The Effect of Diagnostic

Confidence on the Probability of

Optical Colonoscopic
Confirmation of Potential Polyps
Downloaded from www.ajronline.org by 50.109.71.117 on 10/22/15 from IP address 50.109.71.117. Copyright ARRS. For personal use only; all rights reserved

Detected on CT Colonography:
Prospective Assessment in 1,339
Asymptomatic Adults
Perry J. Pickhardt1–3 OBJECTIVE. We sought to evaluate the effect of interpreter confidence on the likelihood
J. Richard Choi3,4 that a lesion detected on CT colonography (CTC) will correspond to a matched polyp seen on
Pamela A. Nugent2 optical colonoscopy.
William R. Schindler5 SUBJECTS AND METHODS. Same-day CTC and optical colonoscopy were performed
on 1,339 asymptomatic adults. A standard matching algorithm for polyp size and location was
used. For each potential polyp detected on CTC, the level of diagnostic confidence was pro-
spectively rated on a 3-point scale (1, least certain; 2, intermediate; and 3, most certain).
RESULTS. For CTC-detected lesions 6 mm or larger, diagnostic confidence levels of 1, 2,
and 3 corresponded to matched polyps on optical colonoscopy in 33.3% (45/135), 50.0% (103/
206), and 66.8% (157/235) of cases, respectively (p < 0.01). Similar trends were present for cat-
egories of lesions that measured 6–7 mm, 8–9 mm, and 10 mm or larger, rising to a match rate
of 82.1% (55/67) for lesions 10 mm or larger that were diagnosed with a level-3 confidence rat-
ing. The likelihood that a matched polyp was adenomatous increased with greater levels of di-
agnostic confidence. Of note, level-3 confidence for lesions measuring 8–9 mm on CTC more
often yielded a matching neoplasm on optical colonoscopy than level-1 or level-2 confidence
for lesions measuring 10 mm or larger (60.3% [35/58] vs 20.8% [10/48]; p < 0.0001).
CONCLUSION. Greater diagnostic confidence for an individual lesion detected on CTC
correlates with a significantly increased likelihood that a matching polyp will be found on op-
tical colonoscopy and that this matched polyp will be neoplastic. Although polyp size represents
the primary criterion for CTC screening algorithms, this data could help guide the decision to
opt for noninvasive CTC surveillance versus optical colonoscopy for polypectomy.
Received March 4, 2004; accepted after revision
May 25, 2004.

Supported in part by Department of Defense Advances in
Medical Practice funds and by the Society of Computed
Body Tomography and Magnetic Resonance.
The opinions and assertions contained herein are the
private views of the authors and are not to be construed as
official or as reflecting the views of the Departments of the
Navy, Army, or Defense.
1
Department of Radiology, University of Wisconsin
Medical School, E3/311 Clinical Science Center,
600 Highland Ave., Madison, WI 53792-3252. Address
correspondence to P. J. Pickhardt.
2
Department of Radiology, National Naval Medical Center,
Bethesda, MD 20889.
3
Department of Radiology and Nuclear Medicine,
Uniformed Services University of the Health Sciences,
Bethesda, MD 20814.
4
Department of Radiology, Walter Reed Army Medical
Center, Washington, DC 20307-5001.
5
Department of Gastroenterology, Naval Medical Center-
San Diego, San Diego, CA 92134-5000.
AJR 2004;183:1661–1665
0361–803X/04/1836–1661
© American Roentgen Ray Society
C
T colonography (CTC), also
known as virtual colonoscopy,
has been shown to be a highly ac-
curate test for the detection of colorectal pol-
yps when state-of-the-art technique is
applied [1]. For CTC to function as a clini-
cally useful and cost-effective screening tool,
patients with certain types of potential pol-
yps detected on CTC need not be referred for
immediate optical colonoscopy for polypec-
tomy [2]. There is general agreement that pa-
tients with polyps 10 mm or larger probably
should undergo immediate optical colonos-
copy for polypectomy, whereas patients with
only diminutive polyps (≤ 5 mm) do not need
to be referred for immediate polypectomy.
However, controversy surrounds the appro-
priate management of intermediate-sized le-
sions (6–9 mm) [1, 2]. Although relatively
little is known about the natural history of
subcentimeter polyps [3], the existing body
of data supports the validity of noninvasive
surveillance [4, 5].
The likelihood that a lesion detected on
CTC will correspond to a polyp on optical
colonoscopy probably depends on a variety of
factors, including size, morphology, quality of
the colonic preparation, and degree of luminal
distention. The purpose of this study was to
assess prospectively the effect of the inter-
preter’s overall diagnostic confidence in le-
sions detected on CTC on the likelihood that
a matching lesion will be subsequently found
on optical colonoscopy. This information
could be useful for clinical decision making
regarding the next appropriate step for CTC-
detected lesions and could affect the develop-
ment of CTC screening algorithms.
Subjects and Methods
Our study protocol for same-day CTC and opti-
cal colonoscopy was approved by the institutional

AJR:183, December 2004 1661


review board at all participating medical centers,
and all subjects provided written informed con-
sent. The primary study group was composed of
asymptomatic adults ages 50 and older referred for
colorectal cancer screening. Exclusion criteria in-
cluded positive results on a stool guaiac test or
iron-deficiency anemia within the previous 6
Downloaded from www.ajronline.org by 50.109.71.117 on 10/22/15 from IP address 50.109.71.117. Copyright ARRS. For personal use only; all rights reserved
months; rectal bleeding, hematochezia, or uninten-
tional weight loss of more than 10 lb (4.54 kg)
within the previous 12 months; optical colonos-
copy within the previous 10 years or barium en-
ema within the previous 5 years; personal history
of adenomatous polyps, colorectal cancer, or in-
flammatory bowel disease; and family history of
familial adenomatous polyposis or nonpolyposis
cancer syndromes. Over a 15-month period, 1,339
asymptomatic adults (mean age, 57.8 years; median
age, 56 years) successfully completed same-day
CTC and optical colonoscopy. CTC performance
characteristics from a large subset of these patients
have been reported previously [1], but the data re-
garding diagnostic confidence levels have not been
evaluated and are the focus of this study.
In our colonic cleansing regimen, patients in-
gested split doses of 45 mL of sodium phosphate
and 10 mg of bisacodyl. Patients also consumed
split doses of 250 mL of dilute barium (2.1%
weight/volume) and 60 mL of water-soluble iodi-
nated contrast material (diatrizoate meglumine
and diatrizoate sodium) for stool tagging and elec-
tronic fluid subtraction, as has been described pre-
viously [6]. Our CT protocol and CTC
interpretation technique have also been detailed
previously [1, 6]. In brief, after patient-controlled
rectal insufflation of room air, breath-hold supine
and prone CT acquisitions were obtained on 4-
and 8-MDCT scanners (LightSpeed Plus and
LightSpeed Ultra, respectively; GE Healthcare).
The CT technique entailed 1.25- to 2.5-mm colli-
mation, 13.5- to 15.0-mm/sec table speed, and 1-
mm reconstruction intervals.
Prospective interpretation of CTC studies was
performed by one of six radiologists using a com-
mercially available CT colonography system
(V3D Colon, version 1.2; Viatronix). The 3D en-
doluminal fly-through display was generally used
for primary polyp detection, and the 2D images
were used mainly for confirmation and problem
solving. For all detected lesions prospectively
identified as polyps on CTC, a level of diagnostic
confidence or certainty was assigned, using a 3-
point scale (1, least certain; 2, intermediate; or 3,
most certain). Although assignment of confidence
levels was subjective and somewhat individual-
ized, typical features of a CTC-detected lesion
with a diagnostic confidence level of 3 could in-
clude a well-circumscribed polypoid lesion clearly
identifiable on both supine and prone views. Fea-
tures associated with a level-1 confidence might
include an ill-defined margin, internal heterogene-
ity, coexisting retained debris or poor luminal dis-
tention, and a lesion not well seen on one of the
images. Confidence levels could also be expected
to increase with larger polyp sizes. Polyp mor-
1662
Pickhardt et al.
phology was prospectively recorded as peduncu-
lated, sessile, or flat. Lesions were measured on
the 3D image using electronic calipers and were
recorded by segment.
Optical colonoscopy was performed immedi-
ately after prospective CTC interpretation using
standard commercial video colonoscopes. The
colonoscope was advanced to the cecum and then
sequentially withdrawn into more distal segments
for polyp detection. Polyps were measured using a
calibrated linear probe, which is more accurate
than either visual or biopsy-forceps estimation [7].
Our polyp-matching algorithm required agreement
between the findings on CTC and on optical
colonoscopy for size (within a 50% margin of er-
ror) and location (within the same or adjacent seg-
ment) [1]. The colonoscopist completed the
evaluation of a given segment before being ap-
prised of the CTC results for the previous seg-
ment. If a polyp 5 mm or larger was detected on
CTC but not on prospective optical colonoscopy,
the colonoscopist closely reexamined that segment
and was allowed to review the CTC images for
guidance. This step provided an enhanced refer-
ence standard that surpassed optical colonoscopy
alone [1].
All retrieved polyps during optical colono-
scopic polypectomy were sent for histologic ex-
amination. Polyps measuring 6 mm and greater,
particularly adenomas, represented the lesions of
primary interest for this study. Given their lack of
clinical significance, diminutive lesions (< 5 mm)
were excluded from the final analysis [8].
The effect of interobserver variability was as-
sessed by having a second radiologist interpret
100 randomly selected CTC studies. The radiolo-
gist was blinded to the results of both optical
colonoscopy and the initial CTC interpretation.
The same diagnostic confidence scale was applied
to all lesions detected at this second interpretation.
For analysis of diagnostic confidence data, we
grouped the polyps into the following categories:
6–7 mm, 8–9 mm, and 10 mm and larger. These
size categories were used in our initial algorithm
for primary CTC screening to help determine the
next appropriate management step. Statistical
analysis for significance testing was made using
chi-square and Fisher’s exact tests, as appropriate.
Results
A total of 1,437 polyps in 656 (49.0%) of
the 1,339 asymptomatic adults were identi-
fied on optical colonoscopy after the initial
CTC findings were revealed. Of the 1,437
total polyps seen on optical colonoscopy,
380 (26.4%) measured 6 mm or larger, of
which CTC prospectively detected 305 (sen-
sitivity, 80.3%). The overall positive predic-
tive value was 53.0% (305/576). The
likelihood that a CTC-detected lesion corre-
sponded to a matching polyp on subsequent
optical colonoscopy increased according to
CTC confidence level: 33.3% (45/135),
50.0% (103/206), and 66.8% (157/235) for
confidence levels 1, 2, and 3, respectively. A
similar trend was found when other polyp
size thresholds between 7 mm and 10 mm
were applied, as well as for polyp size cate-
gories 6–7 mm, 8–9 mm, and 10 mm and
larger (Table 1). At a 10-mm-and-larger
threshold, CTC-detected lesions prospec-
tively identified with a confidence level of 1,
2, and 3 corresponded to matching polyps
on optical colonoscopy in 25.0% (5/20),
60.7% (17/28), and 82.1% (55/67) of cases,
respectively (Table 1).
The differences among confidence levels
were statistically significant for all polyp size
categories (p < 0.05). Pair-wise comparisons
between two individual levels were also sta-
tistically significant (p < 0.05), except for
confidence levels 1 and 2 at 8–9 mm and lev-
els 2 and 3 at 6–7 mm. The relative separa-
tion between match rates for levels 1 and 3
increased for larger polyps (Fig. 1). For
CTC-detected lesions measuring 10 mm or
more, only 25% of lesions diagnosed with
level-1 confidence corresponded to matching
polyps on optical colonoscopy, compared
with 82% of those diagnosed with level-3
confidence (p < 0.001).
Of all 380 polyps 6 mm or larger found on
optical colonoscopy, 234 (61.6%) were ade-
nomatous, 197 of which were prospectively
detected on CTC (sensitivity, 84.2%). Most
CTC-detected adenomas were prospectively
assigned a level-3 diagnostic confidence
level, including 118 (59.9%) of 197 ade-
nomas 6 mm or larger, 79 (71.8%) of 110 ad-
enomas 8 mm or larger, and 44 (81.5%) of
54 adenomas 10 mm or larger. The likeli-
hood that a CTC-detected lesion corre-
sponded to an adenomatous polyp on optical
colonoscopy and at histologic evaluation in-
creased with higher levels of diagnostic con-
fidence at all thresholds of polyp size (Table
1). The differences among confidence levels
were statistically significant for all polyp size
categories (p < 0.05). Pair-wise comparisons
between individual levels were also statisti-
cally significant (p < 0.05), except for levels
1 and 2 at 8–9 mm and at 10 mm or larger,
and levels 2 and 3 at 6–7 mm. The pair-wise
comparison between levels 1 and 3 was
highly significant for all polyp size catego-
ries (p < 0.001). As we found with the total
sample of polyps, we observed greater sepa-
ration between adenoma match rates for lev-
els 1 and 3 with larger polyp sizes (Fig. 2),
AJR:183, December 2004
 Diagnosing Polyps on CT Colonography

Likelihood of a Matching Polyp on Optical Colonoscopy for Lesions Detected on CT Colonography According to Polyp Size
TABLE 1
and Diagnostic Confidence Level

Polyp Size, All Polyps on CT Colonography Sessile Morphology Flat Morphology Pedunculated Morphology
Confidence Match on Adenoma Match on Adenoma Match on Adenoma Match on Adenoma
Levela Colonoscopyb Matchc Colonoscopyb Matchc Colonoscopyb Matchc Colonoscopyb Matchc
Downloaded from www.ajronline.org by 50.109.71.117 on 10/22/15 from IP address 50.109.71.117. Copyright ARRS. For personal use only; all rights reserved

≥ 10 mm
1 5/20 (25.0) 2/20 (10.0) 2/13 (15.4) 0/13 (0) 1/3 (33.3) 0/3 (0) 2/4 (50.0) 2/4 (50)
2 17/28 (60.7) 8/28 (28.6) 13/19 (68.4) 5/19 (26.3) 2/4 (50.0) 1/4 (25.0) 2/5 (40.0) 2/5 (40.0)
3 55/67 (82.1) 44/67 (65.7) 23/32 (71.9) 15/32 (46.9) 3/5 (60.0) 2/5 (40.0) 29/30 (96.7) 27/30 (90)
8–9 mm
1 10/22 (45.5) 2/22 (9.1) 6/14 (42.9) 0/14 (0) 4/6 (66.7) 2/6 (33.3) 0/2 (0) 0/2 (0)
2 22/39 (56.4) 7/39 (17.9) 13/22 (59.1) 4/22 (18.2) 3/10 (30.0) 0/10 (0) 6/7 (85.7) 3/7 (42.9)
3 44/58 (75.9) 35/58 (60.3) 33/45 (73.3) 28/45 (62.2) 3/3 (100) 3/3 (100) 8/10 (80.0) 4/10 (40.0)
6–7 mm
1 30/93 (32.3) 17/93 (18.3) 16/58 (27.6) 9/58 (15.5) 13/32 (40.6) 8/32 (25.0) 1/3 (33.3) 0/3 (0)
2 64/139 (46.0) 43/139 (30.9) 52/108 (48.1) 37/108 (34.3) 6/21 (28.6) 2/21 (9.5) 6/10 (60.0) 4/10 (40.0)
3 58/110 (52.7) 39/110 (35.5) 46/85 (54.1) 31/85 (36.5) 7/11 (63.6) 4/11 (36.4) 5/14 (35.7) 4/14 (28.6)
Note.—See text for discussion of statistical significance. Numbers in parentheses are percentages.
aDiagnostic confidence levels: 1 = least certain, 2 = intermediate, 3 = most certain.

bFraction of lesions detected on CT colonography that matched true polyps on optical colonoscopy, regardless of histology.

cFraction of lesions detected on CT colonography that matched adenomatous polyps on optical colonoscopy.

including increases from 9% to 60% for pol-
yps measuring 8–9 mm and from 10% to
66% for polyps measuring 10 mm or larger.
Important differences in CTC polyp detec-
tion can occur near a given threshold for
polyp size. For instance, a lesion assigned a
CTC confidence level of 3 that measures just
below the 10-mm cutoff (i.e., an 8- or 9-mm
lesion) would be more likely to represent a
true polyp than a lesion measuring 10 mm or
larger that was assigned a CTC confidence
level of 1 (75.9% [44/58] vs 25.0% [5/20]; p <
0.0001). Similarly, an 8- or 9-mm lesion as-
signed a CTC confidence level of 3 would be
more likely to represent a neoplasm (ade-
noma) than a CTC confidence level–1 or –2
lesion measuring 10 mm or larger (60.3%
[35/58] vs 20.8% [10/48]; p < 0.0001).

0.9 0.8

0.8 0.7
Likelihood of Match

0.6
Likelihood of Match

0.7

0.5
0.6
0.4
0.5
0.3
0.4
0.2
0.3 0.1

0.2 0
6–7 8–9 ≥ 10 6–7 8–9 ≥ 10
Polyp Size (mm) Polyp Size (mm)

Fig. 1.—Line graph shows relationship between diagnostic confidence level for le-
sions detected on CT colonography and likelihood that matching polyps will be
found on optical colonoscopy. For each category of polyp size, likelihood of match
increases as diagnostic confidence increases. In addition, relative spread between
match rates generally increases at larger polyp sizes. ◆ = level-1 confidence, ■ =
level-2 confidence, ▲ = level-3 confidence.
Fig. 2.—Line graph shows relationship between diagnostic confidence level for lesions
detected on CT colonography and likelihood that adenomatous (neoplastic) polyps will
be found on optical colonoscopy. For each category of polyp size, likelihood of adenoma
match increases as diagnostic confidence increases. Note that, as in Figure 1, relative
spread between match rates generally increases at larger polyp sizes. ◆ = level-1 con-
fidence, ■ = level-2 confidence, ▲ = level-3 confidence.

AJR:183, December 2004 1663


For sessile, flat, and pedunculated polyp
morphologies, the trend of increasing likeli-
hood of an optical colonoscopy match for le-
sions diagnosed with increasing diagnostic
certainty was generally evident (Table 1). The
match rate for level-3 diagnostic confidence
Downloaded from www.ajronline.org by 50.109.71.117 on 10/22/15 from IP address 50.109.71.117. Copyright ARRS. For personal use only; all rights reserved
was greater than that for level-l confidence for
all polyp morphologies at all size categories.
For sessile lesions, the match rate progres-
sively increased from the least to the most cer-
tain levels, and the overall differences for each
size category were statistically significant (p <
0.01), except for all polyps in the 8–9 mm
group. The pair-wise comparison between
levels 1 and 3 for sessile morphology was sta-
tistically significant (p < 0.01) for all size cat-
egories. Multiple additional pair-wise
comparisons were statistically significant for
various other combinations of polyp size and
morphology but were too numerous to list in-
dividually here. One interesting finding was
that for CTC-detected flat lesions, the level-2
match rates for 6- to 9-mm lesions were rela-
tively low (0–30%), whereas the level-2
match rates for similarly sized pedunculated
lesions were considerably higher (40–86%).
For CTC-detected pedunculated lesions mea-
suring 10 mm or larger, 97% (29/30) matched
with polyps on optical colonoscopy, and 90%
(27/30) were neoplastic.
In the 100 randomly selected cases that
were double-interpreted, 13 lesions 6 mm or
larger were identified on the initial CTC in-
terpretation but not on the second, and seven
lesions were identified on the second inter-
pretation but not on the first. Of these 20 “le-
sions” seen by just one radiologist, none was
assigned a diagnostic confidence level of 3.
By comparison, of the 36 lesions 6 mm or
larger detected and agreed upon by both re-
viewers, 21 (58.3%) were assigned a diag-
nostic confidence level of 3. Over half of
these 36 agreed-upon lesions matched ade-
nomas on subsequent optical colonoscopy,
compared with only a quarter of the 20 le-
sions identified by just one reviewer. Overall,
the second CTC interpretation uncovered
only two additional adenomas in these 100
patients. One patient already had a separate
adenoma in the 6-mm-or-larger category de-
tected prospectively by the first reviewer.
Discussion
CTC, also referred to as virtual colonos-
copy, is a promising tool for colorectal can-
cer screening [1]. For it to be a useful and
cost-effective approach, rational manage-
1664
Pickhardt et al.
ment of subcentimeter polyps is required to
avoid overuse of optical colonoscopy [2].
Physicians generally agree that CTC-de-
tected lesions 10 mm and greater should be
referred for polypectomy and that diminutive
polyps (≤ 5 mm) can be monitored at routine
screening intervals [1, 8]. The appropriate
management for CTC-detected lesions mea-
suring 6–9 mm, however, has not been estab-
lished. Noninvasive surveillance of small
lesions (< 10 mm) would seem prudent be-
cause less than 1% of subcentimeter lesions
(including both adenomatous and nonade-
nomatous polyps) are histologically ad-
vanced. Most never progress to cancer, and
the typical dwell time for even a 10-mm ade-
noma is possibly a decade or longer [1, 9].
Furthermore, in a well-designed optical
colonoscopy trial, follow-up of unresected
subcentimeter polyps was shown to be safe
for as long as 3 years [5].
Polyp size has served as the primary crite-
rion by which CTC performance has been
assessed. A major reason is that polyp size
serves as a rough surrogate for histologic ex-
amination because most diminutive lesions
(≤ 5 mm) are nonadenomatous and most pol-
yps 6 mm or larger are adenomatous [1, 10].
However, polyp detection on CTC is not al-
ways straightforward, and factors beyond
polyp size alone affect the degree of diagnos-
tic certainty.
The diagnostic confidence level that a re-
viewer may assign to a given CTC finding
believed to represent a polyp encompasses a
complex blend of factors that cannot be eas-
ily quantified. Inherent characteristics of the
detected lesion, such as its size, morphology
(sessile, pedunculated, or flat), and relation-
ship to the colonic folds all play a role. Other
factors depend more on the overall study
quality, such as the quality of colonic prepa-
ration and the degree of luminal distention.
Reviewer experience is likely to be another
important contributing factor. Our findings
show that polyp size and morphology have
an effect on diagnostic confidence, but we
did not directly assess the relative contribu-
tion of the various remaining factors. Re-
gardless, our findings show that there is
clearly a reproducible relationship between
the degree of diagnostic certainty and the
likelihood of finding a matching polyp on
optical colonoscopy.
This new data on diagnostic confidence
could potentially play a role in primary CTC
screening. For instance, noninvasive surveil-
lance of a CTC-detected 8- or 9-mm lesion
assigned a level-1 diagnostic confidence
seems quite reasonable because the likeli-
hood of finding a matching adenoma on opti-
cal colonography may be less than 10%. The
likelihood of finding a matching adenoma
rises to 60% for a CTC-detected 8- or 9-mm
lesion assigned a level-3 diagnostic confi-
dence; although noninvasive follow-up with
CTC would still be a reasonable option,
some may consider polypectomy to be ap-
propriate for such cases. Similarly, the differ-
ence between the likelihoods of finding an
adenoma match for a CTC-detected 10-mm
lesion assigned a level-1 confidence rating
(10%) and one assigned a level-3 confidence
rating (66%) could influence management,
particularly in patients with significant co-
morbidities. Fortunately, most polyps 8 mm
or larger identified on CTC were assigned
the highest level of diagnostic confidence,
which bodes well for the positive predictive
value in larger, more clinically significant
polyps. The increased uncertainty seen with
CTC-detected lesions smaller than 8 mm
correlates with the decreased by-patient
specificity that we have reported, which fur-
ther supports surveillance of these possible
lesions instead of immediate invasive
polypectomy [1].
With more accumulated experience in inter-
preting CTC studies and continued improve-
ment in the technique, the number of lesions
assigned to the lowest level of diagnostic cer-
tainty can be expected to decrease. This de-
crease could result from greater specificity
gained both by identifying fewer false-positive
lesions and by learning to assign a higher con-
fidence level to more likely matches.
There are limitations to our study. As we
mentioned, the subjective and multifactorial
nature of what constitutes a certain diagnos-
tic confidence level prevents us from provid-
ing strict definitions at this time, but general
guidelines for each category could evolve.
For instance, specific language could be in-
corporated, such as “likely polyp” for level-3
confidence, “possible polyp” for level-2 con-
fidence, and “unlikely polyp” for level-1
confidence. Another limitation is the use of
an imperfect reference standard (optical
colonoscopy). Through the use of the seg-
ment-by-segment unblinding of findings, we
created an enhanced reference standard and
were able to show that at least some of the
false-positive findings on CTC actually rep-
resented false-negative findings on optical
colonoscopy. However, there were likely ad-
ditional true polyps detected on CTC that
AJR:183, December 2004

simply were not found on optical colonoscopy
despite a second look. Further studies are re-
quired to assure reproducibility in CTC polyp
measurement and also to compare primary 2D
versus 3D polyp measurement techniques. The
fact that our analysis was primarily focused on
Downloaded from www.ajronline.org by 50.109.71.117 on 10/22/15 from IP address 50.109.71.117. Copyright ARRS. For personal use only; all rights reserved
the polyp and not on the patient was also a lim-
itation; per-patient confidence levels were not
prospectively obtained.
Interobserver variability in CTC interpre-
tation has been identified as a potential con-
cern for implementation of a screening
program [11]. However, using a primary 3D
approach for polyp detection, we have shown
good interobserver agreement, despite the
low prevalence of significant polyps in a
screening population [1].
In conclusion, the process of ultimately
deciding whether a detected abnormality
should be called a polyp on CTC is complex
and involves a variety of factors beyond just
lesion size and morphology. Despite this
fact, our findings indicate that increased di-
AJR:183, December 2004
Diagnosing Polyps on CT Colonography
agnostic confidence for an individual lesion
detected on CTC correlates with a signifi-
cantly increased likelihood that a matching
polyp will be found on optical colonoscopy
and that this matching polyp will be neoplas-
tic. This additional information could help
guide radiologists, patients, and referring
physicians as to the next appropriate step
when a potential polyp is detected on CTC.
References
1. Pickhardt PJ, Choi JR, Hwang I, et al. Computed
tomographic virtual colonoscopy to screen for
colorectal neoplasia in asymptomatic adults. N
Engl J Med 2003;349:2189–2198
2. Pickhardt PJ. CT virtual colonoscopy for colorec-
tal screening: transition from validation to imple-
mentation. Abdom Imaging 2004 (in press)
3. Schoen RE. Surveillance after positive and nega-
tive colonoscopy examinations: issues, yields,
and use. Am J Gastroenterol 2003;98:1237–1246
4. Hofstad B, Vatn MH, Larsen S, Osnes M. Growth
of colorectal polyps: recovery and evaluation of un-
resected polyps of less than 10 mm, 1 year after de-
tection. Scand J Gastroenterol 1994;29:640–645
5. Hofstad B, Vatn MH, Andersen SN, et al. Growth
of colorectal polyps: redetection and evaluation
of unresected polyps for a period of three years.
Gut 1996;39:449–456
6. Pickhardt PJ, Choi JR. Electronic cleansing and
stool-tagging in CT colonography: advantages
and pitfalls with primary three-dimensional eval-
uation. AJR 2003;181:799–805
7. Gopalswamy N, Shenoy VN, Choudhry U, et al. Is
in vivo measurement of size of polyps during
colonoscopy accurate? Gastrointest Endosc 1997;
46:497–502
8. Bond JH. Clinical relevance of the small colorec-
tal polyp. Endoscopy 2001;33:454–457
9. Bond JH. Update on colorectal polyps: manage-
ment and follow-up surveillance. Endoscopy
2003;35[suppl]:S35–S40
10. Pickhardt PJ, Choi JR, Hwang I, Schindler WR.
Nonadenomatous polyps at CT colonography:
prevalence, size distribution, detection rates, and
implications for colorectal cancer screening. Ra-
diology 2004 (in press)
11. Johnson CD, Harmsen WS, Wilson LA, et al.
Prospective blinded evaluation of computed to-
mographic colonography for screen detection of
colorectal polyps. Gastroenterology 2003;125:
311–319
1665