Introduction

Crohn disease (CD) is a chronic bowel inflammatory disease mainly affecting the terminal ileum, characterized by relapses and remissions over time [1]. Although the introduction of immunosuppressive therapies has reduced the risk of relapses and complications, surgery is still required in up to 50% of patients, when medical treatment is ineffective [1,2,3]. However, the resection of the affected loops is not curative. Endoscopic recurrence at the anastomotic site (i.e., the neo-terminal ileum) may be detected in up to 90% of asymptomatic patients within one year, while symptomatic recurrence affects up to 30% of patient within three years [4]. Moreover, the continuing process of inflammation and healing of the bowel wall may result in fibrotic changes, which may lead to fibrostenosis of the anastomosis. Unlike inflammatory stenosis, fibrostenosis is not sensitive to the anti-inflammatory treatments and requires surgery or endoscopic treatment.

Ileocolonoscopy is the standard of reference in the evaluation of post-surgical recurrence in CD. However, it is restricted to lumen visualization. Imaging modalities such as computed tomography-enterography (CTE), CT-enteroclysis and magnetic resonance-enterography (MRE) are non-invasive and accurate tools for the assessment of CD activity, allowing an overview of both intestinal and extraintestinal manifestations of the disease [5, 6]. MRE is currently the preferred modality owing to the absence of radiation exposure [5, 6]. However, while the role of CT-enteroclysis in post-operative CD has been established, only a few studies have evaluated whether MRE has reliable performances in the evaluation of the anastomotic site and in particular in distinguishing between inflammatory recurrence and fibrostenosis [7,8,9].

The purpose of this study was to evaluate the accuracy of MRE in the assessment of the anastomotic status in patients with CD who have undergone ileocolic resection.

Materials and methods

Patient population

A retrospective search of the databases of two tertiary referral institutions was performed to retrieve all consecutive patients with CD who underwent MRE from January 2013 to December 2016. This initial search retrieved a total of 279 patients. A search was then performed to identify those who had prior ileocolic resection with ileocolic anastomosis, resulting in 72 patients. Twenty patients were further excluded for the following reasons: (i) MRE and ileocolonoscopy performed more than 10 days of each other (n = 15); (ii) poor quality of MRE images or lack of small bowel distension (i.e., non-compliant patients, discontinued examination) (n = 4); and (iii) absence of contrast-enhanced MRE images (n = 1). Figure 1 shows flowchart of patients into the study.

Fig. 1
figure 1

Study flowchart

Full size image

The study population included 52 patients. For clinical reasons, 6 patients underwent MRE more than once in the study time, with a total of 62 MRE examinations included. On 62 MRE, there were 25 men and 37 women with a mean age of 47 ± 15 (SD) years (range: 25–90 years). Mean time between surgery and MRI was 8 ± 3 (SD) years (range: 5 months–43 years). Mean time between repeated MREs of these six patients was 16 months (range: 11–24 months).

Indications of the 62 MRE examinations were: (i) clinical recurrence in 20/62 (32%); (ii) evaluation of response to therapy in 23/62 (37%); (iii) recurrence as evidenced at endoscopy in 10/62 (16%); (iv) evaluation of fistula and/or abscess in 5/62 (8%); and (v) severe stenosis in 4/62 (6%).

Institutional review board of both institutions approved the study, and a waiver for informed consent was obtained.

MRE protocol

All examinations were performed on 1.5-T magnets (Magnetom®Avanto, BV17Version, Siemens-Healthineers, or Signa®HDx Excite Twin Speed, GE-Healthcare) using a phased-array torso coil. All patients, who were at least 6 h-fasting, were instructed to drink 1000 mL of polyethylene glycol solution (Selg Esse®, Alfasigma or Fortrans®, Ipsen Pharma) in around 40 min at regular intervals to obtain an adequate and homogeneous distension of intestinal loops. To reduce bowel peristalsis, butylscopolamine at a dose of 1 mg i.v. (Buscopan®, Boehrngen) or glucagon at a dose of 1 mg (Glucagen®, Novo-Nordisk) was administrated before MRE examination.

The standard protocol included T2-weighted single-shot, steady state and gradient-echo three-dimensional (3D) T1-weighted sequences in the axial and coronal planes before and after intravenous administration of contrast material (gadoterate meglumine, Dotarem®, Guerbet or Gd-BOPTA Multihance®; Bracco Imaging SpA) at a dose of 0.2 mL/kg body weight and an injection rate of 2 mL/sec at 45 and 75 s after injection. In addition, T2-weighted fat suppressed images were obtained on axial plane. In Institution 1, axial diffusion-weighted images (DWI) were also acquired, representing 37 MRE examinations. MREs were performed with the patient in supine position in Institution 1 and in prone position in Institution 2. No differences in MRE evaluation have been reported using these two different acquisition techniques [10]. Table 1 describes MRE protocols of both institutions.

Table 1 MR-enterography protocols for Institution 1 and Institution 2
Full size table

Imaging analysis

MRE image analysis was performed by two radiologists with 15 (M.B.) and 3 years (C.P.) of experience in MRE working in consensus. Both readers were blinded to clinical history, endoscopic and histologic results. Several MRE findings were evaluated using a standard data collection form, including intestinal and extra-intestinal features [11]. At the end of the reading session, the two radiologists were asked to make a final diagnosis for each MRE examination (i.e., normal anatomosis, inflammatory recurrence or fibrostenosis) using the whole set of sequences for each examination.

Intestinal features included anastomotic wall thickening (> 3 mm) and marked anastomotic wall thickening (> 6 mm) on T2-weighted images, with confirmation on steady state to avoid peristaltic contraction artifacts; wall features on T2-weighted images (wall stratification with hyper-intense intermediate layer or homogenously hypointense wall); presence or absence of wall enhancement on contrast-enhanced T1-weighted images; enhancement pattern (early and mucosal enhancement or progressive and transmural enhancement); degree of bowel wall enhancement, assessed using a 3-point scale as follows: 0) equivalent to that of normal bowel wall; (1) mild enhancement that is greater than that of normal small bowel but markedly less than that of nearby vascular structures; (2) moderate bowel-wall enhancement but somewhat less than that of nearby vascular structures; (3) marked enhancement similar to that of nearby vascular structure [12]; the presence of intestinal complications such as ulcer (a break of the luminal surface of the bowel wall, confined to the bowel wall), pseudo-polyp (projecting lesions of granulation tissue in the luminal surface of the wall bowel due to healing from inflammation), fistula (a simple or complex tract of the bowel wall extending into the mesenteric fat connecting to other loops or structures), or stenosis (a luminal narrowing [i.e., lumen diameter < 12 mm] with dilation of the upstream bowel segment > 30 mm) [11]. When MRE examination included DWI, after identification of the anastomotic site on b0 images and confirmation on T2-weighted, b1000 images were considered for the assessment of restricted diffusion of anastomotic wall and signal intensity was graded using a 3-point scale as follows: 0) no increased diffusion restriction; (1) mild DWI signal intensity that is similar to but lower than that of lymph nodes; (2) moderate DWI signal intensity, similar from that of lymph nodes; (3) marked DWI signal intensity, higher than that of lymph nodes and the spleen [12]. Finally, apparent diffusion coefficient (ADC) value was calculated on the corresponding ADC map by drawing a region of interest (ROI) at the anastomotic wall. ADC and normalized ADC values (i.e., normalized relative to ADC of cerebrospinal fluid present on the same section) were obtained [12,13,14].

Extra-intestinal features included presence of reactive lymph nodes (short axis < 10 mm); comb sign (vasa recta hypervascularity, dilatation and tortuosity); perivisceral inflammatory infiltration; abscess (well-defined fluid collection with an enhancing wall) and inflammatory mass (ill-defined solid mesenteric inflammation without fluid component or wall) [11].

The results of the consensus reading were used for further statistical analysis, including diagnostic capabilities of MRE for the identification of normal anastomosis vs. abnormal anastomosis and identification of inflammation vs. fibrosis.

Standard of reference

The standard of reference for categorizing the status of the anastomotic sites was established by the study coordinator (P.S.) who was not involved in the reading (26 years of experience). The following data were considered as standard of reference: (i) endoscopic features, which were classified using a modified Rutgeerts endoscopic score [15, 16] for 34/62 (55%) examinations; (ii) results of histopathological analysis of biopsy specimens obtained endoscopically at the anastomotic site for 28/62 (45%) examinations; (iii) results of histopathological examination after surgery for 8/62 (13%) examinations. Results of CD activity index (CDAI) and evolution of clinical symptoms, biological variables and endoscopic features after specific treatment were obtained for all examination.

Ileocolonoscopies were performed by board-certified gastroenterologists using standardized procedures, and endoscopic findings were categorized using a modified Rutgeerts score [15]. Histopathological examinations were graded by board-certified pathologists with 20 years of experience in Crohn disease. Inflammation was considered when histopathological analysis or endoscopic examination showed ulcerations in the neo-terminal ileum (Rutgeerts score > 1) [15]. Fibrostenosis was considered when histopathological analysis and endoscopic examination showed anastomotic stricture and no visible ulcerations (Rutgeerts score = 0b) and favorable evolution after endoscopic dilatation of the anastomotic site. Normal anastomosis was considered when histopathological findings or endoscopic features showed no abnormalities (Rutgeerts score = 0a) in combination with CDAI < 150 [16]. For clarity purpose and clinical relevance, anastomotic sites containing a combination of features suggestive of inflammation and fibrosis at histopathological analysis were classified on the basis of the most prominent abnormality (i.e., inflammation vs. fibrosis) [9, 15,16,17].

Statistical analysis

Statistical analysis was performed using “R” statistical software V. 3.2.5 (R Foundation). A primary endpoint was to identify differences among patients diagnosed as either normal, with recurrent disease, or with fibrostenosis.

Descriptive statistics were calculated for all variables evaluated on MRE. Quantitative variables were expressed as means, standard deviations and ranges. Qualitative variables were expressed as raw numbers, proportions and percentages.

To identify MRE variables associated with the diagnosis of normal anastomosis and those associated with abnormal anastomosis, patients with normal anastomosis were compared with those with abnormal anastomosis. To identify MRE variables associated with the diagnosis of recurrence and those associated with the diagnosis of fibrostenotic anastomosis, patients with recurrent disease at the anastomotic site were compared with those with fibrostenotic anastomosis. Categorical (qualitative) variables were compared by the Fisher’s exact test. Differences in wall thickness and normalized ADC values were analyzed using Welch t-test. The relationships between each MRE variable and the status of the anastomotic site were tested by univariate logistic regression analysis. The odds ratio (OR) and 95% confidence intervals (CIs) for proportions were calculated. All statistical tests were two-tailed, and statistical significance was considered for p < 0.05. Sensitivity, specificity, and accuracy were estimated with their 95% CIs for all MRE criteria for the diagnosis of abnormal anastomosis, recurrence, and fibrostenosis.

Results

Patients

Using the standard of reference, 18/62 (29%) anastomoses were normal without inflammation or stenosis and 44/62 (71%) anastomoses were abnormal, including 36/62 (58%) anastomosis with inflammation and 8/62 (13%) anastomoses with fibrostenosis. Histopathological results were available for 28/62 (45%) MRE examinations and revealed pure inflammation in 13 anastomoses (13/62; 21%), mixed inflammatory and fibrotic changes in 6 anastomoses (6/62; 10%), pure fibrosis in 4 anastomoses (4/62; 6%), whereas no pathologic abnormalities in 5 anastomoses (5/62; 8%). Ileocolonoscopy results were available for 34/62 (55%) MRE examinations and showed pure inflammation in 15 anastomoses (15/62; 24%), fibrotic stenosis in 4 anastomoses (4/62; 6%) and no endoscopic abnormalities in 15 anastomoses (15/62; 24%).

Differentiation between abnormal and disease-free anastomosis

The distribution of MRE findings among the three groups of anastomoses are reported in Table 2. Table 3 shows the association between independent categorical MRE criteria and anastomotic site status, as well as the results of univariate analysis with logistic regression. Mean wall thickening was 5.3 ± 2.4 (SD) mm (range: 2–11 mm). Mean wall thickening of abnormal anastomosis was 6 ± 2.4 (SD) mm (range: 3–11 mm). The most discriminative MRE features for differentiating between normal and abnormal anastomoses were anastomotic wall thickening, anastomotic wall stratification, segmental wall enhancement, moderate wall enhancement, early and mucosal enhancement, and moderate/marked hyperintensity on DWI (p < 0.001 for all variables). Anastomotic wall thickening and segmental anastomotic wall enhancement were the two most sensitive and accurate MRE variables for the diagnosis of abnormal anastomosis with sensitivities of 82% (95% CI: 67–92%) and accuracies of 84% (95% CI: 72–92%). The sensitivities, specificities, and accuracies of all MRE variables for correct categorization of the anastomotic site as abnormal or normal are shown in Table 4.

Table 2 Distribution of categorical MR-enterography variables among three groups of patients with Crohn disease and prior ileocolic resection and suspected anastomotic abnormality
Full size table
Table 3 Association between MR-enterography variables and anastomotic site status at univariate logistic regression analysis† in 62 MR-enterography examinations
Full size table
Table 4 Sensitivity, specificity, and accuracy of MR-enterography categorical variables for the diagnosis of abnormal anastomosis (inflammation or stenofibrosis vs. normal) in 62 MR-enterography examinations
Full size table

The two observers correctly identified abnormal anastomosis and disease-free anastomosis in 40/44 (90%) and 17/18 (95%) MREs, respectively, while misdiagnosed as pathologic and normal MRE in 1/18 (5%) and 4/44 (10%) MREs, respectively, achieving sensitivity, specificity, and accuracy of 90% (40/44; 95% CI: 80–96%), 95% (17/18; 95% CI: 84–98%), and 92% (57/62; 95% CI: 82–97%), respectively, for the diagnosis of abnormal anastomosis (Fig. 2).

Fig. 2
figure 2

Histopathologically confirmed inflammatory recurrence of the ileocolic anastomosis in a 45-year-old man with Crohn disease; a T2-weighted MR image in the axial plane shows wall thickening and stratification at the anastomotic site (arrow) and of distal ileum (arrowheads), b T1-weighted gradient-recalled echo MR image in the axial plane obtained after intravenous administration of a gadolinium chelate shows marked mucosal enhancement at the anastomotic site (arrow) and of distal ileum (arrowheads), c Diffusion-weighted MR image in the axial plane reveals marked hyperintensity of the anastomosis wall (arrow) and distal ileum (arrowheads)

Full size image

Regarding DWI, among the 37 MREs with DWI, the pathologic anastomoses were correctly identified by restricted diffusion in 19/25 anastomoses (76%). Hyperintensity on DWI was identified in 4/13 (31%) normal anastomoses. Moderate/marked hyperintensity on DWI had 78% accuracy (29/37; 95% CI: 62–90%) and 100% specificity (12/12; 95% CI: 75–100%) for the diagnosis of abnormal anastomosis. Mean ADC value of the anastomotic wall was 1334 ± 213 (SD) mm2/s (range: 838–1761mm2/s). Mean ADC value of disease free-anastomosis was 1375 ± 224 (SD) mm2/s (range: 1118–1761mm2/s) and that of abnormal anastomosis ADC values was 1620 ± 212 (SD) mm2/s (range: 1118–1761mm2/s). Mean normalized ADC value was 0.47 ± 0.12 (SD) (range: 0.27–0.73). Mean normalized ADC value was 0.5 ± 0.1 (SD) (range: 0.42–0.64) for disease free-anastomosis and 0.47 ± 0.1 (SD) (range: 0.27–0.73) for abnormal anastomosis, respectively. No significant differences in ADC values and normalized ADC values were found (p = 0.34 and p = 0.52, respectively) between normal and pathologic anastomosis.

Inflammatory recurrence vs. fibrosis

The two observers classified the 44 abnormal anastomoses as follows: 32/44 (72%) as inflammatory recurrence, 8/44 (18%) as fibrostenosis, and 4/44 (10%) were misdiagnosed as disease-free anastomoses. Inflammatory recurrence and fibrostenosis were correctly identified on 30/36 (83%) and 6/8 (75%) MREs, respectively, achieving sensitivity, specificity, and accuracy of 83% (30/36; 95% CI: 68%–92%), 75% (6/8; 95% CI: 59%–86%), and 82% (32/44; 95% CI: 66%–91%), respectively, in the diagnosis of anastomotic recurrence against fibrostenosis (Fig. 3). Mean wall thickening of inflammatory recurrence was 5.8 ± 2.4 (SD) mm (range: 3–11 mm), while in fibrostenosis was 7.2 ± 2.5 (SD) mm (range: 3–11 mm). At univariate analysis, hyperintensity of the anastomotic site on DWI was the most sensitive finding for differentiating between recurrence and fibrostenosis (Table 5) and for the diagnosis of recurrent CD at the anastomotic site, achieving sensitivity of 89% (17/19; 95% CI: 67–99%) (Table 6). As a note, in one of the two above-mentioned false positive DWI, hyperintensity was secondary to collapsed bowel; moreover, no intra- or extra-intestinal signs of were present such as wall stratification or hyperintensity on T2, either comb sign, while wall enhancement was slight and transmural (Figs. 4, 5).

Fig. 3
figure 3

Histopathologically confirmed fibrostenosis of the ileocolic anastomosis in a 62-year-old woman with Crohn disease. a T2-weighted MR image in the axial plane shows homogeneous, hypointense wall thickening (arrow). b T1-weighted gradient-recalled echo MR image in the axial plane obtained after intravenous administration of a gadolinium chelate shows “en bloc” transmural enhancement (arrow). c Diffusion-weighted MR image in the axial plane reveals no hyperintensity at the anastomotic site (circle)

Full size image
Table 5 Association between MR-enterography variables and anastomotic site status at univariate logistic regression analysis† (inflammation vs. fibrostenosis) in 44 MR-enterography examinations
Full size table
Table 6 Sensitivity, specificity, and accuracy of MR-enterography categorical variables for the diagnosis of inflammation vs. stenofibrosis in 44 MR-enterography examinations
Full size table
Fig. 4
figure 4

Histopathologically confirmed inflammatory recurrence of the ileocolic anastomosis in a 35-year-old man with Crohn disease; a T2-weighted MR image in the axial plane shows wall thickening and stratification at the anastomotic site (arrow), b T1-weighted gradient-recalled echo MR image in the axial plane obtained after intravenous administration of a gadolinium chelate shows moderate transmural enhancement at the anastomotic site (arrow), c Diffusion-weighted MR image in the axial plane shows moderate hyperintensity of the anastomosis wall (arrow)

Full size image
Fig. 5
figure 5

Histopathologically confirmed fibrostenosis of the ileocolic anastomosis in a 57-year-old woman with Crohn disease. a T2-weighted MR image in the coronal plane shows hypointense wall thickening of the anastomotic wall (arrow) and the distal ileum (arrowheads). Dilatation of the upstream bowel segment is also seen (star). b T2-weighted MR image in the axial plane at the stenotic level (arrow). c T1-weighted gradient-recalled echo MR image in the coronal plane obtained after intravenous administration of a gadolinium chelate shows transmural enhancement at the anastomotic site (arrow) but stratified enhancement at the stenotic level (arrowheads). d Diffusion-weighted MR image in the axial plane reveals no hyperintensity at the stenotic level (circle)

Full size image

ADC values were 1331 ± 218 (SD) mm2/s (range: 838–1636 mm2/s) for recurrent CD and 1407 ± 191 (SD) mm2/s (range: 1227–1628 mm2/s) for fibrostenosis. Mean normalized ADC value was 0.5 ± 0.1 (SD) (range: 0.32–0.73) for inflammatory anastomosis and 0.46 ± 0.1 (SD) (range: 0.27–0.59) for fibrostenosis, respectively. No significant differences between recurrent CD and fibrostenosis were found in ADC and normalized ADC values (p = 0.96 and p = 0.51, respectively).

Discussion

The present study indicates that MRE as a reliable imaging modality in the assessment of the anastomotic status in patients with CD who previously underwent ileal or ileocolic resection with ileocolic anastomosis. Abnormal ileocolic anastomosis was identified with high accuracy (92%) by two readers in consensus. These rates are similar to those previously reported [18]. Moreover, the most discriminative features for differentiating between normal and abnormal anastomoses such as wall thickening, wall stratification, pattern of wall enhancement and hyperintensity on DWI are included in the most used imaging scores in CD [19, 20]. However, our results differ from those reported by Baillet et al. [21]. These researchers have evaluated MREs performed within the first year following surgery, founding a significant correlation between ADC values and the anastomotic status but not for morphological features either enhancement pattern [21]. Conversely, in the present study, the identification of the abnormal anastomosis relied on several findings, such as thickening, stratification and enhancement pattern of the anastomotic wall. A longer time from initial surgery (mean range 8 years) and a consequent longer history of recurrence and remission of CD may account for the discrepancies between the two studies.

In our study, interesting results come from DWI. While hyperintensity on DWI was associated with a specific anastomotic status, moderate/marked restricted diffusion was highly specific in the diagnosis of abnormal anastomosis (100%; p < 0.001). Several artifact effects such as inadequately distended loops, peristaltic contractions or spontaneous T2 shine-through of the bowel may contribute to mild signal intensity of normal bowel wall on DWI [22]. Conversely, pathologic anastomosis was likely depicted as moderate/markedly hyperintense on DWI, suggesting that, while the aforementioned artifacts may contribute only to a low signal, real restricted diffusion of the water molecules within the abnormal bowel wall produces more intense signal on DWI. These results may support the utility of DWI in addition to the morphological sequences, and to consider it when gadolinium-based contrast agents cannot be administrated [22,23,24]. Checking out the corresponding features on T2-weighted and steady state images (i.e., trueFISP and FIESTA) while evaluating the DWI intensity may be helpful in the assessment of the anastomotic status. Differently from the recent study by Strakšytė et al., ADC values were not helpful in this setting, and we partly attribute that to motion artifacts which may occur despite the use of antiperistatical agents [25].

Regarding the performances of MRE in distinguishing between inflammatory recurrence and fibrosis, we found an accuracy of 82% and a specificity of 75%. Nevertheless, these results are in line with those reported in non-previously operated patients with CD [26, 27]. The long history of inflammation and healing affecting the anastomosis over time may lead to the simultaneous presence of active inflammation superimposed upon a fibrotic substrate, this affecting the performance of MRE readings, as suggested by Gee et al., using histologic specimens as reference [27]. In our series, superimposed inflammation upon fibrosis was found in 6/28 (21%) histopathological samples, which had led to a combination of inflammatory and fibrotic features on MRE, precluding a correct differentiation in some MRE examinations. Hyperintensity on DWI showed good performances in distinguishing between inflammatory recurrence and fibrostenosis. In particular, visual assessment of DWI accurately identified inflammatory recurrence in 17/19 (89%) of MREs for which an actual recurrence was deemed present using the standard of reference. No other variables evaluated in the present study could help differentiate between the two entities, including ADC values.

Our results are different from those reported in studies performed on non-previously operated patients. Prior studies found that many features such as wall stratification, wall thickening, lymph nodes, comb-sign, stenosis of the affected loop and wall enhancement pattern were helpful in distinguish between recurrence and fibrostenosis [20, 26,27,28,29,30,31]. Several studies reported lower ADC values in fibrotic bowel wall compared to non-fibrotic bowel wall [14, 21, 26, 31, 32]. It may be possible that these differences rely on differences in standard of references, duration of disease and treatments.

A retrospective design, a small sample size, and the lack of histopathological examination in a subset of patients are the major limitations of our study. Histopathological evaluation of the anastomosis should be routinely performed to guide the treatment and to identify malignant lesions arising in stenotic loops [33]. The lack of interobserver variability assessment is another limitation of this study. On the other hand, this is the first study which evaluated the accuracy of MRE in distinguishing between recurrence and fibrostenosis of the ileocolic anastomosis in CD and, in particular, which analyzed DWI among the different imaging parameters. However, the small sample size did not allow a multivariate analysis to evaluate the potential confounding factors. Further studies with a larger sample size and histopathological correlations are required to investigate the possible role of this sequence in the differentiating between anastomosis recurrence and fibrostenosis.

In conclusion, the present study shows that MRE is a reliable imaging modality in assessing ileocolic anastomosis status with high accuracy. Several features are helpful in differentiating between normal and abnormal anastomoses such as wall thickening, wall stratification, wall enhancement pattern, and DWI hyperintensity. However, performances are lower for differentiating between inflammatory recurrence and fibrostenosis, probably due to the concomitance of inflammatory and fibrotic changes in some patients. Pattern of wall enhancement and DWI can be useful in this setting. DWI may be useful in identifying abnormal anastomosis and may have a role in distinguishing between inflammatory recurrence and fibrostenosis. Further studies are required to deeply investigate the role of DWI in the ileocolic anastomotic status of CD patients.