Advances in Knowledge At enteric phase CT enterography, jejunal attenuation is higher than ileal attenuation, and collapsed bowel loops demonstrate gr

Transcription

1 Fargol Booya, MD Joel G. Fletcher, MD James E. Huprich, MD John M. Barlow, MD C. Daniel Johnson, MD Jeff L. Fidler, MD Craig A. Solem, MD William J. Sandborn, MD Edward V. Loftus, Jr, MD William S. Harmsen, MD Active Crohn Disease: CT Findings and Interobserver Agreement for Enteric Phase CT Enterography 1 Purpose: To retrospectively evaluate small-bowel enhancement characteristics and the sensitivity, specificity, and interobserver agreement of computed tomographic (CT) findings by using histologic and endoscopic results as a reference standard in patients undergoing enteric phase CT enterography. Materials and Methods: The institutional review board approved this retrospective HIPAA-compliant study, which included patients who consented to having their medical records used for research purposes. Enteric phase CT enterographic and ileoscopic findings with or without ileal histologic results were examined in 42 patients (24 women, 18 men). Enteric phase CT enterography was performed after 150 ml of intravenous contrast material was administered at 4 ml/sec, with a 45-second delay. Mural attenuation was measured in the distended and collapsed jejunal and ileal loops and in the terminal ileum. Two radiologists examined CT images for findings of Crohn disease. Mural attenuation for different bowel loops was compared by using a Student t test, with statistics used to measure interobserver agreement and Pearson correleation coefficients used to compare visual and quantitative measures. ORIGINAL RESEARCH GASTROINTESTINAL IMAGING 1 From the Department of Radiology (F.B., J.G.F., J.E.H., J.M.B., C.D.J., J.L.F.), Division of Gastroenterology and Hepatology, Department of Internal Medicine (C.A.S., W.J.S., E.V.L.), and Division of Biostatistics (W.S.H.), Mayo Clinic College of Medicine, 200 First St SW, Rochester, MN From the 2004 RSNA Annual Meeting. Received August 27, 2005; revision requested November 2; revision received February 6, 2006; accepted February 21; final version accepted June 1. Address correspondence to J.G.F. ( fletcher.joel@mayo.edu). Results: Conclusion: Distended jejunal loops had significantly greater attenuation than distended ileal loops (113 HU vs 72 HU; P.001). Attenuation of collapsed jejunal (134 HU) and ileal (108 HU) loops was greater than that of distended jejunal and ileal loops (P.001). Terminal ileal enhancement was the most sensitive visual CT finding of Crohn disease for both radiologists. Mural thickening demonstrated the greatest interobserver agreement ( 0.83). Visual enhancement and quantitative mural attenuation were significantly correlated (P.003). At enteric phase CT enterography, jejunal attenuation is greater than ileal attenuation and collapsed bowel loops demonstrate greater attenuation than distended bowel loops. Mural hyperenhancement and increased mural thickness are the most sensitive CT findings of active Crohn disease. RSNA, 2006 RSNA, 2006 Radiology: Volume 241: Number 3 December

2 Advances in Knowledge At enteric phase CT enterography, jejunal attenuation is higher than ileal attenuation, and collapsed bowel loops demonstrate greater attenuation than distended bowel loops. Mural hyperenhancement and increased mural thickeness are the most sensitive CT findings of active Crohn disease and demonstrate substantial and almost perfect interobserver agreement, respectively. Multidetector computed tomography (CT) has been recognized as a promising modality for the evaluation of Crohn disease and other small-bowel abnormalities (1 4). CT enterography differs from routine abdominal CT in that CT enterography uses multidetector CT (narrow section thickness and reconstruction interval), intravenous contrast material, and large volumes of a neutral oral contrast agent to improve depiction of the small-bowel wall and lumen. In patients with Crohn disease, CT enterography is increasingly used to detect enteric inflammation in addition to extra-enteric complications (5 8). While water is the most readily available neutral enteric contrast agent, other agents that distend the smallbowel wall to a greater extent such as methylcellulose solution, polyethylene glycol solution, and low-attenuation barium suspensions are increasingly being employed. Findings such as mural thickening, mural enhancement, increased attenuation of the perienteric fat, and the comb sign have been reported to indicate active inflammatory Crohn disease (1,2,7,9,10). Crohn disease is a transmural process that may extend to the surrounding perienteric fat and mesentery. While mucosal and submucosal abnormalities (eg, vascular malformations, nonsteroidal anti-inflammatory drug ulcerations, and neoplasms) may be optimally detected by using invasive techniques such as capsule endoscopy (11), we have recently found CT enterography to compare favorably with capsule endoscopy in patients suspected of having Crohn disease a finding that has also been suggested by other authors (11,12). Moreover, CT enterography avoids the risk of capsule retention, which can occur in up to 6.7% of patients with Crohn disease (13). Multiple acquisition parameters affect the appearance of the small bowel at CT enterography. While narrow section thicknesses are uniformly used during CT enterography to visualize the small-bowel wall and lumen, wide variation exists in the results of studies on the phase of contrast material enhancement (ie, the delay in scanning after intravenous injection of contrast material) (5,6,9,11,14,15). Although Wold et al (5) noted that there was no difference in the diagnosis of Crohn disease in a small series of patients who underwent enteric and hepatic phase scanning with CT enterography, Horton et al (16) found that bowel enhancement was greater in the arterial phase (30 seconds after intravenous injection of contrast material) than in the hepatic phase (60 seconds after intravenous injection of contrast material). In early reports of CT enterographic techniques, researchers consequently initiated scanning before the hepatic phase (40 45 seconds after intravenous injection of contrast material) (6,9,15). While these reports estimate the performance of CT enterography for Crohn disease, they do not describe bowel enhancement. Scanning performed somewhat later (40 60 seconds after intravenous injection of contrast material) has also been studied as a means of increasing the conspicuity of colorectal lesions at contrast material enhanced CT colonography (11,17,18). Such timing is similar to timing used for pancreatic phase imaging, during which both the superior mesenteric artery and the superior mesenteric vein are well opacified by intravenous contrast material (19). We began performing CT enterography during the enteric phase (45 seconds after intravenous injection of contrast material) because imaging performed at this time might increase the conspicuity of abnormal small-bowel enhancement or mural stratification. Enhancement characteristics during enteric phase scanning of jejunal versus ileal small-bowel loops, normal versus diseased small-bowel loops, and collapsed versus distended small-bowel loops have not been described. Thus, the purpose of our study was to retrospectively evaluate small-bowel enhancement characteristics and the sensitivity, specificity, and interobserver agreement of CT findings by using histologic and endoscopic results as a reference standard in patients undergoing enteric phase CT enterography. Materials and Methods Inclusion Criteria Our institutional review board approved this retrospective study, which was performed in compliance with the Health Insurance Portability and Accountability Act. Patient inclusion criteria for this study were (a) consent from the patient to use past medical records for research purposes, (b) contrast-enhanced enteric phase CT enterography performed for clinical purposes by using a single outpatient CT scanner between October 2003 and March 2004, and (c) ileocolonoscopy with intubation of the terminal ileum performed within 30 days of the date of CT enterography. Ninety-seven patients underwent enteric phase CT enterography during Published online before print /radiol Radiology 2006; 241: Abbreviation: CI confidence interval Author contributions: Guarantors of integrity of entire study, all authors; 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, F.B., J.G.F., C.D.J., J.L.F.; clinical studies, F.B., J.G.F., J.E.H., J.M.B., C.D.J., J.L.F., C.A.S., W.J.S., E.V.L.; statistical analysis, J.G.F., W.S.H.; and manuscript editing, all authors Authors stated no financial relationship to disclose. 788 Radiology: Volume 241: Number 3 December 2006

3 the catchment period (Fig 1). Forty-five patients underwent CT enterography and terminal ileoscopy within 30 days of CT (median, 1.0 day; mean, 2.9 days; range, 0 28 days). Three patients did not consent to the use of their medical records for research purposes. Thus, 42 patients who underwent enteric phase CT enterography and ileoscopy were included in our study population. Of the 42 patients, 24 (57%) were women and 18 (43%) were men. The mean age for women was 46.6 years (median, 46.0 years; range, years), and the mean age for men was 42.5 years (median, 40.5 years; range, years). Thirty-five patients who were known to have or who were suspected of having Crohn disease were referred for CT enterography, with the remaining seven patients referred for a variety of clinical indications (celiac disease, partial smallbowel obstruction, pain, or gastrointestinal hemorrhage). All CT scanning was performed by using an eight detector row CT scanner (LightSpeed Ultra; GE Healthcare, Milwaukee, Wis). CT Scanning CT enterography was performed in each patient in a similar manner. Each patient was given 10 mg of oral metoclopramide (Reglan; Pharmaceutical Associates, Greenville, SC) 75 minutes prior to CT scanning in order to increase gastric and small-bowel peristalsis. Patients then drank four 375-mL cups of diluted methylcellulose, which was composed of 1000 ml of water and 500 ml of enterocele methylcellulose (Mallinckrodt Pharmaceuticals, St Louis, Mo). The first aliquot was given immediately after metoclopramide administration, with subsequent aliquots given at 25, 50, and 65 minutes after the ingestion of the peristaltic agent. Immediately prior to scanning, patients were given 1 mg of glucagon intravenously. Contrast-enhanced CT was performed by using 270 ma, 120 kvp, 0.5-second tube rotation time, and 150 ml of intravenous contrast material (Omnipaque; Amersham Health, Princeton, NJ) containing 300 mg of iodine per milliliter. The contrast agent was injected at a rate of 4 ml/sec, with scanning initiated after a 45-second delay. A 2.5-mm section thickness was used, and images were reconstructed every 1.5 mm. Semiautomated Quantitative Image Evaluation One unblinded gastrointestinal radiologist (J.G.F., with 6 years of experience as an attending gastrointestinal radiologist), who was aware of the intubation distance at ileocolonoscopy but was not aware of the ileoscopic findings, defined transverse images and selected five bowel segments that were to be examined during quantitative analysis of small-bowel enhancement for each patient. These five bowel segments included (a) normal-appearing distended ileum, (b) collapsed ileum, (c) normalappearing distended jejunum, (d) collapsed jejunum, and (e) the terminal ileal segment (either collapsed or distended, limited to the region visualized at ileocolonoscopy) that demonstrated the greatest enhancement. For our purposes, a distended bowel was defined as a bowel in which the lumen was filled by enteric contrast material, while a collapsed bowel was defined as a bowel that demonstrated an incompletely filled lumen and coaptation of the opposing bowel walls. If hyperenhancement was visualized in any distended jejunal segment, this segment was also included in quantitative analysis. Because the small-bowel wall is only a few millimeters thick, attenuation measurements with conventional ovalshaped regions of interest are fraught Figure 1 with error arising from volume averaging. Consequently, we used the line profile tool of the Analyze 5.0 software package (Analyze Direct, Lenexa, Kan). This tool measures attenuation by sampling the attenuation of each pixel along a line. By placing a line that runs perpendicular through the bowel wall (from the fluid-filled lumen to the perienteric fat), the maximum attenuation of the bowel wall can be measured and automatically recorded. Furthermore, the line profile tool allows for the sampling of adjacent parallel or radial lines at preset intervals. By using this method, two authors (J.G.F. and F.B., a research fellow) obtained a total of 20 measurements for maximal bowel wall attenuation either at 1-mm increments when a 2-cm bowel segment was displayed in the transverse plane or at 10 increments by using a radial tool when the selected bowel segment was transected by the transverse cut plane (Fig 2). The accuracy of this method for measuring bowel wall attenuation has been previously validated in a phantom model across a range of physiologic bowel wall thicknesses (14). For each patient, the first 20 data points for maximal mural attenuation were saved and additional data points, if any, were discarded. In all cases, the analysis of 20 lines spaced at 1-mm intervals or 36 segments spaced at 10 intervals was sufficient to collect 20 data points. The mean of these 20 maximal mural attenuation measurements was calculated for each small-bowel loop in Figure 1: Selection of patients for retrospective study, with all patients undergoing enteric phase CT enterography and ileoscopy within 30 days. Radiology: Volume 241: Number 3 December

4 each patient and was used in subsequent analysis. After the selected image is loaded into Analyze 5.0 (Analyze Direct), mural attenuation measurements can be obtained in a few minutes. Visual Image Evaluation Two gastrointestinal radiologists (J.M.B. and J.E.H., with 10 and 28 years of experience, respectively, as attending gastrointestinal radiologists) reviewed all images and were blinded to all clinical, endoscopic, imaging, and pathologic information, except for the extent of terminal ileal intubation (in centimeters) at ileoscopy. The radiologists were asked to evaluate the corresponding segment of the terminal ileum in each CT enterographic data set for five CT findings of active Crohn disease. These CT findings included mural hyperenhancement, mural thickening, mural stratification, increased attenuation in the perienteric fat, and the comb sign (20,21). Mural hyperenhancement was defined as segmental enhancement in all or part (in the case of mural stratification) of the small-bowel wall that was greater than that in the adjacent smallbowel loops. Mural thickening was defined as a small-bowel wall thickness of more than 3 mm. Mural stratification was defined as a bilaminar or trilaminar appearance of the small-bowel wall, as previously described (22). The comb sign was defined as regional dilation of the vasa recta, as previously described (21). Increased attenuation in the perienteric fat was defined as fluid attenuation in the perienteric fat. Each of these CT findings was subjectively graded as not present, present to a mild degree, present to a moderate degree, or present to a severe degree. In 38 patients, the distance of inspection was reported in the ileocolonoscopy report (mean, 9.8 cm; range, 2 20 cm). In the other four patients, we assumed that the mean distance of inspection at ileocolonoscopy was 9.8 cm. To ensure that terminal ileal distention did not bias the measurements of terminal ileal attenuation in patients with active Crohn disease, a gastrointestinal radiologist (J.G.F.) scored small-bowel distention for the jejunum, ileum, and terminal ileum that was estimated to be seen at ileocolonoscopy. A previously described four-point distention scale was used for this purpose (5). For this scale, adequate distention was defined as separation of the lumen by enteric contrast material without collapse. A grade of 4 was given when more than 90% of the small bowel was adequately distended. A grade of 3 was assigned if 70% 90% of the small-bowel loops were adequately distended, and a grade of 2 was given when 40% 69% of the small-bowel loops were adequately distended. A grade of 1 was given if less than 40% of the small bowel was adequately distended. Reference Standards for Small-Bowel Inflammation and Crohn Disease Because some histologic and endoscopic findings are nonspecific for Crohn disease, a panel of three gastroenterologists (C.A.S., E.V.L., and W.J.S.; with 2, 9, and 11 years of subspecialty experience, respectively, in inflammatory bowel disease) created a combined histologic and endoscopic reference standard. When biopsy of the terminal ileum was not performed, only endoscopic assessment of the terminal ileum was used. One of the gastroenterologists (C.A.S.) reviewed the written pathology and ileos- Figure 2 Figure 2: Active Crohn disease and sampling method used for attenuation measurements in 55-year-old woman with endoscopic and pathologic diagnosis of active Crohn disease. (a) Transverse CT scan shows mural thickening and hyperenhancement (arrows) relative to normal ileum (arrowhead). Attenuation for diseased ileum was 122 HU 12. Attenuation for normal ileum was 55 HU 22. (b) Line profile tool (white line) was placed across the terminal ileal wall to measure attenuation at 1-mm intervals. Sampling is performed across entire bowel wall (from mesenteric fat to water-filled lumen) in the direction of the arrow. (c) Histogram derived from b shows attenuation measurements across small-bowel wall, demonstrating maximal attenuation (arrowhead), perienteric fat (long arrow), and contrast material filled lumen (short arrow). 790 Radiology: Volume 241: Number 3 December 2006

5 copy reports to categorize patients into one of three groups: patients with definitely active Crohn disease, patients with probable active Crohn disease, and patients without Crohn disease. Patients with definitely active Crohn disease had either chronic ileitis, which was diagnosed histologically in conjunction with any endoscopic abnormality, or acute ileitis, which was diagnosed histologically in conjunction with ulceration, stenosis, or erosions that were evident at endoscopy. Patients with probable active Crohn disease could have several combinations of histologic and endoscopic findings, including (a) chronic ileitis at histologic analysis and normal endoscopic results, (b) normal histologic findings but stenosis, ulceration, granularity, or friability at endoscopy, or (c) acute ileitis at histologic analysis, with granularity, friability, or erythema at endoscopy (23). When available, the clinical reports of small-bowel follow-through examinations were reviewed by two authors (F.B., J.G.F.) to correlate any fluoroscopic abnormalities in the jejunum with quantitative measures of jejunal mural attenuation. Statistical Analysis The mean maximal mural attenuation values for the distended jejunum and distended ileum were compared. In all patients, distended small-bowel loops were compared with collapsed smallbowel loops in the same small-bowel region (ileum or jejunum). The normalappearing distended ileum was compared with the terminal ileum in patients with active Crohn disease. Comparisons were made by using a Student t test, with P values of less than.05 indicating a statistically significant difference. A 2 analysis was used to determine differences in the grades of distention by using pairwise comparisons between the jejunum, ileum, and terminal ileum and between patients with Crohn disease and those without Crohn disease for each bowel segment. Statistical Evaluation of CT Findings and Interobserver Agreement Specific CT findings were used by each radiologist to assess the presence (mild, Figure 3 Figure 3: Normal appearance of ileum and jejunum at transverse CT enterography in 26-year-old woman who did not have Crohn disease, as proved during 5-cm ileoscopy and pathologic analysis. (a) Normal-appearing ileum (arrows) and (b) normal-appearing jejunum (arrows) are shown. Jejunal attenuation is greater than ileal attenuation (204 HU vs 56 HU). moderate, or severe) or absence of Crohn disease; these findings were then compared with the reference standard. The sensitivity for active Crohn disease and the specificity for the absence of Crohn disease were assessed and reported with 95% confidence intervals (CIs) for each radiologist. To summarize luminal distention scores, we reported the mean, standard deviation, and range of distention scores for the jejunum, ileum, and terminal ileum as seen at ileoscopy. The agreement between the two radiologists for the presence or absence of each CT finding was also examined by using the statistic, as described by Landis and Koch (23). A value of indicates slight agreement; , fair agreement; , moderate agreement; , substantial agreement; and , almost perfect agreement. Statistical Correlation of Visual and Quantitative Estimates of Mural Enhancement For quantitative data, we performed a receiver operating characteristic analysis for the terminal ileal maximal mural attenuation (absolute number) and for the ratio of terminal ileal attenuation to normal-appearing distended ileal attenuation by comparing it with our reference standard. A cutoff value for the presence of disease was selected to estimate the performance of terminal ileal attenuation and to determine the ratio of terminal ileal attenuation to normal-appearing ileal attenuation for the prediction of active Crohn disease. These cutoff values were selected by maximizing the sum of the sensitivity and specificity values. To assess agreement, quantitative data were dichotomized by using these cutoff values and were compared with the binary visual assessment (presence or absence) of mural hyperenhancement made by radiologists using statistics. We also compared the intensity of mural hyperenhancement, which was rated by each radiologist as not present, mild, moderate, or severe, with the quantitative continuous measures of mean maximal mural hyperenhancement (as measured with the line profile tool) by using Pearson correlation coefficients. For Pearson correlation coefficients, P values of less than.05 were considered to indicate a statistically significant difference. Results Patient Demographic Data Reference standard assessment yielded 12 (29%) of 42 patients with definitely active Crohn disease and an additional three patients (7%) with probable active Crohn disease. Relating to the potential presence of jejunal disease, the results of correlative small-bowel follow through were available in eight patients. Radiology: Volume 241: Number 3 December

6 None of these patients demonstrated active jejunal disease. Small-Bowel Enhancement Patterns The attenuation of the normal distended jejunum (113 HU 26 [standard deviation]) was significantly higher (P.001) than that of the normal distended ileum (72 HU 13) (Fig 3). The attenuation of the collapsed bowel was significantly higher than that of the distended bowel for both the jejunum (mean, 134 HU 23 vs 113 HU 26; P.001) and the ileum (mean, 108 HU 20 vs 72 HU 13; P.001) (Fig 4). No mural hyperenhancement in the jejunum was identified in the eight patients who underwent correlative small-bowel follow Figure 4 Figure 4: Transverse CT enterographic image in 37-year-old man demonstrates increased attenuation in collapsed normal jejunal loop (arrowheads) relative to distended normal jejunal loops (arrows) (134 HU vs 96 HU). Figure 5 through and who had no jejunal disease. Two patients without correlative fluoroscopic results demonstrated jejunal mural hyperenhancement of 218 and 196 HU (18 and 73 HU, respectively, above the attenuation of the normal-appearing distended jejunal loops). One of these patients had endoscopic findings of active duodenal Crohn disease. The attenuation of the terminal ileum in patients with active or probable active Crohn disease (n 15) was higher than that in the terminal ileum of patients without active Crohn disease (n 27) (119 HU vs 96 HU, P.001). Additionally, in patients with active Crohn disease, terminal ileal attenuation was greater than normal-appearing distended ileal attenuation (119 HU vs 74 HU, P.001). In patients without active disease according to the reference standard, terminal ileal attenuation (Fig 5) was higher than distended ileal attenuation (96 HU vs 71 HU, P.001). For patients with definitely active Crohn disease or for those who were suspected of having Crohn disease, receiver operating characteristic analysis demonstrated that the area under the curve for mean maximal terminal ileal attenuation and the ratio of terminal ileal attenuation to normal-appearing ileal attenuation was Distention scores for the terminal ileum did not differ significantly between patients with active or probable Crohn disease and those without Crohn disease (mean, 3.1 for patients with active or probable Crohn disease vs 2.8 Figure 5: Transverse enteric phase CT enterographic image in 25-year-old man with active Crohn disease demonstrates typical increase in terminal ileal mural attenuation. Mural attenuation of terminal ileum (arrows) was higher than that of distended ileal loops (arrowhead) (128 HU vs 78 HU). for patients without Crohn disease; P.40). Distention scores for the jejunum and ileum did not differ significantly between patients with (P.63) and those without (P.46) active Crohn disease. Six (14%) of 42 patients had a distention score of 1 (ie, less than 40% of the bowel segment had inadequate luminal fluid) for the terminal ileum; two of these patients had probable Crohn disease and four had no disease. The distention of the ileum and terminal ileum was significantly greater than that of the jejunum (P.001 and P.001, respectively), but there was no difference between the two ileal distention scores (P.24). Mean distention scores for the jejunum, ileum, and terminal ileum were (range, 1 4), (range, 1 4), and (range, 1 4), respectively. CT Findings and Interobserver Agreement The most sensitive visual marker of Crohn disease was mural hyperenhancement, followed by mural thickening (Table 1). Reader 1 did not identify any cases of mural thickening without mural hyperenhancement, whereas reader 2 identified two cases of mural thickening without mural hyperenhancement. However, neither of these cases of mural thickening without mural hyperenhancement had active or probable Crohn disease in the terminal ileum according to the reference standard. The most specific markers of active disease were the comb sign and increased attenuation in the perienteric fat. Readers had almost perfect agreement ( 0.83) in identifying mural thickening as a sign of Crohn disease. There was substantial agreement ( 0.61) between the two radiologists for the identification of mural hyperenhancement. Correlation of Visual and Quantitative Estimates of Mural Enhancement A threshold value of 110 HU for terminal ileal attenuation and a ratio of 1.3 for terminal ileal attenuation to normalappearing distended ileal attenuation were chosen on the basis of the sum of the sensitivity and specificity values from the receiver operating characteris- 792 Radiology: Volume 241: Number 3 December 2006

7 Table 1 Sensitivity and Specificity of CT Findings and Calculations for Interobserver Agreement CT Finding Sensitivity (%) Specificity (%) Reader 1 Reader 2 Reader 1 Reader 2 Value 95% CI Value 95% CI Value 95% CI Value 95% CI Interobserver Agreement ( ) Mural hyperenhancement* 73 (11/15) 45, (12/15) 52, (19/27) 50, (22/27) 62, Mural thickening 53 (8/15) 26, (11/15) 45, (24/27) 71, (24/27) 71, Mural stratification 53 (8/15) 26, (7/15) 21, (21/27) 58, (24/27) 71, Fat attenuation 40 (6/15) 16, (6/15) 16, (25/27) 56, (26/27) 81, Comb sign 67 (10/15) 38, (8/15) 26, (24/27) 71, (27/27) 89, Note. Data in parentheses were used to calculate percentages. * Defined as enhancement of the entire wall or portions of the wall that is greater than the enhancement of the nearby normal-appearing small-bowel loops. Table 2 Figure 6 Agreement between Quantitative Measures of Ileal Enhancement and Readers Determination of Mural Hyperenhancement Reader Coefficient* Pearson Correlation Value P Value Note. For purposes of agreement, a threshold attenuation of 110 HU was considered to represent disease. * Level of agreement for the presence or absence of disease. Agreement was calculated between readers 1 and 2 for the determination of mural hyperenhancement and terminal ileal attenuation greater than 110 HU. Calculated for the intensity of enhancement. Indicates moderate agreement. Figure 6: Transverse CT enterographic image at level of terminal ileum in 24-year-old woman shows how a collapsed bowel loop may produce spurious hyperenhancement. (a) Collapsed terminal ileum (arrow) with attenuation of 133 HU. (b) Adjacent distended terminal ileum (arrow) with attenuation of 57 HU. At ileoscopy, terminal ileum appeared normal. tic analysis. The sensitivity and specificity of quantitative measurements of terminal ileal attenuation for demonstrating active Crohn disease by using a cutoff threshold value of 110 HU were estimated to be 73% (95% CI: 45%, 92%; 11 of 15) and 78% (95% CI: 57%, 92%; 21 of 27), respectively. The sensitivity and specificity of quantitative measurements by using a ratio of 1.3 for terminal ileal attenuation to normal-appearing distended ileal attenuation were estimated to be 87% (95% CI: 59%, 98%; 13 of 15) and 52% (95% CI: 32%, 71%; 14 of 27), respectively. Statistics, which were used to assess the agreement between binary visual and quantitative data, ranged from 0.42 to 0.51, indicating moderate agreement (Table 2). Pearson coefficients were used to examine the correlation of scaled visual and continuous quantitative measures of mural enhancement and were highly significant for both readers, with P values ranging from.002 to.003 (Table 2). The reasons for false-positive results on the basis of quantitative measures of mural attenuation (and not on the basis of visual CT findings) included collapsed small bowel (n 4), backwash ileitis in a patient with ulcerative colitis (n 1), and short gut syndrome (n 1). Figure 6 shows a collapsed bowel loop that resulted in a false-positive result when quantitative attenuation measurements were used. Four false-negative results would have occurred if only the quantitative measurements were used. With respect to visual CT findings, there were four false-negative findings. Quantitative measures could have facilitated the diagnosis of active Crohn disease in one of these patients who had a terminal ileal attenuation of 125 HU. However, in the remaining three patients, quantitative measures of terminal ileal attenuation were within the normal range (69, 89, and 96 HU). Discussion Our study found that jejunal attenuation is significantly higher than ileal attenua- Radiology: Volume 241: Number 3 December

8 tion (113 HU vs 72 HU, P.001) when examining distended small-bowel loops at enteric phase CT enterography. Furthermore, collapsed jejunal and ileal loops have greater attenuation than their distended counterparts (P.001). These small-bowel attenuation patterns are important for radiologists to recognize and to take into consideration when evaluating patients for active Crohn disease, because segmental mural hyperenhancement at CT and magnetic resonance (MR) imaging is indicative of active inflammation (14,24). Understanding that collapsed bowel segments have greater attenuation than distended bowel segments and that the jejunum has greater attenuation than the ileum should help radiologists avoid overdiagnosing small-bowel inflammation. Because collapsed bowel loops have increased attenuation that is similar to that of inflamed bowel loops, secondary findings of Crohn disease (eg, perienteric fat stranding or the comb sign) must be used to make an accurate diagnosis. Terminal ileal attenuation was higher in patients with active Crohn disease than in patients without Crohn disease (119 HU vs 96 HU, P.001). Furthermore, terminal ileal attenuation was higher than normal-appearing distended ileum attenuation (119 HU vs 74 HU, P.001) in patients with active Crohn disease. This quantitative finding correlated with visual estimates of disease activity by radiologists. The most sensitive visual CT finding of Crohn disease activity was mural hyperenhancement, which had a sensitivity of 73% 80% according to radiologist assessment. Increased mural thickness had slightly lower sensitivity for active Crohn disease but was a more specific finding. There was no incremental benefit in the sensitivity of either reader for the detection of mural thickening and mural hyperenhancement combined compared with the detection of mural hyperenhancement alone. The comb sign and increased attenuation in the perienteric fat were the most specific findings of active Crohn disease. In most studies, researchers have evaluated the performance of CT enterography in comparison with clinical or radiographic criteria (1,2,6,9) and not in comparison with endoscopic and histologic reference standards, as was used in our study. However, Voderholzer et al (25) recently evaluated 56 patients with Crohn disease who underwent CT enteroclysis and capsule endoscopy. Voderholzer et al found that, for terminal ileal inflammation due to Crohn disease, CT enteroclysis had a sensitivity of 67% and capsule endoscopy had a sensitivity of 80%. In a recent study comparing portal phase CT enterography, capsule endoscopy, smallbowel follow through, and ileocolonoscopy, researchers reported sensitivities of 82%, 83%, 65%, and 74%, respectively, for active Crohn disease but found that CT enterography was more specific than capsule endoscopy (26). Hara et al (11) identified only three (50%) of six patients with terminal ileal ulcers by using CT enterography, while Hassan et al (12) identified 26 (87%) of 30 patients with Crohn disease by using CT enteroclysis, with ileoscopy as the true reference standard. In our study, the performance of readers in identifying active inflammation (sensitivity, 73% 80%) is in agreement with the results obtained in these earlier reports that incorporated endoscopy into the reference standard for the assessment of active Crohn disease. In our study, interobserver agreement for specific CT findings of active Crohn disease was high, particularly for mural thickening ( 0.83) and mural hyperenhancement ( 0.61). Such agreement between radiologists demonstrates the robustness of CT enterography as a test for active Crohn disease. In prior studies regarding the performance of CT enterography, researchers have either employed one radiologist or have reported results in consensus rather than individually (5,6,9,15). Each radiologist s visual assessment of mural hyperenhancement (absent, mild, moderate, or severe) correlated significantly with terminal ileal attenuation when the scaled degrees of hyperenhancement provided by radiologists were compared with continuous quantitative data (Pearson correlation coefficient, 0.42; P.002). Given the similar performance of visual and quantitative measures of mural attenuation in facilitating the identification of active Crohn disease, routine use of quantitative measures in identifying small-bowel inflammation is not likely to improve overall test performance. Such quantitative measures may be useful to the radiologist when attempting to classify regions that are questionable for active Crohn disease or, in the future, for providing a quantitative measurement of Crohn disease burden on the basis of reproducible measurements of mural hyperenhancement and thickness. Similar techniques would likely extend to MR enterography, during which differences in signal intensity that indicate small-bowel inflammation may be even greater and the patient is not exposed to ionizing radiation (3,27,28). We identified several patients in whom visual assessment by a radiologist could demonstrate a reason for falsepositive results by using quantitative measurements (ie, collapsed small bowel, backwash ileitis, and short gut), thereby reinforcing the fact that terminal ileal hyperenhancement is a nonspecific finding and must be interpreted with other CT findings in diagnosing active inflammatory disease. There are several potential limitations of our study. Principally, these limitations include the relatively small number of patients and the retrospective nature of the study. While we believe that one of the strengths of our study is the use of an endoscopic and histologic reference standard for active Crohn disease; this reference standard could be applied to the terminal ileum only and not to the proximal ileum or jejunum. In this regard, we cannot state with certainty that the attenuation measurements of normal-appearing jejunal loops were from normal (and not diseased) jejunum. However, eight patients underwent both small-bowel follow through and CT enterography, and both examinations failed to demonstrate active jejunal Crohn disease in these eight patients. Enteric phase CT enterography did demonstrate jejunal mural hyperen- 794 Radiology: Volume 241: Number 3 December 2006

9 hancement in two other patients, one of whom had endoscopically confirmed proximal small-bowel inflammation. Finally, the Analyze 5.0 program (Analyze Direct) that was used in this study measured maximal mural attenuation across a line that bisected the bowel wall and was therefore not susceptible to partial volume averaging, which region of interest measurements are. While these measurements take only a few minutes to obtain per patient, the process of loading the Digital Imaging and Communications in Medicine data into the analysis program can be cumbersome for large data sets; therefore, this process is not currently suitable for incorporation into clinical practice. In conclusion, radiologists should be aware that the enhancement of the jejunum is substantially greater than that of the ileum and that collapsed bowel loops have greater attenuation than their distended counterparts at enteric phase CT enterography. Mural hyperenhancement and increased mural thickness are the most sensitive markers of Crohn disease and are detected with a high degree of reproducibility. References 1. Nagamata H, Inadama E, Arihiro S, Matsuoka M, Torii A, Fukuda K. The usefulness of MDCT in Crohn s disease [in Japanese]. Nippon Shokakibyo Gakkai Zasshi 2002;99: Reittner P, Goritschnig T, Petritsch W, et al. Multiplanar spiral CT enterography in patients with Crohn s disease using a negative oral contrast material: initial results of a noninvasive imaging approach. Eur Radiol 2002;12: Bruel JM, Gallix B, Achard C, Pierredon MA, Molina E. Multidetector CT and MRI in diseases of the gastrointestinal tract [in French]. J Radiol 2003;84: [Published correction appears in J Radiol 2003; 84(6):742.] 4. Boudiaf M, Jaff A, Soyer P, Bouhnik Y, Hamzi L, Rymer R. Small-bowel diseases: prospective evaluation of multi detector row helical CT enteroclysis in 107 consecutive patients. Radiology 2004;233: Wold PB, Fletcher JG, Johnson CD, Sandborn WJ. Assessment of small bowel Crohn disease: noninvasive peroral CT enterography compared with other imaging methods and endoscopy feasibility study. Radiology 2003;229: Mazzeo S, Caramella D, Battolla L, et al. Crohn disease of the small bowel: spiral CT evaluation after oral hyperhydration with isotonic solution. J Comput Assist Tomogr 2001;25: Choi D, Jin Lee S, Ah Cho Y, et al. Bowel wall thickening in patients with Crohn s disease: CT patterns and correlation with inflammatory activity. Clin Radiol 2003;58: Lee SS, Ha HK, Yang SK, et al. CT of prominent pericolic or perienteric vasculature in patients with Crohn s disease: correlation with clinical disease activity and findings on barium studies. AJR Am J Roentgenol 2002; 179: Mako EK, Mester AR, Tarjan Z, Karlinger K, Toth G. Enteroclysis and spiral CT examination in diagnosis and evaluation of small bowel Crohn s disease. Eur J Radiol 2000; 35: Zissin R, Hertz M, Paran H, Bernheim J, Shapiro-Feinberg M, Gayer G. Small bowel obstruction secondary to Crohn disease: CT findings. Abdom Imaging 2004;29: Hara AK, Leighton JA, Sharma VK, Fleischer DE. Small bowel: preliminary comparison of capsule endoscopy with barium study and CT. Radiology 2004;230: Hassan C, Cerro P, Zullo A, Spina C, Morini S. Computed tomography enteroclysis in comparison with ileoscopy in patients with Crohn s disease. Int J Colorectal Dis 2003; 18: Buchman AL, Miller FH, Wallin A, Chowdhry AA, Ahn C. Videocapsule endoscopy versus barium contrast studies for the diagnosis of Crohn s disease recurrence involving the small intestine. Am J Gastroenterol 2004;99: Bodily KD, Fletcher JG, Solem CA, et al. Crohn disease: mural attenuation and thickness at contrast-enhanced CT enterography correlation with endoscopic and histologic findings of inflammation. Radiology 2006;238: Rollandi GA, Curone PF, Biscaldi E, et al. Spiral CT of the abdomen after distention of small bowel loops with transparent enema in patients with Crohn s disease. Abdom Imaging 1999;24: Horton KM, Eng J, Fishman EK. Normal enhancement of the small bowel: evaluation with spiral CT. J Comput Assist Tomogr 2000;24: Morrin MM, Farrell RJ, Kruskal JB, Reynolds K, McGee JB, Raptopoulos V. Utility of intravenously administered contrast material at CT colonography. Radiology 2000; 217: Sosna J, Morrin MM, Kruskal JB, Farrell RJ, Nasser I, Raptopoulos V. Colorectal neoplasms: role of intravenous contrast-enhanced CT colonography. Radiology 2003;228: Fletcher JG, Wiersema MJ, Farrell MA, et al. Pancreatic malignancy: value of arterial, pancreatic, and hepatic phase imaging with multi detector row CT. Radiology 2003; 229: Madureira AJ. The comb sign. Radiology 2004;230: Meyers MA, McGuire PV. Spiral CT demonstration of hypervascularity in Crohn disease: vascular jejunization of the ileum or the comb sign. Abdom Imaging 1995;20: Gore RM, Marn CS, Kirby DF, Vogelzang RL, Neiman HL. CT findings in ulcerative, granulomatous, and indeterminate colitis. AJR Am J Roentgenol 1984;143: Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33: Sempere GA, Martinez Sanjuan V, Medina Chulia E, et al. MRI evaluation of inflammatory activity in Crohn s disease. AJR Am J Roentgenol 2005;184: Voderholzer WA, Beinhoelzl J, Rogalla P, et al. Small bowel involvement in Crohn s disease: a prospective comparison of wireless capsule endoscopy and computed tomography enteroclysis. Gut 2005;54: Solem CA, Loftus EV Jr, Tremaine WJ, Harmsen WS, Zinsmeister AR, Sandborn WJ. Correlation of C-reactive protein with clinical, endoscopic, histologic, and radiographic activity in inflammatory bowel disease. Inflamm Bowel Dis 2005;11: Low RN, Francis IR, Politoske D, Bennett M. Crohn s disease evaluation: comparison of contrast-enhanced MR imaging and singlephase helical CT scanning. J Magn Reson Imaging 2000;11: Masselli G, Brizi GM, Parrella A, Minordi LM, Vecchioli A, Marano P. Crohn disease: magnetic resonance enteroclysis. Abdom Imaging 2004;29: Radiology: Volume 241: Number 3 December