How long does it take for polyps to become cancerous

1From the Imaging Biomarkers and Computer-Aided Diagnosis Laboratory, Department of Radiology and Imaging Sciences, National Institutes of Health Clinical Center, Bldg 10, Room 1C368X, MSC 1182, Bethesda, MD 20892-1182.

Address correspondence to the author (e-mail: vog.hin@smr).

Copyright © RSNA, 2010

Abstract

Polyp size is a critical biomarker for clinical management. Larger polyps have a greater likelihood of being or of becoming an adenocarcinoma. To balance the referral rate for polypectomy against the risk of leaving potential cancers in situ, sizes of 6 and 10 mm are increasingly being discussed as critical thresholds for clinical decision making (immediate polypectomy versus polyp surveillance) and have been incorporated into the consensus CT Colonography Reporting and Data System (C-RADS). Polyp size measurement at optical colonoscopy, pathologic examination, and computed tomographic (CT) colonography has been studied extensively but the reported precision, accuracy, and relative sizes have been highly variable. Sizes measured at CT colonography tend to lie between those measured at optical colonoscopy and pathologic evaluation. The size measurements are subject to a variety of sources of error associated with image acquisition, display, and interpretation, such as partial volume averaging, two- versus three-dimensional displays, and observer variability. This review summarizes current best practices for polyp size measurement, describes the role of automated size measurement software, discusses how to manage the measurement uncertainties, and identifies areas requiring further research.

© RSNA, 2010

Introduction

The size of a colonic polyp is a biomarker that correlates with its risk of malignancy and guides its clinical management (1). The risk of malignancy increases with increasing polyp size (2).

The relationship of polyp size and risk of malignancy was originally reported in the pathology and surgery literature; our knowledge later expanded with the widespread use of endoscopy (2,3). This wealth of data informs the clinical management of polyps found at computed tomographic (CT) colonography. For example, the CT Colonography Reporting and Data System (C-RADS), a consensus recommendation of expert members of the Working Group on Virtual Colonoscopy, specifically incorporates polyp size into its recommendations (4) (Table 1). In this system, polyp size thresholds at 6 and 10 mm have particular significance. A CT colonographic study that demonstrates only one or two polyps 6–9 mm in size is categorized as “C2,” whereas one that depicts a 10 mm or larger polyp is categorized as “C3.” The recommended management in this system is surveillance or colonoscopy for C2 and colonoscopy for C3. A consensus of multiple national medical societies, however, recommends immediate polypectomy for all polyps 6 mm or larger (5).

Table 1

C-RADS Categories

Open in a separate window

Note.—Adapted and reprinted, with permission, from reference 4.

Given this central role of polyp size as a biomarker, the precision and accuracy of polyp measurement is an important issue. Appropriately, polyp size measurement has undergone intense study in the CT colonography literature over the past several years. The purpose of this article is to review the state of knowledge on polyp size measurements, to summarize what we already know and what we need to know, to prevent repetition of work already done, and to facilitate focus on what is needed.

Types of Polyps and Their Likelihood at Different Size Thresholds

According to the adenoma-carcinoma theory, adenomatous polyps are the precursors of most colorectal cancers (2,6). In the National Polyp Study, of 5066 polyps removed in 2362 patients, 66.5% were adenomatous, 11.2% were hyperplastic, and 22.3% were other types (normal colonic mucosa, inflammatory or juvenile polyps, or other less common entities) (7). Evidence indicates that larger adenomas and adenomas with villous histology put patients at greater risk of developing cancer (7,8). For example, in a cohort of 1618 patients observed for an average of 14 years per patient following sigmoidoscopy, colon cancer developed in 3.7% of patients with a rectosigmoid adenoma at study entry that was tubulovillous, villous, or 1 cm or larger (9). In comparison, colon cancer developed in only 0.5% of patients with only small tubular adenomas.

Cancer is known to occur more frequently in larger polyps. In the St. Marks Hospital experience, cancer was present in about 1% of adenomas smaller than 1 cm, 10% of 1–2 cm adenomas, and nearly 50% of adenomas larger than 2 cm (2). The St. Marks study was conducted between 1957 and 1968, the colons were inspected mainly with barium enema rather than colonoscopy, and while not clearly specified, the patient population was likely mainly a symptomatic (not screening) one. As the authors state, the number of unrecognized polyps was likely to be large. Thus, the cancer rate in adenomas is likely inflated in this study. In a screening population, cancerous polyps are uncommon. In the Department of Defense CT colonography screening trial, only two of 1233 asymptomatic adults had cancerous polyps (0.2%) (10).

Two other important polyp characteristics that vary with polyp size are the presence of high-grade dysplasia and villous features. Both high-grade dysplasia and villous features are viewed as intermediate stages in the path to malignancy and have been called “advanced histology.” In the National Polyp Study, 6–10 mm and larger than 10 mm adenomas were 4.8 and 20.3 times more likely to have high-grade dysplasia compared with small adenomas, respectively (7). The frequencies of high-grade dysplasia in 6–10 mm adenomas and adenomas larger than 10 mm were 4.4% and 16.2%, respectively. In a study of 13992 asymptomatic patients who underwent screening colonoscopy at one of 17 practice sites, only nine of 1198 patients (0.8%) with a largest polyp between 6 and 9 mm had a polyp with high-grade dysplasia; 53 (4.4%) had a villous or tubulovillous adenoma (11). The frequencies of high-grade dysplasia and villous features increased to 4.7% and 21.5%, respectively, for polyps in patients whose largest polyp was 10 mm or greater. The much lower frequencies of high-grade dysplasia in this study compared with the National Polyp Study may relate to the inclusion of symptomatic patients in the latter study and to the known variability of pathologists at interpreting the histologic findings. The authors also observed large variability in the prevalence of advanced neoplasia if polyp size measurements were adjusted upward or downward at the 9–10 mm and 5–6 mm size boundaries.

The polyp sizes in each of these studies were determined by using different methods. For example, histologic size after fixation was used in one study (9) and size at endoscopy was used in others (7,11).

Hyperplastic polyps are found frequently at colonoscopy and represent a large fraction of polyps smaller than 1 cm. For patients whose largest polyp was 6–9 mm, the most advanced histology was hyperplastic in 27.9%; the frequency increased to 41.2% in patients whose largest polyp was 1–5 mm (11). In the Polyp Prevention Trial, 437 of 1637 patients (26.7%) had one or more hyperplastic polyps, most of which were diminutive (12).

Most hyperplastic polyps are thought to have little or no malignant potential. However, a small subset of hyperplastic polyps (<3%) may become cancerous through the so-called serrated adenoma pathway (13,14). This subset of polyps bears some features of hyperplastic polyps but may in fact belong to another category of polyps called serrated adenomas. The serrated adenoma pathway is an alternate pathway to cancer in addition to the adenoma-carcinoma sequence described above. Unlike cancers associated with the adenoma-carcinoma sequence, for which genetic mutations are characteristic, carcinomas that develop through the serrated adenoma pathway may demonstrate microsatellite instability, a type of genetic instability caused by disruption of DNA mismatch repair. As many as 20% of sporadic colorectal cancers may develop through the serrated adenoma pathway (14). Our knowledge about the serrated adenoma pathway is in a nascent stage at present. It is not yet possible to distinguish adenomatous and hyperplastic polyps at CT colonography, although hyperplastic polyps may be flatter and hence less conspicuous at CT colonography compared with adenomatous polyps (15,16).

Polyp Growth Rates

Cancerous polyps tend to grow slowly. It is estimated that the polyp dwell time, the time needed for a small adenoma to transform into a cancer, may be on average 10 years (17). Evidence from the heyday of barium enema examinations indicates that most polyps do not grow or grow very slowly (18). For example, Welin et al (19) measured the growth rates of 375 colorectal lesions at serial double-contrast barium enema examinations. By using an exponential growth model, the authors estimated the doubling times of the majority of the polyps to be measured in years. The fastest-growing polyps and cancers had an estimated doubling time of between 138 and 866 days; the fastest growing cancer grew 2.5 mm in 100 days. In a study from the Mayo Clinic (8) of 226 patients studied serially by means of single-contrast barium enema examination and having at least one index polyp 1 cm or larger, only 83 polyps (37%) demonstrated interval growth (defined as diameter increase of > 25%, corresponding to a doubling in volume) during a mean follow-up of 68 months.

There is evidence from endoscopic surveillance of polyps that most polyps do not grow and some may even regress. Hoftad et al (20) followed polyps less than 10 mm in situ for 3 years in 116 patients undergoing annual colonoscopy. Polyps 5–9 mm in size showed a net tendency to regress. The authors concluded that follow-up of patients with unresected polyps up to 9 mm was safe. More recently, Loeve et al (21) compared data from the National Polyp Study to simulation data from an epidemiologic model. They concluded that the high observed adenoma detection rates and low colorectal cancer incidence in the National Polyp Study could be explained by a dynamic process of both formation and regression of adenomas. It is well known that certain anti-inflammatory drugs can shrink or prevent colorectal adenomas (22).

The Endoscopy and Pathology Literature on Colonic Polyp Size

Endoscopists and pathologists have extensively studied polyp size measurements in an effort to determine which most accurately represents truth (Table 2). A number of factors affect such size measurements leading to overestimation, underestimation, or high variability. These factors include overestimation due to the magnification and distortion produced by the wide-angle lens at the tip of the endoscope, underestimation due to piecemeal removal of a polyp and shrinkage after fixation, and variability due to inaccurate measurement tools and observer variation.

Table 2

Summary of Literature Evaluating Polyp Size Measurement at Optical Colonoscopy, Pathologic Examination, and CT Colonography

Open in a separate window

Note.—Blank cells indicate data that were either not provided in the studies or not readily summarized in this table format. CTC = CT colonography, LP = linear probe, MPR = multiplanar reformatted (arbitrary, oblique orientation of the imaging plane), OBF = open biopsy forceps, OC = optical colonoscopy, Orth = orthogonal (transverse, coronal, or sagittal plane that may be “optimized” to obtain maximum measurement), PF = postfixation, PP = postpolypectomy, Tr = transverse, VE = visual estimation.

* Indicates relative accuracy of the tests in comparison to a reference standard, most accurate listed first; reference standards varied among the studies (ie, ruler, pathologic examination, or optical colonoscopy, depending on the study).

† Relative polyp size measurements from smallest to largest.

‡ Indicates polyp sizes were highly variable compared with a reference standard (eg, high standard deviation or wide limits of agreement of the measurements) or assesses relative variability of two or more methods of polyp size measurement (eg, interobserver or intraobserver variability).

In 1996, Morales et al (23) compared polyp size estimates made in vivo at optical colonoscopy using visual comparison with the size of open biopsy forceps to direct measurements of polyps removed by means of snare polypectomy and to measurements made by a pathologist after formalin fixation of the polyps. The authors found that endoscopic estimates were on average 1.6 mm (18%) larger than postpolypectomy measurements. Nearly three-fourths of the colonoscopic estimates were larger than the postpolypectomy measurements, but there was no significant difference between the postpolypectomy and postfixation measurements.

Schoen et al (24) assessed endoscopist and pathologist measurements of polyp size. The reference standard was size measurements made with a millimeter ruler by an independent examiner (distinct from the endoscopists and pathologists being assessed) who measured the polyps immediately after removal but prior to immersion in formalin fixation. The endoscopists used visual estimation of polyp size. The pathologists made their size measurements immediately at removal of the polyps from formalin prior to sectioning. The authors found that formalin fixation did not significantly affect polyp size, that endoscopist estimates of polyp size were often unreliable (20% of measurements differed from the reference standard by more than 3 mm—both underestimation and overestimation), and that pathologist measurements of polyp size were preferable. The maximum polyp size change following formalin fixation was ±2 mm.

Gopalswamy et al (25) compared three endoscopic methods of polyp size measurement: open biopsy forceps, linear probe, and visual estimation. The reference standard was the size of the polyp measured outside the body with a ruler immediately after polypectomy. The authors found that size measured by using a flexible linear probe correlated best with the reference standard, that visual estimation (through the colonoscope but without reference to an object of known size) was next best, and that open forceps was the least accurate. In contrast to the study by Morales et al, they did find significant shrinkage (by 12%–18%) of polyps after formalin fixation compared with the actual size determined by using a ruler immediately following polypectomy. While a flexible linear probe has been used in some studies, unfortunately the open biopsy forceps is the most frequently reported endoscopic method in recent research articles. The authors noted some of the problems with the open biopsy forceps; these include difficulty in aligning the forceps along the largest diameter of the polyp, measuring polyps 8 mm or larger because the diameter of the polyp exceeds that of the forceps, and measuring pedunculated polyps at acute bends or on mucosal folds. The wide-angle lenses typically used in endoscopes can also lead to image distortion and apparent decrease in polyp size.

Fennerty et al (26), in an artificial colon model, found that endoscopists tended to underestimate polyp size. This result contradicted the findings of Morales et al described above. The authors also found significant intraobserver variability when the same polyp was measured 2 weeks later. In another artificial colon model, Schwartz et al (27) also found that endoscopists underestimated polyp size but that the measurement error could be reduced significantly with directed training.

Of interest is the observation that even pathologists cannot agree on polyp size. In a study by Rubio et al (28), 18 pathologists and four surgeons measured a set of phantom polyps by using a ruler. The reference sizes of the “polyps” were 8–28 mm as determined by a micrometer (a highly accurate measuring device). About one-third of the measurements over- or underestimated true size by more than 1 mm.

A synthesis of these articles indicates that size measurements by endoscopists and pathologists can be highly variable. The commonly used comparison to open biopsy forceps is unreliable. Postfixation measurements are up to 18% smaller than those made after polypectomy.

Comparisons of CT Colonography to Endoscopic and Pathologic Measurements

There have been a number of studies addressing the accuracy of polyp size measurement at CT colonography (Table 2). Some studies used phantoms, others used clinical data, and some used a combination of the two.

In a study of 251 polyps in which the authors compared CT colonography to optical colonoscopy with a linear probe, Yeshwant et al (29) found that to ensure identification of all polyps measuring 10 mm at optical colonoscopy, all polyps measuring 8 mm or larger at CT colonography needed to be identified. This result agreed with findings in an observer performance study by Burling et al (30) using 28 observers; in that study, polyp sizes were generally underestimated by 2–3 mm compared with optical colonoscopy.

In a preliminary report of polyp size measurements from the National CT Colonography Trial, Chen et al (31) found that colonoscopy measurements were on average approximately 1.2 mm larger than those made at CT colonography. In contrast, polyp size at pathologic measurement postfixation (C.D. Johnson, written communication, July 16, 2009) was approximately 0.9 mm less than that at CT colonography. The authors found that the optical colonoscopic measurements were consistently 12% higher and the pathologic measurements 25% smaller than those at CT colonography, regardless of polyp size.

In a reanalysis of data from 600 patients from the Cotton CT colonography trial (32), Gupta et al (33) found that both colonoscopy and CT colonography were inaccurate for determining polyp size as compared with prefixation sizes. There were 55 polyps of 6 mm or larger in the study. The variability was on the order of 1–3 mm for both colonoscopy and CT colonography. With CT colonography, polyp sizes were both over- and underestimated. The sensitivity for polyp detection was relatively poor in the initial trial, leading to concerns about inadequate reader expertise that might have adversely affected size measurements also.

In the pig colon model, Park et al (34) found that CT colonography was a more reliable and accurate way to measure polyp size compared with optical colonoscopy. The authors also found that the standard method of matching polyps by size at optical colonoscopy and CT colonography (the ±50% size criterion) was inferior to a fixed margin-of-error criterion (±5 mm for a polyp 10 mm or larger).

A synthesis of these data suggests that (according to Yeshwant et al and Chen et al) polyp size measurements determined at CT colonography lie intermediate to those made at pathologic evaluation and colonoscopy and may be closest to the in vivo size (Fig 1). The results of Gupta et al and Park et al contradict this conclusion to some degree, but their studies have limitations (poor sensitivity or a phantom model, respectively) that reduce the generalizability of their findings.

Open in a separate window

Figure 1:

Hypothetical graph shows polyp sizes measured at optical colonoscopy (OC), CT colonography (CTC), and pathologic evaluation compared with “true” (in vivo) size, estimated from analysis of literature. Although the literature is conflicting, CT colonographic measurements tend to lie between those of optical colonoscopy and pathologic evaluation and may be closest to “true” size as reflected in the graph. The differences, about ±2 mm, are comparable to and potentially sum with other measurement errors, such as observer variability and method of measurement (eg, 2D vs 3D CT colonography). Horizontal thick lines show lower boundaries of C-RADS C2 (6–9 mm) and C3 (10 mm and larger) categories according to the CT colonographic polyp size measurement. If C-RADS size boundaries were applied to optical colonoscopic or pathologic measurements, the boundaries would need to shift up or down, respectively, to keep the “true” polyp size constant. The C-RADS guidelines were determined to some extent by combining knowledge of size-based polyp risk gained from the optical colonoscopy and pathology literature with the ability of CT colonography to help identify polyps of clinically relevant size.

The literature regarding polyp size and risk of carcinoma refers to sizes from pathologic evaluation and colonoscopy. Hence, to align clinical management decisions at CT colonography with risk-based guidelines published in the existing literature, one must adjust the size threshold above which polyps are thought to convey greater risk—that is, polyps reaching the 1 cm threshold or larger at colonoscopy are likely to measure 8 mm or larger at CT colonography. However, to align polyp management decisions with sizes based on pathologic measurements, then polyps 1 cm or larger as measured at pathologic examination allow a higher threshold at colonography to be set, such as 1.2 cm. Size differences of a few millimeters can matter clinically when a polyp is misclassified into an incorrect size category, as will be described in this article. Since CT colonographic measurements tend to lie intermediate to the pathologic and colonoscopic measurements and may be more reliable than both of these, CT colonographic measurements may represent an ideal compromise size.

Overview of Factors That Affect Polyp Size Measurements at CT Colonography

Polyp size measurements can be affected by factors during acquisition and display of the images as well as during the process of obtaining the measurements. Acquisition factors include spatial resolution and partial volume averaging, motion, and image noise. Display factors include the use of two-dimensional (2D) multiplanar versus three-dimensional (3D) endoluminal images, the choice of rendering thresholds and window settings, the quality of the colon segmentation, and the effects of high-attenuation endoluminal contrast agents. Measurement factors include observer variability and choice of measurement tools. This article will examine each of these areas in turn.

Spatial Resolution

Partial volume effects due to a collimation or reconstruction interval that is too thick can lead to inaccurate polyp measurements, particularly of smaller polyps. Polyp morphology is best depicted when the effective section thickness is less than that of the polyp size of interest (35). Distortion of the polyp morphology is expected to lead to inaccurate size measurements. With the widespread use of collimations less than 2.5 mm and reconstruction intervals of 1.25 mm or less for CT colonography, partial volume averaging is likely to be less important for the 6 mm and larger polyp sizes of interest.

Motion Artifact

Bowel motion and bulk patient motion could produce artifacts that distort polyps, decreasing their detectability and adversely affecting polyp size measurement. With multidetector helical CT colonography, scanning of the abdomen and pelvis can be accomplished in a single breath hold, eliminating most respiratory and bulk motion artifacts. Because peristalsis is sufficiently intermittent or of low amplitude on the time scale of the scan duration, motion artifacts due to peristalsis are relatively rare (36). The value of spasmolytics has been controversial and mainly advocated to improve patient comfort or improve distention rather than to diminish motion artifacts (37).

Radiation Dose

Young et al (38) in a colon phantom evaluated CT colonographic data sets obtained at two radiation doses (tube currents of 5 mAs and 65 mAs). The most accurate polyp size measurements were workstation-dependent but for two of the three workstations, size measurements on 2D images were most accurate at both doses. Error values for polyp sizes were comparable for the higher dose and very-low-dose images.

Fletcher et al (39) evaluated a colon phantom by using four different techniques (12–100-mA tube current, 0.5-second tube rotation time) with different radiation doses. There was no significant association between measurement inaccuracy and dose setting either between the two reviewers or between manual and automated measurements. Similarly, in a colon phantom, Bielen et al (40) found maximum automated measurement differences that were comparable at nine combinations of kilovoltage and milliamperage (100–140 kV, 10–55 mAs). Similar results were also reported by Blake et al (41) at 30 and 100 mAs.

These data indicate that, at least in phantoms, polyp size measurement is not compromised by the use of low-dose CT colonography.

Rendering Threshold and Window Settings

When polyp size measurements are made on 3D endoluminal surface reconstructions from CT colonographic examinations, the thresholds used to produce the renderings can affect the measurements. For example, as the threshold is lowered toward that of air in the colonic lumen, the rendered surface moves toward the lumen; alternatively, as the threshold is raised toward that of soft tissue, the rendered surface moves in the direction of the colonic wall. In either situation, the shape of a polyp can change as the threshold is varied; in some cases the polyp can disappear altogether. Similar observations have been made for polypoid airway lesions at CT bronchography (42).

Polyp size measurements are highly sensitive to such threshold changes. Park et al (43) found that polyp size measurements in the pig colon model were most accurately measured when the surface rendering threshold was set at approximately −500 HU, midway between air and water CT attenuation. The mean measurement difference between the observed and reference sizes increased to 0.86 mm for a rendering threshold of −800 HU.

Similarly, when polyp size measurements are made on 2D CT colonographic images, window width and level settings affect size measurements. For example, Young et al (38) found that lung window settings gave more accurate size measurements than soft-tissue window settings. Soft-tissue window settings gave measurements that were on average 1.5–3 mm lower than the true size. The window levels for lung window settings used by two of the workstations in the study (−400 and −600 HU) were closer to the midpoint between air and colonic wall CT attenuations (which is the proper window level for accurate size measurement) while the window levels for soft-tissue window settings (20 and 40 HU) were closer to that of the colonic wall CT attenuation, thereby distorting the polyp shape. Of note, the authors found that 3D measurements were better than 2D measurements for one workstation; while the authors did not provide an explanation for this finding, the window level for lung windows for this workstation was −200 HU, too high to give accurate measurements by using 2D images. These results are consistent with the findings of Park et al in the preceding paragraph.

To summarize, these data indicate that surface rendering thresholds of −500 HU are best for 3D polyp measurement and lung window settings with a window level at or near −500 HU are best for 2D measurement. While window settings are commonly reported in the literature, the window level is often set too high to provide accurate size measurements. It is also notable that the rendering threshold is rarely reported in CT colonography research studies.

Fecal and Fluid Tagging

In the preceding section, a rendering threshold or window level midway between colonic air and colonic wall yielded most accurate polyp size measurements in the absence of intraluminal contrast agents. However, when intraluminal contrast agents (administered for fluid and/or fecal tagging) abut a polyp, the rendering threshold needs to be midway between the CT attenuation of the contrast material and the polyp to avoid distortions. In these situations the window level setting needs to be adjusted upward to accommodate the shift in CT values.

In an assessment of the effect of endoluminal contrast material on polyp size estimation using an anthropomorphic colon phantom, Zalis et al (44) found that when the CT attenuation of the bowel contrast material was in the range of 300–500 HU and electronic cleansing was used, there was no substantial change in readers’ estimations of polyp size. When the CT attenuation of the oral contrast agent was 560 HU or greater, streak artifacts occurred and polyp size measurement inaccuracies of 1–2 mm were recorded.

In a pig phantom, polyp size measurements were affected by the CT attenuation of the endoluminal contrast material (45). Slater et al found that when the CT attenuation was high, bone viewing windows gave the most accurate size measurements.

Another phenomenon that can affect polyp size measurements when intraluminal contrast agents are present is called pseudoenhancement (46). In pseudoenhancement, high CT attenuation adjacent to a polyp causes an artificial elevation of the CT values within the polyp, leading to polyp size measurement inaccuracies. The artificialvation in CT attenuation is due to x-ray scattering and beam hardening that is incompletely corrected by CT image reconstruction algorithms (47–49).

These results indicate that care must be taken in making polyp size measurements in the presence of high-attenuation contrast material used for fluid and fecal tagging. A pseudoenhancement or scatter correction may improve estimation of polyp size in this setting (46,50).

Two- versus Three-dimensional CT

There are conflicting results regarding whether measurement on 2D or 3D images is the best way to assess polyp size at CT colonography.

Pickhardt et al (51) studied a colonic phantom and in vivo clinical CT colonographic images. They used a calibrated linear probe at optical colonoscopy as the reference standard. They made measurements by using 2D multiplanar and 3D endoluminal displays. They found that the 2D measurements (made with a window width of 2000 HU and level of 0 HU) always underestimated actual polyp sizes. The mean error in polyp size was 3 mm using an optimized 2D view (the largest of the three multiplanar views). The error in the 3D size measurement was smaller, 1.9 mm on average. As the authors observed, the error for the 2D measurements could be reduced by lowering the window level. Similar to the study by Young et al (38), which used a different workstation, the window level for the workstation used by Pickhardt et al was too high and explains why the 2D measurements were less accurate than the 3D measurements.

In a pig colon model in which simulated polyps were constructed by puckering the mucosa of an inverted colon and securing it with sutures, Park et al (34) found that 2D measurements on optimized multiplanar reformatted images were more accurate than 3D measurements on 3D endoluminal images. However, optimizing the 2D plane was a cumbersome task and not generally used clinically. The 3D measurements were more accurate than measurements using standard orthogonal 2D planes. The authors concluded that 3D measurements may be the preferred technique.

Punwani et al (52) found that 2D measurements consistently overestimated polyp diameter but had the best repeatability and least effect of observer experience. In agreement with some investigators (26,27), they found that optical colonoscopy measurements underestimated the true polyp size. An important limitation was that the measurements were made in a wooden phantom with attenuation of −370 HU. Earlier work from the same research group (53) found better inter- and intraobserver variability with 2D compared with 3D surface-rendered endoluminal measurements in the presence of suboptimal distension. In both studies, 3D polyp measurement was more accurate than 2D measurement.

Taylor et al (54) found that manual 3D surface-rendered endoluminal and automated measurements were more accurate than 2D measurements. The experiments were performed in a fresh and unfixed human colectomy specimen following subtotal colectomy from a patient with familial adenomatous polyposis. Similar results were obtained by Yeshwant et al (29), who found that manual 3D endoluminal polyp size measurements best approximated true size as determined at optical colonoscopy with a calibrated linear probe.

Two studies showed results contradictory to those just described. Burling et al (55) found that 3D surface-rendered endoluminal polyp measurement led to large errors in estimated polyp size compared with 2D displays. The authors speculated that their results may have been specific to a particular viewing workstation. Jeong et al (56) found that 2D measurements had lower errors than did 3D measurements compared with colonoscopy.

Young et al (38) found that the best measurement method (2D versus 3D) depended on the particular workstation. For the three evaluated workstations, a different method (2D axial, 2D multiplanar, and 3D cube) worked best for each. As shown earlier, the 2D window level settings were inappropriately high for one of the workstations.

The data described above indicate a dependence on the particular workstation and practical issues in optimizing the best 2D measurement plane. In general, measurements at CT colonography tend to approximate those at colonoscopy best when measured on 3D.

Intra- and Interobserver Variability

Variability in polyp size measurement at CT colonography is potentially problematic because clinical management of patients depends in part on the C-RADS categories, which are defined by 6- and 10-mm size thresholds (4). Hence, variations in size measurements of just 1 or 2 mm could cause a polyp to be in either the small (6-9 mm) or large (>10 mm) size category, leading to either a C-RADS 2 or 3 diagnosis. This phenomenon does in fact occur for example, in one study polyp size miscategorization was reported to occur in 19% of measurements made by 28 observers (30).

Burling et al (57) found considerable interobserver variability in CT colonographic measurements made by two experienced observers (a radiologist and a radiographer) either manually or by using automated software. The 95% limits of agreement for the two readers’ measurements spanned a range of approximately ±4 mm. Reader experience is also likely to play an important role in achieving optimal measurement accuracy (55).

Yeshwant et al (29) found that polyp linear and volume measurements demonstrated substantial intra- and interobserver variability. The variabilities were least for measurements made on the 2D images, but even then the 95% limits of agreement spanned ranges of approximately 40%–60%.

These data indicate that, as described above for colonoscopic and pathologic measurements, there is considerable variability in polyp sizes measured at CT colonography by radiologists and other trained observers. Such variability can lead to C-RADS misclassifications in a substantial number of patients.

Intrastudy Variability

At surveillance or subsequent screening CT colonographic examinations, it will be important to assess for change in size of polyps found at earlier CT colonographic examinations. If the “true” polyp size at follow-up CT colonography is larger than at baseline CT colonography, the polyp has grown. However, the assessment of polyp size change is complicated by the fact that both the baseline and follow-up measured polyp sizes have associated uncertainties or errors for the reasons described above (eg, observer variability or factors associated with the image acquisition or display). When computing the size change or difference, these errors propagate such that the uncertainty or error in the polyp size change is usually greater than the uncertainties of the individual baseline and follow-up polyp sizes due to compounding (58). For example, if single polyp size measurements have uncertainties (standard deviations) of ±2 mm, the uncertainty in the difference of two polyp size measurements can be estimated to be ±2√2 mm or approximately ±3 mm. It is easy to see how such errors could lead to inappropriate clinical management such as recommendations for biopsy of polyps that have not actually grown.

Yeshwant et al (29) analyzed polyp size differences between the supine and prone scans from a CT colonographic examination. The authors expected the measured size difference between the two scans to be zero, since polyps should not grow or shrink over the course of a CT colonographic examination that takes only several minutes to complete. They found the least error in size difference estimation (≤2 mm) when they converted the manually assessed polyp volume to an equivalent diameter. Manual estimates of linear size obtained on the 3D endoluminal reconstructions came in a close second, with an error of less than or equal to 3 mm.

Automated Measurements

Automated size measurement might be expected to provide more accurate sizes than those obtained manually. A number of such tools currently exist and are being evaluated.

In a glass colon model it was found that, by using a commercial automated polyp measurement tool, accuracy of polyp size measurements depended on polyp size, morphology, and location with respect to the tip of haustral folds (39). Size measurements were less accurate for 5-mm polyps, flat polyps, and polyps on haustral fold tips. Automated polyp measurements were more precise (reproducible) than were manual measurements. Increased precision of automated polyp size measurement has been confirmed by van Wijk et al (59) and Bielen et al (40), who performed similar experiments.

In the pig colon model, automated measurements were most accurate for well-circumscribed smooth rounded polyps, but errors occurred especially in large, lobulated, or irregularly shaped polyps or polyps located next to a bulbous haustral fold (43).

In a Plexiglas phantom, Blake et al (41) used automated software to measure polyp volume. The software was highly accurate, with errors in volume of only 3%–5%. Measurements were not adversely affected by either the orientation of the phantom or the tube current settings.

Jeong et al (56) found that automated measurements eliminated both intra- and interobserver variability. Burling et al (60), with in vitro experiments, found intra- and interobserver agreement to be superior by using automated measurements compared with manual measurements. In another study by the same research group (57) on real polyps, however, automated measurements demonstrated inaccuracies that were similar to those of manual measurements.

From these experiments, it can be concluded that, at least in phantoms, automated polyp size measurements are highly reproducible and accurate and may reduce or eliminate observer variability. The automated measurement software tends to report the same polyp size regardless of where the radiologist clicks on the polyp to initiate the automated measurement. This is because the software typically identifies the edges of the polyp, which reduces the importance of the location of the radiologist’s initial click as long as the click is inside the polyp. However, problems in measurement can occur for polyps that are either flat, irregularly shaped, or on the tips of haustral folds. There may be a benefit to automated polyp volume measurement if its accuracy is confirmed in the clinical setting.

Summary of Recommendations for Polyp Size Measurement at CT Colonography

To summarize, polyp size measurements may be made on either 2D or 3D images. If 2D images are used, the window level should be approximately −500 HU. If 3D images are used, the rendering threshold should be approximately −500 HU. In general, 3D measurements are preferable, although 2D measurements may be better if the colon is suboptimally distended. If high-attenuation oral contrast material abuts the polyp, then one should either raise the window level or use interpretation software that automatically corrects the CT attenuation of the images. Size increases of more than 2 mm are required to confirm polyp growth. Automated polyp size measurement is not yet adequately proved in the clinic and should not substitute for careful manual measurements. The importance of observer experience in making size measurements is unclear; minimal training to avoid basic pitfalls may be of value.

For flat or sessile lesions, one should measure the greatest diameter. For pedunculated lesions, the greatest diameter of the polyp head should be measured; the measurement should not include the polyp stalk (4).

Which Sizing Method Is Most Accurate?

So given the differences in sizes and variabilities in the different measurement methods, which sizing method should be used (optical colonoscopic, CT colonographic, or pathologic measurement)? If the true size is the size in vivo, a synthesis of the literature suggests that CT colonographic measurements are the most accurate sizes. However, the literature on clinical outcomes is based almost entirely on colonoscopic and pathologic size measurements. Does this mean that the literature on clinical outcomes needs to be updated using CT colonographic measurements? Or should we use our knowledge of the relationship between colonoscopic, pathologic, and CT colonographic measurements to adjust the results from the colonoscopy and pathology literature to guide our management of polyps of different sizes at CT colonography? Given the time it takes to perform these studies, the best solution is to do both. It may help that CT colonographic measurements tend to lie between colonoscopic and pathologic measurements.

What Size Thresholds Should Be Chosen for Intervention and Surveillance?

Given the imprecision and inaccuracy in measured polyp size, how should one choose the size thresholds for intervention and surveillance? The evidence indicates that optical colonoscopy tends to overestimate polyp size by 1–2 mm on average, relative to the size in vivo, and that observer variability adds another 2 mm or so of imprecision. Does this mean that we must remove all 6 mm and larger polyps as measured at CT colonography to ensure removal of all polyps measuring 1 cm or more at optical colonoscopy? Does the extra 2–4-mm margin of error justify the marked increase in number of polypectomies that must be performed? For example, the American College of Radiology Imaging Network study found that the referral rate for colonoscopy would be 12% if using a 6-mm size threshold but would increase to 17% for a 5-mm size threshold, with little improvement in sensitivity (61). The central question is how to choose a practical polyp management scheme that balances the costs and risks of polypectomy against the benefits of removing a lesion with increased risk of cancer and advanced histology (Fig 2). A cost-effectiveness analysis (62) based on size threshold showed a substantial cost savings and marked reduction in colonoscopy-related complications if diminutive lesions (<6 mm) were not reported as recommended in C-RADS. The concept of a size threshold for intervention that is based on CT colonographic findings is in marked distinction to the gastroenterology practice of removing all polyps regardless of size, a pragmatic response in part motivated by the difficulty of accurate polyp size measurement at optical colonoscopy and the inability of optical colonoscopy to monitor polyps without an additional invasive procedure.

Open in a separate window

Figure 2:

Graph of risks and referral rates as a function of polyp size. Risks of high grade dysplasia (HGD) in all polyps and cancer in adenomatous or all polyps are from references (2,7,11). For comparison, test-positive rates for CT colonography are shown as reported in references (10,61). Patients with positive results at CT colonography would be referred for colonoscopy (OC). Risks are estimated from graphs or tables in the cited articles and modified to fit the format of this graph; they should be viewed as approximate. Lower risks of cancer and high-grade dysplasia by Lieberman et al are attributable to differences in patient population and denominators used for calculating risk. Note that small shifts (for example, ±2 mm) in size threshold (along x-axis) greatly affect referral rates (because the referral rate curves have steep slopes) but affect risk relatively less (because the risk curves have shallower slopes).

The best size threshold also depends on polyp growth rate and the surveillance interval. By “surveillance,” I refer here to repeat testing of patients whose polyps are being closely monitored in situ; the term surveillance has also been applied to the monitoring of patients who have had polyps removed and who are at increased risk of developing recurrent polyps. Since the optimal surveillance interval depends on polyp growth rate, the growth rate is a key variable. Another important variable affecting the surveillance interval is the polyp size at which invasion through the colonic wall is likely to occur. This size and the growth rate help determine the width of the therapeutic window after which the risk of metastasis may increase dramatically. Lead-time bias is also important (63). Detecting and removing polyps earlier, when they are smaller, is not necessarily beneficial (and may represent overdiagnosis). This is a well-recognized problem in other types of early cancer detection with imaging, such as in lung and breast screening (64,65). Mathematical models may be helpful for formulating different strategies for determining the optimal surveillance interval (66). However, without knowledge of the growth rate, the best size threshold and surveillance interval is speculative.

The proposal to watch polyps 6–9 mm with surveillance CT colonography in 1–3 years to assess for polyp growth is incorporated in C-RADS and has been advocated by some (67). It should be noted that the concept of polyp surveillance is controversial and is currently being evaluated only in research settings. A consensus of the American Cancer Society, the U.S. Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology currently recommends immediate polypectomy for all polyps 6 mm or larger found at CT colonography (5).

The use of size alone as a biomarker limits our ability to manage polyps. Other important factors are the presence of advanced histology: high-grade dysplasia and villous histology. These additional factors are not identifiable at CT colonography, although villous polyps are known to display more frequently a thin layer or drop of adherent contrast material (68). In the absence of knowledge about a polyp’s histology at CT colonography, current clinical practice is to remove all polyps above a size threshold, the majority of which have little or no malignant potential. Clearly, the use of size alone as a biomarker is inadequate to determine the risk of a polyp. However, unless we can develop new imaging methods to diagnose advanced histologic features, one must rely on size-based strategies to remove larger polyps and watch or ignore smaller polyps. Polyp-specific optical contrast agents are under development; their applicability to CT colonography is currently unknown (69,70). The development of such histologic markers should be a high priority.

While one rational CT colonography surveillance strategy based on polyp size is under evaluation, in which small polyps are observed rather than removed, consensus will be difficult to achieve until more data are available (71). One problem is that colon cancers are rare, so large numbers of patients must be screened to show a statistically significant reduction in cancer rates. Until researchers can compare the incidence of colon cancers after negative CT colonography or optical colonoscopy and different screening intervals, it will be difficult to determine the appropriate surveillance strategy. In one study, the use of CT colonography actually found more cancers than at optical colonoscopy and in another, the use of optical colonoscopy was ineffective in preventing proximal colon cancers (72,73). It may be that CT colonography will become the primary colorectal cancer screening test, so that issues of size thresholds based on CT colonographic measurements will take on increasing importance.

Another problem is the error in measurement of polyp size. To properly implement a surveillance strategy, one must determine which polyps are growing. Polyp size changes of 1 or 2 mm at serial examinations could simply represent measurement error and observer variability. It may not be appropriate for a polyp that “grows” 2 mm to be referred for polypectomy. Note that this error spans nearly half the range of the 6–9-mm small polyp category.

Surveillance (polyp monitoring, or watchful waiting) adds to costs, patient anxiety, and radiation exposure. A potential solution to these problems is to forego the idea of polyp surveillance altogether. Instead, the CT colonography size threshold for polypectomy could be set to 8 mm, to add a safety buffer and account for discrepancies between the optical colonoscopic and CT colonographic size measurements. Rather than report all polyps 6 mm or larger as recommended by C-RADS, the threshold of reporting could be raised to be 8 mm. By harmonizing the thresholds for polypectomy and reporting and eliminating the polyp surveillance pathway, we could reduce confusion (both for patients and referring physicians), simplify management, and reduce costs. It is not clear that any benefit is to be gained by reporting, monitoring, or removing 6–7-mm polyps. This idea leads to a crucial recommendation: CT colonography researchers should stop considering the small 6–9-mm polyps as a single entity or group and should start teasing apart the importance of polyps of different sizes within this size category. There are vastly more 6–7-mm polyps than 8–9-mm polyps. The risk-to-benefit ratio of reporting and removing 6–7-mm polyps may be lower than we think. For example, in the study by Lieberman et al (11), there was little change in prevalence of advanced neoplasia in polyps smaller than 6 mm when 6-mm polyps were downsized to 5 mm in a sensitivity analysis to study the effect of misclassification of polyp size (an increase from 1.7% to 2.0%). Radiologists should be prepared for resistance from gastroenterologists to the concept of further raising the size threshold (74). Gastroenterologists will point to clinical data suggesting this approach is risky. The truth is that the “riskiness” of this approach is not firmly established and data supportive of a higher threshold exist in the literature (75,76). This compromise position or concept needs to undergo debate and hopefully will spur additional data collection to support or refute it.

Is greater accuracy of polyp size measurement needed, and is it feasible? Given the difficulty in getting a reference standard for the true size, except in phantoms, it is difficult to know how to do this. Nevertheless, proof of accuracy of size measurements for more difficult-to-measure polyps is needed, such as for flat and pedunculated polyps and polyps on folds, under tagged fluid or stool, or at air-fluid levels. More evidence on the effect of image resolution on polyp size measurement accuracy is also needed.

More data on the accuracy of automated and manual size measurements for assessing polyp growth is also needed. Guidelines are needed for determining clinical management for size changes according to the baseline size of the polyp and the interval between examinations.

In conclusion, until the development of biomarkers indicative of histology, polyp size will continue to be the most important biomarker for determining management of colonic polyps found at screening and diagnostic CT colonography. It is important to understand the precision and accuracy of polyp size measurements. Ultimately, however, the goal is to determine a feasible scheme for managing polyps of different measured sizes, particularly those numerous polyps measuring 6–9 mm for which the best management strategy is still evolving.

Essentials

• Polyp size measurement at CT colonography, optical colonoscopy, and pathologic evaluation has been studied extensively, with sometimes contradictory results.

• Polyp size measurements at CT colonography tend to lie between those made at colonoscopy and pathologic evaluation (1–2 mm smaller than colonoscopic and 1–2 mm larger than pathologic measurements).

• Polyp size measurements at CT colonography are subject to errors and variability on the order of 2–3 mm, large enough to affect clinical management; radiologists need to consider these errors in their decision-making process.

• Best practices for polyp size measurement at CT colonography include the use of three-dimensional endoluminal displays, two-dimensional displays with a window level near −500 HU, and automated measurement software.

• An important research goal is to ascertain the size threshold that best balances the referral rate for colonoscopy versus the frequency of colorectal cancer following CT colonographic screening; the author proposes a threshold of 8 mm.

Acknowledgments

The author thanks Andrew Dwyer, MD, for critical review of the manuscript and helpful discussions.

Received May 19, 2009; revision requested June 26; revision received July 29; accepted October 1; final version accepted October 15.

The author has pending and/or awarded patents and receives royalty income from iCAD Medical. His laboratory received free research software from Viatronix and receives research support from iCAD Medical.

Funding: This research was supported by the intramural research program of the National Institutes of Health Clinical Center (project Z01 CL040003).

Abbreviations:

C-RADSCT Colonography Reporting and Data System3Dthree-dimensional2Dtwo-dimensional

References

1. Einspahr JG, Alberts DS, Gapstur SM, Bostick RM, Emerson SS, Gerner EW. Surrogate end-point biomarkers as measures of colon cancer risk and their use in cancer chemoprevention trials. Cancer Epidemiol Biomarkers Prev 1997;6(1):37–48 [PubMed] [Google Scholar]

2. Muto T, Bussey HJ, Morson BC. The evolution of cancer of the colon and rectum. Cancer 1975;36(6):2251–2270 [PubMed] [Google Scholar]

3. Grinnell RS, Lane N. Benign and malignant adenomatous polyps and papillary adenomas of the colon and rectum; an analysis of 1,856 tumors in 1,335 patients. Int Abstr Surg 1958;106(6):519–538 [PubMed] [Google Scholar]

4. Zalis ME, Barish MA, Choi JR, et al. CT colonography reporting and data system: a consensus proposal. Radiology 2005;236(1):3–9 [PubMed] [Google Scholar]

5. Levin B, Lieberman DA, McFarland B, et al. Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. CA Cancer J Clin 2008;58(3):130–160 [PubMed] [Google Scholar]

6. Vogelstein B, Fearon ER, Hamilton SR, et al. Genetic alterations during colorectal-tumor development. N Engl J Med 1988;319(9):525–532 [PubMed] [Google Scholar]

7. O’Brien MJ, Winawer SJ, Zauber AG, et al. The National Polyp Study. Patient and polyp characteristics associated with high-grade dysplasia in colorectal adenomas. Gastroenterology 1990;98(2):371–379 [PubMed] [Google Scholar]

8. Stryker SJ, Wolff BG, Culp CE, Libbe SD, Ilstrup DM, MacCarty RL. Natural history of untreated colonic polyps. Gastroenterology 1987;93(5):1009–1013 [PubMed] [Google Scholar]

9. Atkin WS, Morson BC, Cuzick J. Long-term risk of colorectal cancer after excision of rectosigmoid adenomas. N Engl J Med 1992;326(10):658–662 [PubMed] [Google Scholar]

10. 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(23):2191–2200 [PubMed] [Google Scholar]

11. Lieberman D, Moravec M, Holub J, Michaels L, Eisen G. Polyp size and advanced histology in patients undergoing colonoscopy screening: implications for CT colonography. Gastroenterology 2008;135(4):1100–1105 [PMC free article] [PubMed] [Google Scholar]

12. Laiyemo AO, Murphy G, Sansbury LB, et al. Hyperplastic polyps and the risk of adenoma recurrence in the polyp prevention trial. Clin Gastroenterol Hepatol 2009;7(2):192–197 [PMC free article] [PubMed] [Google Scholar]

13. O’Brien MJ. Hyperplastic and serrated polyps of the colorectum. Gastroenterol Clin North Am 2007;36(4):947–968, viii [PubMed] [Google Scholar]

14. East JE, Saunders BP, Jass JR. Sporadic and syndromic hyperplastic polyps and serrated adenomas of the colon: classification, molecular genetics, natural history, and clinical management. Gastroenterol Clin North Am 2008;37(1):25–46, v [PubMed] [Google Scholar]

15. Summers RM, Frentz SM, Liu J, et al. Conspicuity of colorectal polyps at CT colonography: visual assessment, CAD performance, and the important role of polyp height. Acad Radiol 2009;16(1):4–14 [PMC free article] [PubMed] [Google Scholar]

16. Summers RM, Liu J, Yao J, Brown L, Choi JR, Pickhardt PJ. Automated measurement of colorectal polyp height at CT colonography: hyperplastic polyps are flatter than adenomatous polyps. AJR Am J Roentgenol 2009;193(5):1305–1310 [PMC free article] [PubMed] [Google Scholar]

17. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997;112(2):594–642 [PubMed] [Google Scholar]

18. Pickhardt PJ. The natural history of colorectal polyps and masses: rediscovered truths from the barium enema era. AJR Am J Roentgenol 2007;188(3):619–621 [PubMed] [Google Scholar]

19. Welin S, Youker J, Spratt JS., Jr The rates and patterns of growth of 375 tumors of the large intestine and rectum observed serially by double contrast enema study (Malmoe technique). Am J Roentgenol Radium Ther Nucl Med 1963;90:673–687 [PubMed] [Google Scholar]

20. 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(3):449–456 [PMC free article] [PubMed] [Google Scholar]

21. Loeve F, Boer R, Zauber AG, et al. National Polyp Study data: evidence for regression of adenomas. Int J Cancer 2004;111(4):633–639 [PubMed] [Google Scholar]

22. Asano TK, McLeod RS. Non steroidal anti-inflammatory drugs (NSAID) and Aspirin for preventing colorectal adenomas and carcinomas. Cochrane Database Syst Rev 2004;(2):CD004079. [PMC free article] [PubMed] [Google Scholar]

23. Morales TG, Sampliner RE, Garewal HS, Fennerty MB, Aickin M. The difference in colon polyp size before and after removal. Gastrointest Endosc 1996;43(1):25–28 [PubMed] [Google Scholar]

24. Schoen RE, Gerber LD, Margulies C. The pathologic measurement of polyp size is preferable to the endoscopic estimate. Gastrointest Endosc 1997;46(6):492–496 [PubMed] [Google Scholar]

25. Gopalswamy N, Shenoy VN, Choudhry U, et al. Is in vivo measurement of size of polyps during colonoscopy accurate?. Gastrointest Endosc 1997;46(6):497–502 [PubMed] [Google Scholar]

26. Fennerty MB, Davidson J, Emerson SS, Sampliner RE, Hixson LJ, Garewal HS. Are endoscopic measurements of colonic polyps reliable?. Am J Gastroenterol 1993;88(4):496–500 [PubMed] [Google Scholar]

27. Schwartz E, Catalano MF, Krevsky B. Endoscopic estimation of size: improved accuracy by directed teaching. Gastrointest Endosc 1995;42(4):292–295 [PubMed] [Google Scholar]

28. Rubio CA, Grimelius L, Lindholm J, et al. Reliability of the reported size of removed colorectal polyps. Anticancer Res 2006;26(6C):4895–4899 [PubMed] [Google Scholar]

29. Yeshwant SC, Summers RM, Yao JH, Brickman DS, Choi JR, Pickhardt PJ. Polyps: linear and volumetric measurement at CT colonography. Radiology 2006;241(3):802–811 [PubMed] [Google Scholar]

30. Burling D, Halligan S, Altman DG, et al. Polyp measurement and size categorisation by CT colonography: effect of observer experience in a multi-centre setting. Eur Radiol 2006;16(8):1737–1744 [PubMed] [Google Scholar]

31. Chen MH, Blevins M, Herman BA, Johnson CD. National CT Colonography Trial (ACRIN 6664): are CT colonography, colonoscopy and pathology giving the same measures for polyp size? [abstr].. In: Radiological Society of North America Scientific Assembly and Annual Meeting Program Oak Brook, Ill: Radiological Society of North America, 2008; 279 [Google Scholar]

32. Cotton PB, Durkalski VL, Pineau BC, et al. Computed tomographic colonography (virtual colonoscopy): a multicenter comparison with standard colonoscopy for detection of colorectal neoplasia. JAMA 2004;291(14):1713–1719 [PubMed] [Google Scholar]

33. Gupta S, Durkalski V, Cotton P, Rockey DC. Variation of agreement in polyp size measurement between computed tomographic colonography and pathology assessment: clinical implications. Clin Gastroenterol Hepatol 2008;6(2):220–227 [PMC free article] [PubMed] [Google Scholar]

34. Park SH, Choi EK, Lee SS, et al. Polyp measurement reliability, accuracy, and discrepancy: optical colonoscopy versus CT colonography with pig colonic specimens. Radiology 2007;244(1):157–164 [PubMed] [Google Scholar]

35. Whiting BR, McFarland EG, Brink JA. Influence of image acquisition parameters on CT artifacts and polyp depiction in spiral CT colonography: in vitro evaluation. Radiology 2000;217(1):165–172 [PubMed] [Google Scholar]

36. Smith FW, Sleisenger MH. Physiology of the colon. In: Sleisenger MH, Fordtran JS. eds. Gastrointestinal disease: pathophysiology, diagnosis, management 2nd ed.Philadelphia, Pa: Saunders, 1978; 1523–1548 [Google Scholar]

37. Taylor SA, Halligan S, Goh V, et al. Optimizing colonic distention for multi-detector row CT colonography: effect of hyoscine butylbromide and rectal balloon catheter. Radiology 2003;229(1):99–108 [PubMed] [Google Scholar]

38. Young BM, Fletcher JG, Paulsen SR, et al. Polyp measurement with CT colonography: multiple-reader, multiple-workstation comparison. AJR Am J Roentgenol 2007;188(1):122–129 [PubMed] [Google Scholar]

39. Fletcher JG, Booya F, Melton Z, et al. Automated polyp measurement with CT colonography: preliminary observations in a phantom colon model. AJR Am J Roentgenol 2007;188(4):945–952 [PubMed] [Google Scholar]

40. Bielen D, Vanslembrouck R, De Keyzer F, et al. Computed tomographic colonography: automated tool for polyp measurement delivering on patient risk stratification. J Comput Assist Tomogr 2008;32(4):504–510 [PubMed] [Google Scholar]

41. Blake ME, Soto JA, Hayes RA, Ferrucci JT. Automated volumetry at CT colonography: a phantom study. Acad Radiol 2005;12(5):608–613 [PubMed] [Google Scholar]

42. Summers RM, Shaw DJ, Shelhamer JH. CT virtual bronchoscopy of simulated endobronchial lesions: effect of scanning, reconstruction, and display settings and potential pitfalls. AJR Am J Roentgenol 1998;170(4):947–950 [PubMed] [Google Scholar]

43. Park SH, Choi EK, Lee SS, et al. Linear polyp measurement at CT colonography: 3D endoluminal measurement with optimized surface-rendering threshold value and automated measurement. Radiology 2008;246(1):157–167 [PubMed] [Google Scholar]

44. Zalis ME, Perumpillichira JJ, Kim JY, Del Frate C, Magee C, Hahn PF. Polyp size at CT colonography after electronic subtraction cleansing in an anthropomorphic colon phantom. Radiology 2005;236(1):118–124 [PubMed] [Google Scholar]

45. Slater A, Taylor SA, Burling D, Gartner L, Scarth J, Halligan S. Colonic polyps: effect of attenuation of tagged fluid and viewing window on conspicuity and measurement—in vitro experiment with porcine colonic specimen. Radiology 2006;240(1):101–109 [PubMed] [Google Scholar]

46. Näppi J, Yoshida H. Adaptive correction of the pseudo-enhancement of CT attenuation for fecal-tagging CT colonography. Med Image Anal 2008;12(4):413–426 [PMC free article] [PubMed] [Google Scholar]

47. Maki DD, Birnbaum BA, Chakraborty DP, Jacobs JE, Carvalho BM, Herman GT. Renal cyst pseudoenhancement: beam-hardening effects on CT numbers. Radiology 1999;213(2):468–472 [PubMed] [Google Scholar]

48. Rao PS, Alfidi RJ. The environmental density artifact: a beam-hardening effect in computed tomography. Radiology 1981;141(1):223–227 [PubMed] [Google Scholar]

49. Joseph PM, Spital RD. The effects of scatter in x-ray computed tomography. Med Phys 1982;9(4):464–472 [PubMed] [Google Scholar]

50. Liu J, Yao J, Summers RM. Scale-based scatter correction for computer-aided polyp detection in CT colonography. Med Phys 2008;35(12):5664–5671 [PMC free article] [PubMed] [Google Scholar]

51. Pickhardt PJ, Lee AD, McFarland EG, Taylor AJ. Linear polyp measurement at CT colonography: in vitro and in vivo comparison of two-dimensional and three-dimensional displays. Radiology 2005;236(3):872–878 [PubMed] [Google Scholar]

52. Punwani S, Halligan S, Irving P, et al. Measurement of colonic polyps by radiologists and endoscopists: who is most accurate?. Eur Radiol 2008;18(5):874–881 [PubMed] [Google Scholar]

53. Taylor S, Slater A, Honeyfield L, Burling D, Halligan S. CT colonography: effect of colonic distension on polyp measurement accuracy and agreement-in vitro study. Acad Radiol 2006;13(7):850–859 [PubMed] [Google Scholar]

54. Taylor SA, Slater A, Halligan S, et al. CT colonography: automated measurement of colonic polyps compared with manual techniques—human in vitro study. Radiology 2007;242(1):120–128 [PubMed] [Google Scholar]

55. Burling D, Halligan S, Taylor S, et al. Polyp measurement using CT colonography: agreement with colonoscopy and effect of viewing conditions on interobserver and intraobserver agreement. AJR Am J Roentgenol 2006;186(6):1597–1604 [PubMed] [Google Scholar]

56. Jeong JY, Kim MJ, Kim SS. Manual and automated polyp measurement comparison of CT colonography with optical colonoscopy. Acad Radiol 2008;15(2):231–239 [PubMed] [Google Scholar]

57. Burling D, Halligan S, Taylor SA, Honeyfield L, Roddie ME. CT colonography: automatic measurement of polyp diameter compared with manual assessment—an in-vivo study. Clin Radiol 2007;62(2):145–151 [PubMed] [Google Scholar]

58. Bevington PR, Robinson DK. Data reduction and error analysis for the physical sciences Boston, Mass: McGraw-Hill, 2003 [Google Scholar]

59. van Wijk C, Florie J, Nio CY, et al. Protrusion method for automated estimation of polyp size on CT colonography. AJR Am J Roentgenol 2008;190(5):1279–1285 [PubMed] [Google Scholar]

60. Burling D, Halligan S, Roddie ME, et al. Computed tomography colonography: automated diameter and volume measurement of colonic polyps compared with a manual technique—in vitro study. J Comput Assist Tomogr 2005;29(3):387–393 [PubMed] [Google Scholar]

61. Johnson CD, Chen MH, Toledano AY, et al. Accuracy of CT colonography for detection of large adenomas and cancers. N Engl J Med 2008;359(12):1207–1217 [PMC free article] [PubMed] [Google Scholar]

62. Pickhardt PJ, Hassan C, Laghi A, Zullo A, Kim DH, Morini S. Cost-effectiveness of colorectal cancer screening with computed tomography colonography: the impact of not reporting diminutive lesions. Cancer 2007;109(11):2213–2221 [PubMed] [Google Scholar]

63. Kramer BS, Croswell JM. Cancer screening: the clash of science and intuition. Annu Rev Med 2009;60:125–137 [PubMed] [Google Scholar]

64. Reich JM. A critical appraisal of overdiagnosis: estimates of its magnitude and implications for lung cancer screening. Thorax 2008;63(4):377–383 [PubMed] [Google Scholar]

65. Mahnken JD, Chan W, Freeman DH, Jr, Freeman JL. Reducing the effects of lead-time bias, length bias and over-detection in evaluating screening mammography: a censored bivariate data approach. Stat Methods Med Res 2008;17(6):643–663 [PubMed] [Google Scholar]

66. Eddy DM. A mathematical model for timing repeated medical tests. Med Decis Making 1983;3(1):45–62 [PubMed] [Google Scholar]

67. Pickhardt PJ, Taylor AJ, Kim DH, Reichelderfer M, Gopal DV, Pfau PR. Screening for colorectal neoplasia with CT colonography: initial experience from the 1st year of coverage by third-party payers. Radiology 2006;241(2):417–425 [PubMed] [Google Scholar]

68. O’Connor SD, Summers RM, Choi JR, Pickhardt PJ. Oral contrast adherence to polyps on CT colonography. J Comput Assist Tomogr 2006;30(1):51–57 [PubMed] [Google Scholar]

69. Hsiung PL, Hsiung PL, Hardy J, et al. Detection of colonic dysplasia in vivo using a targeted heptapeptide and confocal microendoscopy. Nat Med 2008;14(4):454–458 [PMC free article] [PubMed] [Google Scholar]

70. Roney CA, Jianwu X, Biying X, Jabour P, Griffiths G, Summers RM. Glycoprotein expression by adenomatous polyps of the colon. In: Hu XP, Clough AV, eds. Proceedings of SPIE: medical imaging 2008—physiology, function, and structure from medical images Vol 6916 Bellingham, Wash: SPIE–The International Society for Optical Engineering, 2008; 69161O. [Google Scholar]

71. Pickhardt PJ, Kim DH. CT colonography (virtual colonoscopy): a practical approach for population screening. Radiol Clin North Am 2007;45(2):361–375 [PubMed] [Google Scholar]

72. Kim DH, Pickhardt PJ, Taylor AJ, et al. CT colonography versus colonoscopy for the detection of advanced neoplasia. N Engl J Med 2007;357(14):1403–1412 [PubMed] [Google Scholar]

73. Baxter NN, Goldwasser MA, Paszat LF, Saskin R, Urbach DR, Rabeneck L. Association of colonoscopy and death from colorectal cancer. Ann Intern Med 2009;150(1):1–8 [PubMed] [Google Scholar]

74. Rex DK, Overhiser AJ, Chen SC, Cummings OW, Ulbright TM. Estimation of impact of American College of Radiology recommendations on CT colonography reporting for resection of high-risk adenoma findings. Am J Gastroenterol 2009;104(1):149–153 [PubMed] [Google Scholar]

75. Kisiel JB, Bruining DH, Fletcher JG, Limburg PJ. Computed tomographic colonography and high-risk adenomas: a gap in the guidelines?. Gastroenterology 2009;136(7):2395–2396; discussion 2396–2397 [PubMed] [Google Scholar]

76. Kim DH, Pickhardt PJ, Taylor AJ. Characteristics of advanced adenomas detected at CT colonographic screening: implications for appropriate polyp size thresholds for polypectomy versus surveillance

Can a polyp turn cancerous in 5 years?

How long do cancerous polyps take to develop?

Can colon cancer develop in 3 years?

Can a polyp turn into cancer in 2 years?