Accuracy

Judy Yee

Multiple studies have been performed to evaluate the sensitivity and specificity of computed tomographic colonography (CTC) for the detection of polyps and cancer in various types of patient cohorts. Most published studies have been carried out at a single center in symptomatic or high-risk patients (see Table 7.1). There are also several large multicenter trials that have been published.

Over the last 10 years, significant advances in image acquisition and image display capabilities have occurred and these still continue to evolve. In evaluating the results of performance studies it is important to note that specific study parameters may affect performance results. Variations in these parameters are likely the cause for some of the conflicting results in the published literature.

PATIENT RISK

CTC studies may include average-risk patients, high-risk patients, or both types of patients. Most of the published studies evaluating the performance of CTC have been performed in patients who are at increased risk for developing colorectal cancer. These are patients who may have a personal or family history of colorectal cancer or adenomatous polyp. They may have symptoms or signs of colorectal cancer such as iron-deficiency anemia, positive fecal occult blood test or hematochezia.

The use of CTC as a screening tool for colorectal carcinoma is optimally validated by testing performance in average-risk patients who have a low prevalence of disease (see Figure 7.1). Several studies have been published evaluating the performance of CTC in a screening type patient population.¹, ² and ³ The American College of Radiology Imaging Network (ACRIN) National CT Colonography Trial is the largest study evaluating the performance of CTC in a screening population to date.⁴

PATIENT PREPARATION

The results from most published studies examining the accuracy of CTC for polyp detection are based on patients who have undergone bowel cleansing using a wet preparation method with the ingestion of polyethylene glycol. Polyethylene glycol lavage solution is often the preferred agent by gastroenterologists and study patients typically must undergo both the CT examination and colonoscopy on the same day, often within several hours of each other. A large amount of residual fluid in the colon may cause decreased sensitivity for polyp detection on CTC. Saline cathartics such as sodium phosphate (phosphosoda) and magnesium citrate are osmotic agents that often leave the colon relatively dry. However, these “dry preparations” may leave more adherent solid stool, which can lead to an increased number of false positives.

The colon is distended by retrograde insufflation using either room air or carbon dioxide. Most published studies have included a protocol using air insufflation of the colon for CTC. However, more current published studies include the use of electronic insufflation of the colon with carbon dioxide as a method that provides optimal colonic distention more consistently. Earlier studies of CTC performance included the injection of glucagon as an antispasmodic agent. However, glucagon is no longer used routinely for CTC because it has not been found to be effective in improving colonic distention or lesion detection.⁵, ⁶ and ⁷ This also avoids additional cost and is less invasive to the patient. Hyoscine butylbromide (Buscopan) has been found to be a more effective spasmolytic agent and is often used in Europe to improve colonic distention for CTC.⁸,⁹

COMPUTED TOMOGRAPHY PROTOCOL

Earlier CTC performance studies employed single detector helical CT technology. Single detector computed tomography (SDCT) scan protocols required collimations of 3 mm or larger. Thicker collimation such as 5 mm or larger may make it more difficult to detect lesions, particularly smaller polyps. There are now multiple published studies
evaluating the sensitivity and specificity of CTC using multidetector row CT. Multidetector computed tomography (MDCT) allows for thinner collimation of 1.25 mm or smaller. This may lead to improved spatial resolution and increased sensitivity, particularly for smaller polyps and flat lesions. Additionally, narrower collimation may allow easier differentiation between a true polyp versus stool so that there are fewer false positives, leading to increased specificity. However, the use of thinner collimation includes drawbacks such as higher effective radiation dose to the patient, noisier images, and larger data sets for evaluation and storage.

TABLE 7.1 Polyp Detection: Single-Center Studies

			≥10 mm				6-9 mm
Study	Scan Parameters	No. of Patients and Type	No. of Polyps	Per-Polyp Sensitivity	Per-Patient Sensitivity	Per-Patient Specificity	No. of Polyps	Per-Polyp Sensitivity	Per-Patient Sensitivity	Per-Patient Specificity
Hara et al., 1997	SD/5/120/140	70 high risk	15	70%	75%	91%	21	63% (≥5 mm)	66% (≥5 mm)	63% (≥5 mm)
Dachman et al., 1998	SD/5/120/100	44 high risk	6	83% (>8 mm)	–	100%	3	33%	–	–
Rex et al., 1999	SD/5/120/200	46 asymptomatic	14	25-60%	75-83%	89%	14	43%	43%	–
Fenlon et al., 1999	SD/5/110/110	100 high risk	22	91%	96%	96%	36	82%	94%	92%
Pescatore et al., 2000	SD/5/110/200	50 high risk	11	–	37-62%	74%	8	–	71% (<10 mm)	75% (<10 mm)
Fletcher et al., 2000	SD/5/120/70	180 high risk	121	75%	85%	93%	142	47%	88% (≥5 mm)	72% (≥5 mm)
Morrin et al., 2000	SD+MD/2.5-5/120/120-200	81 high risk	20	90%	87%	100%	20	65%	73%	96%
Spinzi et al., 2001	SD/5/120/230-260	96 high risk	13	62%	–	100%	32	56% (<10 mm)	–	–
Hara et al., 2001	SD+MD/5/120/50-70	283 high risk	19	80-89%	78-100%	90-93%	–	–	–	–
Yee et al., 2001	SD/3/120/120-150	300 (204 high risk)	82	90%	100%	72% (overall)	141	80%	93%	–
Gluecker et al., 2002	MD/5/120/90	50 high risk	11	82%	–	90% (overall)	15	33%	–	–
Macari et al., 2002	MD/1/120/50	105 high risk	14	93%	–	98% (overall)	27	70%	–	–
McFarland et al., 2002	SD/5/120/200	70 high risk	40	68%	88%	60% (overall)	45	36%	71%	–
Iannaccone et al., 2003	MD/2.5/120/50	158 (127 high risk)	13	100%	–	97% (overall)	24	83%	–	–
Pineau et al., 2003	SD/5/120/200	205 (157 high risk)	27	78%	90%	95%	48	75%	84% (≥6 mm)	83% (≥6 mm)
Johnson et al., 2003	SD+MD/5/120/70-80	703 high risk, asymptomatic	59	32,34,73%	35,38,72%	97,98,98%	94	29,35,57%	41,46,69%	88,93,95%
Van Gelder et al., 2004	MD/2.5/120/100	249 high risk	48	75-77%	84%	92%	36	64-75%	–	69-71% (≥6 mm)
Macari et al., 2004	MD/1/120/50	68 asymptomatic	3	100%	100%	99%	17	53%	–	–
Iannaccone et al., 2005	MD/2.5/140/10	88 (76 high risk)	11	100%	100%	99%	26	86% (≥6 mm)	84% (≥6 mm)	82% (≥6 mm)
Yasumoto et al., 2006	MD/1.25/120/120	50 mixed	29	98%	–	–	70	87%	–	–
SD, single detector; MD, multidetector/collimation (mm)/kVp/mAs.

FIGURE 7.1 Large Polyp in an Asymptomatic Screening Patient. A large polyp is visualized on the axial images in abdominal window in both supine (A) and prone (B) images. The appearance of the lesion is more prominent using a colon window (C). The lesion was also identified on the 3D endoluminal view (D).

Almost all published CTC performance studies employ dual position scanning typically in the supine and prone positions. Standard CTC protocol includes the use of dual position scanning because this allows shifting of residual fluid and stool to reveal the underlying colon wall for examination and it can increase segmental distention leading to improved lesion detection.¹⁰, ¹¹ and ¹²

IMAGE DISPLAY

Options for image display include two-dimensional (2D) views, including axial images and multiplanar reformations, as well as various three-dimensional (3D) views, including the traditional endoluminal fly-through and the subvolume cube view. The performance of alternative displays has been studied less extensively and includes the dissection view, unfolded cube display, map projections, and planar virtual pathology view. Currently, most published CTC performance studies have included readers that employ 2D views primarily with the use of 3D views for problem solving. This has been found to be time efficient although a wide range of results has been reported. The 2D axial images may be easily and rapidly correlated with sagittal or coronal multiplanar reformats to identify true lesions. 2D views allow easy orientation from the extraluminal point of view and the inherent ability to evaluate lesion density. The 3D endoluminal views can
provide rapid differentiation of a polyp from a fold and is the preferred interpretation technique by some readers because it may be easier to learn and identify polyps. Primary 3D interpretation has been found in several published studies to be an accurate visualization technique for lesion detection.³,¹³

READER TRAINING

The level of training and experience of readers for CTC performance trials has been variable. This has been partly due to the lack of clear guidelines until the American College of Radiology Practice Guidelines for the Performance of CTC in Adults was published in 2005.¹⁴ These guidelines recommend that a supervising or interpreting physician should have reviewed at least 50 CTC cases either by taking a course with interactive training, double-readings performed by a trained physician, or correlation of CTC cases with colonoscopy.

It has become increasingly apparent that the level of training and experience of the reader is one of the most important factors influencing the diagnostic performance of CTC.¹⁵ Studies with good or excellent performance results are often carried out at single centers using experienced readers. The inclusion of inexperienced readers is considered to be one of the contributing factors to the poor performance of CTC.¹⁶, ¹⁷ and ¹⁸ Some of the more recently published performance studies and large multicenter trials that are currently under way ensure reader training by proficiency testing or documenting adequate reader case experience.⁴ One to three readers or consensus readings has been typically included in most published CTC performance trials.

DETECTION OF POLYPS: 1997 TO 2000

An initial evaluation of the sensitivity and specificity of CTC for polyp detection was performed by Hara et al.¹⁹ CTC was performed in the supine position using 5-mm collimation, pitch 1.3, 120 kVp, and 140 mA followed by colonoscopy in 70 high-risk patients. Images were independently interpreted by two observers who evaluated the axial images or the 3D endoluminal views using proprietary software. Only adenomatous polyps were considered for analysis. The per-patient sensitivity for adenomas measuring 10 mm and larger was 75% (9/12) with a specificity of 91% for both observers. The sensitivity and specificity for patients with adenomas 5 cm and larger averaged 66% and 63% respectively. Patients with adenomas <5 mm were detected with a sensitivity and specificity of 45% and 80% respectively. Interpretation of axial images only yielded lower sensitivity and specificity for 10 mm or larger adenomas (58% and 74%, respectively) compared with use of the 3D endoluminal views. It was concluded in this early study that CTC was a promising test for the detection of polyps 5 mm and larger.

In a prospective study of 44 high-risk patients undergoing CTC followed by colonoscopy, Dachman et al. used axial images for primary interpretation and use of the 3D endoluminal views for problem solving only.²⁰ Dual position scanning was performed using SDCT with 5-mm collimation, pitch 1.5, 120 kVp, and 100 mA. The sensitivity and specificity for detection of polyps >8 mm were 83% (5/6) and 100% for two observers. There were only three polyps measuring 5 to 8 mm and both observers had a sensitivity of 33%. The average interpretation time was 28 minutes. The 3D endoluminal view was employed by the readers in 52% of patients and did not significantly impact on interpretation times. The authors concluded that CTC may be performed using axial images to initially search for polyps of significant size and that the endoluminal views should be used only when needed to distinguish folds from polyps.

Rex et al. published a study of their early experience with CTC cases that had been performed in 1995 to 1996.¹ CTC was performed in 46 asymptomatic patients using dual-slice technique with 5-mm collimation, pitch 1.5, 120 kVp, and 300 mA. Interpretation was performed by two readers in batches using proprietary software. They found that 3D endoluminal views were superior to axial imaging for polyp detection but still did poorly. CTC detected only 1 of 4 (25%) adenomas 2 cm or larger, 6 of 10 (60%) adenomas 1 to 1.9 cm, 6 of 14 (43%) adenomas 6 to 9 mm, and 7 of 65 (11%) 5 mm or smaller. The sensitivity for the detection of patients with adenomas 1 to 1.9 cm was 83% and the specificity for patients with adenomas 1 cm or larger was 89%. Three large, flat adenomas of the right colon were missed. Critical factors for accurate interpretation of CTC were meticulous bowel cleansing and adequate distention. On the basis of these results, the authors concluded that CTC was not yet ready to be used as a screening test.

Only gold members can continue reading. Log In or Register to continue