Colorectal Polyps: Detection with Low-Dose Multi–Detector Row Helical CT Colonography versus Two Sequential Colonoscopies 1

Riccardo Iannaccone, MDCarlo Catalano, MDFilippo Mangiapane, MDTakamichi Murakami, MD,

PhDAntonietta Lamazza, MDEnrico Fiori, MDAlberto Schillaci, MDDaniele Marin, MDItalo Nofroni, MScMasatoshi Hori, MD, PhDRoberto Passariello, MD

Published online10.1148/radiol.2373041747

Radiology 2005; 237:927–937

Abbreviation:CI � confidence interval

1 From the Departments of Radiologi-cal Sciences (R.I., C.C., F.M., D.M.,R.P.), Surgery “Pietro Valdoni,” Endo-scopic Unit (A.L., E.F., A.S.), and Ex-perimental Medicine and Pathology-Medical Biostatistics (I.N.), Universityof Rome-La Sapienza, Rome, Italy; andDepartment of Radiology, Osaka Uni-versity Graduate School of Medicine,Osaka, Japan (T.M., M.H.). ReceivedOctober 11, 2004; revision requestedDecember 21; revision received March3, 2005; accepted March 16. Addresscorrespondence to R.I., Via ArturoGraf 40, 00137 Rome, Italy (e-mail:[email protected]).

Authors stated no financial relation-ship to disclose.

Author contributions:Guarantors of integrity of entire study,R.I., R.P.; study concepts/study designor data acquisition or data analysis/interpretation, all authors; manuscriptdrafting or manuscript revision for im-portant intellectual content, all au-thors; approval of final version of sub-mitted manuscript, all authors; litera-ture research, R.I., D.M.; clinicalstudies, R.I., F.M., A.L., E.F.; statisticalanalysis, R.I., I.N.; and manuscript ed-iting, R.I., C.C., T.M., M.H., R.P.© RSNA, 2005

Colorectal Polyps: Detectionwith Low-Dose Multi–DetectorRow Helical CT Colonographyversus Two SequentialColonoscopies1

PURPOSE: To prospectively evaluate the diagnostic accuracy of low-radiation-dosecomputed tomographic (CT) colonography for detection of colorectal polyps byusing two sequential colonoscopies, with the second colonoscopy as the referencestandard.

MATERIALS AND METHODS: The study was local ethics committee approved,and all patients gave written informed consent. Colonographic images were ac-quired by using a low-dose multi–detector row CT protocol (effective milliampere-second setting, 10 mAs). Three observers interpreted the CT colonographic dataseparately and independently by using a two-dimensional technique. Initial con-ventional colonoscopy was performed by an endoscopist unaware of the CT colono-graphic findings. Second colonoscopy performed within 2 weeks by a colonoscopistaware of both the CT colonographic and the initial colonoscopic findings served asthe reference standard. The sensitivities of CT colonography and initial colonoscopywere calculated on a per-polyp and a per-patient basis. Specificities and positive andnegative predictive values also were calculated on a per-patient basis.

RESULTS: Eighty-eight patients underwent CT colonography and initial conven-tional colonoscopy on the same day. Per-polyp sensitivities were 62% and 83% forCT colonography and initial colonoscopy, respectively. Sensitivities for detection ofpolyps 6 mm in diameter or larger were 86% and 84% for CT colonography andinitial colonoscopy, respectively. Initial colonoscopy failed to depict 16 polyps, six ofwhich were correctly detected with CT colonography. For identification of patientswith polyps 6 mm in diameter or larger, CT colonography and initial colonoscopy,respectively, had sensitivities of 84% and 90%, specificities of 82% and 100%,positive predictive values of 70% and 100%, and negative predictive values of 91%and 95%.

CONCLUSION: Low-dose CT colonography compares favorably with colonoscopyfor detection of colorectal polyps 6 mm in diameter or larger, with markedlydecreased performance for detection of polyps 5 mm in diameter or smaller.© RSNA, 2005

Computed tomographic (CT) colonography is a rapidly evolving imaging procedure inwhich CT data sets are used to produce two- and three-dimensional images of the colon.In recent years, this examination has emerged as a valid diagnostic colorectal cancer test(1–8). A limitation of the current published data on CT colonography is that conventionalcolonoscopy was used as the reference standard. There is compelling evidence that con-ventional colonoscopy, even when meticulously performed by experienced endoscopists,has a substantial miss rate in the detection of polyps (9–13). The limitations of currentlyavailable colonoscopes may hamper the visualization of mucosa in the proximal aspects offolds, flexures, and valves (12).

Gastrointestinal Imaging

927

Ra

dio

logy

In addition, some small polyps may bedifficult to identify because of their size,flat morphologic features, and/or lack ofcolor contrast with the surrounding mu-cosa (14). Although several techniques toreduce the miss rate of colonoscopy (ie,high-magnification chromoscopy withdye spraying [15,16], cap-fitted colonos-copy [17], and wide-angle colonoscopy[12]) have been proposed, to our knowl-edge none has reached widespread usethus far.

Owing to the use of an imperfect refer-ence standard in studies reported in theCT colonographic literature, the diagnos-tic performance of CT colonography pos-sibly is underestimated. Specifically, ithas been recently demonstrated that, inseveral instances, CT colonography de-picts polyps that have been missed atconventional colonoscopy (13). Thus,some of the false-positive findings at CTcolonography could be false-negativefindings at conventional colonoscopy(8,18).

The ideal validation of a test for theidentification of colorectal polyps, ei-ther CT colonography or conventionalcolonoscopy, would involve direct com-parison with the findings at pathologicexamination of the entire large bowel—astudy that is impossible to perform (11).A method to create an enhanced refer-ence-standard examination—a so-calledsegmental unblinding—was proposed byPineau et al (18) and subsequentlyadopted in two recent studies (5,6). Thismethod involves the incremental reveal-ing of the CT colonographic findings tothe endoscopist during colonoscopy asthe colonoscope is withdrawn segmentby segment, after the endoscopist has re-corded his or her own independent ob-servations (19). Therefore, if a polyp isidentified in a given segment at CTcolonography and is not seen at the“first-look” colonoscopy, the endoscopisthas to reexamine the segment to resolvethe discrepancy. However, if the CTcolonographic result is negative, no fur-ther colonoscopic evaluation is per-formed (18).

To our knowledge, no study to investi-gate the performance of CT colonogra-phy with use of two sequential colonos-copies had been performed before thecurrent investigation. With this ap-proach, after complete initial colono-scopic and CT colonographic examina-tions, a complete second colonoscopy isperformed and serves as the referencestandard. Thus, the purpose of our studywas to prospectively evaluate the diag-nostic accuracy of low-radiation-dose

multi–detector row helical CT colonogra-phy for the detection of colorectal polypsby using two sequential colonoscopies,with the second colonoscopy as the ref-erence standard.

MATERIALS AND METHODS

The study was approved by the local eth-ics committee of the University ofRome-La Sapienza and conducted be-tween November 2002 and December2003. Written informed consent was ob-tained from all patients after the purposeand protocol of the study had been fullyexplained to them. This study was per-formed in accordance with the Declara-tion of Helsinki principles (20).

Patient inclusion criteria included thefollowing: average-risk colorectal cancerscreening performed, family history ofcolorectal carcinoma, personal or familyhistory of colorectal polyps, follow-up ofan abnormal screening test result (ie, pos-itive guaiac-based stool test, barium en-ema examination, or sigmoidoscopy re-sult) performed, evaluation of hema-tochezia performed, change in bowelmovement habits, weight loss, abdomi-nal pain, and/or iron deficiency–relatedanemia. Exclusion criteria included thefollowing: history of familial adenoma-tous polyposis or hereditary nonpolypo-sis cancer syndromes, prior colorectalsurgery, suspected inflammatory boweldisease, acute diverticulitis or bowel ob-struction, rejection for conventionalcolonoscopy or CT colonography for anyreason, medical condition that precludedthe use of bowel preparation, inability togive informed consent, and/or preg-nancy. Each patient included in thestudy underwent CT colonography fol-lowed by initial colonoscopy on the sameday and second colonoscopy 7–14 days(average, 9.2 days) later.

CT Colonographic Technique

The study participants underwent co-lonic cleansing by drinking 2 L of poly-ethylene glycol electrolyte solution (Iso-colan; Bracco, Milan, Italy) and 10 mg ofbisacodyl (Dulcolax; Boehringer Ingel-heim, Florence, Italy) the day before thescheduled examinations. Bisacodyl wasused to reduce the amount of fluid in-gested by the patient and thus improvepatient compliance (21).

The CT colonographic examinationswere performed with a multi–detectorrow helical CT scanner (Somatom Plus 4Volume Zoom; Siemens Medical Systems,Forchheim, Germany) 3–6 hours before

initial colonoscopy. CT colonographywas performed with the patient in theprone and supine positions after intrave-nous administration of 20 mg of hyo-scine-N-butylbromide (Buscopan; Boehr-inger Ingelheim). With the patient in theleft lateral decubitus position, the colonwas gently insufflated with room air,with a rubber catheter placed in the rec-tum according to patient tolerance. Toprevent air leakage from the anus, therectal tube was subsequently clampedand left in situ during imaging.

First, with the patient in the prone po-sition, a scout CT image was obtained toestimate the adequacy of colonic disten-tion. Further air insufflation was per-formed when collapsed bowel segmentswere observed. Before supine imaging,the colon was insufflated with additionalair, again according to maximal patienttolerance, and colonic distention wasverified on a second scout CT image. Airinsufflation was performed in all patientsby the same nurse.

The CT examinations were performedby using the following low-dose proto-col, which was previously optimized forthe CT scanner used in the present studyby Iannaccone et al (22): 4 � 2.5-mmsection collimation (effective sectionthickness, 3.0 mm), 1.0-mm reconstruc-tion interval, 17.5 mm/sec table speed,gantry rotation time of 0.5 seconds, 140kV, and effective milliampere-second set-ting of 10 mAs. The ensuing totalweighted CT dose index for the com-bined prone and supine image acquisi-tions was 2.74 mGy (22). The acquisitiontime ranged from 14 to 20 seconds. Aresident who was not involved in CTcolonographic data evaluation docu-mented any complication associatedwith CT colonography.

CT Colonographic Image Analysis

Three gastrointestinal radiologists (R.I.,C.C., F.M.), who were unaware of both thespecific indications for CT colonographyand the results of the initial colonoscopicexamination, separately and independentlyreviewed each case directly on a dedicatedworkstation by using a software packagewith volume-rendering capabilities (Vitrea;Vital Images, Plymouth, Minn). Theseobservers had previously interpreted ap-proximately 400 (R.I.), 200 (C.C.), and100 (F.M.) CT colonographic cases withendoscopic correlation. The findings ineach patient were prospectively recordedby the three observers before secondcolonoscopy was performed.

For image analysis, the observers were

928 � Radiology � December 2005 Iannaccone et al

Ra

dio

logy

asked to use a previously validated time-efficient technique (23,24). In brief, theinitial analysis entailed reviewing magni-fied two-dimensional transverse CT im-ages. When a suspected polyp was de-tected on the two-dimensional transverseCT images, coronal and sagittal CT im-ages and three-dimensional endoluminalviews were evaluated to confirm the find-ing. If no suspected polyp was identifiedat review of the transverse images, nofurther image analysis was performed.The presence, location, size, and mor-phologic features of all suspected polypswere documented. Polyps were measuredon the transverse CT images by using anelectronic ruler.

To specify the location of each polyp,the colon was divided into six segments:cecum, ascending colon and hepatic flex-ure, transverse colon and splenic flexure,descending colon, sigmoid colon, andrectum. With regard to morphologic fea-tures, all polyps were classified as pedun-culated, sessile, or flat. Flat polyps weredefined as those with a base at least twiceas long as the height. Sessile polyps werethose that did not meet the criteria to beclassified as flat and did not have a stalk.Pedunculated polyps had an identifiablestalk. All polyps seen at CT colonographywere photographed, and the colono-graphic images were stored in digital for-mat. The image interpretation time foreach CT colonographic examination wasrecorded with a stopwatch.

Initial Conventional Colonoscopy

All colonoscopies were performed byone of two staff endoscopists (A.L., E.F.)with extensive and comparable colono-scopic experience (each having inter-preted more than 5000 cases prior to thisstudy). Both endoscopists were unawareof the CT colonographic findings. Twoendoscopists participated in the study be-cause the sensitivity of colonoscopy hasbeen shown to vary between examiners(25) and because variation in examina-tion technique is potentially responsiblefor the nondetection of polyps (26). Toavoid bias, the two endoscopists alter-nated between performing the first andsecond examinations. All examinationswere performed while the patients weresedated. Sedation was induced with in-travenous administration of midazolamhydrochloride (Versed; Hoffmann-La Roche,Nutley, NJ).

The endoscopist performed conven-tional colonoscopy with a standard vid-eocolonoscope (model C240 or CV-1;Olympus Optical, Tokyo, Japan). The en-

doscopist was asked to pass the instru-ment tip to the cecum and withdraw theendoscope segment by segment for thedetection of polyps. Cecal intubation wasverified on the basis of the identificationof the appendiceal orifice, triradiate cecalfold, and ileocecal valve and the findingsat biopsy of the small bowel after intuba-tion of the terminal ileum. The presence,location, size, and morphologic featuresof all colorectal polyps were docu-mented. All polyps were photographed.The width of each lesion was estimatedto the nearest millimeter by means ofvisual comparison with the known diam-eter (4 mm) of an opened biopsy forcepsthat was pushed against the polyp; theheight was estimated by placing theclosed forceps tip (2.1 mm) adjacent tothe lesion. This allowed subsequent cal-culation of the height-to-width ratio foreach polyp, which is important in assess-ing the flat morphologic features of pol-yps.

The location of each polyp wasmapped within the same six colonic seg-ments used in CT colonographic analy-sis. The morphologic criteria for polyps atcolonoscopy were the same as those atCT colonography. During the initial con-ventional colonoscopic examination per-formed on the same day as but after CTcolonography, no polyp was sampled atbiopsy or resected. In addition, a residentwho was not involved in CT colono-graphic data evaluation was present todocument any complication associatedwith initial colonoscopy.

Second Conventional Colonoscopy

Each second colonoscopic procedurewas performed by the endoscopist whodid not perform the first examination.The bowel preparation before colonos-copy and the technique used were thesame as those described for the firstcolonoscopic examination. The location,size, and morphologic criteria used forpolyp assessment also were the same asthose used for the first colonoscopy. Theendoscopist who performed the secondcolonoscopic examination was aware ofthe findings of both initial colonoscopyand CT colonography. Specifically, to fa-cilitate polyp-to-polyp matching withthe lesions depicted at second colonos-copy, the initial colonoscopic images andthe CT colonographic images—togetherwith the reference images of the identi-fied polyps—were reviewed before thesecond colonoscopy. To solve any poten-tial discrepancy in the polyp-to-polypmatching, the endoscopist who per-

formed the first colonoscopic examina-tion and one radiology resident who wasaware of the results derived by the threeCT observers also were present. The com-plications associated with the secondcolonoscopic procedure also were re-corded.

Histologic Features of Polyps

All polyps were sampled at biopsy orresected during the second colonoscopy.The histologic appearances of all polypspecimens were evaluated by the samepathologist, who had 11 years of experi-ence. Histologically, polyps were classi-fied as nonneoplastic (including hyper-plastic, inflammatory, and juvenile pol-yps) or neoplastic (including tubular,villous, tubulovillous, serrated, and mi-crotubular adenomas; carcinomas in situ;and invasive carcinomas) (27).

Data Analyses

Results were calculated in two ways: byusing individual polyp detection—thatis, per-polyp analysis—and by using pa-tient detection—that is, per-patient anal-ysis. For per-polyp analysis, the results ofthe second colonoscopy—in terms of thelocation, size, and morphologic featuresof the depicted polyps—were consideredthe reference standard against which thefindings at initial colonoscopy and CTcolonography were compared. For agiven polyp to be considered a true-pos-itive match between second colonoscopyand either CT colonography or initialcolonoscopy, the location, size, and mor-phologic features of the polyp had to beconsidered.

With regard to location and morpho-logic features, a polyp detected at CTcolonography or first colonoscopy wasconsidered to be concordant with oneseen at second colonoscopy if it was lo-cated in the same or an adjacent colonicsegment (with the exception of polyps inthe cecum, for which no margin of errorwas allowed) and had the same morpho-logic features, compared with the polypseen at second colonoscopy. Precisematching of polyps according to size cri-teria was rarely possible because of theknown measurement error with conven-tional colonoscopy and the unknownmeasurement error with CT colonogra-phy (5,18). Thus, for a given polyp tomeet the size criteria, the size measuredat CT colonography or initial colonos-copy had to be within a 50% margin oferror of the size determined at secondcolonoscopy (5,6,18).

Volume 237 � Number 3 Colorectal Polyp Detection with CT Colonography � 929

Ra

dio

logy

Each polyp that was seen at secondcolonoscopy but not documented at CTcolonography or initial colonoscopy wasclassified as a false-negative finding of CTcolonography or initial colonoscopy, re-spectively. In cases in which additionalpolyps that were not seen at first colonos-copy were identified at second colonos-copy, the two endoscopists jointly dis-cussed the possible reason(s) for the ini-tial nonvisualization (ie, a polyp situatedbehind a colonic fold, an area of interestthat was difficult to assess for anatomicreasons, or perceptual error). If a polypwas depicted at CT colonography or ini-tial colonoscopy but not confirmed atsecond colonoscopy, it was deemed to bea false-positive finding of CT colonogra-phy or initial colonoscopy, respectively.

After the comparisons, a retrospectivereview of all false-negative and false-pos-itive CT colonographic findings wasjointly performed by the three observersand a study supervisor (R.P.) to deter-mine the reasons for the diagnostic er-rors. In addition, because of the potentialrange of size measurements for a givenpolyp among the three observers, a con-sensus reading was performed to deter-mine the CT colonographic size of eachpolyp. Therefore, the average resultsgiven throughout this report refer to thepolyp sizes estimated during this consen-sus reading.

For per-patient analysis, the overallresults of CT colonography or initialcolonoscopy were compared with theoverall results of second colonoscopyfor each patient. A patient was consid-ered to have true-positive findings atCT colonography or initial colonoscopyif the given examination depicted atleast one polyp that was seen at secondcolonoscopy on the basis of the loca-tion, size, and morphologic criteria de-scribed earlier.

Statistical Analyses

All statistical analyses were conductedby using commercially available software(SPSS for Windows, version 11.0.0; SPSS,Chicago, Ill). For per-polyp analysis, thesensitivity and corresponding 95% confi-dence interval (CI) of both CT colonog-raphy and initial colonoscopy were cal-culated on the basis of the total numberof polyps (ie, neoplastic and nonneoplas-tic polyps combined) and on the basis ofthe number of neoplastic polyps only.The specificity of the examinations forthe detection of individual polyps could

not be calculated because the total num-ber of true-negative polyps could not beassessed (5).

For per-patient analysis, the sensitiv-ity, specificity, positive predictive value,and negative predictive value of both CTcolonography and initial colonoscopy,

with corresponding 95% CIs, were calcu-lated.

In this study, per-polyp and per-pa-tient analysis data were calculated ac-cording to size thresholds (6,19), withseparate evaluations for polyps 5 mm indiameter or smaller, 6 mm in diameter or

TABLE 1Demographic Characteristics and Indications for Colonoscopy in 88 Patients

Characteristic or Indication Value

Men 55 (62)Women 33 (38)Mean age (y)* 62.4 (50–74)Indication for colonoscopy

Asymptomatic patients 38 (43)Screening 12 (14)Personal history of polyps 11 (12)Family history of colorectal cancer 9 (10)Abnormal screening test result† 6 (7)

Symptomatic patients 50 (57)Hematochezia 13 (15)Change in bowel movement habits 12 (14)Weight loss 10 (11)Abdominal pain 8 (9)Iron deficiency–related anemia 7 (8)

Note.—Unless otherwise noted, data are numbers of patients and numbers in parentheses arepercentages.

* Numbers in parentheses are the age range.† Includes positive result of guaiac-based stool test (n � 4), barium enema examination (n � 1),

or sigmoidoscopy (n � 1).

TABLE 2Distribution of Polyps according to Size, Location, and Histologic Type

Polyp Location and Histologic Type

Polyp Size*

�5 mm 6–9 mm �10 mm Total

RectumNeoplastic 3 1 1 5Nonneoplastic 5 3 1 9Total 8 4 2 14

Sigmoid colonNeoplastic 4 5 3 12Nonneoplastic 10 3 2 15Total 14 8 5 27

Descending colonNeoplastic 4 2 1 7Nonneoplastic 7 1 0 8Total 11 3 1 15

Transverse colon and splenic flexureNeoplastic 1 2 1 4Nonneoplastic 6 1 0 7Total 7 3 1 11

Ascending colonic and hepatic flexureNeoplastic 5 1 0 6Nonneoplastic 5 1 1 7Total 10 2 1 13

CecumNeoplastic 1 3 1 5Nonneoplastic 6 3 0 9Total 7 6 1 14

All colonic segmentsNeoplastic 18 14 7 39Nonneoplastic 39 12 4 55Total 57 26 11 94

* Data are numbers of polyps.

930 � Radiology � December 2005 Iannaccone et al

Ra

dio

logy

larger, 7 mm in diameter or larger, 8 mmin diameter or larger, 9 mm in diameteror larger, and 10 mm in diameter orlarger.

Interobserver reliability for both thedetection of individual polyps and thedetection of patients with polyps at CTcolonography was evaluated by calculat-ing � statistics for multiple observers withuse of nonweighted binary � statistic val-ues. A � value of 0.01–0.20 was consid-ered to indicate minor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderateagreement; 0.61–0.80, high agreement;and 0.81–1.00, excellent agreement.

RESULTS

From November 2002 to December 2003,a total of 503 patients were referred toour institution for conventional colonos-copy and met the criteria for enrollmentin the study. Of these 503 patients, 398declined to participate. Of the 105 re-maining patients, 11 were subsequentlyexcluded because of failure of the CTscanner on the day of the scheduled CTcolonographic examination and six wereexcluded because the cecum was notreached at initial colonoscopy; thus, thecompletion rate (ie, cecal intubation rate)was 94% (88/[88 � 6]). The remaining 88patients underwent a complete CTcolonographic examination and two se-quential colonoscopic examinations and

therefore constituted the final study pop-ulation. The demographic characteristicsof these patients and their indications forcolonoscopy are given in Table 1. Over-all, 38 (43%) patients were asymptomaticand the remaining 50 (57%) patientswere scheduled to undergo conventionalcolonoscopy for the evaluation of symp-toms.

Second Conventional Colonoscopy

Second colonoscopy (Table 2) was per-formed 7–14 days (average, 9.2 days) af-ter the first colonoscopic examination.Cecal intubation was achieved in all pa-tients. No complication occurred duringsecond colonoscopy. Of the 88 patientsenrolled in the study, 42 (48%) showedno evidence of having polyps. A total of94 polyps were identified in the remain-ing 46 (52%) patients: 23 patients had asingle polyp; 11, two polyps; and 12,three or more polyps. Of the 94 polyps,57 (61%) were 1–5 mm in diameter; 26(28%), 6–9 mm in diameter; and 11(12%), 10 mm or greater in diameter.With regard to morphologic features, 36(38%) polyps were categorized as pedun-culated; 53 (56%) polyps, as sessile; andfive polyps (5%), as flat.

The histologic types of all 94 polypswere successfully classified. Nonneoplas-tic histologic features were found in 55(58%) polyps, all of which proved to behyperplastic. The 39 (42%) remaining

polyps were neoplastic and included 26(28%) tubular adenomas, six (6%) tubu-lovillous adenomas, two (2%) serratedadenomas, three (3%) carcinomas in situ,and two (2%) invasive carcinomas.

Initial Conventional Colonoscopy

No complications were associated withthe initial colonoscopic examination.Overall, initial colonoscopy (Table 3)yielded a per-polyp sensitivity of 83%,with a sensitivity of 87% for the detec-tion of neoplastic polyps. Initial colonos-copy “missed” 16 polyps (diameterrange, 4–14 mm) that were depicted atsecond colonoscopy, for a total miss rateof 17%. Five (four tubular adenomas andone tubulovillous adenoma; diameterrange, 5–14 mm) of these 16 missed pol-yps were neoplastic. Thus, the neoplasticpolyp miss rate of initial colonoscopywas 13% (five of 39 polyps). The reasonsfor missed polyps at initial colonoscopywere a blind spot (ie, the polyp was lo-cated behind a colonic fold) in 12 casesand perceptual error in the remainingfour cases. All polyps found at initialcolonoscopy were confirmed and re-sected at second colonoscopy—that is,there were no false-positive findings atinitial colonoscopy.

Table 4 shows the results of initialcolonoscopy for the identification of pa-tients with colorectal polyps according topolyp size. At per-patient analysis, initial

TABLE 3Sensitivity of CT Colonography (for Three Observers) and Initial Colonoscopy for Individual Polyp Detection

Parameter

Polyp Size

�5 mm �6 mm �7 mm �8 mm �9 mm �10 mm All

Observer 1Sensitivity, all polyps 51 (29/57) 84 (31/37) 88 (23/26) 94 (17/18) 100 (14/14) 100 (11/11) 64 (60/94)95% CI 38%, 63% 69%, 92% 71%, 96% 74%, 99% 79%, 100% 74%, 100% 54%, 73%Sensitivity, neoplastic polyps 50 (9/18) 81 (17/21) 93 (14/15) 100 (11/11) 100 (9/9) 100 (7/7) 67 (26/39)95% CI 29%, 71% 60%, 92% 70%, 99% 74%, 100% 70%, 100% 65%, 100% 51%, 79%

Observer 2Sensitivity, all polyps 47 (27/57) 86 (32/37) 88 (23/26) 94 (17/18) 100 (14/14) 100 (11/11) 63 (59/94)95% CI 23%, 43% 72%, 94% 71%, 96% 74%, 99% 79%, 100% 74%, 100% 53%, 72%Sensitivity, neoplastic polyps 44 (8/18) 81 (17/21) 87 (13/15) 100 (11/11) 100 (9/9) 100 (7/7) 64 (25/39)95% CI 25%, 66% 60%, 92% 62%, 96% 74%, 100% 70%, 100% 65%, 100% 48%, 77%

Observer 3Sensitivity, all polyps 42 (24/57) 86 (32/37) 88 (23/26) 89 (16/18) 93 (13/14) 100 (11/11) 60 (56/94)95% CI 30%, 55% 72%, 94% 71%, 96% 67%, 97% 69%, 99% 74%, 100% 50%, 69%Sensitivity, neoplastic polyps 39 (7/18) 71 (15/21) 80 (12/15) 100 (11/11) 100 (9/9) 100 (7/7) 56 (22/39)95% CI 20%, 61% 50%, 86% 55%, 93% 74%, 100% 70%, 100% 65%, 100% 41%, 71%

Initial colonoscopySensitivity, all polyps 82 (47/57) 84 (31/37) 85 (22/26) 83 (15/18) 86 (12/14) 91 (10/11) 83 (78/94)95% CI 71%, 90% 69%, 92% 67%, 94% 61%, 94% 60%, 96% 62%, 98% 74%, 89%Sensitivity, neoplastic polyps 83 (15/18) 91 (19/21) 87 (13/15) 82 (9/11) 89 (8/9) 86 (6/7) 87 (34/39)95% CI 61%, 94% 71%, 97% 62%, 96% 52%, 95% 57%, 98% 49%, 97% 73%, 94%

Note.—Data are percentages. For sensitivity data, the numbers used to calculate the percentages are in parentheses. Each 95% CI corresponds to thedirectly preceding sensitivity value. Data are presented separately for all polyps (ie, neoplastic and nonneoplastic polyps combined) and for neoplasticpolyps alone.

Volume 237 � Number 3 Colorectal Polyp Detection with CT Colonography � 931

Ra

dio

logy

colonoscopy had a sensitivity of 94%, aspecificity of 100%, a positive predictivevalue of 100%, and a negative predictivevalue of 93%.

CT Colonography

No complication occurred during theCT colonographic examinations. Themean time required for data interpreta-tion was 10 minutes (range, 8–15 min-utes).

The � values for agreement among thethree observers indicated moderate tohigh agreement on a per-polyp basis andhigh to excellent agreement on a per-patient basis (Table 5).

CT colonography (Table 3) yielded amean per-polyp sensitivity for the threeobservers of 62% ([60 � 59 � 56]/[94 �94 � 94]), with a mean sensitivity forneoplastic polyp (Fig 1) detection of 62%([26 � 25 � 22]/[39 � 39 � 39]). With theanalysis focused on polyps 6 mm in di-

ameter or larger, CT colonographyyielded a substantially increased detec-tion rate: The mean sensitivity for thethree observers was 86% ([31 � 32 �32]/[37 � 37 � 37]), with a mean sensi-tivity for neoplastic polyp detection of78% ([17 � 17 � 15]/[21 � 21 � 21]).Notably, all 11 neoplastic polyps 8 mmin diameter or larger were correctly iden-tified by the three observers, yielding asensitivity of 100%.

In addition, CT colonography cor-rectly depicted six of the 16 polyps (three�5 mm polyps, two 8-mm polyps, andone 14-mm polyp) that were not seen atinitial colonoscopy but were confirmedand resected at second colonoscopy. Fourof these six polyps were neoplastic (threeadenomas, one tubulovillous adenoma)(Fig 2).

CT colonography yielded 34, 35, and38 false-negative findings for observers 1,

TABLE 4Performance Values for CT Colonography (for Three Observers) and Initial Colonoscopy in Identifying Patients withColorectal Polyps

Parameter

Polyp Size

�5 mm �6 mm �7 mm �8 mm �9 mm �10 mm All

SensitivityObserver 1 76 (13/13�4) 86 (25/25�4) 91 (20/20�2) 94 (16/16�1) 100 (14/14) 100 (10/10) 83 (38/38�8)

53%, 90% 69%, 94% 72%, 98% 73%, 99% 78%, 100% 72%, 100% 69%, 91%Observer 2 76 (13/13�4) 83 (24/24�5) 91 (20/20�2) 94 (16/16�1) 100 (14/14) 100 (10/10) 80 (37/37�9)

53%, 90% 66%, 92% 72%, 98% 73%, 99% 79%, 100% 72%, 100% 67%, 89%Observer 3 71 (12/12�5) 83 (24/24�5) 91 (20/20�2) 94 (16/16�1) 100 (14/14) 100 (10/10) 78 (36/36�10)

47%, 87% 66%, 92% 72%, 98% 73%, 99% 79%, 100% 72%, 100% 64%, 88%Initial colonoscopy 100 (17/17) 90 (26/26�3) 86 (19/19�3) 82 (14/14�3) 86 (12/12�2) 90 (9/9�1) 94 (43/43�3)

82%, 100% 74%, 96% 67%, 95% 59%, 94% 60%, 96% 60%, 98% 83%, 98%Specificity

Observer 1 100 (71/71) 83 (49/49�10) 91 (60/60�6) 96 (68/68�3) 99 (73/73�1) 100 (78/78) 76 (32/32�10)95%, 100% 72%, 91% 82%, 96% 88%, 99% 93%, 100% 95%, 100% 62%, 87%

Observer 2 100 (71/71) 83 (49/49�10) 92 (61/61�5) 97 (69/69�2) 99 (73/73�1) 100 (78/78) 76 (32/32�10)95%, 100% 72%, 91% 84%, 97% 90%, 99% 93%, 100% 95%, 100% 62%, 87%

Observer 3 100 (71/71) 80 (47/47�12) 89 (59/59�7) 94 (67/67�4) 97 (72/72�2) 99 (77/77�1) 71 (30/30�12)95%, 100% 67%, 88% 80%, 95% 86%, 98% 91%, 99% 93%, 100% 56%, 83%

Initial colonoscopy 100 (71/71) 100 (59/59) 100 (66/66) 100 (71/71) 100 (74/74) 100 (78/78) 100 (42/42)95%, 100% 94%, 100% 94%, 100% 95%, 100% 95%, 100% 95%, 100% 92%, 100%

Positive predictive valueObserver 1 100 (13/13) 71 (25/25�10) 77 (20/20�6) 84 (16/16�3) 93 (14/14�1) 100 (10/10) 79 (38/38�10)

77%, 100% 55%, 84% 58%, 89% 62%, 94% 70%, 99% 72%, 100% 66%, 88%Observer 2 100 (13/13) 71 (24/24�10) 80 (20/20�5) 89 (16/16�2) 93 (14/14�1) 100 (10/10) 79 (37/37�10)

77%, 100% 54%, 83% 61%, 91% 67%, 97% 70%, 99% 72%, 100% 65%, 88%Observer 3 100 (12/12) 67 (24/24�12) 74 (20/20�7) 80 (16/16�4) 88 (14/14�2) 91 (10/10�1) 75 (36/36�12)

75%, 100% 50%, 80% 55%, 87% 58%, 92% 64%, 97% 62%, 98% 61%, 85%Initial colonoscopy 100 (17/17) 100 (26/26) 100 (19/19) 100 (14/14) 100 (12/12) 100 (9/9) 100 (43/43)

82%, 100% 87%, 100% 83%, 100% 79%, 100% 75%, 100% 70%, 100% 92%, 100%Negative predictive value

Observer 1 95 (71/71�4) 92 (49/49�4) 97 (60/60�2) 99 (68/68�1) 100 (73/73) 100 (78/78) 80 (32/32�8)87%, 98% 82%, 97% 89%, 99% 92%, 100% 95%, 100% 95%, 100% 65%, 90%

Observer 2 95 (71/71�4) 91 (49/49�5) 97 (61/61�2) 99 (69/69�1) 100 (73/73) 100 (78/78) 78 (32/32�9)87%, 98% 80%, 96% 89%, 99% 92%, 100% 95%, 100% 95%, 100% 63%, 88%

Observer 3 93 (71/71�5) 90 (47/47�5) 97 (59/59�2) 98 (67/67�1) 100 (72/72) 100 (77/77) 75 (30/30�10)86%, 97% 79%, 96% 89%, 99% 92%, 100% 95%, 100% 95%, 100% 60%, 86%

Initial colonoscopy 100 (71/71) 95 (59/59�3) 96 (66/66�3) 96 (71/71�3) 97 (74/74�2) 99 (78/78�1) 93 (42/42�3)95%, 100% 87%, 98% 88%, 99% 89%, 99% 91%, 99% 93%, 100% 82%, 98%

Note.—Data are percentages. The numbers used to calculate the percentages are in parentheses. The corresponding 95% CI is on the second line.

TABLE 5Agreement between Observers Regarding Presence or Absence of IndividualColorectal Polyps and Identification of Patients with Colorectal Polyps

Agreement AnalysisObserver 1 vs

Observer 2Observer 2 vs

Observer 3Observer 1 vs

Observer 3

Per polyp 0.63 0.59 0.51Per patient 0.79 0.79 0.81

Note.—Data are � values.

932 � Radiology � December 2005 Iannaccone et al

Ra

dio

logy

2, and 3, respectively. At retrospectiveanalysis, no clear cause of error (Table 6)could be determined for 26, 26, and 28 ofthese false-negative findings, respec-tively, because the polyps could not beseen in a well-distended and adequatelycleansed colonic segment (Fig 3). CTcolonography also yielded 19, 20, and 24false-positive findings for observers 1, 2,and 3, respectively. At retrospective anal-ysis, the most frequent cause of error wasresidual fecal material (Fig 4).

The results of the three observers andCT colonography in the identification ofpatients with colorectal polyps accordingto polyp size are summarized in Table 4.At per-patient analysis, CT colonographyyielded a mean sensitivity for the threeobservers of 80% ([38 � 37 � 36]/[46 �46 � 46]), a mean specificity of 75%([32 � 32 � 30]/[42 � 42 � 42]), a meanpositive predictive value of 78% ([38 �37 � 36]/[48 � 47 � 48]), and a mean

negative predictive value of 78% ([32 �32 � 30]/[40 � 41 � 40]).

DISCUSSION

Our study results show that low-dose CTcolonography has an excellent diagnos-tic performance in the detection of colo-rectal polyps 10 mm in diameter orlarger, with a sensitivity of 100%: 11 of11 polyps were correctly detected on CTcolonographic images by the three CTobservers. Even when the cutoff polypsize was reduced to 6 mm, CT colonogra-phy had a mean sensitivity of 86% ([31 �32 � 32]/[37 � 37 � 37]) for the threeobservers, which was slightly superior tothe sensitivity of initial colonoscopy(84%) in our study and in the upperrange of values for CT colonography re-ported in the literature (1–8). This find-ing is of pivotal clinical importance be-cause polyps 6 mm in diameter or larger

have the highest probability of being ma-lignant (28).

In contrast, although the sensitivity ofinitial colonoscopy remained stable (sen-sitivity for detection of polyps �5 mm,82%), the diagnostic performance of CTcolonography decreased substantiallywhen polyps 5 mm in diameter orsmaller were considered (mean sensitiv-ity, 47% [29 � 27 � 24]/[57 � 57 � 57]for the three observers). This finding is inagreement with several previous reports(1,7,29). In our study, 26, 26, and 28 pol-yps 5 mm in diameter or smaller couldnot be detected with CT colonography,even at retrospective analysis. Because asubstantial proportion of these small pol-yps missed at CT colonography were hy-perplastic, one possible explanation forthe nonvisualization of such lesionscould be their soft consistency, whichfavors the effacement against the colonicmucosa once the colon is distended withair (1,30).

In our study, however, the sensitivitiesof CT colonography for the detection ofall polyps (ie, neoplastic and nonneo-plastic polyps combined) and for the de-tection of neoplastic polyps only wererelatively stable. This indicates that di-minutive lesions, regardless of their his-tologic type, are difficult to detect at CTcolonography. However, the minority ofsuch small polyps are malignant (31–33),and the adenoma-to-carcinoma transfor-mation is believed to take at least 10years on average (34,35). Thus, the clini-cal importance of identifying such smallpolyps is controversial.

It is perhaps more important that all ofthe flat polyps in our series were missedat CT colonography. Although only fivesuch lesions were identified, there is in-creasing evidence that flat adenomas areextremely difficult to detect at CTcolonography (7,8,36–38) because theycause only minimal alterations of the co-lonic mucosa. In this regard, it should beemphasized that even conventionalcolonoscopy has limited capability to de-pict flat lesions without the supplemen-tary use of special techniques, such ashigh-magnification chromoscopy anddye spraying (16). In addition, the clini-cal importance of flat adenomas is stillcontroversial because their prevalenceand malignant potential are unclear (39).

Another important issue that emergedfrom our research is that false-positive CTcolonographic findings are relatively fre-quent: The three observers had 19, 20,and 24 false-positive findings in ourstudy. In agreement with the findings oftwo recent studies (40,41), the most fre-

Figure 1. Seven-millimeter-diameter polyp (arrow) in 68-year-old woman. (a) Transverse CTcolonographic image shows the polyp with round borders in the sigmoid colon. (b) Coronal CTcolonographic image clearly shows the polyp. (c) Three-dimensional volume-rendered endolu-minal CT image clearly demonstrates the polyp’s sessile structure. (d) Initial conventionalcolonoscopic image shows the polyp, which was removed at second colonoscopy and found tobe a tubular adenoma at histologic analysis.

Volume 237 � Number 3 Colorectal Polyp Detection with CT Colonography � 933

Ra

dio

logy

quent cause of the false-positive cases inour study was the presence of residualfecal material. Small fecal residue oftenadheres to the colonic mucosa and is toosmall to contain gas bubbles (a findingconsistent with fecal material). Thus,such residue is often indistinguishablefrom polyps and thus has the potential toprompt many unnecessary colonoscopicexaminations (42).

Although per-polyp analysis is impor-tant for determining which polyps willmost likely be identified or missed with CTcolonography, from a clinical point ofview per-patient analysis is far more impor-tant because it enables the use of CTcolonography to preselect those patientswith polyps who might benefit fromcolonoscopy (5,8). In this regard, the diag-nostic performance of CT colonographywas comparable to that of initial colonos-copy in terms of sensitivity, specificity, andpositive and negative predictive values forthe identification of patients with polyps10 mm in diameter or larger. Moreover, thenegative predictive value of low-dose CTcolonography—that is, its effectiveness inthe identification of patients who did notneed to undergo colonoscopy—was ap-proximately equal to the negative predic-tive value of initial colonoscopy in theidentification of patients with polyps 6mm in diameter or larger (91% for CTcolonography vs 95% for initial colonos-copy). This finding has important clinicalimplications for CT colonography as a po-tential screening tool because 46%–85% ofscreening colonoscopies do not reveal clin-ically important abnormalities (43,44).

Moreover, our study results demon-strate that low-dose CT colonographicimages are interpreted with moderate tohigh interobserver agreement on a per-polyp basis and with high to excellentagreement on a per-patient basis amongobservers with different levels of experi-ence. The three observers had previouslyinterpreted approximately 400, 200, and100 CT colonographic cases with endo-scopic correlation in our series. The re-producibility of our results was thus dem-onstrated. These results correlate withthe findings of other authors (6,45) butdiffer from those reported by Johnson etal (38). It is possible that the interob-server variability in our study was re-duced, compared with that in the John-son et al (38) study, owing to the use ofthin-section multi–detector row helicalCT image acquisitions.

One of the major strengths of our studywas the use of two sequential colonosco-pies to evaluate the performance of CTcolonography. The performance of a sec-

ond complete colonoscopic examinationby an experienced endoscopist who wasaware of the CT colonographic and initialcolonoscopic findings yielded a greatly im-proved reference standard compared withthe reference standards used in the major-ity of studies reported in the CT colono-graphic literature. It is conceivable that, al-though second colonoscopy may depictsome polyps missed at the first examina-tion, it may fail to depict other missed pol-yps (10,13). However, it can be assumedthat after two colonoscopic proceduresperformed by two experienced endosco-pists and CT colonography, the probabilityof having missed polyps in our cohort wasminimal.

In addition, the use of two sequentialcolonoscopies yielded important infor-mation about the performance of con-ventional colonoscopy. In agreementwith previous study (9,10,12,13) find-ings, our study results show that conven-

tional colonoscopy has a substantial missrate in the identification of colorectalpolyps. The overall miss rate was 17% inour study: 16 of 94 polyps were not seenprospectively. In our series, 16 polyps(five neoplastic, 11 nonneoplastic) thatwere missed at initial colonoscopy wereidentified at second colonoscopy. Six ofthese polyps, four of which were neoplas-tic, were seen at CT colonography. Thus,if initial colonoscopy had been the refer-ence standard, these six polyps wouldhave been wrongly classified as false-pos-itive findings at CT colonography. Inagreement with the recent findings ofPickhardt et al (13), the most frequentreason for missed polyps at initialcolonoscopy in our study was the loca-tion of polyps behind colonic folds (in 12[75%] of 16 cases).

Because both endoscopists in our studyhad extensive colonoscopic experience(as indicated by cecal intubation rates of

Figure 2. Fourteen-millimeter-diameter polyp (arrow) in 62-year-old man. (a) Transverse CTcolonographic image shows the polyp with round borders in the transverse colon. (b) Coronal CTcolonographic image findings confirm the presence of the polyp. (c) Three-dimensional volume-rendered endoluminal CT image clearly demonstrates that the polyp is located behind a colonicfold (arrowheads). (d) Second conventional colonoscopic image shows the polyp. Despite itsrelatively large size, this polyp was missed at initial colonoscopy and difficult to detect even atsecond colonoscopy owing to its location behind the colonic fold. Histologic evaluation revealedthis polyp to be a tubulovillous adenoma.

934 � Radiology � December 2005 Iannaccone et al

Ra

dio

logy

94% and 100% at initial and secondcolonoscopy, respectively), the miss rateof conventional colonoscopy confirmsthat colonoscopy is an imperfect refer-ence standard. Therefore, an improvedreference standard is needed for accurate

assessment of the performance of CTcolonography (18,19).

As a possible way to create an im-proved reference standard, so-called seg-mental unblinding has been described(18) and adopted (5,6). This method has

the important advantage of enablingboth initial evaluation and reevaluationof a colonic segment during the sameendoscopic session, so only one colono-scopic procedure is needed. However,this approach also has certain limita-tions: First, if a polyp is not seen at CTcolonography and at first-look colonos-copy, a second colonoscopic look is notperformed (18). Second, owing to a learn-ing effect, this technique may hamperthe comparison of diagnostic perfor-mance between colonoscopy and CTcolonography, because if the endoscopistmisses a polyp in a specific segment, heor she may unintentionally adjust theexamination approach during assess-ment of the remaining segments (8).Third, segmental unblinding with same-day colonoscopy may be difficult to per-form at many centers owing to logisticreasons, because the CT colonographicreport must be available within a fewhours. This can be difficult to achieve inroutine clinical practice, especially ifthree CT observers are needed to prospec-tively assess interobserver variability.

The limitations of our study requirecomment. First, similar to what has beenreported regarding colonoscopy and theexpertise of endoscopists (44), there isincreasing evidence that greater observerexperience is associated with increasedCT colonographic performance (4). In ourstudy, the three observers had greater-than-average experience in CT colono-graphic data interpretation, having inter-preted approximately 400, 200, and 100cases with endoscopic correlation. There-fore, further studies are needed to con-firm the diagnostic performance of low-dose CT colonography when its findingsare interpreted by less experienced read-ers.

In addition, instead of monobasic so-dium phosphate monohydrate com-

Figure 3. Two polyps, 8 and 4 mm in diameter, in 63-year-old man. (a) Initial conventionalcolonoscopic image shows the 8-mm polyp (white arrow) within the cecum. The 4-mm polyp (blackarrow) can also be seen. These polyps were not detected at CT colonography by the three observers,and both were confirmed and removed at second colonoscopy. At histologic analysis, these lesionswere shown to be hyperplastic polyps. (b) Coronal CT colonographic image obtained at retrospectiveanalysis shows no abnormality in the well-distended cecum (C). (c) Three-dimensional volume-rendered endoluminal CT image obtained at retrospective analysis clearly shows the anatomy of thececum, but no polyp is depicted. (d) Another three-dimensional volume-rendered endoluminal CTimage obtained at retrospective analysis reveals no abnormality of the cecal mucosa.

TABLE 6Causes of False-Positive and False-Negative Findings in Individual Polyp Detection at CT Colonography

Finding TypeTotal No. of

FindingsPerceptual

ErrorFecal

MaterialResidual

FluidCollapsed

BowelThickFold

MotionArtifact

No ClearCause*

False-negativeObserver 1 34 3 1 3 1 0 0 26Observer 2 35 3 2 2 2 0 0 26Observer 3 38 4 2 2 2 0 0 28

False-positiveObserver 1 19 0 12 0 0 7 0 0Observer 2 20 0 12 0 0 8 0 0Observer 3 24 0 14 0 0 9 1 0

Note.—Data are numbers of false-negative or false-positive findings due to the given causes.* Polyp could not be seen in a well-distended and adequately cleansed colonic segment retrospectively.

Volume 237 � Number 3 Colorectal Polyp Detection with CT Colonography � 935

Ra

dio

logy

bined with dibasic sodium phosphateheptahydrate, polyethylene glycol elec-trolyte solution and bisacodyl were usedfor cathartic preparation. The monobasicsodium phosphate monohydrate–dibasicsodium phosphate heptahydrate prepa-ration has been demonstrated to leave adrier colonic surface compared with thepolyethylene glycol preparation (46).However, in our study, residual fluidswere rarely responsible for diagnostic er-rors at CT colonography.

In conclusion, our study results showthat the performance of low-dose multi–detector row helical CT colonographycompares favorably with that of conven-tional colonoscopy in the detection ofcolorectal polyps 6 mm in diameter orlarger and has markedly decreased perfor-mance in the detection of polyps 5 mmin diameter or smaller. The high negativepredictive value of CT colonography canbe used to exclude the presence of clini-cally important colorectal polyps in themajority of patients and thus potentiallyto reduce the number of negative-resultcolonoscopies performed. Further studiesare needed to determine which lesioncutoff size is clinically acceptable and theappropriate time interval for repeat CTcolonography when lesions smaller thanthis cutoff size are detected.

References1. Fenlon HM, Nunes DP, Schroy PC 3rd, Bar-

ish MA, Clarke PD, Ferrucci JT. A compar-ison of virtual and conventional colonos-copy for the detection of colorectal polyps.N Engl J Med 1999;341:1496–1503.

2. Fletcher JG, Johnson CD, Welch TJ, et al.Optimization of CT colonography tech-nique: prospective trial in 180 patients. Ra-diology 2000;216:704–711.

3. Yee J, Akerkar GA, Hung RK, Steinauer-Gebauer AM, Wall SD, McQuaid KR. Colo-rectal neoplasia: performance characteris-

tics of CT colonography for detection in300 patients. Radiology 2001;219:685–692.

4. Johnson CD, Toledano AY, Herman BA, etal. Computerized tomographic colonogra-phy: performance evaluation in a retro-spective multicenter setting. Gastroenter-ology 2003;125:688–695.

5. Pineau BC, Paskett ED, Chen GJ, et al. Vir-tual colonoscopy using oral contrast com-pared with colonoscopy for the detectionof patients with colorectal polyps. Gastro-enterology 2003;125:304–310.

6. Pickhardt PJ, Choi JR, Hwang I, et al. Com-puted tomographic virtual colonoscopy toscreen for colorectal neoplasia in asymp-tomatic adults. N Engl J Med 2003;349:2191–2200.

7. Macari M, Bini EJ, Jacobs SL, et al. Colorec-tal polyps and cancers in asymptomaticaverage-risk patients: evaluation with CTcolonography. Radiology 2004;230:629–636.

8. Van Gelder RE, Nio CY, Florie J, et al. Com-puted tomographic colonography com-pared with colonoscopy in patients at in-creased risk for colorectal cancer. Gastro-enterology 2004;127:41–48.

9. Hixson LJ, Fennerty MB, Sampliner RE,McGee D, Garewal H. Prospective study ofthe frequency and size distribution of pol-yps missed by colonoscopy. J Natl CancerInst 1990;82:1769–1772.

10. Rex DK, Cutler CS, Lemmel GT, et al.Colonoscopic miss rates of adenomas de-termined by back-to-back colonoscopies.Gastroenterology 1997;112:24–28.

11. Bensen S, Mott LA, Dain B, Rothstein R,Baron J; for the Polyp Prevention StudyGroup. The colonoscopic miss rate andtrue 1-year recurrence of colorectal neo-plastic polyps. Am J Gastroenterol 1999;94:194–199.

12. Rex DK, Chadalawada V, Helper DJ. Wideangle colonoscopy with a prototype in-strument: impact on miss rates and effi-ciency as determined by back-to-backcolonoscopies. Am J Gastroenterol 2003;98:2000–2005.

13. Pickhardt PJ, Nugent PA, Mysliwiec PA,Choi JR, Schindler WR. Location of adeno-mas missed by optical colonoscopy. AnnIntern Med 2004;141:352–359.

14. Brooker JC, Saunders BP, Shah SG, et al.Total colonic dye-spray increases the de-tection of diminutive adenomas duringroutine colonoscopy: a randomized con-trolled trial. Gastrointest Endosc 2002;56:333–338.

15. Rembacken BJ, Fujii T, Cairns A, et al. Flatand depressed colonic neoplasms: a pro-spective study of 1000 colonoscopies inthe UK. Lancet 2000;355:1211–1214.

16. Hurlstone DP, Cross SS, Adam I, et al. Aprospective clinicopathological and endo-scopic evaluation of flat and depressedcolorectal lesions in the United Kingdom.Am J Gastroenterol 2003;98:2543–2549.

17. Matsushita M, Hajiro K, Okazaki K,Takakuwa H, Tominaga M. Efficacy of totalcolonoscopy with a transparent cap incomparison with colonoscopy without thecap. Endoscopy 1998;30:444–447.

18. Pineau BC, Paskett ED, Chen GJ, DurkalskiVL, Espeland MA, Vining DJ. Validation ofvirtual colonoscopy in the detection ofcolorectal polyps and masses: rationale forproper study design. Int J GastrointestCancer 2001;30:133–140.

19. Dachman AH, Zalis ME. Quality and con-sistency in CT colonography and researchreporting. Radiology 2004;230:319–323.

20. 41st World Medical Assembly Declarationof Helsinki. Recommendations guidingphysicians in biomedical research involv-ing human subjects. Bull Pan Am HealthOrgan 1990;24:606–609.

21. Adams WJ, Meagher AP, Lubowski DZ,King DW. Bisacodyl reduces the volume ofpolyethylene glycol solution required forbowel preparation. Dis Colon Rectum1994;37:229–233.

22. Iannaccone R, Laghi A, Catalano C, et al.Detection of colorectal lesions: lower-dosemulti-detector row helical CT colonogra-phy compared with conventional colonos-copy. Radiology 2003;229:775–781.

23. Dachman AH, Kuniyoshi JK, Boyle CM, etal. CT colonography with three-dimen-sional problem solving for detection of co-lonic polyps. AJR Am J Roentgenol 1998;171:989–995.

24. Macari M, Milano A, Lavelle M, Berman P,Megibow AJ. Comparison of time-efficientCT colonography with two- and three-di-mensional colonic evaluation for detect-

Figure 4. Four-millimeter-diameter false-positive lesion (arrow) in 64-year-old woman. (a) Transverse CT colonographic image shows the lesionin the sigmoid colon. (b) Sagittal CT colonographic image findings confirm the presence of the lesion. (c) Three-dimensional volume-renderedendoluminal CT image clearly shows the lesion. Although all three observers reported this lesion to be a polyp, it was not seen at first or secondcolonoscopy. This lesion is believed to represent fecal residue.

936 � Radiology � December 2005 Iannaccone et al

Ra

dio

logy

ing colorectal polyps. AJR Am J Roentge-nol 2000;174:1543–1549.

25. Rex DK, Rahmani EY, Haseman JH, LemmelGT, Kaster S, Buckley JS. Relative sensitivityof colonoscopy and barium enema for de-tection of colorectal cancer in clinical prac-tice. Gastroenterology 1997;112:17–23.

26. Rex DK. Colonoscopic withdrawal tech-nique is associated with adenoma missrates. Gastrointest Endosc 2000;51:33–36.

27. Rubio CA, Jaramillo E, Lindblom A, Fogt F.Classification of colorectal polyps: guide-lines for the endoscopist. Endoscopy 2002;34:226–236.

28. Rex DK. Virtual colonoscopy: time forsome tough questions for radiologists andgastroenterologists. Endoscopy 2000;32:260–263.

29. Macari M, Bini EJ, Xue X, et al. Colorectalneoplasms: prospective comparison of thin-section low-dose multi-detector row CTcolonography and conventional colonos-copy for detection. Radiology 2002;224:383–392.

30. Bertoni G, Sassatelli R, Conigliaro R, et al.Visual “disappearing phenomenon” canreliably predict the nonadenomatous na-ture of rectal and rectosigmoid diminutivepolyps at endoscopy. Gastrointest Endosc1994;40:588–591.

31. Nusko G, Mansmann U, Altendorf-Hof-mann A, Groitl H, Wittekind C, Hahn EG.Risk of invasive carcinoma in colorectaladenomas assessed by size and site. Int JColorectal Dis 1997;12:267–271.

32. Bond JH. Clinical relevance of the smallcolorectal polyp. Endoscopy 2001;33:454–457.

33. Aldridge AJ, Simson JN. Histological as-sessment of colorectal adenomas by size:are polyps less than 10 mm in size clini-cally important? Eur J Surg 2001;167:777–781.

34. Winawer SJ, Zauber AG, Ho MN, et al; forthe National Polyp Study Workgroup. Pre-vention of colorectal cancer by colono-scopic polypectomy. N Engl J Med 1993;329:1977–1981.

35. Hofstad B, Vatn N. Growth rate of colonpolyps and cancer. Gastrointest EndoscClin N Am 1997;7:345–363.

36. Rex DK, Vining D, Kopecky KK. An initialexperience with screening for colon pol-yps using spiral CT with and without CTcolography (virtual colonoscopy). Gastro-intest Endosc 1999;50:309–313.

37. Fidler JL, Johnson CD, MacCarty RL,Welch TJ, Hara AK, Harmsen WS. Detec-tion of flat lesions in the colon with CTcolonography. Abdom Imaging 2002;27:292–300.

38. Johnson CD, Harmsen WS, Wilson LA, etal. Prospective blinded evaluation of com-puted tomographic colonography forscreen detection of colorectal polyps. Gas-troenterology 2003;125:311–319.

39. Hurlstone DP, Brown S, Cross SS. The roleof flat and depressed colorectal lesions incolorectal carcinogenesis: new insightsfrom clinicopathological findings in high-

magnification chromoscopic colonoscopy.Histopathology 2003;43:413–426.

40. Yee J, Kumar NN, Hung RK, Akerkar GA,Kumar PR, Wall SD. Comparison of supineand prone scanning separately and incombination at CT colonography. Radiol-ogy 2003;226:653–661.

41. Edwards JT, Mendelson RM, Fritschi L, etal. Colorectal neoplasia screening with CTcolonography in average-risk asymptom-atic subjects: community-based study. Ra-diology 2004;230:459–464.

42. Ferrucci JT. Colon cancer screening withvirtual colonoscopy: promise, polyps, pol-itics. AJR Am J Roentgenol 2001;177:975–988.

43. Lieberman DA, Weiss DG, Bond JH, AhnenDJ, Garewal H, Chejfec G; for VeteransAffairs Cooperative Study Group 380. Useof colonoscopy to screen asymptomaticadults for colorectal cancer. N Engl J Med2000;343:162–168.

44. Imperiale TF, Wagner DR, Lin CY, LarkinGN, Rogge JD, Ransohoff DF. Risk of ad-vanced proximal neoplasms in asymptom-atic adults according to the distal colorectalfindings. N Engl J Med 2000;343:169–174.

45. McFarland EG, Pilgram TK, Brink JA, et al.CT colonography: multiobserver diagnos-tic performance. Radiology 2002;225:380–390.

46. Macari M, Lavelle M, Pedrosa I, et al. Effectof different bowel preparations on residualfluid at CT colonography. Radiology 2001;218:274–277.

Volume 237 � Number 3 Colorectal Polyp Detection with CT Colonography � 937

Ra

dio

logy