Magnetic Resonance Imaging of the Small Bowel in Crohn's Disease: A Systematic Review and Meta-Analysis

Introduction. Crohn's disease is most commonly found in the terminal ileum and colonic region. Magnetic resonance has become a useful modality for assessing small bowel activity. In this study, we performed a systematic review and meta-analysis on the use of MR in detecting small bowel activity as well as extramural complications in Crohn's patients. Methods. Two independent reviewers sorted through articles until October 2, 2014. We included both studies providing raw data for pooling and studies without raw data. Sensitivity, specificity, likelihood ratios, and 95% confidence intervals were calculated for each study. Results. There were 27 included studies, of which 19 were included in the pooled analysis. Pooled analysis of the 19 studies (1020 patients) with raw data revealed a sensitivity of 0.88 (95% CI 0.86 to 0.91) and specificity was 0.88 (95% CI 0.84 to 0.91). In regard to detecting stenosis, pooled sensitivity was 0.65 (95% CI 0.53 to 0.76) and specificity was 0.93 (95% CI 0.89 to 0.96). Conclusion. MR imaging provides a reliable alternative in detecting small bowel activity in patients with Crohn's disease. Its advantages include high diagnostic accuracy and no radiation exposure while its disadvantages include high cost and limited availability.


Introduction
Crohn's disease is an inflammatory bowel disease that may present with systemic symptoms such as fever, fatigue, and weight loss, as well as abdominal symptoms including pain and diarrhea [1]. Unlike ulcerative colitis, Crohn's disease can manifest anywhere in the gastrointestinal tract, though it most commonly affects the terminal ileum and colon. It is estimated that almost 50% of patients with Crohn's disease will have involvement of the small bowel, and up to 30% will have small bowel involvement only [2]. Crohn's disease is most often diagnosed by a combination of clinical features, endoscopy, and histopathology. Although the exact pathogenesis is still unclear, the disease manifests itself endoscopically as focal ulcerations with skip lesions (normal appearing bowel along with areas of inflammation) [1].
Recently, there has been an increase in the use of imaging modalities in assisting the diagnosis of Crohn's disease as well as assessing disease severity. Conventional enteroclysis, ultrasound (US), computed tomography (CT), and magnetic resonance (MR) imaging have all been used to detect inflammation in the bowel [3,4]. The use of imaging for diagnosing small bowel activity has become even more relevant since traditional methods of diagnosis (e.g., endoscopy) are not able to visualize the small bowel reliably, and up to 10% of patients will have small bowel involvement not amenable to visualization by endoscopy [1]. Additionally, endoscopic methods of assessment carry the risk of procedural complications and can cause patient discomfort [5].
Multiple studies have investigated the use of imaging to diagnose small bowel activity in Crohn's patients. Although the use of conventional enteroclysis and CT have shown good diagnostic accuracy, they are limited by exposure to ionizing radiation [6]. Ultrasound is a nonradiating form of imaging but is limited because image quality is dependent on technician expertise [7].
Magnetic resonance has become a useful modality for assessing small bowel activity in Crohn's disease, with multiple studies showing great sensitivity and specificity. The use of enteral contrast agents using MR enterography protocols has allowed for better distention as well as visualization of the small bowel [8]. Previous studies and reviews have looked at the use of MR in Crohn's disease; however these studies were limited in only assessing small bowel activity or extraluminal complications [9,10]. Multiple new studies have since emerged looking at this field. In this study, we

Results
The initial search yielded 3981 studies. These studies were initially sorted by title yielding 332 studies, which were then limited to 29 studies based on abstract review. After retrieving the full articles and contacting authors for missing information, only 27 studies met the inclusion criteria (Supplementary Figure 1). Of the 27 included studies, 19 provided sufficient raw data to be included in the pooled analysis; eight studies provided only summaries of the results . All studies had a QUADAS-II score equal to or greater than 9, indicating low-risk.

Study Characteristics.
Individual study characteristics are reported in Tables 1 and 2. A total of 19 studies with 1020 patients were included in the meta-analysis. Eight studies with a total of 650 patients were only included in the systematic review (Table 3) .
In regard to imaging characteristics, most studies used a magnetic field strength of 1.5 T, with 2 studies using 1.0 T Canadian Journal of Gastroenterology and Hepatology     [12,23] and 2 studies using 3.0 T [34,35]. Three studies did not mention the magnetic field strength used [13,31,37]. Six of the 27 studies used enteroclysis as a method of introducing oral contrast [13,14,16,18,20,31], while 19 studies used standard enterography. One study used both enterography and enteroclysis [19], while 1 study did not mention how oral contrast was given [15]; neither of these studies were included in the meta-analysis. Radiologist experience with abdominal MR and time interval between MR and the reference standard varied widely between the studies (Table 2). In regard to disease activity, unless otherwise specified below, we considered positive small bowel activity to be when the individual study considered the disease active (no specific parameters were used). Maccioni et al. [17] provided raw data for both T1-weighted and T2-weighted imaging. We chose to include T2-weighted results as they have been shown previously to be more accurate for small bowel activity [39]. Oto et al. selected for patients with active disease and thus had no results for specificity [30]. Because Seiderer et al. used an anterograde endoscopic approach, the jejunum was used for analysis as very few patients had their terminal ileum intubated [20]. For Adamek et al., we used histopathology as the reference standard rather than ileocolonoscopy (both were provided) [34]. Alternatively, for Siddiki et al., we used ileocolonoscopy as the reference standard since not as many patients had histopathology results [21]. For Kumar et al., we used bowel thickening as representative of small bowel activity [37]. Jensen et al. published two studies in 2011. Because there was no overlap in the study populations, both were included in our meta-analysis [28,29].

Nonpooled Studies Summary. Of the 8 studies which
were not included in the pooled analysis, by far the largest was Grand et al. with 310 patients (out of 650 total patients) [33]. Of these, 162 underwent MR and endoscopy within 30 days and the per-patient analysis for the distal ileum revealed a sensitivity of 85% and a specificity of 79% in diagnosing Crohn's disease activity. The results of the remaining studies are summarized in Table 3.

Per-Patient Pooled Analysis.
Forest plots for the sensitivity, specificity, and sROC for the use of MR in diagnosing small bowel Crohn's disease activity are presented (Figure 1,  Supplementary Figure 2). Pooled analysis of the 19 studies with raw data revealed a sensitivity of 0.88 (95% CI 0.86 to 0.91) with a heterogeneity of 2 = 80.38 and 2 of 77.6% (Figure 1). The pooled specificity was 0.88 (95% CI 0.84 to 0.91) with a heterogeneity of 2 = 55.11 and 2 of 67.3% (Figure 1). Using a random effects model, the positive likelihood ratio was 5.2 (95% CI 2.62 to 10.29) and the negative likelihood ratio was 0.17 (95% CI 0.11 to 0.27) (Supplementary Figure 3). Using Moses' constant linear model, we were able to construct an sROC with an AUC of 0.93 (Supplementary Figure 2). As expected, most values reside in the left upper corner, suggesting high sensitivity and specificity.

Subgroup Analysis.
Analysis of only prospective studies revealed a pooled sensitivity of 0.89 (95% CI 0.86 to 0.92) and a pooled specificity of 0.90 (95% CI of 0.86 to 0.93) (Figure 2). Five studies used enteroclysis as the method of administrating oral contrast [14,16,18,20,31]. Pooled analysis of these 5 studies gave a sensitivity of 0.84 (95% CI of 0.74 to 0.91) and a pooled specificity of 0.89 (95% CI of 0.78 to 0.96) as demonstrated in the forest plots ( Figure 3).

Discussion
Our study represents the largest systematic review (in terms of patients and number of studies) of MR imaging for the detection of small bowel activity in Crohn's disease. Like previous studies and reviews, we demonstrate that MR imaging possesses high sensitivity and specificity in detecting small bowel activity [3,9,10]. Along with a relatively high positive likelihood ratio and relatively low negative likelihood ratio, it can be used in combination with pretest probabilities to determine small bowel activity in the appropriate clinical setting. The results from this meta-analysis are similar to those previously reported [9]. Of note, many of the studies differ from those included in previous reviews, not only because we focused only on the small bowel, but also because we included both studies with per-patient and per-segment analysis (though only per-patient analysis was pooled). Additionally, differences in response rates in regard to contacting authors likely explain the discrepancy in studies included in our review as compared to others.  One of the many research areas in MR imaging is the use of enteroclysis in inflammatory bowel disease. Enteroclysis has been proposed to provide better small bowel distension because the contrast is provided directly through nasojejunal intubation rather than orally [40]. The limitations of enteroclysis are that it is not as widely available, requires fluoroscopic insertion of a nasojejunal tube (thus exposing patients to radiation), and is less well tolerated by patients [41]. A subgroup analysis of studies using enteroclysis did not demonstrate higher sensitivity and specificity. Similarly, Negaard et al. directly compared use of oral contrast and enteroclysis and also did not find any significant difference [19]. One explanation for the lack of increment benefit of enteroclysis is because small bowel Crohn's disease usually affects the terminal ileum distally, rather than proximally, where the advantages of enteroclysis are more apparent. Overall, enteroclysis has not yet been shown to have any significant difference in diagnostic accuracy, and its role in Crohn's patients is still uncertain. Further study is warranted in determining potential benefits of enteroclysis in Crohn's disease proximal to the ileum.
In regard to intra-and extraluminal complications such as fistulas, abscesses, and stenosis, MR has been theorized to be the gold standard, since ileocolonoscopy can only assess luminal disease, and surgery is too invasive and not a feasible diagnostic modality. Our analysis revealed fairly high specificity in detecting stenosis, but only moderate sensitivity. However, our analysis was limited due to the small number of studies included. Previous studies looking at both small and large bowel have shown relatively high detection rates for stenosis [26,42]. Nevertheless, a metaanalysis by Qiu et al. showed that CT imaging may be better at detecting fistulas and stenosis. However, the results were not statistically significant, and the sensitivities and specificities were comparable to more recently published studies [43,44].
No differences were noted in detecting abscesses. The results for pooled analysis by Qiu et al. for stenosis revealed similar numbers to our study (sensitivity 65.3%, specificity 94.4%). Current European guidelines recommend MR, US, and CT enterography or enteroclysis for the detection of intestinal involvement and penetrating lesions in CD. Additionally, the use of small bowel follow-through or small bowel enteroclysis is acceptable for detection of stenosis. They are also the recommended techniques for detection of extramural complications of CD [1,45]. American guidelines are less specific but do include the use of MR, amongst a multitude of other imaging modalities, to delineate and discriminate intra-abdominal masses/abscesses and in the evaluation of small bowel pathology in patients with CD [46].
One of the difficulties in replacing the gold standard with MR imaging is the lack of standardization of the imaging signs suggestive of active disease, especially with the growing number of sequences available. Previous studies have demonstrated that the most accurate signs of inflammation for MR were wall enhancement, mucosal lesions, and wall T2 hyperintensity [39,47]. In one study, Maccioni et al. looked at the difference using T1-versus T2-weighted imaging. They found that T2-weighted images provided greater sensitivity and specificity in diagnosing ileal lesions [17]. Additionally, a study by Udayasankar et al. found similar results in both the small and large bowel [48]. Previous studies have also found that sequences using diffusion-weighted imaging had high sensitivity and specificity [30,49,50]. Recently, there have been development of validated scoring systems including the MaRIA (Magnetic Resonance Index of Activity) score for assessment of disease activity and severity, the Lemann score, or the Crohn's Disease Digestive Damage Score, which takes into account many factors (clinical, endoscopic, and imaging findings) and attempts to measure cumulative damage [51,52].   An important consideration is whether the results of MR imaging change clinician management. A study by Mendoza et al. showed that MR helped in decision-making in more than half of patients, especially those involving the use of biological therapies and surgery [53]. Messaris et al. showed that 69% of patients had changes to medical and/or surgical management after clinicians were given MR imaging results [54]. Similar results have shown that MR findings influence surgical approaches to managing Crohn's patients [55].
Some of the limitations of our study include the varied length of time between the reference standard and MR imaging. Since clinical activity can change quite drastically, especially with the use of medications, some of the results might have been inaccurate in determining disease activity. Secondly, since we used per-patient data, we were not able to differentiate severity in regard to determining small bowel activity. Other studies have suggested that MRI has good correlation with Crohn's severity indices [56]. Similarly, we were unable to perform per-segment analysis which might have led to an overestimation of the accuracy of MR imaging. This is likely due to the use of endoscopy as a reference standard and its inadequacy in assessing more proximal small bowel. Other limitations include the fact that we were only able to analyse one complication (stenosis), and that others (such as abscess and fistulas) are not well visualized on endoscopy. Similarly, due to the small number of studies, we were not able to determine whether more advanced MR (such as MR with 3.0 T magnetic field strength) had any additional benefit. Finally, the large heterogeneity amongst the studies, including reference standards, radiologists experience, and results, suggests that more definitive studies might still be required. Sources of heterogeneity include inclusion of studies of different sample size and different criteria for disease activity, as well as inclusion of studies using different MR enterography/enteroclysis protocols (including different magnetic field strengths, oral contrast, and radiologist experience).
In regard to other modalities of imaging, many studies assessed the use of ultrasound and computed tomography. The benefit of ultrasound is that it does not involve ionizing radiation and is relatively inexpensive [3]. Previous studies assessing ultrasound have demonstrated high sensitivities and specificities. There is one large-scale trial comparing US and MR currently in progress: the UK-based MR Enterography or Ultrasound in Crohn's disease (METRIC) trial [3,57]. The use of US is thought to be limited by operator-experience; however MR has also been shown to have interobserver variability. However, the use of scoring systems such as MaRIA has been proven to improve interobserver agreement [58]. Similarly, one meta-analysis has shown similar accuracy between CT and MR. CT has the benefit of being widely available and cost-effective. It, however, also carries the risk of ionizing radiation, especially amongst patients who might require multiple scans throughout the course of their life-long disease [43].
In conclusion, MR imaging provides a reliable alternative to ileocolonoscopy in detecting small bowel activity in patients with Crohn's disease. Its advantages include high diagnostic accuracy, favorable safety profile, and the ability to assess intra-and extraluminal complications. Its disadvantages include high cost and limited availability. Nevertheless, with the rapid expansion in MR accessibility, it will likely play a greater role in the future in both the diagnosis and management of patients with Crohn's disease.