cancer lesion that was not detected was 7mm in size. A multicenter trial of a symptomatic high risk patient population with colorectal neoplasm was undertaken by Cotton et al (2004) to help clarify this single institution study. A low sensitivity of 75% was recorded as two of eight colorectal cancers were missed. Sizes were again offered for the two missed lesion with one being a 17mm and the other 7mm. Despite an impressive design, this study appeared to have abnormally low sensitivity rates for all lesion types and sizes. A contributing factor that may have lead to CTC scan interpretation discrepancies in this study was the potential inexperience of investigator radiologists.

The last two studies examine the ability of CTC in high risk, symptomatic patients to detect colorectal cancers. Rockey et al (2005), a multicenter study with a high risk patient sample, had reported a sensitivity of 78% for 7 out of 9 carcinomas detected. This study had used an advanced slice MDCT as an appropriate standard for assessment. The two missed carcinomas were not specifically mentioned, but eight of the nine cancers were Dukes stage A or B which has a high probability of being smaller in size. The study by Chung et al (2005) had used the highest channel MDCT scanner and lowest collimation at 0.75 of any CTC comparison study published. All other prominent methodologies employed by other recent studies were included in the study significantly improving their internal validity. The sensitivity was 100% for detection of all 21 colorectal cancers. These results are encouraging but a patient sample of 51 participants is lower than what would be favorable for ruling out statistical bias from patient variability. Nonetheless these data suggest an acceptable and high sensitivity for detection of cancerous colorectal lesions with almost any detector CT colonoscopy.

Implications or Unresolved Questions

In summary, computed tomographic colonoscopy appears to have significant clinical advantages over conventional colonoscopy. Collective sensitivities and specificities were fairly high at ~ 85% for the larger lesion size group. There was an expected but noticeable drop in specificity to 68% in the 6-9mm lesion size group. Studies that were included in this review passed fairly strict inclusion criteria from a methodological standpoint. However, even well designed and internally valid studies may be inherently flawed in some way. Two studies in particular had unusually low specificities for both the 10mm or greater and 6-9mm lesion group. Results from these two trials brought the overall sensitivities down below 90%. Both studies highlighted that a potential reader bias may have confounded the study when interpreting the CT colonoscopy scan. Poor educational and lack of standardized training in identifying abnormalities have been suggested to contribute to a poor sensitivity (Fidler et al, 2004; Pescatore et al 2000). CTC image quality can also have an affect on a reader’s ability to interpret the scan. Increasing radiation usually correlated with less image noise and better picture quality. Radiation doses used in studies for this review were not indicated as specific issues for poor image quality. Inadequate bowel preparation or inadequate air insufflations have been implicated. However all studies had acceptable bowel preparation and air insufflations greater then 75% for patients. Other studies mention that improvements in MDCT technology, software and technique would be able to compensate for poor bowel prep and air insufflations (Barish et al, 2005).

Aside from colorectal lesion size, there was a consistently high rate of flat or sessile lesions missed with CTC compared to CC. The clinical significance of a flat colorectal lesion is still unclear. One recent study examining the issue of flat colorectal lesions argued they are rare in a Western screening population as well as advanced flat neoplasms (Pickhardt et al, 2004). This study also reported that the sensitivity of detecting flat lesions overall was similar to that of polypoid lesion detection, and does not represent a significant drawback for CTC screening.

When comparing CTC to CC in the studies selected, previous research suggested that the CTC Scanner, collimation size, and different dimensional imaging were primary factors to explain poor sensitivities. In this review, there was a correlation of improved CTC sensitivity with both lesion size groups when a higher channel MDCT scanner, collimation size of 1mm or lower and primary 3D scanning or fly through technology were used. Current CTC studies appear closer to utilizing all three of these techniques together in newer studies (Chung et al, 2005). Newer techniques incorporating fecal tagging to reduce the number of false positives on CTC scans are currently suggested.

An important implication that was brought up in virtually every study was that “gold” standard of conventional colonoscopy was not 100% accurate for lesions detection. Problems with incomplete scans and blind spots for lesion detection on colonic folds are important drawbacks of CC. There were many lesions noticed on CTC that were recorded as false positives, when in fact they may have been true positives and thus increasing the specificity of CTC and decreasing that of CC. This observation would also impact the data for which positive and negative predictive values are based. Extracolonic findings may be another benefit of CTC over CC.

Although no US dollar amount was listed for CTC in any of the studies, Marcari et al (2004) listed the average reimbursement for a colonoscopy at $1730. This amount was insinuated to be much higher than a CTC would cost per patient. An implication for widespread use of CTC as a primary screening technique would be reimbursement by managed care organizations. A review newsletter by The Medical Letter on Drugs and Therapeutics in February 2005 reported that virtual colonoscopy has an expected cost similar to that of traditional colonoscopy. This article also stated that costs could be higher if CTC screening detects removable lesions and a follow up colonoscopy would have to be performed.

Future of Computed Tomographic Colonoscopy

Future studies comparing CTC to CC as a primary screening tool for colorectal cancer should focus on clinically significant lesion sizes as these bear the most important decision for removal or not. More standardized training and education should be available for radiologists interpreting CTC scans. This training should encompass specific characteristics of abnormality relating to lesion shape and size. A possible solution is computer automated interpretation software that is currently developed but not clinically tested (Fidler et al, 2004; Pescatore et al, 2000). Trials should continue to employ the highest channel MDCT available with a low collimation size and primary 3D or fly through computer imaging as this was correlated with higher sensitivities.

Improving CTC as a cost-effective technology while maintaining the convenience as a primary diagnostic test for colorectal lesion screening will help influence demand for reimbursement from third party payers.

Conclusion

The goal of primary colorectal screening should be identification of early malignant carcinomas, but also precursor lesions of any clinically relevant size. The review paper illustrates that computer tomographic colonoscopy is a highly sensitive and specific test for detecting malignant colorectal lesions. In addition, CTC has made progress in demonstrating equivalence to CC as primary screening tool for colorectal lesions of various sizes. Unfortunately, more consistency in CTC sensitivity data and optimizing study designs with the latest technology is needed before definitive implementation of CTC as a widespread primary screening tool over CC can be assured.

References

Barish, M. S. J., Ferrucci JT. (2005). Consensus on current clinical practice of virtual colonoscopy. American Journal of Roentgenology, 184, 786-792.

Bernstein, D. R. A. (1996). Malignancy in crohn's disease. American Journal of Gastroenterology, 91(3), 434-440.

Brenna, E. (1990). The benefit of colonoscopy. scand Journal of Gastroenterology, 25(1), 81-88.

Callstrom MR., J. C., Fletcher JG. Reed JE. Ahlquist DA. Harmsen WS. Tait K. Wilson LA. Corcoran KE. (2001). Ct colonography without cathartic preparation: Feasibility study. Radiology, 219(3): (Jun), 693-698.

Cappell, M. (2002). Recent advances in gastroenterology. The Medical Clinics of North America, 86(6), 13-15.

Chen SC., L. D., Hecht JR. Kadell BM. (1999). Ct colonography: Value of scanning in both the supine and prone positions. American Journal of Roentgenology, 172(3):(Mar), 595-599.

Chung, D. H. K., Choi WJ, Kim JK. (2005). Ct colonography using 16-mdct in the evaluation of colorectal cancer. American Journal of Roentgenology, 184, 98-103.

Cohnen M., V. C., Beck A. Andersen K. Heinen W. vom Dahl S. Aurich V. Haeussinger D. Moedder U. (2004). Feasibility of mdct colonography in ultra-low-dose technique in the detection of colorectal lesions: Comparison with high-resolution video colonoscopy. American Journal of Roentgenology, 183(5): 1355-1359.

Cotton PB., D. V., Pineau BC. Palesch YY. Mauldin PD. Hoffman B. Vining DJ. Small WC. Affronti J. Rex D. Kopecky KK. Ackerman S. Burdick JS. Brewington C. Turner MA. Zfass A. Wright AR. Iyer RB. Lynch P. Sivak MV. Butler H. (2004). Computed tomographic colonography (virtual colonoscopy): A multicenter comparison with standard colonoscopy for detection of colorectal neoplasia. JAMA, 291(14): (Apr 14), 1713-1719.

Dachman, A. a. K. J., Boyle CM, Samara Y, Hoffmann KR, Rubin DT and Hanan I. (1998). Ct colonography with three-dimensional problem solving for detection of colonic polyps. American Journal of Roentgenology, 171(4), 989-995.

Edwards JT., M. R., Fritschi L. Foster NM. Wood C. Murray D. Forbes GM. (2004). Colorectal neoplasia screening with ct colonography in average-risk asymptomatic subjects: Community-based study. 230(2):(Feb), 459-464.

Elwood, J. A. G., Schulp MM, McNoe B, Barbeza Go, North F, Sutton K, Parry B, Chadwick VS. (1995). Flexible sigmoidoscopy or colonoscopy for colorectal screening: A randomized trail of performance and acceptability. Cancer Detection and Prevention, 19(4), 337-347.

Fenlon HM., N. D., Schroy PC 3rd. Barish MA. Clarke PD. Ferrucci JT. (1999). A comparison of virtual and conventional colonoscopy for the detection of colorectal polyps. New England Journal of Medicine., 341(20):(Nov 11), 1496-1503.

Fidler, J. J. F., CD Johnson, JE Huprich, JM Barlow, F Earnest, BJ Bartholmal. (2004). Understanding interpretive errors in radiologists learning computed tomography colonography. Academic Radiology, 11(7), 750-756.

Ginnerup Pedersen B., C. T., Bjerregaard NC. Ljungmann K. Laurberg S. (2003). Colonoscopy and multidetector-array computed-tomographic colonography: Detection rates and feasibility. Endoscopy, 35(9): (Sep), 736-742.

Gluecker T., D. G., Keller W. Jornod P. Meuli R. Schnyder P. (2002). Performance of multidetector computed tomography colonography compared with conventional colonoscopy. Gut, 51(2): (Aug), 207-211.

Hara AK., J. C., Reed JE. Ahlquist DA. Nelson H. MacCarty RL. Harmsen WS. Ilstrup DM. (1997). Detection of colorectal polyps with ct colography: Initial assessment of sensitivity and specificity. Radiology, 205(1):(Oct), 59-65.

Hoppe H., N. P., Spreng A. Quattropani C. Mattich J. Dinkel HP. (2004). Prospective comparison of contrast enhanced ct colonography and conventional colonoscopy for detection of colorectal neoplasms in a single institutional study using second-look colonoscopy with discrepant results. American Journal of Gastroenterology, 99(10):(Oct), 1924-1935.

Iannaccone R., L. A., Catalano C. Brink JA. Mangiapane F. Trenna S. Piacentini F. Passariello R. (2003). Detection of colorectal lesions: Lower-dose multi-detector row helical ct colonography compared with conventional colonoscopy. Radiology, 229(3):(Dec), 775-781.

Itzkowitz, S. (2002). Cancer prevention in patients with inflammatory bowel disease. Gastroenterology Clinics of North America, 31(4), 1133-1144.

Kay CL., K. D., Hawes RH. Young JW. Cotton PB. (2000). Virtual endoscopy--comparison with colonoscopy in the detection of space-occupying lesions of the colon. Endoscopy, 32(3): (Mar), 226-232.

Laghi A., I. R., Carbone I. Catalano C. Panebianco V. Di Giulio E. Schillaci A. Passariello R. (2002). Computed tomographic colonography (virtual colonoscopy): Blinded prospective comparison with conventional colonoscopy for the detection of colorectal neoplasia. Endoscopy, 34(6):(Jun), 441-446.

Macari M., B. E., Jacobs SL. Naik S. Lui YW. Milano A. Rajapaksa R. Megibow AJ. Babb J. (2004). Colorectal polyps and cancers in asymptomatic average-risk patients: Evaluation with ct colonography. Radiology, 230(3):(Mar), 629-636.

Mendelson RM., F. N., Edwards JT. Wood CJ. Rosenberg MS. Forbes GM. (2000). Virtual colonoscopy compared with conventional colonoscopy: A developing technology. Medical Journal of Australia, 173(9):(Nov 6), 472-475.

Neri E., G. P., Battolla L. Vagli P. Boraschi P. Lencioni R. Caramella D. Bartolozzi C. (2002). Colorectal cancer: Role of ct colonography in preoperative evaluation after incomplete colonoscopy. Radiology, 223(3):(Jun), 615-619.

Ng CS., D. T., Pinto EM. Courtney HM. Bull RK. Prevost AT. Campbell GA. Freeman AH. Dixon AK. (2002). Evaluation of ct in identifying colorectal carcinoma in the frail and disabled patient. European Radiology, 12(12):(Dec), 2988-2997.

Pescatore P., G. T., Delarive J. Meuli R. Pantoflickova D. Duvoisin B. Schnyder P. Blum AL. Dorta G. (2000). Diagnostic accuracy and interobserver agreement of ct colonography (virtual colonoscopy). Gut, 47(1):(Jul), 126-130.

Pickhardt PJ., C. J., Hwang I. Butler JA. Puckett ML. Hildebrandt HA. Wong RK. Nugent PA. Mysliwiec PA. Schindler WR. (2003). Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults. New England Journal of Medicine., 349(23):(Dec 4), 2191-2200.

Pickhardt PJ., N. P., Choi JR. Schindler WR. (2004). Flat colorectal lesions in asymptomatic adults: Implications for screening with ct virtual colonoscopy. American Journal of Roentgenology, 183(5):(Nov), 1343-1347.

Pineau BC., P. E., Chen GJ. Espeland MA. Phillips K. Han JP. Mikulaninec C. Vining DJ. (2003). Virtual colonoscopy using oral contrast compared with colonoscopy for the detection of patients with colorectal polyps. Gastroenterology, 125(2):(Aug), 304-310.

Rockey DC., P. E., Niedzwiecki D., Davis W. Bosworth HB. Sanders L. Yee J. Henderson J. Hatten P. Burdick S. Sanyal A. Rubin DT. Sterling M. Akerkar G. Bhutani MS. Binmoeller K. Garvie J. Bini EJ. McQuaid K. Foster WL. Thompson WM. Dachman A. Halvorsen R. (2005). Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy: Prospective comparison. Lancet, 365(9456):(Jan 22), 305-311.

Royster AP., F. H., Clarke PD. Nunes DP. Ferrucci JT. (1997). Ct colonoscopy of colorectal neoplasms: Two-dimensional and three-dimensional virtual-reality techniques with colonoscopic correlation. American Journal of Roentgenology, 169(5):(Nov), 1237-1242.

Summers RM., Y. J., Johnson CD. (2004). Ct colonography with computer-aided detection: Automated recognition of ileocecal valve to reduce number of false-positive detections. Radiology, 233(1):(Oct), 266-272.

Van Gelder RE., N. C., Florie J. Bartelsman JF. Snel P. De Jager SW. Van Deventer SJ. Lameris JS. Bossuyt PM. Stoker J. (2004). Computed tomographic colonography compared with colonoscopy in patients at increased risk for colorectal cancer. Gastroenterology, 127(1):(Jul), 41-48.

Vogt C., C. M., Beck A. vom Dahl S. Aurich V. Modder U. Haussinger D. (2004). Detection of colorectal polyps by multislice ct colonography with ultra-low-dose technique: Comparison with high-resolution videocolonoscopy. Gastrointestinal Endoscopy, 60(2):(Aug), 201-209.

Winawer S, F. R., Rex D, Bond J, Burt, R, Ferrucci J, Ganiats T, Levin T, Woolf S, Johnson D, Kirk L, Litin S, and Simmang C for US Multisociety Task Force on Colorectal Cancer. (2003). Colorectal cancer screening and surveillance: Clinical guidelines and rationale- update based on new evidence. Gastroenterology, 124(2), 544-560.

Yee J., A. G., Hung RK, Steinauer-Gebauer, Wall SD, McQuaid KR. (2001). Colorectal neoplasia: Performance characteristics of ct colonography for detection in 300 patients. Radiology, 219(3):(June), 685-692.

Appendix

Appendix A

TABLE 1 - Sensitivities, Specificities, +/- Predictive Values of lesions 10mm or greater

Study Author	Study Year	Sample Size & Age (yrs)	Overall 10mm Sensitivity*	Overall 10mm Specificity*	+ Predict Value*	- Predict Value*	Coll. (mm)	CTC Scanr #
Fenlon et al	1999	N=100 Age= 62.0	CTC-96% CC-100%	CTC-96%	96%	96%	5	2Dw/3D single
Mendelson et al	2000	N=100 Age=65	CTC-73% CC-NS	CTC- 94.1	NS	NS	5	2Dw/3D single
Yee et al	2001	N=300 Age=62.6	CTC-90% CC-NS	CTC-74.5%	81%	97.2%	3	2Dw/3D single
Gluecker et al	2002	N=51 Age=62.5	CTC-82% CC-NS	CTC- 90%	NS	NS	5	2Dw/3D multi
Pickhardt et al	2003	N=1233 Age= 57.8	CTC-93.8% CC-87.5%	CTC-96.0%	48.9%	99.7%	1.25-2.5	3Dw/2D 4/8multi
Iannaccone et al	2003	N= 158 Age=63.5	CTC-96% CC-NS	CTC- 96.6%	94.1%	97.7	3	2Dw/3D 4-multi
Hoppe et al	2004	N=92 Age=66	CTC-95% CC-NS	CTC-98%	95%	98%	2	2Dw/3D 4multi
Cotton et al	2004	N=615 Age=61	CTC-55% CC-100%	CTC-96.0% CC-100%	50.0%	99.8%	2.5 & 5	2Dw/3D 2/4multi
Vogt et al	2004	N=115 Age=58	CTC-100 CC-NS	CTC-82	NS	NS	1	3Dw/2D 4 multi
Cohnen et al	2004	N=137 Age= 57.1	CTC-78.6% CC-NS	CTC-100%	76.3%	75.6%	1	2Dw/3D 4multi
Macari et al	2004	N=68 Age= 55	CTC-100% CC-NS	CTC-98.5%	NS	NS	1	2Dw/3D 4multi
Rockey et al	2005	N=614 Age=57	CTC- 59% CC- 98.4%	CTC-96.0% CC- 99.6%	NS	NS	2.5	2Dw/3D 4/8multi
Chung et al	2005	N=51 Age=63	CTC-100% CC-78%	CTC-NS	NS	NS	0.75	3Dw/2D 16 multi

Note: * are given as per-patient/polyp analysis

NS = Not Stated

2Dw/3D = primary 2 Dimensional analysis with 3 Dimensional verification for discrepancy

#multi = # of row multidetector;

single = single row detector

TABLE 2 - Sensitivities, Specificities, +/- Predictive Values for lesions 6-9mm in size

Study Author	Study Year	Sample Size & Age (yrs)	6-9mm Sensitivity*	6-9mm Specificity*	+ Predict Value*	- Predict Value*	Coll. (mm)	CTC Scanr #
Fenlon et al	1999	N=100 Age= 62.0	CTC-94% CC-100%	CTC-92%	92%	94%	5	2Dw/3D single
Mendelson et al	2000	N=100 Age=55+	CTC-19% CC-NS	CTC-NS%	88%	89%	5	2Dw/3D single
Yee et al	2001	N=300 Age=62.6	CTC-80.1 CC- NS	CTC- 72.0%	62.9%	92.2%	3	2Dw/3D single
Gluecker et al	2002	N=51 Age=62.5	CTC-33% CC-NS	CTC- 90%	NS	NS	5	2Dw/3D multi
Pickhardt et al	2003	N=1233 Age= 57.8	CTC-88.7% CC-92.3%	CTC-77.0%	40.7%	97.8%	1.25-2.5	3Dw/2D 4/8-multi
Iannaccone et al	2003	N= 158 Age=63.5	CTC-83.3% CC-NS	CTC-NS	NS	NS	3	2Dw/3D 4- multi
Hoppe et al	2004	N=92 Age=66	CTC-76% CC-94%	CTC-88%	79%	86%	2	2Dw/3D 4multi
Cotton et al	2004	N=615 Age=61	CTC-39% CC-99%	CTC-90.5% CC-100%	46.6%	87.7%	2.5 & 5	2Dw/3D 2/4multi
Vogt et al	2004	N=115 Age=58	CTC-91% CC-NS%	CTC-83%	NS	NS	1	3Dw/2D 4 multi
Cohnen et al	2004	N=137 Age= 57.1	CTC-85.7% CC-NS	CTC-92.8%	76.3%	75.6%	1	2Dw/3D 4multi
Macari et al	2004	N=68 Age= 55	CTC-52.9% CC-NS	CTC-89.7%	NS	NS	1	2Dw/3D 4multi
Rockey et al	2005	N=614 Age=57	CTC- 51% CC- 99%	CTC-96.0% CC- 99.6%	NS	NS	2.5	2Dw/3D 4/8multi
Chung et al	2005	N=51 Age=63	CTC-94% CC-78%	CTC-NS	NS	NS	0.75	3Dw/2D 16 multi

Note: * are given as per-patient/polyp analysis

NS = Not Stated

2Dw/3D = primary 2 Dimensional analysis with 3 Dimensional verification for discrepancy

#multi = # of row multidetector;

single = single row detector

Appendix B

ALL EBM REVIEWS- Cochrane DSR, ACP Journal Club, DARE, and CCTR

Database timeline search from <1991 to March 2005>

1) colonoscop$.mp. (767)

2) (CT or virtual or computed tomographic).mp.

[mp=ti, ot, ab, tx, kw, ct, sh, hw] (13003)

3) (specifi$ and sensitiv$).mp. (9240)

4) 1 and 2 and 3 (20)

5) from 4 keep 3-6, 13, 15-16, 18 (8)

OVID SEARCH- MEDLINE and PUBMED

Ovid MEDLINE(R) <1996 to March Week 1 2005>

SEARCH #1

1 exp colonoscopy/ or exp sigmoidoscopy/ (10300)

2 exp tomography, x-ray computed/ or exp colonography, computed tomographic/

or exp tomography, spiral computed/ or colonography, computed tomographic/ (155918)

3 1 and 2 (500)

4 exp aged/ or exp middle aged/ (2388417)

5 3 and 4 (291)

6 exp "Sensitivity and Specificity"/ (173799 )

7 5 and 6 (75)

8 "Predictive Value of Tests"/ (58398)

9 7 and 8 (19)

10 from 9 keep 1-2, 4-6, 8, 10, 12, 14-19 (14)

SEARCH #2

1. colonoscop$.tw. (7784)

2. virtual.mp. or Computed Tomograph$.tw. or CT.mp. [mp=title, original title, abstract,

name of substance word, subject heading word] (147603)

3. aged/ or exp "aged, 80 and over"/ or exp middle aged/ (2388405)

4. (specif$ or sensitiv$).tw. (1487303)

5. 1 and 2 and 3 and 4 (117)

6. predict$ and value$).tw. (78847)

7. 5 and 6 (23)

8. from 7 keep 8, 10, 21 (3)

Secondary or Intergraded Sources and Clinical Practice Guidelines:

1) Preparation of patients for GI endoscopy. American Society for Gastrointestinal Endoscopy - Medical Specialty Society. 2003 Apr. 5 pages.

2) Modifications in endoscopic practice for the elderly. American Society for Gastrointestinal Endoscopy - Medical Specialty Society. 2000 Dec. 3 pages.

3) Colorectal cancer screening and surveillance: clinical guidelines and rationale-update based on new evidence. American College of Gastroenterology - Medical Specialty Society
American College of Physicians - Medical Specialty Society
American Gastroenterological Association - Medical Specialty Society
American Society for Gastrointestinal Endoscopy - Medical Specialty Society. 1997 Feb (revised 2003 Feb). 48 pages.