Computed tomographic (CT) colonography has been evaluated as a screening technique for colorectal cancer, with good results (1,2). Unfortunately, CT c

Transcription

1 ORIGINAL RESEARCH GASTROINTESTINAL IMAGING Jasper Florie, MD Sebastiaan Jensch, MD Rutger A. J. Nievelstein, MD, PhD Joep F. Bartelsman, MD Lubbertus C. Baak, MD, PhD Rogier E. van Gelder, MD, PhD Brigitte Haberkorn, MSc Adrienne van Randen, MD Michiel M. van der Ham, MSc Pleun Snel, MD, PhD Victor P. M. van der Hulst, MD, PhD Patrick M. M. Bossuyt, PhD Jaap Stoker, MD, PhD MR Colonography with Limited Bowel Preparation Compared with Optical Colonoscopy in Patients at Increased Risk for Colorectal Cancer 1 Purpose: Materials and Methods: To prospectively evaluate the diagnostic performance of magnetic resonance (MR) colonography by using limited bowel preparation in patients with polyps of 10 mm or larger in diameter in a population at increased risk for colorectal cancer, with optical colonoscopy as the reference standard. The institutional review boards of all three hospitals approved the study. All patients provided written informed consent. In this multicenter study, patients undergoing colonoscopy because of a personal or family history of colorectal cancer or adenomatous polyps were included. Two blinded observers independently evaluated T1- and T2-weighted MR colonographic images obtained with limited bowel preparation (bright-lumen fecal tagging) for the presence of polyps. The limited bowel preparation consisted of a low-fiber diet, with ingestion of lactulose and an oral gadolinium-based contrast agent (with all three major meals) starting 48 hours prior to imaging. Results were verified with colonoscopic outcomes. Patient sensitivity, patient specificity, polyp sensitivity, and interobserver agreement for lesions of 10 mm or larger were calculated for both observers individually and combined. 1 From the Departments of Radiology (J.F., S.J., R.E.v.G., B.H., A.v.R., M.M.v.d.H., J.S.), Gastroenterology and Hepatology (J.F.B.), and Clinical Epidemiology and Biostatistics (P.M.M.B.), Academic Medical Center, PO Box 22660, 1100 DD Amsterdam, the Netherlands; Departments of Radiology (S.J., V.P.M.v.d.H.) and Gastroenterology and Hepatology (L.C.B.), Onze Lieve Vrouwe Gasthuis, Amsterdam, the Netherlands; Department of Radiology, University Medical Center Utrecht, Utrecht, the Netherlands (R.A.J.N.); and Department of Gastroenterology and Hepatology, Slotervaart Ziekenhuis, Amsterdam, the Netherlands (P.S.). From the 2005 RSNA Annual Meeting. Received December 21, 2005; revision requested February 10, 2006; revision received May 13; accepted June 5; final version accepted September 1. Supported by grant from ZONMW (the Netherlands Organization for Health Research and Development). Address correspondence to J.F. ( j.florie@amc.uva.nl ). Results: Conclusion: Two hundred patients (mean age, 58 years; 128 male patients) were included; 41 patients had coexistent symptoms. At colonoscopy, 12 patients had 22 polyps of 10 mm or larger. Per-patient sensitivity was 58% (seven of 12) for observer 1, 67% (eight of 12) for observer 2, and 75% (nine of 12) for both observers combined for polyps of 10 mm or larger. Per-patient specificity was 95% (178 of 188) for observer 1, 97% (183 of 188) for observer 2, and 93% (175 of 188) for both observers combined. Per-polyp sensitivity was 55% (12 of 22) for observer 1, 50% (11 of 22) for observer 2, and 77% (17 of 22) for both observers combined. Interobserver agreement was 93% for identification of patients with lesions of 10 mm or larger. In patients at increased risk for colorectal cancer, specificity of MR colonography by using limited bowel preparation was high, but sensitivity was modest. RSNA, 2007 RSNA, Radiology: Volume 243: Number 1 April 2007

2 Computed tomographic (CT) colonography has been evaluated as a screening technique for colorectal cancer, with good results (1,2). Unfortunately, CT colonography requires ionizing radiation, and use of this method poses a significant drawback for largescale use in patients at average risk (3,4). An additional drawback is the extensive bowel preparation used in most protocols, which is regarded as a major burden by many patients (5 7). Although some studies have shown that a substantial dose reduction in CT colonography is feasible (8,9), an alternative imaging method that does not require ionizing radiation would be preferable for screening. In recent years, several researchers have investigated the use of magnetic resonance (MR) colonography in symptomatic populations (10 14); most of these researchers concluded that MR colonography has diagnostic value (10 13). Investigators in a few studies have shown that a combination of a low-fiber diet, an oral contrast agent (ie, fecal tagging or fecal masking), and sometimes lactulose results in a more limited and potentially patient-friendly bowel preparation (15 17). For fecal tagging, either a brightlumen strategy (gadolinium for fecal tagging and a water-gadolinium mixture for rectal filling) or a dark-lumen strategy (barium for tagging and water for Advances in Knowledge Sensitivity of MR colonography with limited bowel preparation in a low-prevalence population with polyps of 10 mm or larger is 58% 67% per patient and 50% 55% per polyp. The specificity of MR colonography with limited bowel preparation in a low-prevalence population was 95% 97% for patients without polyps of 10 mm or larger. Interobserver agreement for MR colonography with limited bowel preparation in a low-prevalence population for patients with lesions of 10 mm or larger was 93%. rectal filling) can be used. Lactulose (a mild laxative) is used for stool softening and homogeneous contrast agent distribution. The advantage of using barium for fecal tagging is that it is cheaper; however, investigators have shown less favorable results in regard to patient acceptance of this strategy (18). Though a more limited bowel preparation would make MR colonography an attractive alternative, its usability will also be determined by its test characteristics. However, no studies have been performed, to our knowledge, in large patient populations with the use of bright-lumen MR colonography with fecal tagging, which is supposed to be a patient-friendly strategy primarily because it does not require radiation. The purpose of our multicenter study was to prospectively evaluate the diagnostic performance of MR colonography by using limited bowel preparation for patients with polyps of 10 mm or larger in a population at increased risk for colorectal cancer (19), with optical colonoscopy as the reference standard. Materials and Methods Study Group Between February 20, 2003, and October 27, 2004, consecutive patients with a personal or family history of colorectal polyps or cancer, scheduled for optical colonoscopy, were asked to participate in this study in three hospitals (one university hospital and two secondary referral centers). Exclusion criteria were as follows: age younger than 18 years, presence of a colostomy after colorectal surgery, oral or intravenous administration of another contrast medium within 48 hours prior to MR colonography, inability to hold a breath for 25 seconds, and contraindications for MR colonography (including claustrophobia and pregnancy). Coexistent symptoms were not an exclusion criterion, as long as colonoscopy was not performed (or scheduled earlier) because of symptoms. All patients provided written informed consent. The institutional review boards of all three hospitals approved the study. Patient Preparation prior to MR Colonography MR colonography was to be performed within 2 weeks prior to optical colonoscopy. All patients started preparation 48 hours prior to MR colonography with a low-fiber diet (only well-cooked, nonfibrous vegetables and meat, no fibrous fruit, no whole-wheat cereal products, no nuts), together with ingestion of 12 g of lactulose powder in 6-g packets (Lactulose CF; Centrafarm, Etten-Leur, the Netherlands) dissolved in water once per day (in the morning) for stool softening. An oral contrast agent that contained 10 ml of gadolinium in a dose of 0.5 mmol/ml (gadopentetate dimeglumine, Magnevist; Schering, Berlin, Germany) was added to all major meals during this period (6 meals over 2 days) for stool tagging. The cost was U.S. $90 per patient, converted from euros. If the stool became too soft (diarrhea), patients were allowed to reduce the amount of lactulose. In a questionnaire, patients were asked about stool consistency prior to imaging and whether they reduced the amount of lactulose. MR Colonographic Imaging Protocol MR colonography was performed at the Academic Medical Center or Onze Lieve Vrouwe Gasthuis in Amsterdam, the Netherlands, by a research fellow or a specially trained technician. Prior to imaging, a flexible balloon-tipped catheter Published online before print /radiol Radiology 2007; 243: Abbreviations: 3D three-dimensional 2D two-dimensional Author contributions: Guarantor of integrity of entire study, J.F.; study concepts/study design or data acquisition or data analysis/ interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, J.F., R.E.v.G., A.v.R., M.M.v.d.H., J.S.; clinical studies, S.J., R.A.J.N., J.F.B., L.C.B., R.E.v.G., B.H., A.v.R., M.M.v.d.H., P.S., V.P.M.v.d.H., P.M.M.B., J.S.; statistical analysis, J.F., S.J., R.E.v.G., P.M.M.B., J.S.; and manuscript editing, all authors Authors stated no financial relationship to disclose. Radiology: Volume 243: Number 1 April

3 was placed in the rectum. A spasmolytic agent, 20 mg of butylscopolamine bromide (Buscopan; Boehringer-Ingelheim, Ingelheim, Germany), or, when use of a spasmolytic agent was contraindicated, 1 mg of glucagon hydrochloride (Glucagen; Novo-Nordisk, Bagsvaerd, Denmark) was administered intravenously just prior to moving the table into the imager. Subsequently, the colon was distended with a mixture of water and gadolinium-based contrast agent (10 mmol/l) by using 80 cm of hydrostatic pressure. The amount of fluid administered was estimated on the basis of the markings on the bag. This preparation renders the colonic lumen hyperintense on T1-weighted MR images and hypointense on T2-weighted images. Filling of the colon was not monitored but was stopped when patients reported pain or were uncomfortable. If necessary (on the basis of the scout images), additional fluid was added. MR colonography was performed with 1.5-T (Signa, GE Healthcare, Milwaukee, Wis; Intera, Philips Medical Systems, Best, the Netherlands) or 3-T (Intera; Philips Medical Systems) imagers by using phased-array coils. The protocol consisted of coronal three-dimensional (3D) T1-weighted fast fieldecho and coronal and transverse twodimensional (2D) T2-weighted fast spinecho sequences, which were performed in both prone and supine positions (Table 1). Sequences were performed by using multiple breath holds to reduce imaging time per breath hold. Only the transverse T2-weighted fast spin-echo sequence used with the 1.5-T imager (Signa; GE Healthcare) was performed with free breathing because it would have taken too long to perform it in multiple breath holds with this imager. No overlapping sections were used. All breath holds took seconds. Since imaging took minutes per position, patients had to retain the enema for minutes in total. Although patients were asked to hold the enema during the entire procedure, when necessary the lock of the enema bag was opened to prevent severe leakage. The amount of leakage was noted (a spot of leakage was designated as no or minor leakage and a substantial amount was designated as considerable leakage). Total in-room time was approximately 45 minutes, on average. Evaluation of MR Colonographic Data All images were independently evaluated by two observers who were blinded to the results of colonoscopy. The first observer, an abdominal radiologist (R.A.J.N.) with 11 years of clinical experience with MR imaging, had evaluated images from 20 CT colonographic examinations and 40 MR colonographic examinations, with feedback of the (videotaped) optical colonoscopic results for learning purposes. The second observer (S.J.), a second-year resident in radiology, had read images from more than 160 CT colonographic examinations and had read images from 40 MR Table 1 Parameters for All Three Imagers Sequence and Parameters A colonographic examinations, with feedback of optical colonoscopic results. Both observers evaluated images obtained with all sequences at a workstation (Extended Brilliance Workspace; Philips Medical Systems) or picture archiving and communication system (IM- PAX-SP4-SU4-DS3000; Agfa, Mortsel, Belgium). Since the T2-weighted data set was obtained in two directions, no multiplanar reformatted images were used. The observers assigned scores for presence, size (diameter), and location (cecum, ascending colon, transverse colon, descending colon, sigmoid colon, rectum) of any polyps or colorectal cancer at MR colonography and the sequences in which they were visible. Lesions were measured by using calipers, and screen shots were obtained. The 1.5-T Unit B 3-T Unit Coronal 3D T1-weighted fast field echo Repetition time msec/echo time msec 5.4/ / /2.0 Flip angle (degrees) Section thickness (mm)* Field of view (mm) Matrix No. of sections No. of breath holds Sensitivity encoding factor NA D T2-weighted fast spin echo Coronal Repetition time msec/echo time msec 1050/62 867/ /60 Section thickness (mm) Field of view (mm) Matrix No. of sections No. of breath holds Sensitivity encoding factor NA Transverse Repetition time msec/echo time msec 1354/64 676/ /60 Section thickness (mm) Field of view (mm) Matrix No. of sections No. of breath holds Free breathing 2 4 Sensitivity encoding factor NA Note. A Signa, GE Healthcare; B Intera, Philips Medical Systems; NA not applicable. * Interpolated. 124 Radiology: Volume 243: Number 1 April 2007

4 location was marked on a schematic drawing of the colon. The first observer also evaluated image quality with a fivepoint scale (excellent, polyps 2 mm are probably visible; good, polyps 6 mm are probably visible; moderate, polyps 10 mm are probably visible; fair, polyps 10 mm could be missed; and poor, not diagnostic), distention with a four-point scale (good, polyps 6 mm are probably visible; sufficient, polyps 10 mm could be missed due to suboptimal distention; insufficient, polyps 10 mm could be missed due to poor distention; and collapsed bowel segments), the presence of stool (yes or no stool at all), and the presence of artifacts (if present, the type was specified). Colonoscopy Patients ingested 4 6 L of polyethylene glycol electrolyte solution (KleanPrep; Helsinn Birex Pharmaceuticals, Dublin, Ireland) for bowel preparation on the day before the examination (in patients who ingested 6 L, the last 2Lofthat amount was ingested on the examination day). Optical colonoscopy was used as the reference standard. Optical colonoscopy was performed with a standard colonoscope (CF-Q160AL or CFQ- 160L; Olympus, Tokyo, Japan) at the Endoscopy Department of Academic Medical Center, Onze Lieve Vrouwe Gasthuis, or Slotervaart Ziekenhuis, Amsterdam, the Netherlands. Optical colonoscopy was performed by an experienced staff member (gastroenterologist or gastrointestinal surgeon, with average experience as a staff member of 15 years and a range of experience of 3 25 years) or by a gastroenterology fellow with the direct supervision of the attending gastroenterologist. The endoscopy was recorded on videotape. If polyps were present, the endoscopist scored their exact location (cecum, ascending colon, transverse colon, descending colon, sigmoid colon, rectum), morphology (pedunculated, sessile, flat), and size. Size was estimated prior to removal on the basis of comparison with an open-biopsy forceps of known size (8 mm). A polyp was considered flat if its height was less than one-half of the diameter of the lesion (20). This information was filled out on a case record form. Mean examination time was 33 minutes 21 (standard deviation). Statistical Analysis and the Determination of Lesion Status All polyps were categorized according to their size at endoscopy (medium, 6 9 mm; large, 10 mm). We calculated sensitivity, specificity, predictive values, and interobserver agreement for patients with polyps of 10 mm or larger and those of 6 mm or larger. No consensus reading was performed in our study. Because double reading is sometimes used in screening, we calculated the detection parameters again after combining the results. For this purpose, lesions (true-positive and false-positive lesions) identified at MR colonography by observers 1 and 2 were added. A research fellow who was not involved in reading the images from the MR colonographic examinations (J.F., whose experience consisted of reading images from CT colonography performed in 400 patients and matching 80 CT colonographic evaluations with colonoscopy and matching 50 MR colonographic evaluations with colonoscopy) matched all polyps of 10 mm or larger that were identified at endoscopy with the polyps on MR colonographic images; thus, a frame of reference for true-positive findings at MR colonography was provided. By defining these reference polyps, it was possible to determine whether false-negative MR colonographic findings were caused by interpretation problems or visualization problems. A polyp detected at MR colonography was labeled as true-positive if three criteria were met: The segmental location and location within the segment corresponded with those at optical colonoscopy (when the polyp was situated near the segmental border, localization in the adjacent segment was also tolerated), the polyp size as estimated by the endoscopist (open forceps) corresponded with the size as measured on MR colonographic images (a 50% margin on the basis of the endoscopically determined size was allowed), and the appearance (morphology) closely resembled that of the corresponding polyp at videotaped optical colonoscopy. False-positive findings for lesions of 10 mm or larger. After all MR colonographic examinations had been reviewed and findings were matched, the research fellow who was not involved in reading the images from the MR colonographic examinations (J.F.) determined the nature of false-positive findings of 10 mm or larger by reviewing the videotaped optical colonoscopy, the findings from that procedure, and MR images. If a false-positive lesion of 10 mm or larger was assumed to be a true polyp that was missed at optical colonoscopy, second-look endoscopy was performed. Detection parameters. Patient sensitivity was defined as the number of patients with at least one true-positive lesion relative to the number of patients with polyps at optical colonoscopy. For the 6- and 10-mm thresholds, findings were considered to be true-positive if at least one matched polyp in the respective size range was detected, and findings were considered to be false-negative when no true-positive polyps or only those of a lower size category had been detected. Polyp sensitivity was the proportion of polyps detected at optical colonoscopy with a true-positive MR colonographic result. From some patients, more than one polyp was obtained in this study. Thus, we used generalized estimating equations to revise the data clustering and dependency (21). In the generalized estimating equations, we assessed the adjusted confidence interval for per-polyp sensitivity. Patient specificity was the proportion of patients without polyps at optical colonoscopy with true-negative findings. For the 6- and 10-mm thresholds, findings were considered true-negative if patients without polyps of 6 mm or larger or of 10 mm or larger had no false-positive findings that were of those respective size ranges. Predictive values. The positive predictive value was defined as the proportion of patients with a true-positive finding of all patients with findings of that size range at MR colonography. The negative predictive value was defined as the proportion of patients with a true-negative finding of all patients without findings of Radiology: Volume 243: Number 1 April

5 that size range at MR colonography. Perpolyp, the positive predictive value was defined as the proportion of true-positive polyps of all findings of that size range at MR colonography. Interobserver agreement. For calculation of the interobserver variability, the statistic is, in general, the accepted method. However, beside the fact that this measure strongly depends on the disease prevalence, which was 6% in our study, determination of interobserver variability per patient is hampered if both observers point out different lesions in one patient. We therefore performed interobserver agreement analysis by calculating the agreement in percentages on a perpatient level. The reviewers were considered to agree if they both recorded one or more corresponding lesions or if they both recorded no findings. For corresponding lesions, the largest size measurement of both observers was used to determine the size category, with a 50% margin on the basis of the largest size measurement allowed. Statistical software. Software (SPSS for Windows, version ; SPSS, Chicago, Ill) was used for all statistical analyses except for the calculation of the confidence intervals. Computer software (SAS, version 9.1; SAS Institute, Cary, NC) was used to apply the generalized estimating equations (polyp sensitivity Figure 1 Figure 1: Flow diagram shows patient participation. Table 2 Baseline Characteristics of the Study Group Characteristic confidence intervals); the Proc Genmod command was used. In regard to other confidence intervals, if the number of true-positive, false-negative, false-positive, or the number of true-negative findings was smaller than five, two-sided confidence limits to the binomial distribution were used to calculate the confidence intervals (SEMSTAT Statistical Software Package; Sematech, Austin, Tex); if the number of true-positive, false-negative, false-positive, or the number of true-negative findings was five or larger, a normal approximation to the binomial distribution was used. Results In total, 227 patients were included (Fig 1); 27 patients had to be excluded because of planning problems (n 7, no imaging time was available or there were changes in the schedule); because they did not show up (n 4); because No. of patients enrolled per hospital 227 Academic Medical Center 163 Onze Lieve Vrouwe Gasthuis 53 Slotervaart Ziekenhuis 11 No. of patients included per hospital 200 Academic Medical Center 145 Onze Lieve Vrouwe Gasthuis 46 Slotervaart Ziekenhuis 9 Sex Male patients 128 Female patients 72 Age (y)* (23 84) No. of patients with indication History of colorectal polyps 110 (55) History of colorectal carcinoma 48 (24) Family history of colorectal polyps or cancer 42 (21) No. of patients with coexistent symptoms Abdominal pain 24 (12) Hematochezia 6 (3) Altered bowel habits 11 (6) Optical colonoscopic results No. of polyps 10 mm 22 (12) Mean (mm) 12.6 Median (mm) 10 Range (mm) mm 44 (26) Mean (mm) 7.3 Median (mm) 8 10-mm lesions No. of adenomas 11 No. of polyps with histologic characterization 15 6-mm lesions No. of adenomas 22 No. of polyps with histologic characterization 33 * Data are the mean standard deviation, and numbers in parentheses are the range. Numbers in parentheses are percentages. Number in parentheses is the number of patients with polyps. For polyps 6 9 mm. Value 126 Radiology: Volume 243: Number 1 April 2007

6 they had claustrophobia (n 2); because they withdrew from the study (n 2); because MR colonographic data were lost (n 2); because of a technical MR imaging problem (n 1); because of unrelated health problems Figure 2 (n 1); and because, at optical colonoscopy, only the sigmoid colon was reached (the patient had no lesions [n 1]). In seven other patients, both observers rated MR colonography as not diagnostic (mostly because of motion Figure 2: Example of a 10-mm polyp (arrow) in the cecum of 83-year-old patient with multiple polyps at MR colonography on (a) coronal 3D T1-weighted fast field-echo (5.4/1.6; flip angle, 25 ) and (b) coronal (1050/64) and (c) transverse (1354/64) 2D T2- weighted single-shot fast spin-echo images. Polyp was confirmed at optical colonoscopy and was seen by one observer but missed by the other. Arrowhead shows layer of intermediate signal intensity visible in many patients on the T2-weighted images mostly in the sigmoid colon. artifacts or insufficient distention). In six patients in whom initial endoscopy was not successful because of insufficient bowel preparation (n 1) or incomplete visualization (n 5), repeat endoscopy was successful. Therefore, data from 200 patients were available for analysis (Table 2). Forty-one patients had symptoms associated with colorectal carcinoma just prior to MR colonography. Results of Colonoscopy In 107 (54%) patients, a polyp was detected at optical colonoscopy (Table 2); 12 of these patients had one or more polyps that were 10 mm or larger (total, 22 polyps). Five of the 22 large polyps were flat. Only one symptomatic patient harbored a large polyp, and two symptomatic patients harbored a polyp that was 6 9 mm. No carcinomas were detected. No second-look endoscopic procedures had to be performed because the 17 large false-positive lesions were explained by a large ileocecal valve (two lesions) or air bubbles or residual stool (ascending colon, two lesions; transverse colon, five lesions; descending colon, two lesions; sigmoid colon, six lesions). Results of MR Colonography One hundred sixty-eight patients were imaged at 1.5-T MR colonography; 127 patients were imaged with one unit Table 3 Performance Characteristics for Observers according to Polyp Size for MR Colonography Statistic 10-mm Polyps 6-mm Polyps Observer 1 Observer 2 Both Observers Observer 1 Observer 2 Both Observers Value 95% CI Value 95% CI Value 95% CI Value 95% CI Value 95% CI Value 95% CI Patient analysis Sensitivity 58 (7/12) 30, (8/12) 35, (9/12) 43, (12/26) 27, (16/26) 43, (17/26) 47, 84 Specificity 95 (178/188) 91, (183/188) 95, (175/188) 89, (128/174) 67, (153/174) 83, (116/174) 60, 74 PPV 41 (7/17) 18, (8/13) 35, (9/22) 20, (12/58) 10, (16/37) 27, (17/75) 13, 32 NPV 97 (178/183) 95, (183/187) 95, (175/178) 95, (128/142) 85, (153/163) 90, (116/125) 88, 97 Polyp analysis Sensitivity 55 (12/22) 33, (11/22) 28, (17/22) 48, (20/44) 28, (24/44) 36, (32/44) 53, 86 PPV 48 (12/25) 28, (11/16) 46, (17/34) 33, (20/88) 14, (24/55) 31, (32/125) 18, 33 Note. Values and 95% confidence intervals are percentages. Numbers in parentheses were used to calculate the percentages. CI confidence interval, NPV negative predictive value, PPV positive predictive value. Number of false-positive findings for polyps 10 mm or larger were as follows: observer 1, 13; observer 2, five; and both observers, 17. Those for polyps 6 mm or larger were as follows: observer 1, 68; observer 2, 31; and both observers, 93. Radiology: Volume 243: Number 1 April

7 (Signa, GE Healthcare), and 41 patients were imaged with another unit (Intera, Philips Medical Systems). Thirty-two patients were imaged at 3.0 T. Seventyseven (38%) patients had diarrhea sometime during the bowel preparation; in 48 of these patients, diarrhea occurred just prior to imaging. Thirtyseven patients reduced the amount of lactulose (all but one did so only on the last day). Prior to imaging in 191 patients, a spasmolytic agent was administered. One hundred forty-nine patients received butylscopolamine bromide, 42 received glucagon, and nine received no spasmolytic agent. On average, 1.9 L of water-gadolinium mixture was used to fill the colon. Two patients had considerable leakage; in 11, leakage was minor. MR colonography was well tolerated. Image quality was excellent in 24 patients, good in 90 patients, moderate in 46 patients, and fair in 40 patients. Most commonly seen artifacts were those caused by movement of the bowel on T1-weighted images in 88 patients and inhomogeneity (mostly a layer of intermediate signal intensity) (Fig 2) of luminal content on T2-weighted images in 73 patients. Fecal material was visible in 70 patients on the T1-weighted images and in two patients on both coronal and transverse T2-weighted images. Distention was insufficient in the ascending colon, transverse colon, descending colon, sigmoid colon, and rectum in, respectively, eight patients, nine patients, zero patients, 22 patients, and one patient. In 19 (86%) of 22 polyps that were 10 mm or larger (but including all patients with a large polyp), a reference polyp could be identified on the MR colonographic image by the independent research fellow (J.F.) who was not blinded to the colonoscopic results. All three polyps that were not visible at MR colonography were flat at optical colonoscopy (two were adenomas). Two polyps, which were flat at optical colonoscopy, were visible (one adenoma) in retrospect at MR colonography; both were identified by observer 1 but were missed by observer 2. The flat lesions were located in either the cecum (2) or the ascending colon (3). Figure 3 Detection Parameters and Predictive Values With MR colonography, for observer 1, observer 2, and both observers combined, respectively, sensitivity for detection of polyps was 58% (seven of 12), 67% (eight of 12), and 75% (nine of 12) for large polyps (Table 3; Figs 2,3). For smaller lesions, fewer polyps were detected (Fig 4). Table 4 shows with which sequences polyps were detected. Interobserver Analysis Interobserver agreement with regard to patients with large polyps ( 10 mm) was 93%. For polyps that were 6 mm or larger, interobserver agreement was 70%. Discussion The results of this study indicate that the sensitivity of MR colonography by using limited bowel preparation for detection of patients with polyps in a population with low prevalence of colorectal polyps is modest. When results of Figure 3: Example of a 15-mm polyp (arrow) in the transverse colon of a 56-year-old patient at MR colonography on (a) coronal 3D T1-weighted fast field-echo (5.4/1.6; flip angle, 25 ) and (b) coronal (1050/64) and (c) transverse (1354/64) 2D T2- weighted single-shot fast spin-echo images. both observers were combined, the sensitivity for detection of polyps was 75% for large polyps, whereas sensitivity values for observers 1 and 2 were 58% and 67%, respectively. The specificity was high. For polyps that were of intermediate size, the results were somewhat lower. Interobserver agreement for patients with polyps of 10 mm or larger was 93%. Several limitations of our study must be considered. Only 22 polyps that were 10 mm or larger were detected; this finding limits the precision of our sensitivity estimates. However, the prevalence of patients with large polyps (6% irrespective of histologic findings) is relatively close to that of an asymptomatic population (3.9% in patients with large adenomas only) (1). Both observers were experienced readers of abdominal images and had evaluated at least 40 patients for colorectal polyps with MR colonography by using fecal tagging (with feedback of the optical colonoscopic results) prior to this series. However, this may not be 128 Radiology: Volume 243: Number 1 April 2007

8 Figure 4 Table 4 Figure 4: Example of an 8-mm polyp (arrow) in the ascending colon in same patient as in Figure 1 at MR colonography on (a) coronal 3D T1-weighted fast field-echo (5.4/1.6; flip angle, 25 ) and (b) coronal (1050/64) and (c) transverse (1354/64) 2D T2- weighted fast spin-echo images. Number of Polyps Detected with Combinations of Sequences according to Observer 1 Sequence and Position True-Positive Findings (n 12) Polyps 10 mm False-Positive Findings (n 13) True-Positive Findings (n 20) Polyps 6mm Coronal 3D T1-weighted fast field echo Supine Prone Supine and prone combined D T2-weighted fast spin-echo Coronal Supine Prone Supine and prone combined Transverse Supine Prone Supine and prone combined Supine Prone False-Positive Findings (n 68) sufficient for optimal polyp detection; findings in a study by Burling et al (22) suggested that there is a linear correlation between prior experience and accuracy without a plateau for CT colonography. We anticipated, in our protocol, a second-look optical colonoscopic examination if false-positive lesions of 10 mm or larger were labeled as potentially missed polyps at optical colonoscopy by the reviewer who was not blinded to the colonoscopic results. No second-look endoscopic examinations were deemed necessary. Since the performance of MR colonography was suboptimal, segmental unblinding would have been preferable, as it allows verification of all false-positive lesions. In our study, sensitivity was low compared with that in other studies (10 13). Although this result could have been caused by the limited bowel preparation used, the bowel preparation used seems to be an unlikely explanation because most polyps were visible and were not obscured by nontagged stool. Bowel motion artifacts and air bubbles, which could have rendered identification of true lesions more difficult for readers, were present in many patients. However, studies of MR colonography performed with extensive cleansing probably are hampered similarly. Another explanation for higher sensitivity values found in other studies may be related to the patients who were examined. Both the higher polyp prevalence and the difference in spectrum (ie, larger or more obvious polyps, a higher frequency of cancer) may explain the higher diagnostic accuracy reported in previous series concerning high-prevalence or primarily symptomatic populations. Results of another MR colonographic study with extensive cleansing (14) in a low-prevalence population indicate a lower sensitivity than that in our study. Differences in protocols and fecal tagging strategies may also provide an explanation; dark-lumen MR colonography with intravenously administered gadolinium-based contrast agent has the theoretical advantage that polyps, Radiology: Volume 243: Number 1 April

9 including flat polyps, enhance and are thereby better visualized. However, in an explorative study (23), image quality with dark-lumen MR colonography was not better than that with bright-lumen MR colonography, whereas bowel preparation with barium administered orally was considered burdensome (18) and a detriment to its possible use for screening in low-prevalence populations. Unfortunately, the bright-lumen technique is more costly. Since most of the large polyps (19 of 22) were visible in retrospect, the problem in identifying polyps is an issue of interpretation rather than one of visualization. The specificity was high, even though 17 large false-positive lesions were found. This is the consequence of the fact that the study was performed in a low-prevalence population. For smaller polyps, results were even less favorable than they were for large polyps. Although for screening purposes of patients at average risk these lesions might not be the primary focus, for screening of patients at increased risk these polyps probably should not be neglected. When we compared our study with two recently published CT colonographic studies in asymptomatic patients (1) and patients at increased risk (2), our results are less encouraging. Differences may be a result of the fact that CT colonography is less subject to motion artifacts and the spatial resolution of CT colonography is better than that of MR colonography. Moreover, the MR colonographic images in our study were read with 2D display modes, whereas the images in the two aforementioned CT colonographic studies primarily were read with 3D display modes. This difference in evaluation method has frequently been quoted to explain results that are less promising in three published CT colonographic studies in which 2D display methods were used (24 26). Unfortunately, 3D evaluation of MR colonographic data is not available yet for large-scale use because of technical problems, such as field inhomogeneity, proximity of the tissue being examined to the coil, and nonlinearity of MR imaging values. Technical advances in imaging (eg, shorter acquisition times, improvement of spatial and contrast resolution) or in molecular imaging (enhancing adenomatous lesions by using selective labeling) may make MR colonography competitive with the recently studied lowdose CT colonographic protocols with fecal tagging (27). In summary, although MR colonography with limited bowel preparation might be considered a reasonable option for triaging patients in populations with a low prevalence of colorectal polyps, at present the sensitivity of this method is modest. We believe implementation of MR colonography in colorectal cancer screening has to await further technical improvements. Acknowledgments: The authors thank the following members of the Department of Radiology, Academic Medical Center, Amsterdam, the Netherlands: Cristina Lavini, MPhil, for her valuable help in optimizing imaging protocols; Karin van Gemert-Horsthuis, MD, for critical review of the manuscript; and Shandra Bipat, MSc, for her valuable help in statistical analysis. 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: van Gelder RE, Nio CY, Florie J, et al. Computed tomographic colonography compared with colonoscopy in patients at increased risk for colorectal cancer. Gastroenterology 2004;127: Recommendations of the International Commission on Radiological Protection. Smith H, ed. Ann ICRP 1991;21(1-3): Brenner DJ, Georgsson MA. Mass screening with CT colonography: should the radiation exposure be of concern? Gastroenterology 2005;129: Gluecker TM, Johnson CD, Harmsen WS, et al. Colorectal cancer screening with CT colonography, colonoscopy, and double-contrast barium enema examination: prospective assessment of patient perceptions and preferences. Radiology 2003;227: van Gelder RE, Birnie E, Florie J, et al. CT colonography and colonoscopy: assessment of patient preference in a 5-week follow-up study. Radiology 2004;233: Ristvedt SL, McFarland EG, Weinstock LB, Thyssen EP. Patient preferences for CT colonography, conventional colonoscopy, and bowel preparation. Am J Gastroenterol 2003;98: Iannaccone R, Laghi A, Catalano C, Mangiapane F, Piacentini F, Passariello R. Feasibility of ultra-low-dose multislice CT colonography for the detection of colorectal lesions: preliminary experience. Eur Radiol 2003;13: van Gelder RE, Venema HW, Florie J, et al. CT colonography: feasibility of substantial dose reduction comparison of medium to very low doses in identical patients. Radiology 2004;232: Luboldt W, Bauerfeind P, Wildermuth S, Marincek B, Fried M, Debatin JF. Colonic masses: detection with MR colonography. Radiology 2000;216: Ajaj W, Pelster G, Treichel U, et al. Dark lumen magnetic resonance colonography: comparison with conventional colonoscopy for the detection of colorectal pathology. Gut 2003;52: Pappalardo G, Polettini E, Frattaroli FM, et al. Magnetic resonance colonography versus conventional colonoscopy for the detection of colonic endoluminal lesions. Gastroenterology 2000;119: Hartmann D, Bassler B, Schilling D, et al. Colorectal polyps: detection with dark-lumen MR colonography versus conventional colonoscopy. Radiology 2006;238: Leung WK, Lam WW, Wu JC, et al. Magnetic resonance colonography in the detection of colonic neoplasm in high-risk and average-risk individuals. Am J Gastroenterol 2004;99: Weishaupt D, Patak MA, Froehlich J, Ruehm SG, Debatin JF. Faecal tagging to avoid colonic cleansing before MRI colonography. Lancet 1999;354: Lauenstein T, Holtmann G, Schoenfelder D, Bosk S, Ruehm SG, Debatin JF. MR colonography without colonic cleansing: a new strategy to improve patient acceptance. AJR Am J Roentgenol 2001;177: Lauenstein TC, Goehde SC, Ruehm SG, Holtmann G, Debatin JF. MR colonography with barium-based fecal tagging: initial clinical experience. Radiology 2002;223: Goehde SC, Descher E, Boekstegers A, et al. Dark lumen MR colonography based on fecal tagging for detection of colorectal masses: accuracy and patient acceptance. Abdom Imaging 2005;30: Winawer S, Fletcher R, Rex D, et al. Colorectal cancer screening and surveillance: 130 Radiology: Volume 243: Number 1 April 2007

10 clinical guidelines and rationale update based on new evidence. Gastroenterology 2003;124: Rembacken BJ, Fujii T, Cairns A, et al. Flat and depressed colonic neoplasms: a prospective study of 1000 colonoscopies in the UK. Lancet 2000;355: Pepe MS. The statistical evaluation of medical tests for classification and prediction: Oxford statistical science series 28. New York, NY: Oxford University Press, 2003; Burling D, Halligan S, Atchley J, et al. CT colonography: interpretive performance outside of a research environment [abstr]. Eur Radiol 2005;15(suppl 3):C15 C Florie J, Van Gelder RE, Haberkorn B, et al. Magnetic resonance colonography with limited bowel preparation: a comparison of three strategies. J Magn Reson Imaging (in press). 24. Johnson CD, Harmsen WS, Wilson LA, et al. Prospective blinded evaluation of computed tomographic colonography for screen detection of colorectal polyps. Gastroenterology 2003;125: 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: Rockey DC, Paulson E, Niedzwiecki D, et al. Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy: prospective comparison. Lancet 2005;365: Iannaccone R, Laghi A, Catalano C, et al. Computed tomographic colonography without cathartic preparation for the detection of colorectal polyps. Gastroenterology 2004; 127: Radiology: Volume 243: Number 1 April