CDo

CW*D

Bssc(tutswDwretrDoptDm[0nrc

CEtrste

mia

CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2004;2:314–321

omparison of the Relative Sensitivity of CT Colonography andouble-Contrast Barium Enema for Screen Detectionf Colorectal Polyps

. DANIEL JOHNSON,* ROBERT L. MACCARTY,* TIMOTHY J. WELCH,* LYNN A. WILSON,*ILLIAM S. HARMSEN,‡ DUANE M. ILSTRUP,‡ and DAVID A. AHLQUIST§

Department of Radiology, ‡Department of Health Sciences Research, Section of Biostatistics, and §Department of Internal Medicine,ivision of Gastroenterology and Hepatology, Mayo Clinic Rochester, Rochester, Minnesota

sotsapmcpliton

tDnlobdDtoogai

pcp

d

ackground & Aims: In a population reflective of acreening setting, our aim was to compare the relativeensitivity and specificity of computed tomography (CT)olonography with double-contrast barium enemaDCBE) for detection of colorectal polyps and to assesshe added value of double reading at CT colonography,sing endoscopy as the arbiter. Methods: This prospec-ive, blinded study comprised 837 asymptomatic per-ons at higher than average risk for colorectal cancerho underwent CT colonography followed by same-dayCBE. Examinations with polyps >5 mm in diameterere referred to colonoscopy. Results: CT colonography

eaders detected 56%–79% of polyps >10 mm in diam-ter. In comparison, the sensitivity at DCBE varied be-ween 39% and 56% for the 31 polyps >1 cm. All of theeaders detected more polyps at CT colonography thanCBE, but the difference was statistically significant fornly a single reader (P � 0.02). Relative specificity forolyps >10 mm on a per-patient basis ranged from 96%o 99% at CT colonography, and 99%–100% at DCBE.oubly read CT colonography detected significantlyore polyps than DCBE (81% vs. 45% for polyps >1 cm

P � <0.01], and 72% vs. 44% for polyps 5–9 mm [P <

.01]). Conclusions: Double-read CT colonography is sig-ificantly more sensitive in detecting polyps than single-ead double contrast barium enema. DCBE was signifi-antly more specific than CT colonography.

olorectal cancer is the most common cause of malig-nant death in the United States among nonsmokers.1

vidence exists today from prospective trials,2–4 case-con-rol studies,5–10 and predictive models11–13 to support colo-ectal screening intervention. Despite recommendations forcreening of the entire population,14,15 current screeningests are limited in performance, comfort, safety, and/orxpense. Better screening tools are needed.

Computed tomographic (CT) colonography, a mini-ally invasive new imaging tool, uses advanced visual-

zation technology to produce 2- and 3-dimensional im-ges that permit evaluation of the entire colorectal

tructure. Several promising reports of the performancef CT colonography have been published in the litera-ure.16–26 However, most of these have been based onelected patient groups with high lesion prevalence, suchs symptomatic patients or those with known or sus-ected colorectal neoplasms. Because of selection bias, itay not be appropriate to extrapolate such early out-

omes to the screening setting. It is desirable that theerformance of CT colonography is evaluated critically inarge asymptomatic populations (adults at average andncreased risk15) before a widespread screening applica-ion can be justified. Its comparative performance withther established screening tools in this setting also iseeded to guide practice.Theoretically, CT colonography should be a better

ool than the double-contrast barium enema (DCBE).CBE is limited as a 2-dimensional radiographic tech-ique with a finite number of images to inspect. Over-apping lines and shadows can contribute to distraction,bscuration, and interpretive error. Recent concern haseen raised about the effectiveness of the DCBE for polypetection by the National Polyp Study. In this study, theCBE detected only 48% of polyps �1 cm. Alterna-

ively, CT colonography can generate an infinite numberf views of the colorectum to display the mucosal surfaceptimally. Both 2- and 3-dimensional images can beenerated without confounding overlapping structures,nd with visualization of the entire bowel wall andnternal features of lesions.

The purpose of this study was as follows: (1) to com-are the relative sensitivity and specificity of CTolonography with DCBE for detection of colorectalolyps in an asymptomatic low-prevalence population,

Abbreviations used in this paper: CT, computed tomography; DCBE,ouble-contrast barium enema.

© 2004 by the American Gastroenterological Association1542-3565/04/$30.00

PII: 10.1053/S1542-3565(04)00061-8

ac

DafioriadiOpsHsec�ht(tT2

d(tiopbItiS1i

scrptciB

apfna

StShttttikthpbcvron

(da(le3otnlEssp

f�iFraavtf

S

April 2004 CT COLONOGRAPHY VS. DCBE 315

nd (2) to assess the added value of double reading of CTolonography.

MethodsA total of 837 outpatients who were prescheduled for

CBE, 50 years or older, and at increased risk for developingcolorectal neoplasm (a prior history of colorectal neoplasia, arst-degree family member with a history of colorectal cancer,r new onset of asymptomatic iron-deficiency anemia) wereecruited for CT colonography. Primary care physicians at ournstitution rely heavily on the combination of barium enemand sigmoidoscopy for routine screening of asymptomatic in-ividuals. Sigmoidoscopy usually was performed before bar-um enema (at the discretion of the primary care physician).nly patients at increased risk were recruited to improve therevalence of large adenomas (�1 cm) and to ensure thatcreening colonoscopy was covered by third-party payers.owever, because lesion prevalence was found to be low, the

tudy group reflects an average-risk population akin to thatncountered in a screening setting. All patients had CTolonography before same-day DCBE. Patients with polyps0.5 cm by either radiographic study were recommended to

ave colonoscopy. Exclusion criteria comprised melena, hema-ochezia, inflammatory bowel disease, and familial polyposissymptomatic patients). Signed informed consent was ob-ained. The Institutional Review Board approved the study.he study was conducted between January 1998 and February001.

Preparation

Preparation for the examinations included: clear liquidiet for 24 hours before the examination, magnesium citrate300 mL) and bisacodyl tablets (20 mg) were administered inhe afternoon before the procedure, a single bisacodyl suppos-tory at bedtime, plus two 1000-mL tap water enemas the dayf the examination in all but 2 patients. The remainingatients received polyethylene glycol oral lavage (1 gallon) andisacodyl tablets (10 mgm), Dulcolax suppository (Boehringerngelheim Pharmaceuticals, Inc., Germany) at bedtime andwo 1000-ml tap water enemas on the morning of the exam-nation in one patient, and X-prep (Purdue Pharma L.P.,tamford, CT; extract of Senna concentrate) 5 ounces and two000-mL tap water enemas on the morning of the examinationn one patient.

CT Colonography

A total of 742 patients (89%) received glucagon 1 mgubcutaneously 10 minutes before the CT examination unlessontraindicated or refused by the patient; 95 (11%) did noteceive glucagon. Patients were placed in the lateral decubitusosition for enema tip insertion (with inflated balloon reten-ion cuff) and slow manual insufflation of approximately 2 L ofarbon dioxide was administered (until the patient verballyndicated air administration had reached maximal tolerance).oth supine and prone data acquisitions were obtained and

dditional carbon dioxide was added (as tolerated by theatient) before scanning in the prone position to compensateor any carbon dioxide lost during position changes. A trainedurse monitored the examination. A radiologist was immedi-tely available for questions or assistance.

All examinations were performed using either a G.E. Hi-peed Advantage (105/845 [12%]) or Lightspeed (4-row mul-idetector) helical CT scanner (740/845 [88%]) (GE Medicalystems, Milwaukee, WI). After colon insufflation, a breathold anteroposterior scout was obtained before each acquisitiono assess luminal distention and to prescribe axial sliceshrough the entire large bowel. Images were acquired usinghe single-slice helical CT scanner with 5-mm collimation,able speed of 6.5 mm/s (pitch of 1.3), 3-mm reconstructionntervals, matrix 512 � 512, field-of-view to fit, 70 mA/s, 120ilovolt (peak), 1-second rotation time, standard reconstruc-ion algorithm, and 20-second breath holds. Three to 4 breatholds were performed within 2 minutes for 150 axial imageser acquisition. Three-centimeter acquisition overlap betweenreath holds was obtained. The multislice scanner used 5-mmollimation, pitch of 15 mm/sec, 3-mm reconstruction inter-als, 80 mA, 120 kV(p), 0.7-second rotation time, standardeconstruction algorithm, and a single approximately 20-sec-nd breath hold. The 80 mAs setting was chosen to match theoise of the single-slice scanner using 70 mAs.The image data were networked to an off-line workstation

SUN Microsystems, San Jose, CA) using customized softwareeveloped and extensively tested26 within our laboratory. Thexial images were magnified and reviewed using both lungwidth � 1500, level � �600) and soft-tissue (width � 520,evel � �4) window settings. Suspected abnormalities werevaluated further by using multiplanar reformatted images and-dimensional endoluminal images. Lesion location, size, andbserver confidence (doubtful, somewhat doubtful, conserva-ively confident, moderately confident, highly confident) wereoted for each abnormality. Lesion size was determined as theongest lesion diameter from magnified 2-dimensional images.ach data set (supine, prone) was evaluated sequentially, anduspicious regions were compared between the 2 datasets usingimultaneous synchronized views (axial, 2-dimensional multi-lanar, and 3-dimensional endoluminal) of both.Diagnostic review of each CT colonography study was per-

ormed by 2 of 3 experienced radiologists (Board certified,10 years practice experience, �150 proven CT colonography

nterpretations) blinded to the results at DCBE and endoscopy.or each patient, examinations for interpretation were assignedandomly to 2 of the 3 radiologists using statistical casessignment. Results are reported for each individual reviewernd for double reading (the combined reports of the 2 indi-idual readers). Reviewers were instructed to ignore findingshat were �5 mm in diameter. Colonoscopy was recommendedor any lesion detected �5 mm in diameter.

DCBE

DCBE was performed in routine fashion following thetandard of the American College of Radiology.27 High-den-

swhwttid(brrvvrr

uwrgrCm2o

m[afepe1duwcvgteipesl

prw

waDrsflwsrrscspriaeowtvai�ii

s((Hwa(co

tepmcppeoa

316 JOHNSON ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2, No. 4

ity barium (80% wt/vol) was instilled through an enema tipith the patient in the prone position until the barium columnad reached the distal or midtransverse colon. Excess bariumas drained and carbon dioxide was administered gently over

he course of approximately 5 minutes until full colon disten-ion was achieved (full feeling without pain). During thensufflation process and using fluoroscopic guidance spot ra-iographs of the rectum (anteroposterior and lateral), sigmoidboth left posterior oblique and right posterior oblique views),oth colonic flexures (semi-upright), and the cecum wereoutinely obtained of the air-distended colon. Large overheadadiographs also were obtained including prone and supineiews of the entire colon, both right and left lateral decubitusiews using a horizontal beam, and a prone angled view of theectosigmoid colon. The films were reviewed while the patientemained in the department.

Questionable areas of concern were rechecked at fluoroscopy,sually with additional spot radiographs. DCBE examinationsere never performed in the same patient by any of the 3

adiologists interpreting their CT colonography. The radiolo-ist interpreting the DCBE examination was blinded to theesults at CT colonography and any endoscopic procedure.olonoscopy was recommended for any lesion detected �5m in diameter. Lesions were measured on overhead film, and

0% was subtracted to allow for magnification. The long axisf the lesion was measured.

Colonoscopy

A total of 691 (82.6%) of the 837 patients had one orore confirmatory examinations performed (i.e., colonoscopy

n � 116], surgery [n � 3], flexible sigmoidoscopy [n � 581],nd/or rigid proctoscopy [n � 89]). Colonoscopy was per-ormed in 116 (13.9%) patients in the standard fashion toxamine the entire colorectum. All of the examinations wereerformed by staff gastroenterologists or colorectal surgeonsxperienced with this technique. Incomplete studies in 10 of16 examinations (8.6%) were complete to the sigmoid (1),escending (1), splenic flexure (1), transverse (4), hepatic flex-re (1), and ascending colon (2). The findings in these patientsere compared between CT colonography, DCBE, and

olonoscopy only in the segments of the colon that wereisualized endoscopically. Three patients underwent colon sur-ery. Flexible sigmoidoscopy (581/837, 69%) and rigid proc-oscopy (89/837, 11%) was performed in the usual manner byither a staff gastroenterologist, colorectal surgeon, or special-zed colorectal surgical nurse. A staff physician checked allositive examinations. The extent of insertion was noted onach report. Three patients (0.1%) had lesion confirmation aturgery. Lesion measurement was performed by comparing theesion with a closely held forceps of known size.

Analysis

Endoscopy (colonoscopy, flexible sigmoidoscopy, orroctoscopy) or surgery was considered the gold standard withespect to presence or absence of polyps. The size of the lesionas determined from the pathology report unless the polyp

as removed in pieces. In this latter instance, the size reportedt colonoscopy was used. The reports at CT colonography andCBE were compared with the endoscopy and pathology

eports. If a similar size lesion was discovered within one colonegment (cecum, ascending, hepatic flexure, transverse, splenicexure, descending, sigmoid, rectum) of the reported lesion, itas considered a match. The flexible sigmoidoscopy was con-

idered to be a confirmatory test for polyps located in theectum, sigmoid, and descending colon, and a proctoscopy forectal polyps only. Results are reported in terms of relativeensitivity and relative specificity because confirmatoryolonoscopy was not performed for all patients. Relative sen-itivity per polyp and per patient, and relative specificity peratient, were estimated for each reviewer, and for the doubleeading at CT colonography. Ninety-five percent confidencentervals for these estimates also were calculated. Sensitivitiesnd specificities were compared with McNemar’s test using thexact binomial (i.e., the sign test). Note that in the comparisonf specificities for CT colonography and DCBE the sample sizeas large, resulting in power to detect statistical differences

hat may or may not be clinically important. Only segmentsisualized at endoscopy were compared with CT colonographynd DCBE. Those polyps �5 mm in diameter were consideredn the analysis, stratifying by size category: 5–9 mm and those

10 mm. Double reading included all detections from bothndependent interpretations. No discussion of individual find-ngs or consensus reporting occurred.

ResultsDemographics of Participants

A total of 837 patients were recruited for thetudy. Gender mix included 425 (51%) men and 41249%) women. Ethnic backgrounds included Caucasians825), Native Americans (2), Asians (1), Africans (5), andispanics (4). The mean age was 63.4 years (SD � 7.2)ith a range of 50–86 years. A first-degree relative withhistory of colorectal cancer accounted for 537 of 837

64%) patients, prior personal history of a polyp orancer for 272 of 837 (33%) patients, and the new onsetf asymptomatic anemia in 28 of 837 (3%) patients.

Procedures Performed

A total of 581 flexible sigmoidoscopic examina-ions, 89 proctoscopic examinations, 116 colonoscopicxaminations, and 3 surgeries were performed in 691atients. Of the 52 patients with at least one polyp �5m in diameter, 45 patients had colonoscopy (42 were

omplete), 38 had flexible sigmoidoscopy, and 7 hadroctoscopy. In many cases, flexible sigmoidoscopy orroctoscopy was ordered in conjunction with the bariumnema (standard care by many primary care physicians atur institution). Some patients refused sigmoidoscopynd only a barium enema was ordered. Colonoscopic

f(lttfe

wp�bcToa7l

hh(l23

wdat(i

e

ewmR(ncs�c6(hcoBtD

Dttpt

istao

oir

T

RRRD

a

r

April 2004 CT COLONOGRAPHY VS. DCBE 317

ollow-up examination was performed for all but one2%) lesion detected at DCBE, and all but 20 (11%)esions detected at CT colonography (lack of follow-up inhese lesions probably was owing to delayed interpreta-ion of these research studies). Colonoscopy was per-ormed in 86 patients in which one (68) or both (18)xaminations were interpreted as negative.

Characteristics of Polyps Identified

A total of 70 polyps 5 mm in diameter or largerere present in 52 patients; 39 polyps 5–9 mm wereresent in 30 patients; and 31 polyps in 29 patients were1 cm in diameter (7 patients had one or more polyps of

oth sizes). Median size of the 70 polyps �5 mm was 0.8m (range, 0.5–10.0 cm), mean of 1.4 cm (SD � 1.7).here were 17 adenomas �1 cm. Median and mean sizef these larger polyps was 1.5 cm (range, 1.0–10.0 cm)nd 2.5 cm (SD � 2.1 cm), respectively. Forty-five of the0 polyps (64%) were adenomas. The prevalence of ateast one adenomatous polyp �1 cm in diameter was 2.7%.

The 39 colorectal lesions 5–9 mm in diameter hadistologic findings as follows: 28 adenomas (72%), 4yperplastic (10%), 5 mixed (13%), and 2 reported none5%). The 31 lesions �1.0 cm had the following histo-ogic findings: 17 adenomas (55%), 6 carcinomas (19%),

hyperplastic (6%), 1 lipoma (3%), 2 mixed (6%), andreported none (10%).Distribution of the 39 lesions 5–9 mm in diameter

as as follows: 5 rectum (13%), 9 sigmoid (23%), 3escending (8%), 9 transverse (24%), 6 ascending (15%),nd 7 cecum (18%). Lesions �1.0 cm in diameter hadhe following distribution: 4 rectum (13%), 9 sigmoid29%), 1 descending (3%), 4 transverse (13%), 6 ascend-ng (19%), and 7 cecum (23%).

Technical Description at CT Colonography

Overall examination quality for both same-dayxaminations was judged globally for the presence of

able 1. DCBE Versus CT Colonography: Per Polyp Sensitivityof Examination

5 to 9-mm polyps

No. polypsreviewed

CTcolonography D

eviewer 1 n (%) 30 19 (63%) 13eviewer 2 n (%) 24 7 (29%) 7eviewer 3 n (%) 24 19 (79%) 14ouble-read CT colonography 39 28 (72%) 17

Barium enema examinations were only read by a single radiologist. Coeviewer, and in the full cohort (double-read CT colonography).

xcessive fluid or stool at CT. Residual fluid in the colonas judged to be none in 80 (10%), some in 667 (80%),oderate in 87 (10%), and nondiagnostic in 3 (0.4%).esidual stool in the colon was judged to be none in 493

59%), mild in 278 (33%), moderate in 58 (7%), andondiagnostic in 8 (1%). Colon distention at CTolonography was judged to be optimal in 679 (81%),uboptimal in 1 segment 141 (17%), and suboptimal in1 segment 17 (2%). Breath hold artifacts at CT

olonography were judged as none in 764 (91%), mild in5 (8%), moderate in 7 (1%), and nondiagnostic in 10.1%). Thirty-one patients (4%) at CT colonographyad a nondiagnostic examination by either fluid, stool, orolonic distention result. Of these, 2 patients had a totalf 4 polyps, 3 sized 0.5–0.9 and 1 sized �1.0 cm.ecause of their small numbers, they were included in

he study analysis. None of the patients with polyps atCBE had nondiagnostic examinations.

Performance

The per polyp sensitivity at CT colonography andCBE is summarized in Table 1. The per polyp sensi-

ivity at CT colonography was the same or higher thanhe sensitivity at DCBE for all comparisons, regardless ofolyp size. For a single reader, this difference was statis-ically significant.

Table 2 summarizes per patient sensitivity and specific-ty. In nearly all comparisons, CT colonography had betterensitivity than DCBE, regardless of polyp size. None ofhese differences reached statistical significance. Specificitylways favored DCBE, although specificity was always 90%r higher regardless of the imaging modality.

Double Reading CT Colonography

The sensitivity at double-read CT colonographyn a per polyp basis for the detection of polyps 5–9 mmn diameter increased 9% and 43% for reviewers 1 and 2,espectively, and did not improve sensitivity for reviewer

Proven Polyps in 52 Patients. Endoscopic Proof to Extent

�10-mm polyps

a PNo. polypsreviewed

CTcolonography DCBEa P

) 0.07 28 22 (79%) 12 (43%) 0.02) 1.0 16 11 (69%) 9 (56%) 0.73) 0.13 18 10 (56%) 7 (39%) 0.51) �0.01 31 25 (81%) 14 (45%) 0.01

rison is made in the same group of patients for each CT colonography

: 70

CBE

(43%(29%(58%(44%

mpa

3cficpm

fid9rrsCCgb

tu9rft

s(

Dd1ds(t5on

Ddids4Oi7

T

RRRD

RRRD

Bra

318 JOHNSON ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2, No. 4

. Overall, double-read CT colonography was signifi-antly more sensitive than single-read DCBE (P � 0.01)or the detection of 5–9 mm polyps. Double readingmproved the sensitivity for the detection of polyps �1m for all 3 readers by 2% to 25%. For these largeolyps, double-read CT colonography was significantlyore sensitive than single-read DCBE (P � 0.01) (Table 1).Per patient analysis showed the sensitivity (95% con-

dence interval) at double-read CT colonography for theetection of at least one 5–9 mm polyp to be 83% (65%–4%), P � 0.04 compared with single-read DCBE. Thisepresents an improved sensitivity of 8% to 45% foreviewers 2 and 1, respectively. There was no improvedensitivity for reviewer 3. Specificity with double-readT colonography decreased 4% to 7% over single-readT colonography when detecting polyps �5 mm. Sin-le-read DCBE was significantly more specific than dou-le-read CT colonography (P � 0.001) (Table 2).

Per patient sensitivity (95% confidence interval) forhe detection of at least one polyp �1 cm in diametersing double-read CT colonography was 79% (60%–2%). This was significantly more sensitive than single-ead DCBE (P � 0.04). Specificity at CT colonographyor these larger polyps decreased with double reading 1%o 4% to 95%. Single-read DCBE was significantly more

able 2. DCBE Versus CT Colonography: Per Patient SensitivProof to Extent of Examination

Patients ha

Sensitivity

No. of patientswith �1 polyp

5–9 mmCT

colonography DCBE

eviewer 1 n (%) 24 18 (75%) 14 (58%)eviewer 2 n (%) 16 6 (38%) 7 (44%)eviewer 3 n (%) 20 18 (90%) 15 (75%)ouble-read CTcolonography

30 25 (83%) 18 (60%)

Patients hav

Sensitivity

No. of patientswith �1 polyp

�10 mmCT

colonography DCBE

eviewer 1 n (%) 26 20 (77%) 12 (46%)eviewer 2 n (%) 16 11 (69%) 9 (56%)eviewer 3 n (%) 16 9 (56%) 7 (44%)ouble-read CTcolonographya

29 23 (79%) 14 (48%)

arium enema examinations were only read by a single radiologist. Coeviewer, and in the full cohort (double-read CT colonography). SevenA total of 691 patients had either colonoscopy, flexible sigmoidosco

pecific than double-read CT colonography P �0.001Table 2).

Detection Rates and Polyp Location

Detection rates of 5–9 mm polyps by location forCBE were: rectum 3/5 (60%), sigmoid 6/9 (67%),escending 2/3 (67%), transverse 3/8 (38%), ascending/6 (17%), and cecum 2/7 (29%). Detection rates atouble-read CT colonography were: rectum 4/5 (80%),igmoid 8/9 (89%), descending 2/3 (67%), transverse 6/967%), ascending 3/6 (50%), and cecum 5/7 (71%). Ofhe 39 polyps evaluated at CT colonography between–9 mm in diameter, 17 (44%) were detected by bothbservers, 11 (28%) by one observer, and 11 (28%) byeither.Detection rates of �10-mm polyps by location for

CBE were: rectum 1/4 (25%), sigmoid 2/9 (22%),escending 1/1 (100%), transverse 2/2 (100%), ascend-ng 3/6 (50%), and cecum 5/7 (71%). Detection rates atouble-read CT colonography were: rectum 3/4 (75%),igmoid 9/9 (100%), descending 1/1 (100%), transverse/4 (100%), ascending 4/6 (67%), and cecum 4/7 (57%).f the 31 polyps examined at CT colonography �1 cm

n diameter, 18 (58%) were detected by both observers,(23%) by one observer, and 6 (19%) by neither.

d Specificity: 52 Patients With Proven Lesions. Endoscopic

or more polyps sized 5–9 mm

Specificity

No. of patientswithout any polyp(s)

5–9 mmCT

colonography DCBE P

585 529 (90%) 568 (97%) �0.001324 302 (93%) 317 (98%) �0.001413 371 (90%) 401 (97%) �0.001661 566 (86%) 643 (97%) �0.001

or more polyps sized �10 mm

Specificity

No. of patientswithout any polyp(s)

�10 mmCT

colonography DCBE P

583 562 (96%) 578 (99%) �0.001324 320 (99%) 323 (100%) 0.38417 402 (96%) 411 (99%) 0.04662 630 (95%) 656 (99%) �0.001

ison is made in the same group of patients for each CT colonographynts had �1 polyp.roctoscopy, or surgery, which served as the reference standard.

ity an

ving 1

P

0.211.00.380.04

ing 1

P

0.060.730.730.04

mparpatiepy, p

s5tr(

mtspCpbairmsosd

ltraarcttlDtDdrwmctheTir

pp

cpnsetst(wmvhae(tcbrrpOwD

ttaicims

ocatmairwccht

April 2004 CT COLONOGRAPHY VS. DCBE 319

Missed Lesions

Retrospective review of all cases with missed le-ions showed identification of 7/19 (37%) and 3/9 (33%)–9 mm polyps at DCBE and CT colonography, respec-ively. Among missed lesions �1 cm in diameter, ret-ospective detection was possible in 11/18 (61%) and 5/771%) at DCBE and CT colonography, respectively.

DiscussionThis study prospectively compares the perfor-

ance of CT colonography with DCBE in a large asymp-omatic population with a low disease prevalence. Asuch, the study population represents a comparison ofatients likely to undergo screening with either DCBE orT colonography. CT colonography did not detect allolyps �1 cm detected by endoscopy, the gold standard,ut single-read CT colonography sensitivity estimates byll 3 readers were at least as high as DCBE, although thisncreased detection was significant only for a singleeader for polyps �10 mm. DCBE was significantlyore specific than CT colonography in most compari-

ons. By using double reading at CT colonography, 79%f patients with polyps �1 cm were detected with apecificity of 95%. Only 48% of these larger polyps wereetected at DCBE.

The results should be interpreted in light of 2 studyimitations. The first, favoring CT colonography, is thathe DCBE was read only once, whereas the CT colonog-aphy results are reported for both single reading as wells for a double-reading scenario. The sensitivity results inll cases favored CT colonography with the exception ofeader 2 for the 5–9 mm polyps for which the CTolonography and DCBE results were equal. The lowerhan expected sensitivity for DCBE is nearly identical tohat reported from the National Polyp Study. The secondimitation favoring DCBE is verification bias.28 TheCBE was read clinically at the actual time of examina-

ions. In the event that a polyp �5 mm was detected atCBE, further testing by colonoscopy was recommendeduring the same patient visit. Because the CT colonog-aphy examinations were read later, not all positive testsere followed-up with colonoscopy (because the patientay have returned home or chose not to undergo

olonoscopy for a research test finding). The confirmatoryests may have been preferentially performed in patientsaving a positive finding at DCBE, leading to an under-stimation of false-positive results at CT colonography.his is unlikely because a recent study at our institution

n a large screening population having both CT colonog-aphy and colonoscopy had an identical specificity for

olyps �1 cm of 95% at double-read CT colonogra-hy.29

The wide variation in polyp detection rates at CTolonography (45% to 71%) between observers in theresent study was difficult to explain. All of the exami-ations were performed in a blinded fashion by endo-copic and radiologic investigators with similar levels ofxperience. The protocol for CT colonography interpre-ation was consistent throughout the study. During thetudy, we met to ensure continued uniformity of prac-ice. The interpretation protocol at CT colonographyprimary axial image review using lung and soft-tissueindows to identify suspected lesions with the use ofultiplanar reformatted and 3-dimensional endoluminal

iews to improve reader confidence and to problem solve)as been accepted by most experts as clinically relevantnd state of the art. Because all of the readers werexperienced radiologists and trained in CT colonographyhaving read at least 150 cases before these interpreta-ions), it is unlikely that experience and training areausative factors. Interpretation times were not recorded,ut informal discussion of these times among the 3eaders did not uncover any significant differences ineading times. Identical software was used for all inter-retations. Overall examination quality was very good.nly 23 of 703 (3.3%) CT colonography examinationsere considered to be nondiagnostic. No examinations atCBE with polyps were considered nondiagnostic.Technical errors (those polyps that could not be iden-

ified retrospectively) accounted for approximately twohirds of the misses for polyps 5–9 mm in diameter, andpproximately one third of the misses for polyps �1 cmn diameter at both DCBE and CT colonography. Per-eptive errors, polyps that were missed prospectively anddentified retrospectively, accounted for the remaining

isses. These findings are consistent with those of othertudies evaluating the causes of error at DCBE.30–33

Perceptive errors at DCBE have been attributed toverlapping lines and complex anatomic structures thatannot be displayed fully without overlap. Reader fatiguend data overload may have been responsible for many ofhe perceptive errors at CT colonography. The readingethod used in this study required the reader to examinehuge set of data for each patient. A minimum of 1200

mages was reviewed for each patient. All CT colonog-aphy examinations were interpreted in a research settingithout the constraints of a busy clinical schedule. Be-

ause the DCBE examinations were performed as a usuallinical examination, radiologists may have felt moreurried to complete the interpretation than those readinghe CT colonography examination.

dmoicseSrrTets

rbiTbodcp5Dl9pcs(badqcctp

iTocrsctp

scp

ricpbac

fsigrrs

1

1

1

1

1

320 JOHNSON ET AL. CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 2, No. 4

Several measures may counter perceptual errors. First,ouble reading appears to compensate for perceptiveisses, and it requires time but no additional technol-

gy. We found that double reading at CT colonographyncreased sensitivity for polyps by 25% to 43%, but at aost of lost specificity. Although not performed in thistudy, double reading also has been shown to be anffective tool to reduce perceptive errors at DCBE.30

econd, image displays such as virtual pathology, byeducing the number of images to be reviewed by theadiologist, also could improve reader performance.34

hird, many of the errors and interobserver variabilityncountered at CT colonography may be correctable inhe future with the assistance of computer-aided diagno-is.35–39

DCBE is significantly more specific than CT colonog-aphy in nearly all comparisons, although verificationias inflates DCBE specificity estimates. The specificityn this study is higher than those in other reports.40–42

his likely is caused by an overestimate of specificity foroth examinations because full colonoscopy was not rec-mmended unless either test detected a lesion �5 mm iniameter. The lack of double reading at DCBE alsoontributed to a specificity higher than CT colonogra-hy. Specificity differences are largest when analyzing–9 mm polyps (86% vs. 97% at CT colonography andCBE, respectively). The differences are quantitatively

ess when large (�1 cm) polyps are analyzed (95% vs.9% at CT colonography and DCBE, respectively), butroportionately the same. Radiologists interpreting CTolonography should carefully scrutinize suspected le-ions 5–9 mm in diameter for characteristics of stoolinternal heterogeneity, angular surfaces, and movementetween supine and prone images). Stool tagging agentslso may be potentially helpful in reducing false-positiveetections. False positives at CT colonography will re-uire patients to have subsequent and unnecessaryolonoscopy. This will translate into higher medicalosts, additional time from work, added inconvenience tohe patient and their driver, and added risk for colonerforation and complications from sedation.CT colonography has several additional test character-

stics that are desirable for a screening examination.hese include safety (there have been no reported deathsr colon perforations), patient acceptance,43 minimal dis-omfort without the need for sedation,43 the ability toeturn to work immediately after the examination, as-essment of abdominal and pelvic organs,44,45 and theapability of performing colonoscopy immediately afterhe procedure if a lesion is discovered (no additionalreparation required). Because a full bowel preparation

till is required, the discomfort and inconvenience asso-iated with the preparation is similar for CT colonogra-hy and DCBE.43

DCBE shares many of the advantages of CT colonog-aphy including its safety and ability to return to workmmediately after the examination. Patients are less ac-epting of DCBE than CT colonography,43 and detectedolyps can be removed only at colonoscopy after repeatowel preparation. The ability to survey the abdomennd pelvis for extracolonic abnormalities is unique to CTolonography.

CT colonography and DCBE are not significantly dif-erent full structural examinations when interpreted by aingle examiner. Double-read CT colonography is signif-cantly more sensitive than single-read DCBE. Practiceuidelines should be considered to optimize CT colonog-aphy performance by using double reading. Technologicefinements to enhance the accuracy of CT colonographyhould be pursued.

References1. Cancer Facts and Figures 2002. American Cancer Society, 2002.2. Mandel JS, Bond JH, Church TR, Snover DC, Bradley GM, Schu-

man LM, Ederer F. Reducing mortality from colorectal cancer byscreening for fecal occult blood. N Engl J Med 1993;328:1365–1371.

3. Winawer SJ, Flehinger BJ, Schottenfeld D, Miller DG. Screeningfor colorectal cancer with fecal occult blood testing and sigmoid-oscopy. J Natl Cancer Inst 1993;85:1311–1318.

4. Winawer SJ, Zauber AG, Ho MN, O’Brien MJ, Gottlieb LS, Stern-berg SS, Waye JD, et al. Prevention of colorectal cancer bycolonoscopic polypectomy. N Engl J Med 1993;329:1977–1981.

5. Muller AD, Sonnenberg A. Protection by endoscopy against deathfrom colorectal cancer. A case-control study among veterans.Arch Intern Med 1995;155:1741–1748.

6. Muller AD, Sonnenberg A. Prevention of colorectal cancer byflexible endoscopy and polypectomy–a case control study of32,702 veterans. Ann Intern Med 1995;123:904–910.

7. Selby JV, Friedman GD, Quesenberry CP, Weiss NS. A case-control study of screening sigmoidoscopy and mortality fromcolorectal cancer. N Engl J Med 1992;326:653–657.

8. Selby JV, Friendman GD, Quesenberry CP, Weiss NS. Effect offecal occult blood testing on mortality from colorectal cancer: acase control study. Ann Intern Med 1993;118:1–6.

9. Newcomb PA, Norfleet RG, Storer BE, Surawicz TX, Marcus PM.Screening sigmoidoscopy and colorectal cancer mortality. J NatlCancer Inst 1992;84:1572–1575.

0. Lazovich D, Weiss NS, Stevens NG, White E, McKnight B, WagnerEH. A case-control study to evaluate efficacy of screening forfaecal occult blood. J Med Screen 1995;2:84–89.

1. Eddy DM, Nugent TW, Eddy JF, Coller J, Gilbertsen V, Gottlieb LS,Rice R, et al. Screening for colorectal cancer in a high-risk pop-ulation: results of a mathematical model. Gastroenterology1987;92:682–692.

2. Eddy DM. Screening for colorectal cancer. Ann Intern Med 1990;113:373–384.

3. Wagner JL, Tunis S, Brown M, Ching A, Almeda R. The cost-effectiveness of colorectal cancer screening in average riskadults. In: Young G, Levin B, eds. Prevention and early detectionof colorectal cancer. Philadelphia: Saunders, 1996:321–356.

4. Winawer SJ, Fletcher R, Rex DK, Bond JH, Burt RW, Ferrucci JT,

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

RM

April 2004 CT COLONOGRAPHY VS. DCBE 321

Ganiats T, et al. Colorectal cancer screening and surveillance:clinical guidelines and rationale: update based on new evidence.Gastroenterology 2003;124:544–560.

5. Smith RA, von Eschenbach AC, Wender R, Levin B, Byers T,Rothenberger D, Brooks D, et al. American Cancer Society guide-lines on screening and surveillance for the early detection ofadenomatous polyps and colorectal cancer. CA Cancer J Clin2001;51:44–54.

6. Hara AK, Johnson CD, Reed JE, Ahlquist DA, Nelson H, EhmanRL, McCollough CH, et al. Detection of colorectal polyps bycomputed tomographic colography: feasibility of a novel tech-nique. Gastroenterology 1996;110:284–290.

7. Hara AK, Johnson CD, Reed JE, Ehman RL, Ilstrup DM. Colorectalpolyp detection using computed tomographic colography: two-versus three-dimensional techniques. Radiology 1996;200:49–54.

8. Royster AP, Fenlon HM, Clarke PD, Nunes DP, Ferrucci JT. CTcolonoscopy of colorectal neoplasms: two-dimensional andthree-dimensional virtual-reality techniques with colonoscopiccorrelation. AJR Am J Roentgenol 1997;169:1237–1242.

9. Dachman AH, Kuniyoshi JK, Boyle CM, Samara Y, Hoffmann K,Rubin DT, Hanan I. CT colonography with three-dimensional prob-lem solving for detection of colonic polyps. AJR Am J Roentgenol1998;171:989–995.

0. Macari M, Berman P, Dicker M, Milano A, Megibow A. Usefulnessof CT colonography in patients with incomplete colonoscopy. AJRAm J Roentgenol 1999;173:561–564.

1. Fenlon H, Clarke PD, Ferrucci JT. Virtual colonoscopy: imagingfeatures with colonoscopic correlation. AJR Am J Roentgenol1998;170:1303–1309.

2. Fenlon H, McAneny DB, Nunes DP, Clarke PD, Ferrucci JT. Occlu-sive colon carcinoma: virtual colonoscopy in the preoperativeevaluation of the proximal colon. Radiology 1999;210:423–428.

3. Morrin MM, Kruskal JB, Farrell RJ, Goldberg SN, McGee JB,Raptopoulos V. Endoluminal CT colonography after an incompleteendoscopic colonoscopy. AJR Am J Roentgenol 1999;172:913–918.

4. Chen SC, Lu DSK, Hecht JR, Kadell BM. CT colonography: valueof scanning in both the supine and prone positions. AJR Am JRoentgenol 1999;172:595–599.

5. Yee J, Akerkar GA, Hung RK, Sreinauer-Gebauer AM, Wall SD,McQuaid K. Colorectal neoplasia: performance characteristics ofCT colonography for detection in 300 patients. AJR Am J Roent-genol 2001;219:685–692.

6. Fletcher JG, Johnson CD, Welch TJ, MacCarty RL, Ahlquist DA,Reed JE. Optimization of CT colonography technique: a prospec-tive trial in 180 patients. Radiology 2000;216:704–711.

7. American College of Radiology. ACR standard for the perfor-mance of a barium enema examination in adults. In: ACR stan-dards 2000–2001. Reston, VA: American College of Radiology,2002:163–166.

8. Begg CB. Biases in the assessment of diagnostic tests. Stat Med1987;6:411–423.

9. Johnson CD, Harmsen WS, Wilson LA, MacCarty RL, Welch TJ,Ilstrup DM, Ahlquist DA. Prospective blinded evaluation of com-puted tomographic colonography for screen detection of colorec-tal polyps. Gastroenterology 2003;125:311–319.

0. Markus JB, Somers S, O’Malley BP, Stevenson GW. Double-contrast barium enema studies: effect of multiple reading on

perception error. Radiology 1990;175:155–156.

1. Kelvin F, Gardiner R, Vas W, Stevenson GW. Colorectal carcinomamissed on double contrast barium enema study: a problem inperception. AJR Am J Roentgenol 1981;137:307–313.

2. Brady AP, Stevenson GW, Stevenson L. Colorectal cancer over-looked at barium enema examination and colonoscopy: a con-tinuing perceptual problem. Radiology 1994;192:373–378.

3. Gluecker TM, Fletcher JG, Welch TJ, MacCarty RL, Harmsen WS,Ilstrup DM, Wilson LA, et al. Characterization of missed andidentified lesions at CT colonography. AJR 2004;182 (in press).

4. Fletcher JG, Johnson CD, Reed JE, Garry J. Feasibility of planarvirtual pathology: a new paradigm in volume-rendered CT colonog-raphy. J Comput Assist Tomogr 2001;25:864–869.

5. Yoshida H, Masutani Y, MacEneaney P, Rubin DT, Dachman AH.Computerized detection of colonic polyps at CT colonography onthe basis of volumetric features: pilot study. Radiology 2002;222:327–336.

6. Summers RM, Beaulieu CF, Pusanik LM, Malley JD, Jeffrey RB Jr,Glazer DI, Napel S. Automated polyp detector for CT colonogra-phy: feasibility study. Radiology 2000;216:284–290.

7. Paik DS, Beaulieu CF, Jeffrey RB, Karadi C, Napel S. Detection ofpolyps in CT colonography: a comparison of a computer-aideddetection algorithm to 3D visualization methods (abstr). Radiol-ogy 1999;213:193.

8. Vining DJ, Ge Y, Ahn DK, Stelts DR. Virtual colonoscopy withcomputer-assisted polyp detection. In: Doi K, MacMahon H,Gieger ML, Hoffmann KR, eds. Computer-aided diagnosis in med-ical imaging. Amsterdam, The Netherlands: Elsevier Science,1999:445–452.

9. Summers RM, Johnson CD, Pusanik LM, Malley JD, Youssef AM,Reed JE. Automated polyp detection at CT colonography: feasibilityassessment in a human population. Radiology 2001;219:51–59.

0. Jaramillo E, Slezak P. Comparison between double-contrast bar-ium enema and colonoscopy to investigate lower gastrointestinalbleeding. Gastrointest Radiol 1992;17:81–83.

1. Steine S, Stordahl A, Lunde OC, Loken K, Laerum E. Doublecontrast barium enema versus colonoscopy in the diagnosis ofneoplastic disorders: aspects of decision making in general prac-tice. Fam Pract 1993;10:288–291.

2. Jensen J, Kewenter J, Aszteely M, Lycke G, Wojciechowski J.Double contrast barium enema and flexible rectosigmoidoscopy:a reliable diagnostic combination for detection of colorectal neo-plasm. Br J Surg 1990;77:270–272.

3. Gluecker TM, Johnson CD, Harmsen WS, Offord KP, Harris AM,Wilson LA, Ahlquist DA. Colorectal cancer screening by CTcolonography, colonoscopy, and double-contrast barium enema:prospective assessment of patient perceptions and preferences.Radiology 2003;227:378–384.

4. Hara AK, Johnson CD, MacCarty RL, Welch TJ. Incidental extraco-lonic findings at CT colonography. Radiology 2000;215:353–357.

5. Gluecker TM, Johnson CD, Wilson LA, MacCarty RL, Welch TJ,Vanness DJ, Ahlquist DA. Extracolonic findings at CT colonogra-phy: evaluation of prevalence and cost in a screening population.Gastroenterology 2003;124:911–916.

Address reprint requests to: C. Daniel Johnson, M.D., Department ofadiology, Mayo Clinic Rochester, 200 First Street S.W., Rochester,innesota 55905.

Supported by National Institutes of Health grant RO1 CA075333.