DRAFT: Not for Citation
Test characteristics for CT colonography were assessed from studies in which subjects receive both CT colonography and colonoscopy. As CT colonography is a rapidly evolving technology, many of the older studies are generally outdated in assessing test characteristic for CT colonography in use today. Early studies were conducted in polyp-rich cohorts using 2D technology with generally encouraging results (Fenlon 1999,, Yee 2001). However, studies using these technologies in lower prevalence polyp cohorts, such as seen in screening, had less promising results (Johnson 2003, Cotton 2004, Rockey 2005). Mulhall (2005) conducted a systematic review and meta-analysis of 33 CT colonography studies in 6393 patients published from January 1975 to February 2005 and analyzed the findings by mode of imaging, collimation, reconstruction, type of scanner, use of contrast material, the gold standard for comparison, and software used. However, most of those studies were of higher-risk patients and therefore not applicable for an average-risk screening population. Whitlock and colleagues (2008) conducted a structured systematic literature review of CT colonography to inform the USPSTF in their assessment of whether to recommend CT colonography screening for the average-risk population. They found that only 4 of the studies in the Mulhall analysis were among average-risk patients. Of these, 3 studies were quite small and used older, less accurate scanning technologies. The fourth study, the Department of Defense (DoD) study (Pickhardt 2003), was included in the Whitlock assessment along with studies by Johnson (2007), Kim (2007) and the newly published study reporting the results of the National CT Colonography Trial (NCTC) (Johnson 2008). We used the Whitlock evidence review (2008) to identify studies for our consideration.
We used the two large scale multi-site CT colonography studies conducted in the US using current technology and procedures as our main comparators: the DoD study by Pickhardt (2003) and the NCTC (Johnson 2008). These studies represent the current most promising assessments of CT colonography compared to optical colonoscopy in clinical practice. We did not combine the results of these two studies but rather used each study as a separate base-case scenario. We also used a retrospective analysis by Pickhardt (2007a) on his original DoD study and a single institution study by Johnson (2007) to assess primary 2D versus 3D readings. We did not include the study by Kim (2007) in our comparisons due to its small size (n = 96) and the fact that it reported sensitivity and specificity for all polyps rather than for adenomas.
This study was intended to be proof-of-principal that CT colonography could have high test performance in CRC screening. The study accrued 1233 asymptomatic subjects from military facilities from May 2002 and June 2003 for a same-day CT colonography and optical colonoscopy. Subjects completed a rigorous bowel preparation including a standard 24-hour oral administration of sodium phosphate and bisacodyl. Subjects also had a clear-liquid diet plus barium for solid-stool tagging and diatrizoate meglumine and diatrizoate sodium for the opacification of luminal fluid. Three-dimensional endoluminal display was used for the initial detection of polyps on CT colonography, with 2 dimensional views used in assessing suspected abnormalities. Room air was used to insufflate the colon. A 4-channnel or 8-channel CT scanner was used. Polyps were measured with electronic calipers on the 3D view. Extracolonic findings were also reported. The CT scans were read by one of six board-certified radiologists prior to the optical colonoscopy, all of whom had read a minimum of 25 CT scans prior to the study. Optical colonoscopy was performed by 17 experienced endoscopists (14 gastroenterologists and 3 colorectal surgeons). Polyps were photographed and measured using a calibrated linear probe. The study protocol used segmental unblinding for the optical colonoscopy. The endoscopist reported the clinical findings by segment and then was told the CT colonography results for that segment. At this point the endoscopist could go back to review the segment to see if any polyps were missed. The polyps detected were recorded for optical colonoscopy before and after the CT colonography results were revealed. All polyps were sent for histological review. A polyp matching algorithm was used to compare CT colonography and optical colonoscopy with matching criteria of polyps being in the same segment or adjacent segments with polyp dimensions within a 50% margin of error.
The test characteristics were given both per patient and per adenoma, with 92% sensitivity of CT colonography for adenomas 10 mm or larger and 86% sensitivity for adenomas 6 mm or larger. Specificity was 96% for patients with adenomas 10 mm or larger and 80% for patients with adenomas 6 mm or larger. Results were not reported for lesions measuring less than 6 mm. Extracolonic findings deemed to be of high clinical importance were found in 4.5% of subjects. More patients reported greater discomfort with CT colonography (54%) than with optical colonoscopy (38%), while 8% reported equivalent discomfort. General level of satisfaction with CT colonography was rated excellent by 41% of respondents; only 6% and 2% rated their level of satisfaction as fair or poor. Subjects were slightly more likely to state that of the two tests they preferred CT colonography (49% vs. 41%); 9% reported having no preference.
This study, sponsored by the American College of Radiology Imaging Network (ACRIN) and the NCI, was intended to assess the performance of high-quality CT colonography in general community practice. The study accrued 2600 asymptomatic subjects from 15 study centers from February 2005 to December 2006. Ninety-seven percent (2531) of those accrued completed same-day CT colonography and optical colonoscopy. Bowel preparation included stool tagging, laxative purgation, and fluid tagging. Glucagon was administered prior to CT acquisition and carbon dioxide was used for colon insufflation. Each participating radiologist had interpreted at least 500 CT scans or had participated in a 1.5 day course. All radiologists chosen to participate had to complete a qualifying examination in which they achieved a detection rate of 90% or more for polyps measuring 10 mm or larger. All CT scans were performed with multi-detector scanners with a minimum of 16 rows. The study data were randomly assigned to be read independently with the use of a primary two-dimensional search method (2D image display with 3D endoluminal problem solving) of a primary 3D search method with the addition of 2D display of multiplanar images. Only lesions of size 5 mm or larger were recorded. Same day colonoscopy was performed or supervised by experienced endoscopists without knowledge of the CT colonography findings. Segmental unblinding was not employed. For cases in which CT colonography had detected a polyp 10 mm or larger that was not detected on optical colonoscopy, the patient was advised to have an additional colonoscopy. All lesions 5 mm or larger were centrally reviewed by one experienced gastrointestinal pathologist. Lesion size was determined from the pathology report, unless piecemeal removal was performed, in which case colonoscopy-derived size estimates were used. An algorithm similar to that used in the DoD study was used to match polyps.
Sensitivity was reported both by patient and by adenoma. The per-adenoma sensitivity of CT colonography for adenomas or CRC 10 mm or larger was 84%, which was slightly less than the estimate from the DoD study (92%). Sensitivity for adenomas 6 mm or larger was 70%. Specificity was 86% for patients with adenomas 10 mm or larger and 88% for patients with adenomas 6 mm or larger. Extracolonic findings were observed in 66% of subjects, but only 16% were considered of clinical importance requiring either additional evaluation or urgent care.
The DoD study was performed using primary 3D reading. Earlier studies using 2D reading had not obtained as good test performance as that of the DoD study with 3D readings. Ten radiologists, blinded to polyp findings, conducted a retrospective interpretation of 730 CT scans from the original DoD study using a primary 2D approach (Pickhardt 2007a). The primary 2D results were compared with the primary 3D results from the original trial of 1233. Sensitivity for adenomas 6 mm or larger was 44% with the primary 2D approach, compared with 86% for the primary 3D approach. Sensitivity for adenomas 10mm or larger was 75% versus 92% for primary 2D and primary 3D reads, respectively. With a primary 2D approach, per-patient specificity for 2D at the 10 mm threshold for referral was 98% compared to 97% for the 3D evaluation (NB: these specificity estimates are for all polyps, not for adenomas only).
Johnson (2007) conducted a study of 452 asymptomatic subjects with CT scans interpreted using both a primary 2D and a primary 3D approach. The sensitivity of CT colonography for neoplasms 10 mm or larger using a 1.25mm slice thickness were comparable for primary 2D and primary 3D reads (72% versus 73% respectively). However, the range across three readers was wider for the primary 3D reads (67%-78% for primary 2D reads versus 50-83% for primary 3D reads). Specificity for patients with adenomas 10mm or larger was 97-99% for both reading approaches.
All studies of CT colonography characteristics were for a one-time test. No studies to date evaluate repeat screening with a CT colonography. Therefore, we do not have information on the degree to which false-negative test results are random or systematic.