| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Evidence-based Practice |
1 From the Departments of Radiology (K.H., S.B., J.S.) and Nuclear Medicine (R.J.B.), Academic Medical Center, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands. Received April 3, 2007; revision requested June 1; revision received June 27; accepted July 19; final version accepted September 24. Address correspondence to K.H. (e-mail: k.horsthuis{at}amc.uva.nl).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Materials and Methods: MEDLINE, EMBASE, CINAHL, and Cochrane databases were searched for studies on the accuracy of US, MR imaging, scintigraphy, CT, and PET, as compared with a predefined reference standard, in the diagnosis of IBD. Sensitivity and specificity estimates were calculated on per-patient and per-bowel-segment bases by using a bivariate random-effects model.
Results: Thirty-three studies, from a search that yielded 1406 articles, were included in the final analysis. Mean sensitivity estimates for the diagnosis of IBD on a per-patient basis were high and not significantly different among the imaging modalities (89.7%, 93.0%, 87.8%, and 84.3% for US, MR imaging, scintigraphy, and CT, respectively). Mean per-patient specificity estimates were 95.6% for US, 92.8% for MR imaging, 84.5% for scintigraphy, and 95.1% for CT; the only significant difference in values was that between scintigraphy and US (P = .009). Mean per-bowel-segment sensitivity estimates were lower: 73.5% for US, 70.4% for MR imaging, 77.3% for scintigraphy, and 67.4% for CT. Mean per-bowel-segment specificity estimates were 92.9% for US, 94.0% for MR imaging, 90.3% for scintigraphy, and 90.2% for CT. CT proved to be significantly less sensitive and specific compared with scintigraphy (P = .006) and MR imaging (P = .037)
Conclusion: No significant differences in diagnostic accuracy among the imaging techniques were observed. Because patients with IBD often need frequent reevaluation of disease status, use of a diagnostic modality that does not involve the use of ionizing radiation is preferable.
Supplemental material:
http://radiology.rsnajnls.org/cgi/content/full/247/1/64/DC1
http://radiology.rsnajnls.org/cgi/content/full/247/1/64/DC2
© RSNA, 2008
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Although there is increasing interest in determining the degree of inflammatory activity, the clinically most important factor for patients suspected of having IBD is whether disease is present. For symptomatic patients known to have IBD, meanwhile, it is important to determine whether the symptoms are functional or are due to inflammatory activity or residual fibrotic stenosis.
Ileocolonoscopy with tissue sampling generally has been considered the most valuable tool for diagnosing disease in the colon and terminal ileum (2,3). For years, the reference standard for involvement of the small bowel in CD has been small-bowel barium examination performed by using an enteroclysis technique or small-bowel follow-through (3,4). Both ileocolonoscopy and small-bowel barium examination have the advantage of enabling the detection of disease at an early stage owing to their capability to depict the mucosal surface. However, both procedures are time-consuming, and patient tolerance of ileocolonoscopy is low because of the extensive bowel preparation necessary and the discomfort experienced during the procedure (5). Ionizing radiation and extensive bowel preparation are additional drawbacks of small-bowel barium examinations. Moreover, in patients with severe UC, ileocolonoscopy is relatively contraindicated because of the increased risk of perforation.
In recent years, many studies have been performed to investigate the diagnostic potential of less-invasive and more patient-friendly imaging modalities—namely, ultrasonography (US), magnetic resonance (MR) imaging, scintigraphy (both planar and single photon emission computed tomography [SPECT]), computed tomography (CT), and to a lesser extent positron emission tomography (PET). However, published studies vary widely in terms of the reported sensitivities and specificities, from those in which the reported diagnostic capacity approaches that of ileocolonoscopy or small-bowel barium examination (6–10) to those yielding considerable underestimations or overestimations of disease (11–18). To our knowledge, only one meta-analysis focused on imaging IBD has been published to date (19); however, only the role of US in the detection of CD was investigated. Thus, the purpose of our study was to compare, by performing a meta-analysis, the accuracies of US, MR imaging, scintigraphy, CT, and PET in the diagnosis of IBD.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
The title and/or abstract of all retrieved articles was assessed by one observer (K.H.) to determine the eligibility of the articles for inclusion. The reference lists of the review articles and eligible studies were checked manually to identify other relevant articles. Hand searching of major journals was not performed. Only data that were presented as full-text articles were eligible for inclusion. If from reading the abstract it became evident that the article did not describe a prospective study, described a study involving MR imaging at a field strength of 0.5 T or lower, and/or described a study involving fewer than 15 patients, the article was considered ineligible. Although we would have preferred to include only large studies to increase the statistical power of the analyses, a minimal sample size of 15 patients was chosen because most studies involving patients with IBD have small samples. All other eligible articles were retrieved as full-text articles.
Study Selection
Two reviewers (K.H., S.B.) independently checked all retrieved articles to determine whether they satisfied the following criteria: (a) 15 or more patients were involved; (b) the study design was prospective; (c) US, MR imaging, scintigraphy, CT, or PET was used to diagnose IBD; (d) the patient population comprised patients suspected of having IBD, both patients suspected of having IBD and patients known to have IBD, or patients known to have IBD but suspected of having recurrence; (e) histopathologic, ileocolonoscopic, and/or intraoperative findings were used as the reference standard for examination of the colon and terminal ileum, and histopathologic, small-bowel barium examination, and/or intraoperative findings were used as the reference standard for examination of the small bowel; (f) the criteria for positive US, MR, scintigraphy, CT, or PET findings were defined; and (g) the data necessary to calculate 2 x 2 contingency tables were reported. If all criteria were met, the article was included in the study. Disagreements between the two reviewers regarding study inclusion were resolved by consensus. Authors of the primary research were not approached for additional information.
Study Characteristics
Both reviewers independently assessed study characteristics and extracted relevant data (described in detail in the following paragraphs) by using a standardized form. The reviewers were not blinded to the authors' information (eg, authors' affiliation) or the journal title. Inconsistencies in assessment between the reviewers were resolved by consensus.
Patient characteristics.—The following patient characteristics were recorded: (a) number of patients, (b) numbers of male and female patients, (c) mean patient age and patient age range, (d) prevalence of IBD (CD and UC), and (e) inclusion criteria used (eg, based on medical history, physical examination, and laboratory findings) to select patients for the study.
Study quality assessment.—To assess study quality characteristics, the QUADAS (quality assessment of diagnostic accuracy studies) tool was used as a guideline. The QUADAS tool enables reviewers to evaluate the quality of studies, especially investigations of diagnostic accuracy (20,21). The following characteristics were assessed: (a) whether the index test (ie, the radiologic examination being evaluated) was assessed without knowledge of the reference-standard test results, (b) whether the index test findings were interpreted without clinical information, (c) whether the criteria used to diagnose IBD with the index test were clearly described, (d) the time interval between the index test and the reference-standard examination, (e) whether the reference-standard examination findings were interpreted without index test information and results, and (f) whether the reference-standard examination was described correctly.
Imaging features.—The following US features were recorded, if available: (a) type (linear or curved) and frequency of probe(s) used; (b) use of bowel preparation and, if so, type of bowel preparation (bowel cleansing, fasting, and/or diet); (c) type of scanning (conventional gray scale, pulsed, color, or power Doppler); and (d) amount and type of luminal contrast medium (enteroclysis, oral, and/or rectal), if administered.
The following MR imaging features were recorded, if available: (a) magnetic field strength; (b) type of coil used (body or surface); (c) use of bowel preparation and, if so, type of bowel preparation (bowel cleansing, fasting, and/or diet); and (d) amount and type of intravenous and/or luminal contrast medium (enteroclysis, oral, and/or rectal), if administered.
The following radionuclide imaging features were recorded, if available: (a) whether leukocytes were labeled in vitro or intravenously injected labeled antigranulocyte antibodies were used, (b) type and dose of labeling agent, (c) timing of scanning, and (d) scanning technique (SPECT or planar).
The following CT imaging features were recorded, if available: (a) type of scanner (single-section helical, multisection helical, or nonhelical); (b) use of bowel preparation and, if so, type of bowel preparation (bowel cleansing, fasting, and/or diet); and (c) amount and type of intravenous and/or luminal contrast medium (enteroclysis, oral, and/or rectal), if administered.
The following PET imaging features were recorded, if available: (a) type of scanner (dedicated full-ring or other), (b) type and amount of tracer, and (c) timing of scanning.
Imaging criteria and reference standard.—For each study, the imaging criteria (eg, abnormal bowel wall thickening) used to diagnose IBD with the given imaging test were noted. The reference-standard examination used to verify the imaging findings (ie, surgery, histopathology, ileocolonoscopy, or small-bowel barium examination) was also recorded for each study.
Data Synthesis and Analysis
For each study, 2 x 2 contingency tables consisting of true-positive, false-positive, false-negative, and true-negative results were constructed.
Sensitivity and specificity estimates.—Summary sensitivity and specificity estimates were calculated on a per-patient and/or per-segment basis, depending on the way the data were presented: In some studies, patient-based data were reported, whereas in others, segmental data were provided without the possibility of adapting the data to calculate patient-based accuracy values. To calculate summary sensitivity and specificity values, a bivariate approach was used. With this approach, it is assumed that the true values of logit-transformed sensitivity and logit-transformed specificity of the included studies follow an approximately bivariate normal distribution. Mean logit-transformed sensitivity and specificity values, with corresponding standard errors, were calculated by using random-effects and/or fixed-effects (mixed) models, depending on the best fit. After the antilogit transformation of the mean logit-transformed sensitivity and specificity, sensitivity and specificity estimates, with corresponding 95% confidence intervals, were obtained. This bivariate model was analyzed by using linear and nonlinear mixed-model techniques (SAS proc Nlmixed; SAS Institute, Cary, NC).
Subgroup analyses.—To determine possible explanations for the heterogeneity in diagnostic accuracy, a subgroup analysis of predefined factors was performed. The predefined factors regarding patient characteristics were disease type (CD, UC, or both), patient age (patients younger than 18 years vs adult patients), and disease location (small bowel, colon, or both). In addition, a subgroup analysis of the imaging features and imaging criteria used to diagnose IBD was performed. Subgroup analyses were performed on a per-patient basis and only when enough data were available (ie, three or more studies per subgroup).
A z test for unpaired groups was performed to evaluate differences between the imaging modalities and between the subgroups; P < .05 was considered to indicate a significant difference. Microsoft Excel 5.0 (Microsoft, Redmond, Wash), SPSS 12.0 (SPSS, Chicago, Ill), and SAS 9.1 (SAS Institute) were used to perform the statistical analyses.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
US was evaluated in 11 studies (18,22–31); MR imaging, in 11 (6,8,9,11,12,30,32–36); scintigraphy, in nine (13,15,17,29,37–41); and CT, in seven (13,36,39,42–45). No studies in which the accuracy of PET for the diagnosis of IBD was assessed were selected.
Study and Patient Characteristics
The three cases of inconsistency in the reviewers' assessments of the included studies were resolved by consensus. Only one of the 33 included studies solely involved patients with UC; in the remaining 32 studies, patients with CD (n = 20) or both patients with CD and patients with UC (n = 12) were included (Table 1). In 11 of the 33 included studies, no criteria for inclusion were described, whereas in five studies, no clear-cut criteria were given.
|
|
|
|
|
|
|
|
Enough data on imaging features were available to calculate differences in accuracy between MR enterography and MR enteroclysis. The sensitivity of MR enterography for diagnosing IBD was significantly lower than the sensitivity of MR enteroclysis (P = .046), whereas specificity values were comparable (Fig E5, http://radiology.rsnajnls.org/cgi/content/full/247/1/64/DC2).
Regarding the US criteria used to diagnose IBD, a comparison between studies involving a wall thickness cutoff value of 3 mm and those involving a cutoff value of 4 mm or greater (including studies in which the cutoff value was 5 mm) could be performed. No significant differences were observed (Fig E5, http://radiology.rsnajnls.org/cgi/content/full/247/1/64/DC2). When subgroups were defined according to the exact cutoff value provided (ie, 3, 4, or 5 mm), sensitivity was lowest with a threshold of 5 mm (79.5%) compared with the sensitivities calculated with thresholds of 3 mm (91.0%) and 4 mm (94.1%). For CT and scintigraphy, insufficient data were available to perform subgroup analysis of the patient characteristics or the test features and imaging criteria used.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
MR imaging performed better when bowel filling was performed by using an enteroclysis technique than when contrast medium was administered orally. On a per-patient basis, US performed better in the prediction of IBD absence when both the small bowel and the colon were examined than when only the small bowel was evaluated.
We believe there were advantages to our meta-analysis: Only prospective studies involving the specific patient population of interest were included, and data analysis was performed on both per-patient and per-segment bases. We chose to include both studies with accuracy values reported on a per-patient basis and those with accuracy values reported on a per-bowel-segment basis because information about the capability of a given test in localizing disease and determining the extent of disease, in addition to its capability in diagnosing disease, can be obtained from segmental data. Analysis on a per-patient basis probably leads to overestimation of sensitivity values, because any patient with disease is considered to have true-positive findings without consideration of whether the localization of disease is correct or not. However, the analyses in our study were restricted to a dichotomous diagnostic level (ie, disease present or absent), meaning that no conclusions regarding the capability of the imaging modalities in determining the degree of disease activity could be drawn from our data.
An important potential bias of our study was that both patients suspected of having disease and those known to have disease were included, although these two populations have inherently different thresholds for the diagnosis of abnormality. In a patient known to have IBD, the threshold for diagnosing disease probably would be lower owing to a higher index of suspicion. Although this bias probably influenced our study results, we tried to limit this influence by using mixed models for data analysis, which accounts for the heterogeneity between studies caused by different threshold settings or other forms of residual heterogeneity.
An important source of heterogeneity across studies was the imaging criteria used to diagnose IBD. Although subgroup analysis could have provided some guidance as to which criteria should have been used for diagnosis, owing to the scarce data available and the fact that precise definitions of parameters were often lacking, analysis of only one parameter—bowel wall thickening—and for only one modality—US—could be performed.
Cutoff values to differentiate a normal from a thickened bowel wall varied between 3 and 5 mm in the US studies included in our meta-analysis. Although both the sensitivity and the specificity for diagnosing IBD were highest with use of a cutoff value of 4 mm, the limited amount of data prevented us from investigating whether accuracy estimates differed significantly between the different cutoff values. However, the sensitivity achieved with use of the 5-mm threshold was clearly lower than the sensitivity estimates achieved with lower thresholds. In the meta-analysis of US performed by Fraquelli et al (19), different sensitivity and specificity estimates were also seen, depending on the chosen bowel wall thickness threshold: Sensitivity decreased and specificity increased when the threshold changed from 3 to 4 mm. However, as was the case in our study, no test was performed to determine if this difference was significant.
It would have been interesting to determine the diagnostic performance of US and MR imaging in the diagnosis of UC and CD separately. However, in only three of the seven MR imaging and US studies involving both patients with CD and patients with UC was an attempt made to differentiate between these two disease subtypes (8,27,34). In the other studies, accuracy values could not be calculated for the diagnosis of CD and UC individually, because separating the data was not possible. Because the inflammation with UC is exclusively mucosal whereas transmural inflammation is seen with CD, theoretically, more bowel wall thickening should be expected with CD than with UC, while other wall features also can differ (47). Thus, combined analysis of the findings in patients with CD and those in patients with UC with use of identical radiologic criteria does not necessarily reflect with total correctness the accuracy of MR imaging and US in the diagnosis of IBD.
An important limitation of our meta-analysis was the heterogeneity in the reference-standard examinations used to compare the imaging tests. However, only those modalities that are widely accepted as adequate and objective in the diagnosis of IBD were included. Although one might argue that small-bowel barium examination is increasingly being found to be an imperfect reference standard, established superior reference tests for evaluation of the small bowel, such as double-balloon endoscopy (48,49) and video capsule endoscopy (50,51), were not commercially available until recently and at present have limited availability, which precluded the inclusion of these techniques as reference standards in this meta-analysis.
Heterogeneity in the different features of each imaging technique was also seen. Subgroup analysis of the imaging features could be performed only for the different bowel-filling methods used in MR imaging. Oral contrast medium intake was associated with lower sensitivity compared with contrast medium administration by means of enteroclysis. A collapsed bowel can hide lesions or mimic disease by suggesting an abnormally thickened and/or enhancing bowel wall (52–55). The fact that the bowel distention achieved with oral contrast medium can be inadequate in portions of the small bowel might explain the lower sensitivity of MR imaging with oral contrast medium. However, a side-by-side comparison of accuracy between MR enterography and MR enteroclysis was performed in a small prospective study, and no significant differences between the two methods of luminal contrast medium administration were observed (56). It might be advisable to perform a side-by-side comparison of these two methods in a large population to determine which one to use for the diagnosis of IBD.
Finally, heterogeneity in the study design characteristics was also observed. Because the study design can influence the reporting of diagnostic test accuracy and limit the (internal) validity of study results (57,58), it is important to evaluate study quality when reviewing articles. Although patient and study design characteristics were assessed in this meta-analysis, the effects of these factors could not be examined because of incomplete data reporting. Although we attempted to analyze these heterogeneous data by using appropriate analytic approaches, publication bias remained an issue. We did not study publication bias for the following reasons: (a) A recent study (59) in which different statistical methods (60–62) of detecting publication bias were compared revealed that the methods were diverse and, when compared with one another, yielded different estimates; (b) to our knowledge, no registry of diagnostic accuracy studies as opposed to clinical trials exists; and (c) all of the studies had small samples, so it was impossible to determine whether there was an association between sample size and diagnostic performance.
Because the accuracy values for US, MR imaging, scintigraphy, and CT were comparable in this meta-analysis, it might be justified to make a well-considered choice for either of these techniques based on their specific advantages and disadvantages. Because of the relapsing nature of IBD and the young age at which it usually develops, frequent reevaluation of disease is necessary in many patients. Therefore, it might be preferable to use a technique that does not involve ionizing radiation, and the patient-friendliness of the modality also should be considered for this specific patient population. Although the costs of the respective examinations are not to be neglected in a cost-benefit analysis, these have not been taken into account here.
The ionizing radiation used and the long duration of the examination are drawbacks of scintigraphy. Noninvasiveness, low cost, and widespread availability make US a useful modality for imaging IBD. However, US has limitations: The effectiveness of this technique depends on the experience of the operator performing the examination, and the gastrointestinal tract cannot be visualized in its entirety. It would have been valuable to study the effect of operator experience on diagnostic performance. However, exact data on observer experience were available in only one article (23). The advantages of MR imaging include the possibility for cross-sectional imaging in any plane, the absence of ionizing radiation, and the easier follow-up of disease status with MR images than with US images because with US, only selected images from an essentially dynamic examination are available at a later time, while MR examinations are completely standardized and all data can be fully saved.
Although CT is widely used to evaluate IBD, the findings of our meta-analysis, as well as the relatively large radiation dose and the intravenous iodine-based contrast medium needed for CT, favor the use of US or MR imaging. However, a critical point is that although US and MR imaging are often used to evaluate the abdomen in Europe, in the United States, CT is more often used for this purpose. This inclination is reflected by the fact that all of the US studies included in this meta-analysis and the majority of the MR imaging studies included were conducted in Europe. Thus, for radiologists in the United States, the accuracy of US and MR imaging in the diagnosis of IBD might be lower because these techniques are less frequently used, just as the accuracy of CT in the diagnosis of IBD might be lower in Europe owing to less frequent use.
Nevertheless, before any of these imaging tests can have a large role in the diagnosis of IBD, the imaging criteria that are consistent with CD and UC should be clearly established. If standardized criteria were available internationally, larger trials would be possible and comparisons between studies would be simplified. For these purposes, a more standardized technical imaging approach also would be advisable. Therefore, future research should be focused on standardization of the preparation, imaging technique, and imaging criteria used to diagnose IBD, besides including larger numbers of patients. In conclusion, we propose US or MR imaging as the imaging modality of first choice for the diagnosis of IBD; however, more research is needed to improve the accuracy of these techniques.
|
ADVANCES IN KNOWLEDGE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
IMPLICATION FOR PATIENT CARE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
FOOTNOTES |
|---|
Abbreviations: CD = Crohn disease • IBD = inflammatory bowel disease • UC = ulcerative colitis
Author contributions: Guarantor of integrity of entire study, K.H.; 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, K.H., S.B., R.J.B.; statistical analysis, K.H., S.B.; and manuscript editing, all authors
Authors stated no financial relationship to disclose.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. A. Siddiki, J. L. Fidler, J. G. Fletcher, S. S. Burton, J. E. Huprich, D. M. Hough, C. D. Johnson, D. H. Bruining, E. V. Loftus Jr., W. J. Sandborn, et al. Prospective Comparison of State-of-the-Art MR Enterography and CT Enterography in Small-Bowel Crohn's Disease Am. J. Roentgenol., July 1, 2009; 193(1): 113 - 121. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. M. Bierig and A. Jones Accuracy and Cost Comparison of Ultrasound Versus Alternative Imaging Modalities, Including CT, MR, PET, and Angiography Journal of Diagnostic Medical Sonography, May 1, 2009; 25(3): 138 - 144. [Abstract] [PDF]
|
|
|
|
|
|
S. Kuzmich, D. C. Howlett, A. Andi, D. Shah, and T. Kuzmich Transabdominal Sonography in Assessment of the Bowel in Adults Am. J. Roentgenol., January 1, 2009; 192(1): 197 - 212. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J. Oates, D. J. M. Tolan, K. Horsthuis, S. Bipat, and J. Stoker Standard of Reporting for Studies on Inflammatory Bowel Disease Radiology, October 1, 2008; 249(1): 390 - 391. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
